Home

DEVELOPMENT OF A PORTABLE OPTICAL STRAIN - K-REx

1. a Histogram of a typical speckle image b Histogram after equalization Figure 4 2 Histogram equalization 32 e Reduction of FFT spectral leakage using hanning window The phase correlation technique relies heavily on the Fourier transform to transform the image to the frequency domain However Fourier transform inherently is supposed to be applied to infinite periodic signals instead of the finite non periodic signals like a digitalized image When the Fourier transform of a finite non periodic signal is computed the resulted complex spectrum suffers from artifacts Figure 4 3 shows a speckle image and its frequency spectrum It shows a horizontal line and a vertical line across the frequency spectrum The two artifact lines are from the abrupt intensity change at the vertical boundary and horizontal boundary respectively of the speckle image This phenomenon is also called spectral leakage These artifacts will cause the software to report a false positive matching of the speckle image pairs if it is not removed These artifacts can be reduced by elementwise multiplication of the original image by a window that tapers the intensity value at the image boundary to zero Many types of windows have been proposed to reduce the effect of spectral leakage Their performances vary depending on if the signal is random or sinusoidal For a random signal like speckle hanning window has been shown to provide better frequency
2. c F 1000 w o 900 2 E 800 700 N v 2 600 2 a 500 E o o 400 S 300 At transfer of prestress a 3 days after transfer 200 6 days after transfer o s 100 d 14 days after transfer J o 28 days after transfer 0 0 20 40 60 80 100 Distance along beam in Figure 6 13 Optical surface strain measurements during the first 28 days after de tensioning The surface strain of the prestressed concrete member was further monitored for 28 days after the detensioning This is to investigate the capability of the laser speckle for long term strain measurement As shown in Figure 6 13 the laser speckle method works very well during the first month after de tensioning Note in this figure that the peak strains vary along the length of the member producing an asymmetric shape This was due to a slight horizontal eccentricity of the strand in the small trapezoidal cross section which produced bi axial bending in the member 89 6 4 2 Surface strain measurement using the fourth generation prototype A prestressed concrete beam strain measurement experiment that is similar to that described in section 6 2 1 was conducted using the fourth generation prototype of the laser speckle strain sensor The fourth generation prototype as discussed in Chapter 3 is based on the same laser speckle technology as the second generation prototype but has different hardware design dual module
3. void CStrainDlg phase_correlation_halfsize IplImage ref IplImage tpl IplImage poc int Ly Jr K float tmp int step ref gt widthStep int fft_size bs bs uchar tpl_data uchar tpl gt imageData float poc_ data float poc gt imageData load images data to FFTW input for i 0 k 0 i lt bs i fort Gg 0 5 lt bs p jt Krb 4 img2_half k float uchar tpl_data start_y i step start_x j l obtain the FFT of img2 fftwf execute fft_img2_halfsize for i 0 i lt bs bs 2 1 i res_fft_half i 0 img2_fft_half i 0 imgl_fft_half i 0 img2_fft_half i 1 imgl_fft_half i 1 res fft half i LI img2 fft half i 0 lt lt imel fr half ilLt 2 yo img2_fft_half i 1 img1_fft_half i 0 tmp sgrt pow res_fft_half i 0 2 pow res_fft_half i 1 2 res_fft_half i 0 tmp res_fft_half i 1 tmp 142 fftwf execute ifft res halfsize normalize and copy to result image for i 0 i lt fft size itr poc_data i res_half i float fft_size void CStrainDlg cvShiftDFT CvArr src_arr CvArr dst_arr CvMat tmp CvMat glstub q2stub CvMat g3stub q4stub CvMat dlstub d2stub CvMat d3stub d4stub CvMat gl g2 g3 g4 CyMats dily r d2 d3 dae CvSize size cvGetSize
4. Gl lt EnableWindow TRUE EnableWindow TRUE EnableWindow TRUE EnableWindow TRUE case 5 GetDlgItem IDC_ ROUND6 gt EnableWindow TRU break Gl lt void CStrainDlg phase_correlation IplImage ref poc int I Yip K float tmp int width ref gt width int height ref gt height int step ref gt widthStep int fft_size width height float ref data float tpl_data float poc_data f float fl loa LOa for i 0 k O i lt height for j 0 j lt width jt t imgl k 0 z ref data k imgl k 1 0 img2 k 0 tpl_data k img2 k 1 0 fftwf_execute fft_imgl fftwf_execute fft_img2 Hx Hx ex r k ref gt imageData tpl gt imageData poc gt imageData i 4 obtain the cross power spectrum for i 0 i lt height width res_fft i 0 img2_fft imgl_fft i 1 res_fft i 1 img2_fft imgl_fft i 0 tmp sgrt pow res_fft i res_fft i 0 tmp i i 0 0 r2 i imgl_fft i 0 imgl_fft i 1 IplImage tpl IplImage img2_fft i 1 pow res_fft i 1 2 141 img2_fft i 1 ges EDD tme fftwf execute ifft res for i 0 i lt fft size i poc_data i res i 0 float fft_size
5. SetCurrentDirectory defaultfolder CComboBox pCB CComboBox GetDlgItem IDC_DataSetList CStringArray folders CFileFind finder BOOL bWorking finder FindFile defaultfolder while bWorking bWorking finder FindNextFile if finder IsDirectory amp amp finder IsDots 128 folders Add finder GetFilePath now cycle through the array for int i 0 i lt folders GetSize i pCB gt InsertString 1 folders i Right folders i GetLength defaultfolder GetLength 1 int nCount pCB gt GetCount if nCount gt 0 pCB gt SetCurSel nCount 1 OnSelchangeDataSetList UpdateData FALSE return TRUE return TRUE unless you set the focus to a control void CStrainDlg OnPaint if IsIconic CPaintDC dc this device context for painting SendMessage WM_ICONERASEBKGND WPARAM dc GetSafeHdc 0 Center icon in client rectangle int cxIcon GetSystemMetrics SM_CXICON int cylcon GetSystemMetrics SM_CYICON CRect rect GetClientRect amp rect int x rect Width cxIcon 1 2 int y rect Height cylcon 1 2 Draw the icon dc DrawlIcon x y m_hIcon else CDialog OnPaint 129 HCURSOR CStrainDlg OnQueryDragIcon return HCURSOR m_hIcon void CStrainDlg OnButtonConnect CString csTemp int iCamera if 1m_bConnected iCamera 1 m_hCameraA LucamC
6. temp Open defaultfoldert datasetnamet temp txt CFile modeCreate CFile modeWrite TRUE Fil Fil le SeekToBegin le ReadString ReadMeasurementString temp WriteString ReadMeasurementString temp WriteString n while File ReadString ReadMeasurementString FALSE if ReadMeasurementString measurementname temp WriteString ReadMeasurement String temp WriteString n File Close temp Close CFile Remove default foldert datasetname index txt CFile Rename defaultfoldert datasetnamet temp txt defaultfoldert datasetnamet index txt UpdateMeasurementList BOOL CStrainDlg Measure long 1PixelSize 1 8 bits int i jJyk CString cstemp UpdateData TRUE m_Spin SetPos m_pointnumber 151 CFileFind f CString cfilenameL cfilenameR cfilenameL Empty cfilenameL Format s_ dL bmp m_strPath m_pointnumber cfilenameR Empty cfilenameR Format s_ dR bmp m_strPath m_pointnumber hCameras 0 m_hCameraA hCameras 1 zm_hCameraB if m_measurementlist GetSel 0 gt 0 for i 0 i lt CAMNUM i params i format height height params i format pixelFormat LUCAM_PF_8 params i format subSampleX 1 params i format subSampleY 1 params i format width width params i format xOffset O params i format yOffset 0
7. 1 Clicking Create New Data Set will pop up a dialog Figure 5 that allows the user to define new data set There are three input fields in the dialog Data Set Name Data Point Number and Description Alternatively the user can choose the existed data set from the dropdown list The information of the data set created or chosen will be displayed in the Data Point ff and Description textboxes in the main screen of the software Define Data Set Data Set Name Concrete beam 1 Cancel Data Point Number 15 rete beam transfer length measurement Description Figure C 5 Data Set Definition Dialog 2 Clicking Create New Measurement button will pop up a dialog Figure 6 where the user can input the name of the new measurement For example a concrete beam might need to be inspected several times during a month Before starting each round of the inspection a new measurement needs to be created 115 Input the name of new measurement initiat reading Figure D 6 Define the Name of New Measurement 3 Position the sensor on the subject surface where the user desires to make strain measurement as shown in Figure 7 Mark a half circle around the edges of the top contact points Figure D 7 Making Measurement 4 Click the Measure button to start making the readings For the initial readings the programs will wait a period before making the actual reading The durat
8. Substituting Equation 2 17 into Equation 2 19 S 0 H 0 U 0 H e e U o o 2 20 19 Observing that u X Y is the reflected laser light beam intensity from the rough object surface due to the random feature of the surface profile both the amplitude term and phase term are modulated randomly Therefore u X Y can be regarded as a random function Therefore Eguation 2 20 is simplified as S 0 H 0 H 0 0 2 21 Since h x y is symmetric we have H 0 H S 0 H 0 H 0 0 2 22 Thus s x y h x y 2 23 Substituting Equation 2 9 into Equation 2 23 we get the intensity distribution of the subjective speckle pattern J Orbea con Mh y sn nDri Ad where r x y and x yare the coordinates at the image plane J is the standard Bessel function of the first kind D is the lens pupil diameter d is the image distance A is the light wavelength The intensity distribution function of the subjective speckle pattern is plotted in Figure 2 2 According to the Rayleigh criterion Hecht 1998 the average speckle size is determined as the value of r where the value of the function s r drops to its first local minimum By setting J cr 0 the solution is found to be i DE 563 2 25 Ad which in turn gives r 1 22 Mb 2 26 D This is the average size of the speckle on the camera plane 20 Intens
9. phase_correlation reffiltered tplfiltered poc_halfR cvMinMaxLoc poc_halfR amp minvalR amp maxvalR amp minlocR amp maxlocR O if maxlocR xx ol else xx ol if maxlocR yy ol else yy ol dstart dstart dstart dstart x gt bs 2 R x tbs maxlocR x R x maxlocR x y gt bs 2 R ytbs maxlocR y R y maxlocR y x maxlocR x y maxlocR y real cvCreateMat bs bs CV_32FC1 195 im cvCreateMat bs bs CV _32FCl for i 0 k 0 i lt bs i for j 0 j lt bs j k cvmSet real i j float res_fft k 0 cvmSet im i j float res_fft k 1 upsampling in factor y 1 factor x l factor out 2 cvReleaseMat amp real cvReleaseMat amp im real cvCreateMat factor 2 factor 2 CV_32FC1 im cvCreateMat factor 2 factor 2 CV_32FC1 cvSplit out real im 0 O Compute the magnitude of the spectrum Mag sgrt Re 2 Im 2 cvPow real real 2 0 cvPow im im 2 0 cvAdd real im real NULL cvPow real real 0 5 cvReleaseMat amp in cvReleaseMat amp out cvMinMaxLoc real amp minvalR amp maxvalR amp minlocR amp maxlocR 0 cvReleaseMat amp real cvReleaseMat amp im subxR width 2 bs 2 xx 17 float maxlocR x factor subyR yy height 2 bs 2 1 float maxlocR y factor if maxvalL lt THRESHOLD maxvalR lt THRESH
10. amp im float subxL subyL subxL width 2 bs 2 xx 1 float maxlocL x factor subyL yy height 2 bs 2 1 float maxlocL y factor float subxR subyR rect cvRect oldstartR x oldstartR y bs bs cvSetImageROI refR rect cvCopy refR refblockR NULL filterproduct refblockR blockfilter reffiltered filterproduct tplR backup blockfilter tplfiltered phase_correlation reffiltered tplfiltered poc_halfR O ys cvMinMaxLoc poc_halfR amp minvalR amp maxvalR amp minlocR amp maxlocR 0 if maxlocR x gt bs 2 xx oldstartR x bs maxlocR x else xx oldstartR x maxlocR x if maxlocR y gt bs 2 yy oldstartR y bs maxlocR y else yy oldstartR y maxlocR y x maxlocR x y maxlocR y real cvCreateMat bs bs CV_32FC1 im cvCreateMat bs bs CV_32FC1 for i 0 k 0 i lt bs i 162 for j 0 j lt bs jt k cvmSet real i j fl T cvmSet im upsampling in Ly je El factor y l factor loat res_fft k 0 loat res_fft k 1 x 1l factor out 2 cvReleaseMat amp real cvReleaseMat amp im real cvCreateMat factor 2 factor 2 CV_32FC1 im cvCreateMat factor 2 factor 2 CV_32FC1 cvSplit out real im 0 O cvPow real real 2 0 cvPow im im 2 0 cvAdd real im real NULL cvPow real real 0 5 cvReleaseMat amp in
11. prototype Manufacturer or Part name retailer part 15mm Dia Unmounted Linear Glass Polarizing Filter Edmund optics NT54 925 C MOUNT 15MM THIN LENS MT Edmund optics NT54 616 C Mount Plate Beamsplitter 50R 50T VIS Edmund optics NT49 685 C MOUNT 12 5MM THICK LENS MT Edmund optics NT54 623 C MOUNT 12 5MM THIN LENS MT Edmund optics NT55 246 C MOUNT 20MM THICK LENS MT Edmund optics NT54 626 Plano Convex Lens 12 7 x 12 7 VIS 0 Coating Edmund optics NT62 561 TECHSPEC High Efficiency Anti Reflection Coated Windows Edmund optics NT48 924 C Mount Double Male Rotating Barrel Edmund optics NT53 865 Triplets lens Rolyn optics 23 0600 Mono CCD camera Lumenera Corp Lu130M Diode Laser 1112A2 High Performance Laser Module Concept 0001 103 Appendix B Specifications and SolidWorks model of the laser speckle strain sensor the fourth generation prototype e Resolution 20 microstrain e Positioning tolerance 2mm e Response time 0 2 second e Transfer length measurement time for a typical 10 feet prestressed concrete member 3 minutes e Mass 2 6 lbs 2 modules combined e Dimensions 4 x3 x2 one module Figure B 2 3D SolidWorks model of an individual model with interior view 104 Appendix C Uncertainty Analysis This section evaluates the uncertainty of the strain measurement by the laser speckle strain sensor The optical strain sensor measures the surface strain by detecting the in p
12. radio button is used to load predefined camera settings for the subject with concrete surface The Steel radio button is used to load predefined camera settings for the subject with steel surface The Apply button is used to apply the gain value and exposure time set by the user to the strain sensor D 3 2 Measurement Settings In this section the user can configure the output unit delay time for the initial reading gauge length and the desired accuracy The Output Unit dropdown listbox is used to choose the unit of the output The available options include mm inch and micro strain If mm or inch is selected the measurement output is the absolute distance change of the two illuminated points on the subject surface If micro strain is selected the output is the strain that is obtained by dividing the measured distance change by the Gauge Length The Reference Delay textbox is used to set the delay time during which the user can position the sensor appropriately for the initial readings The purpose of this item will be discussed in detail in Section 3 3 of this manual 114 The Gauge Length textbox is used to set the gauge length of the sensor i e the distance between the two illuminated points on the subject surface D 3 3 Measurement This section consists of the controls that are used to manage the application profile and conduct the measurement
13. 0 020 Optical handheld 5 0 010 4 0 000 4 pa ressu S1 2000 p 4000 Figure 6 5 Surface deflection measurement obtained by Whittemore gauge and laser speckle strain sensor 83 6 3 Comparison with an electrical resistance strain gauge during compressed concrete beam strain measurement An experiment was conducted to measure the surface strain of a concrete beam under different compressional loads by using the optical strain sensor the second generation prototype and using the electrical resistance strain gauge ESG for direct comparison An experimental setup used was that shown in Figure 6 6 A concrete beam of length 300mm and 90mm by 90mm sguare cross section was mounted on the compression test setup and the optical strain sensor was positioned alongside the tested concrete beam The beam was loaded with compressional loads ranging from 3 000 bs to 21 000 bs in 2 000 bs increments This setup was successfully used to assist in isolating the longitudinal axial strain component from the other distortions that are inherently present due to varying degrees of bending and torsion At each load level the strain between two fixed points on the beam surface was measured by the optical strain sensor An electrical resistance strain sensor was also used to measure the surface strain at the same time for comparison purposes Figure 6 6 Experiment setup of the comparison of laser speckle strain sensor and electrical resista
14. IPL_DEPT cvSize width height IPL_DEPT Gl lt GetTickCount keSynchronousSnapshots hSynchronousSnapshots rL gt imageData char ppFrames 0 rR gt imageData char ppFrames 1 185 ppFrames hCameras H_8U H_8U Yg if no_blur 1 rh Blurcheck rL rL_blurcheck 0 bl cvCopy rL cvCopy rR PlaySound MA r OURCE SND_NO pThread gt Pos LucamSavel LucamSavel Sleep 100 cvReleasel cvReleasel cvReleasel cvReleasel if LucamDi mage amp rL mage amp rR_b urcheck R_blurcheck amp amp KEINTRESOURC Blurcheck rR rR_blurcheck DEFAULT mage width height mage width height mage amp rL mage amp rR bl sabl MessageBox Fail if free pAllFra m cbPreview EnableWindow T m_cbPreview SetWindowText m_cbApply EnableWindow TRUI return true pAllFrames mes tThreadMessage WM CLOS NULL NULL E IDR_WAVE1 AfxGetResourceHandl EDIALOG NULL NULL LUCAM_PF_8 ppFrames 0 LUCAM PF_8 ppFrames 1 urcheck lurcheck led to unsetup synchronous snapshots RUI E _T Preview i GI BOOL CStrainDlg CalibrationReadingSpeckle double subxL do long lPixelsS ize l int i gk CS tring cste
15. Strain RB gt SetWindowText 25 I Strain p p p PRB CButton GetDlgItem IDC ACCURACY3 p p RB gt SetCheck BST_CHECKED CButton pCBmode CButton GetDlgItem IDC_HANDHOLD pCBmode gt SetCheck BST_CHECKED PRB CButton GetDlgItem IDC Concrete pRB gt SetCheck BST_CHECKED CStrainDlg OnConcrete CComboBox pCBunit CComboBox GetDlgItem IDC UNIT pCBunit gt SetCurSel 1 m_gl 7 125 m_duration 3 127 CEdit pEgl CEdit GetDlgItem IDC GL pEgl gt SetWindowText 8 m_bStop 0 m_Spin SetRange l 50 m_Spin SetPos l m_hCameraA NULL m_hCameraB NULL Gl m_bConnected FALSI Tj m_bPreviewing FALSE m_bSnapping FALSE m_cbPreview EnableWindow FALSE m_cbMeasure EnableWindow FALSE m CalibrateIntensity EnableWindow FALS Gl lt m_cbApply EnableWindow FALSE m_cbLock EnableWindow FALSE TCHAR szDirectory MAX PATH GetCurrentDirectory sizeof szDirectory 1 szDirectory m_strPath szDirectory m_path SetWindowText szDirectory m dlgSnap new CSnapPreviewDlg m_dlgSnap gt Create IDD DIALOG SNAP_PREVIEW defaultfolder C Program Files Strain Data if _access defaultfolder 0 1 CreateDirectory defaultfolder 0
16. already tensioned steel strands After the casting process is complete and the concrete has hardened a detensioning procedure is undertaken by cutting the strands at both ends of the concrete beam to release the tension The shrinking strands grip the concrete and keep it in a compressed state The tension force that the concrete beam will experience during service will be offset by this pre established compression thus producing a minimum chance of crack as long as loads are such that the concrete remains in compression state Naaman 1982 Typically the transfer length defined as the distance reguired to transfer the fully effective prestress force in the reinforcing strand to the concrete is used to evaluate the guality and performance of concrete members The method that is most commonly used by industry for the transfer length measurement is accomplished manually using a Whittemore gauge that has been discussed in Section 1 2 1 To measure the transfer length points are typically mounted using epoxy onto the concrete beam as shown in Figure 6 9 The Whittemore gauge is used to measure the distance change between the points which in turn gives the surface strain profile of the concrete beam A typical strain profile is a curve that varies approximately linearly from zero at the member end to a constant value at a distance from the end of the beam This distance is egual to the transfer length Figure 6 9 Metal points bonded onto th
17. bs IPL DEPTH 8U 1 tplRtemp cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 poc cvCreateImage cvSize bs bs IPL_DEPTH_32F 1 refblockL cvCreateImage cvSize bs bs IPL DEPTH 8U 1 refblockR cvCreateImage cvSize bs bs IPL DEPTH 8U 1 reffiltered cvCreateImage cvSize bs bs IPL_DEPTH_32F 1 tplfiltered cvCreateImage cvSize bs bs IPL DEPTH 32F 1 tplL_backup cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 tplR backup cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 pThread gt PostThreadMessage WM_SHOWDIALOG NULL NULL load reference image refL cvLoadImage cRefFilenameL CV_LOAD IMAGE GRAYSCALE create an image to store phase correlation result poc halfL cvCreateImage cvSize bs Eal bs IPL DEPTH 32F 1 refR cvLoadImage cRefFilenameR CV_LOAD IMAGE GRAYSCALE create an image to store phase correlation result poc_halfR cvCreateImage cvSize bs bs IPL DEPTH 32F 1 create a tplL cvCreateImag int stepsize l92 int colnum width bs stepsizetl int rownum height bs stepsizetl CvRect rect fftwf complex fft_halfL 100 100 for image template for new imge cvSize width height colindex 0 colindex lt colnum colindex IPL DEPTH 8U 1 r fft_halfR 100 100 0 gt 6 for rowindex 0 rowindex lt rownum rowindexr r rect cvRect col
18. imgl imgl_fft FE TW_FORWARD FE TW_ESTIMATE fft_img2 fftwf_plan_dft_2d bs bs img2 img2_fft FE TW_FORWARD FFTW ESTIMATE ifft_res fftwf_plan dft_2d bs bs res_fft res FFTW_BACKWARD FFTW_ESTIMATE planfile fopen if planfile NULL plan wisdom ng fftwf export_ wisdom to file planfile fclose planfile 133 w else MessageBox Unable to connect to the camera else if m_bSnapping if IDYES AfxMessageBox Currently running snapshot captures to stop MB_YESNOCANCEL Sleep 1000 else return if m_bPreviewing if LucamCameraClose m_hCameraA MessageBox Unable to disconnect to the camera Disconnect MB OK else m_bConnected FALSI Gl m_cbConnect SetWindowText _1 m_hCameraA NULL m_hCameraB NULL m_lffFormat height 0 m lffFormat width 0 m_cbPreview EnableWindow FA m_cbPreview SetWindowText _1 m_cbMeasure EnableWindow FA SE Connect SE m_ cbApply EnableWindow FALSE i Preview 134 Connect Do you wish LucamCameraClose m_hCameraB Closing application OnEditchangeDataSetList void CStrainDlg OnButtonPreview if m_bP
19. int nUnit pCBunit gt GetCursSel switch nUnit case 2 cstemp Format Distance change 5 4fmm deflection break case 0 cstemp Format Distance change 5 5finch deflection 25 4 break case l CEdit pEgl CEdit GetDlgItem IDC_GL pEgl gt UpdateData TRUE cstemp Format strain 5 lf micro Strain deflection 25 4 m_ gl 1000000 break if m duration gt 0 CDelayMessageBox mbox this mbox MessageBox cstemp m_duration TRUE CDelayMessageBox MBIcon MBICONNONE cstemp Format 5 4f deflection 164 SS AddCell cstemp m_pointnumber 1 m_measurementlist GetCurSel 1 SS Commit LucamSaveImage width height LUCAM PF_8 ppFrames 0 cfilenameL LucamSaveImage width height LUCAM PF_8 ppFrames 1 cfilenameR cvReleaseMat amp real cvReleaseMat amp im cvReleaseMat amp in cvReleaseMat amp out else LucamSavelImage width height LUCAM_PF_8 ppFrames 0 cfilenameL LucamSavelImage width height LUCAM_PF_8 ppFrames 1 cfilenameR for colindex 0 colindex lt colnum colindex for rowindex 0 rowindex lt rownum rowindex fftwf_free fft_halfL colindex rowindex fftwf_free fft_halfR colindex rowindex cvReleaseImage amp poc cvReleaseImage amp refblockL cvReleaselImage amp refblockR cv
20. params i exposure m_exposure 50 ms exposure params i gain 23 n params i gainGrn1 1 0 n params i gainGrn2 1 0 n params i gainRed 1 0 params i strobeDelay 0 0 unused params i timeout 3000 0 3000 ms params i useHwTrigger FALSE Set this to true for hardware triggered setup with daisy chaining params i useStrobe FALSE Set this to true if daisy chaining cameras params i exposureDelay 0 params i shutterType LUCAM_SHUTTER_TYPE_GLOBAL pParams i amp params i params 0 exposure m_exposureA params 1 exposure m_exposureB pAllFrames UCHAR malloc CAMNUM width height if pAllFrames NULL 152 MessageBox No memory for frames for i 0 i lt CAMNUM i ppFrames i pAllFrames i width height hSynchronousSnapshots LucamEnableSynchronousSnapshots CAMNUM hCameras pParams m_cbPreview EnableWindow FALSE m_cbPreview SetWindowText _T Preview m_cbApply EnableWindow FALSE UpdateWindow pThread gt PostThreadMessage WM_SHOWDIALOG NULL NULL CEdit pEB CEdit GetDlgItem IDC DELAY pEB gt UpdateData TRUE if m_delay lt 0 m_delay 0 for int k 0 k lt m_delay k Sleep 1000 if b_keepgoing FALSE Il b_redo TRUE break if b_keepgoing TRUE amp amp b_redo FALSE LucamTakeS ynchronousSnapshots hS ynchronousSnapshots ppFrames PlaySound MAKEINTRESOURCE IDR_WAVE1 Af
21. rowindex if maxlocRhalf x gt bs 2 else if maxlocRhalf y gt bs 2 tbs maxlocRhalf startR x colindex_store stepsiz startR x colindex_store stepsize maxlocRhalf x startR y rowindex_store stepsiz if maxvalRhalf gt THRESHOLD if maxvalLhalf gt oldLpeak maxvalLhalf gt thresholdhalf peakflag TRUE f GI oldLpeak maxval half oldRpeak maxval Rhalf oldstartL start oldstartR start cvCopy tplRtemp cvCopy tplLtemp m tStartTime if b_keepgoing if oldstartL x oldstartL x 0 TRUE L R amp amp tplR backup tplL backup tbs maxlocRhalf startR y rowindex_store stepsize maxlocRhalf y amp amp maxvalRhalf gt oldRpeak maxvalRhalf gt thresholdhalf NULL NULL GetTickCount clock lt 0 if oldstartL y lt 0 oldstartL y 0 if oldstartR x lt 0 oldstartR x 0 if oldstartR y lt 0 amp amp b_redo FALSI GI lt 160 Xi Yi oldstartR y 0 if oldstartL xtbs gt width oldstartL x width bs 1 if oldstartL y bs gt height oldstartL y height bs l oldstartR x width bs 1 if oldstartR ytbs gt height oldstartR y height bs l rect cvRect oldstartL x oldstartL y bs bs cvSetImageROI refL rect cvCopy refL refblockL NUL
22. with max temperature more than 100 F in the summer The sensor must be able to withstand harsh environments and maintain high functional performance in operation e Minimal training required The digital speckle photography DSP technique that has been reviewed in Chapter 1 was chosen to be used for the development of the optical strain sensor DSP technique has large dynamic range e g the maximum deformation or displacement that the technique can measure which makes it more robust than other optical strain measurement techniques based on laser speckle DSP technique generally has relatively lower resolution which is limited by the speckle 28 size that typically ranges at the micrometer level But the resolution is high enough for the concrete beam strain measurement application During the development of the current laser speckle strain sensor four 4 generations of successively improved designs have been manufactured and tested A prototype was fabricated for each design either on an optical breadboard for concept validation or in a portable form for use in field testing For each design improvements were made based on the knowledge learned through the testing and analysis of the prototype based on the previous generation design The four 4 generations of designs and their prototypes are described below in chronological order of their development 3 1 5 axis motion measurement system An optical system capable of measuring the
23. 1 degree It is shown that at a misalignment angle of 0 4 degree the error is 24 34 microstrain or about the same order as the nominal resolution of the strain sensor At a mislaignment angle of 1 degree the error increases to 152 microstrain Thus the effect of the rotation with respect to Z axis should not be overlooked and must be controlled or corrected Table 5 2 Error caused by the sensor misalignment Misalignment angle degree Error caused by misalignment microstrain 0 1 1 52 0 2 6 08 0 3 13 69 0 4 24 34 0 5 38 03 0 6 54 77 0 7 74 55 0 8 97 37 0 9 123 24 1 0 152 15 5 1 2 2 Misalignment between the two modules of the sensor The modular design of the sensor provides the advantages of easy fabrication and flexible gauge length adjustment but an important issue must be taken care during the calibration Since the sensor consists of two cameras ideally the x and y axis of the two cameras coordinate systems should be parallel with each other and have no orientation difference It is important because the distance change or strain can be calculated only when all the displacement vectors 66 are with respect to the same coordinate system However in practice the orientation of the two cameras in the two modules can hardly perfectly parallel due to the nature of the assembling of the device The actual relationship between the two cameras coordinate is shown in Figure 5 4 Even
24. 698 04 1057 00 1721 39 1593 70 width am 11 22 2126 19 1878 50 2101 41 2832 32 Corelalion _ 46 27 78 62 oo Power mw 0 000 o 868 4921 57894m 12 21 035 1815PM Size 320x240 Average Of Nukor E mse 2 2D oe RE CCD Profiler ESS s18Pm Figure 3 15 Saw tooth laser beam profile CCD Profiler 5 x File View Options Settings Help S 2 0 Sa A gt Jo oe 6 amp io A d gt ioo A ssf FA BS amp 2 Xx Vertical Profile Horizontal Profile 80 0 60 0 Horizontal Vertical Beam Gaussian Beam Gaussian 40 0 Centroid um 2098 65 2144 85 Beam Peak um 2261 48 2098 65 2124 81 2144 85 Width um 80 0 510 44 486 84 634 33 560 05 J20 0 Width um 5005 882 55 858 04 974 32 987 07 Width pm 11 2 1642 74 1524 90 1711 24 1754 22 Comelation A 90 23 92 46 n Power ni 1 000 0 888 4821 57834 115 1041412PM Size 320x240 Average Of NUEDE 58 Asam e 2 E2 o gv Sf nongaussia My Data My Documents 2 CCD Profiler Z Ced Notepad B eM Figure 3 16 Gaussian laser beam profile 45 The other important issue of the laser head is the collimating lens it uses Since the laser light emitting from the laser diode exhibits high divergence and astigmatism a collimating lens usually aspherical is
25. Figure 6 9 Metal points bonded onto the concrete surface eee ee eeeeeeeeeestneeeeeeenaeeeeeeenaeeeeees 86 Figure 6 10 Cross section of the pretensioned concrete MeMbe eeeeseeeeeessneeeeeeeeneeeeeees 87 Figure 6 11 Experiment setup for transfer length measurement of prestressed concrete member using the second generation prototype eeeeeeeesneececeeeeeeeeeenneeeeeceeeeeseseeaeeeeeeceeeeeeeeaaeees 88 Figure 6 12 Comparison of strain measurements immediately after de tensioning of a pre tensioned species see da a DEN FO 88 Figure 6 13 Optical surface strain measurements during the first 28 days after de tensioning 89 Figure 6 14 Experiment setup for the transfer length measurement of prestress concrete member using fourth generation prototype sa GG Ae AWA Gees 90 Figure 6 15 Concrete surface strain measurements immediately after de tensioning of a pre tensioned specimen using laser speckle strain sensort ssccccceeeeeeeeeenneeeeeeeeeeeeeenaeees 91 Figure 6 16 Laser speckle strain sensor mounted on a rail at CXT concrete cross tie plant 92 Figure 6 17 Severe abrasions to the cross tie surface at the saw cutting machine 94 Figure 6 18 Cross tie surface bonded with microscopic reflective particles 1 94 Figure 6 19 Cross tie surface strain measurement Tie 1 Side A o eee eeeeeeeeesneeeeeeenneeeeeees 95 Figure 6 20 Cross tie surface strain measur
26. Fy wag geg 48 Figure 3 19 Rosette setup for two dimensional strain measureMent ee eeeeeseeeeeeeeeneeeeeees 49 Figure 3 20 Visible markings and supporting legs used as the alignment mechanism 50 Figure 4 1 Image processing diagram eesseeeessseceeeessnceeceessseeecesssaeeeeesssaeeecesssaeeesenseaeeeeeses 51 Figure 4 2 Histag rain gua 2AL i Ola neu Rue HD dyddyn a rd Sean dm dyd Fn DUO o sea 52 Figure 4 3 A typical speckle image and its frequency SpectrulD 9999999 9999Ynnnnnnnnnnnnon 23 Figure 4 4 Hanning wind0wWsei ei Gi AG eR ate ON 54 Figure 4 5 Filtered speckle image and its frequency SPeCtruM eeeeeeeeesneeeeesesneeeeeeeeaeeeeeees 54 Figure 4 6 Pyramid scheme aen sree eee Yd DI DA DEU 56 Pieure 127 Sub imdee scheme ae we ad WE A Y WD FA AA 57 Figure 4 8 Zero padding interpolation eesesseseseseeeessretesssreessserersssretsrsereessssrersssrrereserreerseree 59 Figure 5 1 Camera image distortion sc da a AG do dea CY YO aU ydd UCD da dias 63 Figure 5 2 Image pairs for the camera distortion experiment eee eeeeeeeesseeeeeeenneeeeeeenaeeeeeees 64 Figure 5 3 Misalignment between the initial reading and the second reading 99 9 65 Figure 5 4 Misalignment between the two modules of the sensor 9 99999 999999 nnnnnnnnnnen 67 Figure 5 5 Transformation from object coordinates to camera coordinates 9999 n 68 Figure 5 6
27. IDC_LOCK OnButtonLock Z ED IDC STOP OnButtonStop w w Ww w Uu uw Z ED IDC CreateDataSet OnCreateDataset N_CBN_EDITCHANGE IDC_DataSetList OnEditchangeDataSetList N_CBN_SELCHANGE IDC_DataSetList OnSelchangeDataSetList N_BN_CLICKED IDC_BUTTON2 OnNewMeasurement N_BN_CLICKED IDC_DeleteMeasurement OnDeleteMeasurement N_LBN_SELCHANGE IDC_MeasurementList OnSelchangeMeasurementList z Ww Z Q a AN TJ ED IDC Concrete OnConcrete z tU Z Q Ei a AN zI ED IDC STEEL OnSteel 126 ON_BN_CLICKED IDC_OPENDATAFILE OnOpendatafile ON_BN_CLICKED IDC_CalibrateIntensity OnCalibrateIntensity ON_BN_CLICKED IDC_SETREFERENCE OnSetreference ON_BN_CLICKED IDC_Calibration OnCalibration AFX_MSG_MAP END MESSAGE MAP TTTTTTDTTTTTDTDTTDTTTDTTTDTDTDTDTDTTDTDTDTDTDTDTTDTDDTDTDTDTDTTTDTTDTDTDDTDTDTDTDTDTDTTTDDTDTTDTDTDTDTDTDTDTDDDDDDUD CStrainDlg message handlers BOOL CStrainDlg OnInitDialog CDialog OnInitDialog m_surfacetype 0 m_configurationtype 0 SetIcon m hIcon TRUI Gl Setlcon m hIcon FALSE CButton pRB CButton GetDlgItem IDC ACCURACY RB gt SetWindowText 5 I Strain RB CButton GetDlgItem IDC ACCURACY2 RB gt SetWindowText 10 I
28. IplImage refblockL 0 IplImage refblockR 0 IplImage reffiltered 0 IplImage tplfiltered 0 IplImage tplLtemp 0 IplImage tplRtemp 0 IplImage tplL_backup 0 IplImage tplR_backup 0 tplLtemp cvCreateImage cvSize bs bs IPL DEPTH 8U 1 tplRtemp gt cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 poc cvCreateImage cvSize bs bs IPL_DEPTH_32F 1 refblockL cvCreateImage cvSize bs bs IPL DEPTH 8U refblockR cvCreateImage cvSize bs bs IPL_DEPTH_8U reffiltered cvCreateImage cvSize bs bs IPL_DEPTH_32F tplfiltered cvCreateImage cvSize bs bs IPL_DEPTH_32F tplL_backup cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 tplR_backup cvCreateImage cvSize bs bs IPL_DEPTH_8U 1 pThread gt PostThreadMessage WM_SHOWDIALOG NULL NULL load reference image refL cvLoadImage cRefFilenameL CV_LOAD_IMAGE_GRAYSCALE poc_halfL cvCreateImage cvSize bs bs IPL_DEPTH_32F 1 refR cvLoadImage cRefFilenameR CV_LOAD_IMAGE_GRAYSCALE poc_halfR cvCreateImage cvSize bs bs IPL_DEPTH_32F tplL cvCreateImage cvSize width height IPL_DEPTH_8U 1 tpIR cvCreateImage cvSize width height IPL_DEPTH_8U I int stepsize 192 int colnum width bs stepsize 1 int rownum height bs stepsize 1 CvRect rect for colindex 0 colindex lt colnum colindex 156 for rowindex 0 rowindex lt rownum rowindex re
29. Length inch 8 Measurement Choose Working Data Set concrete 4 x Create New Data Set Data Point sq Description Point n Create New Measurement Delete Measurement Figure D 4 Strain Application Main Window D 3 1 Camera Control Before measuring with the sensor the user must configure parameters for the camera in the sensor so that the camera can capture valid speckle images for the analysis The Connect button is used to initialize the strain sensor and prepare the communication between the strain sensor module and the computer 113 The Preview button is used to let user see the images being captured in real time The two images that pop up are captured by the two modules respectively If the images are either too dark or too bright the user can use the controls Gain and Exposure that are described next to adjust accordingly Stopping the preview can be achieved by clicking the same button The Gain textbox is used to set the gain value for the cameras in the strain sensor The range of the accepted gain value is 1 to 23 in integer The Exposure textbox is used to set the time between the start of image capture and the data read out for a snapshot The unit of the value is millisecond If the sensor can t obtain bright enough speckle image with the maximum available gain value the user can increase the brightness further by increasing the Exposure time The Concrete
30. SS Commit 166 void CStrainDlg PostNcDestroy CleanUp CDialog PostNcDestroy BOOL CStrainDlg CreateHanning int M int N CvMat filter BOOL flag int d CvMat largefilter CvRect rect largefilter cvCreateMat M 2 3 N 2 3 CV_32FC1 CreateLargeHanning M 2 3 N 2 3 largefilter IplImage img img_header img cvGetImage largefilter amp img_header rect cvRect M 4 N 4 M N cvSetImageROI img rect cvCopy img filter NULL cvResetImageROI img cvReleaseMat amp largefilter return TRUE void CStrainDlg filterproduct IplImage img CvMat filter IplImage productimg int Ty pe kK float tmp CvScalar s get image properties int width img gt width int height img gt height uchar img_data uchar img gt imageData float productimg_data float productimg gt imageData for i 0 k 0 i lt height i for j 0 j lt width jt k tmp cvmGet filter i j s cvGet2D img i j 167 productimg_data k tmp s val 0 void CStrainDlg converttofloat IplImage img CvMat filter IplImage productimg int shiftx int shifty int Tye ee kK float tmp get image properties int width img gt width int height img gt height uchar img_data uchar img gt imageData float productimg_data float productimg gt imageData for i 0 k 0 i lt height
31. The pixel shifts of these 3 regions are listed in Table 5 1 It is shown that maximum standard deviation of the pixel displacement among the 5 image pairs is 0 06 pixel corresponding to 1 4 microstrain for the 8 inches gauge length configuration Thus it can be concluded that the image distortion has an insignificant effect on the measurement of the optical strain sensor Observing this the distortion effect is neglected in the camera calibration model used in the later analysis associated with this chapter 63 Initial image Post 0 2mm translation image Figure 5 2 Image pairs for the camera distortion experiment Table 5 1 Data of the camera distortion experiment object displacement mm 020 0 20 0 20 0 20 0 20 Ax pixel 5921 5 96 5 96 5 92 5 96 Axxpixel 5 96 6 00 5 96 6 00 5 92 Axa pixel 6 04 5 96 5 92 5 92 5 96 standard deviation 06 02 02 05 02 64 5 1 2 Error due to misalignment As shown in Figure 3 5 the surface strain between point A and point B was measured AB AA by g where L is the gauge length 203 2 mm 8 inches for the current setup and AA AB are the surface displacements at point A and point B The strain calculation scheme described above is suitable for the ideal situation that the sensor is aligned perfectly However in the real situation significant errors could occur either due to the sensor misalignment
32. a metal or plastic housing In choosing the laser head for the sensor there are several factors to consider Ideally the laser beam intensity profile is a Gaussian shape due to several beneficial properties that come with it A Gaussian profile beam always remains Gaussian along its path of propagation through the optical system Hecht 1998 Thus the propagation of Gaussian beams through an optical system can be treated almost as simply as geometric optics However a low guality laser head 44 might produce a saw tooth type of laser beam profile Figure 3 15 It is hard to predict the behavior of a non Gaussian shape laser beam when it goes through the lens In addition the inconsistency of the laser beam profile will introduce artificial noise to the speckle image captured by the camera and this interferes with repeatability of the strain measurement Figure 3 16 shows the beam profile from the high guality laser module used in the strain sensor whose beam profile is close to Gaussian shape File View Options Settings Help Freeze Mode Aa 2 9 e AIIM as c elo Fl seTiooe A efi es Si e 2 Vertical Profile lt M 100 oo Ee 0 4089 4811um 2 Horizontal Profile 100 Horizontal Vertical Beam Gaussian Beam Gaussian 40 0 Centroid um 3057 52 2144 85 Beam Peak um 2496 67 3057 52 2124 81 2144 85 Width um 60 0 206 56 599 73 1297 05 904 25 l2oo Width um 50 0
33. amp minlocLtemp amp maxlocLhalftemp 0 if maxvalLhalftemp gt maxvalLhalf maxvalLhalf maxvalLhalftemp 158 maxlocLhalf maxlocLhalftemp colindex_store colindex rowindex_store rowindex if maxlocLhalf x gt bs 2 startL x colindex_store stepsize bs ma else xlocLhalf x startL x colindex_store stepsize maxlocLhalf x if maxlocLhalf y gt bs 2 startL y rowindex_store stepsize bs maxlocLhalf y else startL y rowindex_store stepsize maxlocLhalf y if maxvalLhalf gt THRESHOLD tplR gt imageData char ppFr rect cvRect width 2 bs 2 ames 1 height 2 bs 2 bs bs cvSetImageROI tplR rect cvCopy tplR tplRtemp NULL filterproduct tplRtemp bloc memcpy imgl_half float fftwf_execute fft_imgl_hal maxvalRhalf 0 colindex_store 0 rowindex_store 0 for colindex 0 colindex lt co for rowindex 0 rowindex lt row phase_correlation_block ff poc_halfR cvMinMaxLoc poc_halfR amp mi amp maxlocRhalftemp O if maxvalRhalftemp gt maxvalR maxvalRhalf maxvalRhalftemp maxlocRhalf maxlocRhalftemp kfilter tplfiltered tplfiltered gt imageData bs bs sizeof float fsize lnum colindex t num rowindex t_halfR colindex rowindex imgl_fft_half nvalRtemp amp maxvalRhalftemp amp minlocRtemp half f 159 colindex_store rowindex_store colindex
34. associated with surface motion during the exposure period of the camera e Fast image capturing speed The optical strain sensor is designed for both manual operation and on track operation For both operation configurations it is inevitable that the sensor undergoes movement or suffers vibration during operation A fast image capturing speed will help minimize the blur of the image 47 Adjustable shutter A versatile strain sensor should be able to measure strain on various material surfaces from concrete and metal to fiber glass In addition the sensor should also function in different ambient light environments including daylight indoor and night condition The adjustable shutter feature enables the sensor to adjust the intensity of the captured image according to surface reflectivity and illumination condition Expansibility Although the current sensor design consists of two modules it should be possible to build a multiple module system to capture the strain in multiple locations at the same time as suggested in Figure 3 18 Furthermore instead of only measuring the strain in one direction a rosette setup could be implemented using four individual modules to measure the three independent components of surface strain Figure 3 19 The expansion of the system in this manner could reguire that multiple cameras be attached to one computer and capture the images at the same time Specimen Surface a A TC Sensor Sensor Se
35. between before and after deformation readings or due to the misalignment between the two cameras in the sensor 5 1 2 1 Misalignment between the initial reading and the second reading During the operation of the sensor when taking the second reading it is inevitable that the sensor will be misaligned to some extent with respect to the sensor position during the initial reading This is mainly represented by a rotation about Z axis the axis perpendicular to the object surface as shown in Figure 5 3 A B A IN B gt lt gt a AB Figure 5 3 Misalignment between the initial reading and the second reading Suppose point A and point B represent the measurement locations for the two modules of the sensor for during initial reading Assume the specimen surface does not undergo any deformation and assume the sensor is misaligned by an angle of 6 relative to the initial reading 65 setup When taking the second reading the measurement locations of the two modules of the sensor will move to point A and point B The sensor will report a non zero strain reading ae AB AA _ L 1 cosg L L 1 cos which is actually an error introduced by the misalignment of the sensor between the initial reading and the second reading The error is essentially independent of the gauge length and primarily dependent on the misalignment angle Table 5 2 shows the measurement error caused by the misalignment angle from angle 0 1 degree to
36. by tape strip spacing and usually low 1 2 6 Laser speckle strain measurement Speckle is generated by illuminating a rough surface with coherent light as shown in Figure 1 7 The random reflected waves interfere with each other resulting in a grainy image as shown Figure 1 8 The speckle pattern could be thought of as a fingerprint of the illuminated area in the sense that the speckle pattern produced by every surface area is unigue Furthermore when the surface area undergoes movement or deformation the speckle pattern in the image plane will also move or deform accordingly Most optical speckle methods for in plane displacement or deformation measurements are based on the same principle That is the grainy speckle pattern image is recorded before the 9 surface is deformed and after the surface deformation The deformation or displacement components can then be extracted by comparing the speckle patterns before and after a surface deformation This is typically done statistically using a cross correlation technigue to measure the speckle displacement See Section 2 4 for detailed discussion of cross correlation technigue Coherent Light Source In Phase Bright Speckle y a Out of Phase NN Dark Speckle Surface Film Plane or Detector Figure 1 7 Speckle Generation Principle There exist two basic categories of speckle technigue for surface strain measurement electronic speckle pattern interferometry ESPI and digita
37. colindex_s rowindex_s for colindex 0 colindex lt coln for rowindex 0 rowindex lt rownum rowindex phase_correlation_block fft_halfR colindex rowindex poc_halfR cvMinMaxLoc poc_hal ute ff f 0 tore 0 tore 0 amp maxlocRhalftemp O if maxvalRhalftemp gt maxvalRhalf lfR amp minvalRtemp Rtemp blockfilter tplfiltered t_imgl_halfsize um colindex t 192 X y h 2 bs 2 height 2 bs 2 bs bs float tplfiltered gt imageData bs bs sizeof amp maxvalRhalftemp float imgl_fft_half amp minlocRtemp Ye maxvalRhalf maxval LRhalftemp maxlocRhalf maxlocRhalftemp colindex_store colindex rowindex_store rowindex if maxlocRhalf x gt bs 2 else if maxvalRhal startR x colindex_store stepsiz startR x colindex_s if maxlocRhalf y gt bs 2 startR y rowindex_s startR y rowindex_s tore stepsiz f gt THRESHOLD if maxval Lhal f gt oldLpeak amp amp maxvalLhalf gt thres peakflag TRUE r Gl hal oldLpeak maxval oldRpeak maxval Rhal oldstartL start oldstartR start cvCopy tplRtemp cvCopy tplLtemp m_tStartTime if b_keepgoing TRUE amp amp b_redo FALS L R GetTickCount clock if oldstartL x lt 0 oldstar
38. cvReleaseMat amp out cvMinMaxLoc real amp minvalR amp maxvalR amp minlocR amp maxlocR 0 cvReleaseMat amp real cvReleaseMat amp im subxR width 2 bs 2 xx 1 float maxlocR x factor subyR yy height 2 bs 2 1 float maxlocR y factor if maxvalL lt THRESHOLD maxvalR lt THRESHOLD AfxMessageBox might be fake peak measure again float deflection float secondorder deflection pl subxLtfp2 subyLrp3 subxRrp4 subyR SECO p4 subxRrp3 subyR 2 25 4 m gl CStdioFile DataFile cstemp Format f f S5 4f 5 4f oldRpeak subxL s ndorder p2 subxLrpl subyL p4 subxRrp3 subyR p2 subxLrpl subyL 5 4f 5 4 f f oldLpeak ubyL subxR subyR deflection secondorder txt CFile modeWrite Creat TRU if DataFile Open c data_fitting CFile modeNoTruncate CFile mod DataFile SeekToEnd DataFile WriteString cstemp ma F E 163 DataFile WriteString n DataFile Close deflection deflectionrsecondorder CSpreadSheet SS defaultfolderr AWW rdatasetnamer xls Sheetl SS BeginTransaction if ref index 0 SS ReadCell cstemp m_pointnumber 1 ref_index 1 float ref_deflection ref_deflection atof LPCSTR cstemp deflection deflectiontref_deflection CComboBox pCBunit CComboBox GetDlgItem IDC UNIT pCBunit gt UpdateData TRUE
39. design vs single module design and is more accurate and flexible The pretensioned members were each 9 6 long with a rectangular cross section as shown in Figure 6 14 In order to facilitate the laser speckle measurements three small 14 inch diameter inserts were cast into the pre tensioned concrete members to allow an aluminum rail to sit on the top of the member surface The sensor was then installed on the rail and was able to traverse freely on it Figure 6 14 Experiment setup for the transfer length measurement of prestress concrete member using fourth generation prototype Surface strain measurements were obtained using the fourth generation laser speckle strain device The results from this beam test are shown graphically in Figure 6 15 It is clear 90 that the Laser Speckle Strain sensor results show scatter below 10 microstrain much less than that of the similar experiment described in Section 6 4 1 in which the second generation laser speckle strain device was used This is realized because of several improvements in the design of the fourth prototype over the previous versions including the dual module design less sensitivity to thermal expansion effect and better camera hardware as discussed in Chapter 3 250 7 200 150 100 Concrete surface strain measured by LSI microstrain o 5 10 15 20 25 30 Distance along beam in Figure 6 15 Concrete surface strain measurements immediately after de ten
40. distance between the lens and the object plane is d and the distance between the 16 camera plane and the lens is d For surface deformation measurement the speckle images are recorded twice by the camera before and after the surface deformation Object plane Lens Camera plane Figure 2 1 Imaging system of recording the speckle pattern The amplitude response of a point source at coordinate 0 0 on the object plane XY which is also the impluse response function of the optical imaging system is defined as Goodman 1996 h x y j P Ad x Ad y expl j2z xx r yy dx dy 2 1 where P Ad x Ad y is the pupil function 1 P Ad x Ad y 2Ad 2 2 0 otherwise where D is the diameter of the lens pupil and x y is the coordinate on the lens plane Denoting p x y we have x pcos0 y psin 2 3 Further define r 4 x y x rcos y rsing 2 4 17 Substituting Eguation 2 3 and Eguation 2 4 into Eguation 2 1 and implementing the polar integral A 22d 27 h x y j expl j2xprcos 0 pd0dp 2 5 0 0 Since the Bessel function of the first kind is defined as J u L expl ju cos 0 ld0 2 6 2m we have P 22d h x y 2z J 2mpr pdp OF Further using the integral identity of the Bessel function of the first kind fur docu du uJ u 2 8 0 Eguation 2 7 can be simplified to be J 2zDr Ad HY VED cad 2 9 Since the
41. double amp min CvPoint amp location void MaxLoc Ipllmage image double amp max CvPoint amp location BOOL Blurcheck Ipllmage img1 IplImage img2 BOOL oldsensor void AdaptiveHist IplImage input Ipllmage output void PreProcess IplImage input Ipllmage output void Subpixel IplImage in double x double y int factor void ShowImage Ipllmage img void filterproduct IplImage img CvMat filter Ipllmage productimg void INTfilter Ipllmage img CvMat filter Ipllmage productimg int shiftx int shifty void converttofloat Ipllmage img CvMat filter Ipllmage productimg int shiftx int shifty BOOL Measure CString measurementname CString datasetname CStdioFile File int datapointnumber CStrainDlg CWnd pParent NULL standard constructor CSnapPreviewDlg m_dlgSnap 120 Dialog Data AFX_DATA CStrainDlg enum IDD IDD_STRAIN_DIALOG CStatic m_refname CColorListBox m_measurementlist CButton m_CalibrateIntensity CSpinButtonCtrl m_Spin CButton m_cbRound CStatic m_path CButton m_cbLock CButton m_cbApply float m_exposure float m_exposureA float m_exposureB float m_gain CButton m_cbConnect CButton m_cbPreview CButton m_cbMeasure int m round CString m_DataSetList int m_accuracy int m_delay double m_gl int m_surfacetype BOOL m invert int m_configurationtype int m_duration AFX_DATA ClassWizard generated virtual
42. eae 3 Figure 1 2 Prestressed concrete with metal points mounted on the surface 4 Figure 1 3 Electrical resistance strain Eatlo eu 2 212 auon da godd oa dadol gedeigaloede 4 Figure I Fiber Bragg Gratings nu A ee A a Ca eee eed eee dai 7 Figure 1 5 Transmission and reflection of Fiber Bragg Grating Merzbacher 1996 7 Figure 1 6 Video extensometer confieuration isie ui Gei a COG DL eee 9 Figure 1 7 Speckle Generation PrincIpl amp s3 3 1 11 uc de cadseidstaaatolas coteniuaiteclassdiechieresedasalgewacedee 10 Figure 1 8 Speckle Pattern ii s s ccsisdscicsivedicceted sa ideuee jaciesed ssid niaii eaii dei aietdaieedeed 10 Figure 1 9 Microstar Strain gauge ui eie ae dyg Ga YF gd embeds FU gd geg Fd ng Bn F a8 11 Figure 1 10 ME 53 laser speckle extensometer 00 0 eeeeseeceessneceeeessneeeceessneeeeeessaeeesessnaeeeseses 13 Figure 2 1 Imaging system of recording the speckle pattern ee eeeeeessneeeeeessneeeeeeenaeeeeeees 17 Figure 2 2 Plot of speckle intensity distribution function eee eee eeseeeeeeesneeeeeeenaeeeeesesaeeeeeee 21 Figure 2 3 Tilt Only Plane and Translation Only Plane eeeeseeeeeeeeneeeeeeesaeeeeeeenaeeeeeees 22 Figure 3 1 5 Axis Measurement Imaging System ecccccccccseeesssseeeccceeeeeeeesessneacceeeeeseeeeees 30 Figure 3 2 Breadboard prototype for the 5 Axis motion measurement system 30 Figure 3 3 Single module design grain Nid dn in RF FN ANN inks NI
43. extensometer ME53 33 A new calibration method called Auto Calibration is proposed It does not reguire any specific eguipment and can be done very fast by the end user Suppose the coordinate system OXY in Figure 5 8 is an arbitrary coordinate system attached to the sensor Now the orientation differences of the two camera coordinates with respect to this arbitrary coordinate are and whose values are unknown Assume that camera A detects a displacement vector x y between the reference image and the post deformation image Similarly suppose camera B detects a displacement vector of x y Note that xX YX Y are in the unit of pixels The displacement vectors x y and x y can be converted to the displacement vectors in physical unit with respect to the arbitrary coordinate system OXY using Eguations 4 20 and Eguation 4 21 as described below Expanding Eguation 4 20 and Eguation 4 21 into the algebra form yield X x Bi y Y px y X 70 X D y gt Y B x tr 0 y 4 25 The distance change of the two observation spots on the object surface can be expressed as 75 A Pal 0 0 L X X FE Y L 2 EE n T eee a L 4 26 L X X 2 T L LXER L X X Fo dM Now by Taylor expansion ys 2 ahead by E 4 27 2 L X X 2 _yy2 Since eh a Y L X X L Y Yy A xX X 4 28 d 2 1 2L Expanding this equation by using the equation
44. function overrides IH AFX_VIRTUAL CStrainDlg protected virtual void DoDataExchange CDataExchange pDX virtual void PostNcDestroy AFX_VIRTUAL Implementation DDX DDV support 121 protected int ref index void CreateLargeHanning int M int N CvMat filter void phase_correlation_block fftwf_complex fftl fftwf_complex fft2 IplImage poc double thresholdhalf double THRESHOLD double concrete_threshold1 concrete_threshold2 steel_threshold1 steel_threshold2 int start_x start_y float deflection CWinThread pThread void fftwcopy fftwf_complex source fftwf_complex target int size BOOL m_bStop void cvShiftDFT CvArr src_arr CvArr dst_arr int height half height int width half_width fftwf_plan ifft_res fftwf_plan fft_img2 fftwf_plan fft imgl fftwf_plan ifft_res_halfsize fftwf_plan fft_img2 halfsize fftwf_plan fft_img1_halfsize fftwf_complex res_fft fftwf_complex img2_fft fftwf_complex imgl_fft fftwf complex res fftwf complex img2 fftwf_complex imgl double p1 p2 p3 p4 double gl g2 93 g4 95 double gbbl gbb2 gbb3 gbb4 gbb5 fftwf_complex res_fft_half fftwf_complex img2_fft_half fftwf_complex img1_fft_half img1_fft_halfL img1_fft_halfR float res_half float img2_ half float img1_half void phase_correlation IplImage ref IplImage tpl Ipllmage poc 122 void phase_correlation_halfsize Ipllmage ref Ipllmage tpl Ip
45. have confirmed this behaviour In the range of 6mm variation of object distance the systematic sensitivity has no significant variation 23 170 168 166 Sensitivity pixels mm 3 a a a a a a a 158 156 154 t t t t t t t 97 98 99 100 101 102 103 1 105 object distance mm Figure 2 5 Invariance of system Sensitivity Zhao 2006 2 4 Digital correlation technique As discussed previously the shift of speckle image captured by the camera is proportional to the translation of the object surface under a specific imaging setup The issue of detecting the surface motion is thus actually one of evaluating the relative shift of the speckle image pairs taken before and after the surface deformation This is done by using the digital correlation technique Suppose we have two speckle images of the object surface taken before and after the surface deformation The traditional cross correlation function is defined by Corr x y yD LL xxi y j 2 27 i l j l By varying the values of x and y the maximum value of the correlation function Corr x y can be found and its coordinates give the relative components of the image displacement The disadvantage of the function above is that it is subjective to changes in image intensity amplitude generally caused by change in lighting conditions across the image recording sequence which are very likely to happen during a typical concret
46. if it is possible to conduct the installation of the cameras carefully and achieve a minimum orientation difference between the two coordinate systems considering the fact that a 0 4 degree misalignment angle introduces as much as 24 microstrain error as described in the previous analysis the coordinate orientation difference between the two cameras of the sensor causes about the same order of error that should not be overlooked Therefore to be able to calculate the displacement change or the strain accurately the orientation difference between the two camera coordinate systems must be taken care of in other word the surface displacements detected by the two cameras must be converted into the same coordinate Y Yb Ya Xb j 1 Ob a Oa Xa Camera coordinate of the left module Camera coordinate of the rightmodule Figure 5 4 Misalignment between the two modules of the sensor 5 2 Homography projection To correct the misalignment errors that were discussed above various information related to the camera orientation must be collected during the calibration procedure For this purpose the homography projection technigue is used For an image of the object surface that is taken by the camera the pose of the object relative to the camera coordinate system can be described using rotation and translation matrix 67 Object coordinate Camera coordinate Figure 5 5 Transformation from object coordinates to camera coordinates A rotati
47. image CvPoint location float data int step CvSize size int x y i J int iNumPoints 1 double dMax 1 cvGetRawData image uchar amp data amp step amp size step sizeof data 0 for i O i lt iNumPoints i dMax i 0 for y O y lt size height y data step for x 0 x lt size width x for i O i lt iNumPoints i if data x gt dMax i 177 for j iNumPoints 1 j gt i j dMax j dMax j 1 location j location j 1 dMax i double data x location i x x location i y y break void CStrainDlg MultipleMinLoc IplImage image CvPoint location float data int step CvSize size int x y i J int iNumPoints 1 double dMin 1 cvGetRawData image uchar amp data amp step amp size step sizeof data 0 for i O i lt iNumPoints i dMin i 1 for y O y lt size height y data step for x 0 x lt size width x for i O i lt iNumPoints i if data x lt dMin i for j iNumPoints 1 j gt i j 178 dMin j dMin j 1 location j location j 1 dMin i double data x location i x x location i y y break void CStrainDlg OnCalibration double subxL subyL subxR subyR CString cstemp int N 15 double baseb 0 004447
48. itself whenever it shifts by an integer multiple of 27 phase There is no easy way to determine the phase change uniquely Another limitation is in the upper bound of the deformation that can be measured due to the limited number of visible fringes on the detector typically a CCD array C Joenathan 1998 Figure 1 9 Microstar Strain gauge Fig 1 9 shows a commercial full field strain sensor named Microstar based on ESPI technique It has the ability to automatically analyze the geometry and the deformation of the surface area of interest within 100nm resolution Due to its miniature size the sensor can be easily attached to the components during testing and has received a wide acceptance in 11 automobile and aerospace industries L X Yang 1999 R Wegner 1999 The Microstar strain sensor requires stringent alignment with the specimen surface and must be fixed onto the specimen surface throughout the measurement process to prevent rigid relative movement This drawback makes it impractical to be used for strain measurement for civil engineering structure where either long term monitoring of the surface strain is required or enormous shock happens to the subject such that the sensor must be removed to avoid damage For instance a prestressed concrete beam is subjected to a nominal 30 000Ib sudden force during the detensioning of the steel strands and this may damage the sensor if left on the concrete surface The DSP technigue
49. laser head that emits laser light of 635 nanometers wavelength which is then expanded by lens L1 12 7mm focal length and lens L2 35mm focal length into a collimated laser light of 10mm diameter The expanded laser light is then directed to the specimen surface by a beam splitter B1 The reflected waves are imaged at a magnification of unity by the lens L2 onto a CCD image sensor of 1392x1024 pixels whose output signal is then sent to a PC for data processing The image sensor is a monochrome Lu130M CCD camera powered by 5v DC with 4 Watts power consumption The whole optical system for individual module was rigidly mounted on single metal base to keep the relative position of the optics components fixed The two modules were then rigidly attached together to Keep the gauge length of the sensor constant 36 Specimen Surface Figure 3 8 Dual module design It is notable that the lens L2 serves two purposes in the system It collimates the laser light and also images the reflected speckle pattern to the camera Since the imaging system is a conjugate configuration a double convex lens is required However regular convex lenses with spherical surface suffer spherical aberration which distorts the speckle image and introduces measurement error The lens actually used in the optical strain sensor is a triplet lens that consists of three single lenses one convex lens and two concave lenses The positive aberration from the convex lens and the
50. method demonstration Image displacement of camera A Ma LU Ot camera B Reading x1 pixel yl pixel x2 pixel y2 pixel 1 31 08 63 16 31 96 23 92 2 2 08 39 20 2 32 5 00 3 85 12 134 80 87 00 185 68 4 29 00 32 92 28 08 65 36 5 66 72 19 16 66 20 106 04 6 19 08 58 28 17 20 134 84 7 14 04 125 92 15 88 36 36 8 37 92 128 44 39 08 26 96 9 87 04 44 16 87 60 24 00 10 215 84 97 44 213 92 38 24 11 18 00 110 40 18 36 153 08 12 46 88 86 72 45 88 5 96 13 42 04 55 92 42 32 45 96 14 63 72 139 00 64 88 7 00 The N displacement vectors obtained by the above procedure are denoted as xa yn u yn x2 Yz 2 Yo w Yin ON yn Since throughout the experiment the specimen surface is free of deformation the relative distance change A are all zero for all the N observations Therefore by Equation 4 30 we have 77 Mo Y2 Xn Xn B 0 4 31 a Xw Ow Xn Yow LBs A Fir Ay Ta a Denoting A Xiz z Xa Ya and h Bi a N YN N Y y gives Ah 0 4 32 The trivial solution of the homogenous system of linear equations 4 32 is zero which is useless And since the row number N is larger than the column number 4 there is no exact non zero solution Instead the approach is to find the least square solution for the homogenous system of linear equations That is to find h that minimize Ah subject to n Using Singular V
51. negative aberration from the concave lens are cancelled out This setup dramatically reduces the distortion of the imaging system A schematic of the dual module design with current nominal dimensions labeled is shown in Figure 3 9 37 70mm CCD camera Figure 3 9 Schematic of the dual module design with dimensions labeled Since the basic methodology of single module design and dual module design are similar it would appear that strain sensors based on either design should be able to measure the strain with comparable resolution However in terms of the practical functionality and performance the dual module design is more preferable for the following reasonings e Assembly Unlike the single module design that requires tedious adjustment of multiple mirrors during the assembly process the dual module design does not require any mirrors so it is almost adjustment free e Gauge length With the dual module design the adjustment of the gauge length is very easy All the user needs to do is to unscrew the two modules from the rods or bars that connect the two modules together adjust the distance between the two modules to achieve the desired gauge length and then re tighten the screws to fix the locations of the two modules With the auto calibration feature that will be discussed in Chapter 5 the calibration of the system parameters takes less than five minutes and can be done in the field without the need for special additi
52. of the 5 axis motion measurement system is shown in Figure 3 1 A prototype of this optical system was also built on breadboard as shown Figure 3 2 Experiments were conducted to confirm that the system is able to separately and accurately resolve 5 axis motion X Y tilt yaw and roll of the specimen with the expected insensitivity to out of plane displacement Zhao 2006 z cib Pitch amp X Roll translation Diode Laser Beam Expander Polarization Beam Splitter ADORNA Speckle Pattem Images Mirror Beam Spitter CCD Camera M2 Figure 3 2 Breadboard prototype for the 5 Axis motion measurement system 30 3 2 Single module design Based on the 5 axis breadboard motion measurement system a portable laser speckle strain sensor was designed and fabricated utilizing one laser head and one camera Since only the in plane displacement components X Y displacements are of interest in the measurement of surface strain an optical strain sensor was built to measure the in plane displacement components of two nearby surface points on the object surface by detecting the speckle shifts at the corresponding translation only planes only The configuration of the imaging system makes the sensor insensitive to any surface motion other than the in plane displacement A schematic diagram of the sensor is shown in Figure 3 3 The sensor is an integration of two identical displacement measurement systems tha
53. of the gauge length easy expansion to rosette 2D strain measurement and high accuracy Extensive testing has been conducted in the laboratory environment for validation of the sensor s capability for concrete surface strain measurement The experimental results from the laboratory testing have shown that the measurement precision of this new laser speckle strain measurement technigue can easily achieve 20 microstrain Furthermore the laser speckle strain sensor was applied to the transfer length measurement of typical prestressed concrete beams for both short term and long term monitoring The measurement of transfer length by the sensor was unprecedented since it appears that it was the first time that laser speckle technigue was applied to prestressed concrete inspection and particularly for use in transfer length measurement In the following field application of the laser speckle strain sensor in the CXT rail cross tie plant the technigue reached 50 microstrain resolution comparable to what could be obtained using mechanical gauge technology It was also demonstrated that the technigue was able to withstand extremely harsh manufacturing environment The accuracy and robustness of the device presents great potential for various civil engineering applications such as crack propagation monitoring and bridge health monitoring It can also be used to measure the surface strain of materials other than concrete These include any materials with rough
54. optical imaging system is shift invariant meaning that the shifting of the input in some direction shifts the output by the same distance and direction for a point source location at the coordinate X Y on the object plane the amplitude response at coordinate x y on the camera plane is h x y X Y x X y Y 2 10 Since the optical imaging system is a linear system the complex amplitude at coordinate x y on the camera plane is the superposition of amplitude response of all the point light sources X Y where eo lt X lt 0 lt Y lt 0 Assuming u X Y f X Y expl j2z0 X Y 2 11 18 The convolution of the object intensity with the point spread function is u x y f h x X y Y u X Y dXdY 2 12 Denoting the Fourier transform ofu x y u x y and h x y U o F u X Y 2 13 U F u x y 2 14 H F h x y P Ad o Ad o 2 15 where 1 ON ON lt c2 P Ad o Ad 2Ad 2 16 t 0 otherwise The convolution form in Eguation 2 12 can be expressed in freguency domain as U H o 0 U o 0 2 17 The intensity at point coordinate x y on the camera plane is u x y u x y u x y 2 18 s x y where u x y is the conjugate of u x y Therefore the Fourier transform of s x y can be represented as the convolution of U x y and U x y S F s x y F u x yu x y 2 19 U U
55. ppFile SeekToBegin int cc 0 CString tempString while ppFile ReadString tempString FALSE fvar float atof tempString CC H switch cc case 1 THRESHOLD fvar break case 2 thresholdhalf fvar break case 3 THRESHOLD_glass fvar break case 4 pl fvar break case 5 p2 fvar break case 6 p3 fvar break case 7 p4 fvar break case 8 gl fvar break case 9 g2 fvar break case 10 g3 fvar break case 11 g4 fvar break case 12 g5 fvar break 182 ppFile SeekToBegin CString cstemp_thres cstemp_speckle cstemp_ paint cstemp_thres Format 5 4f n 5 4f n 5 4f n THR _glass ESHOLD thresholdhalf THRESHOLD cstemp_speckle Format 8 8f n 8 8f n 8 8f n pl1 p2 p3 p4 cstemp_paint Format 8 8f n 8 8f n 8 8f n S8 8f n 8 8f n gl1 g2 93 g94 0 switch m_surfacetype case 0 cstemp_speckle Format 8 8f n 8 8f n 8 8f n 8 8f n cvmGet b 0 1 cvmGet b 0 2 break case l baseb cvmGet b 0 0 cstemp_paint Format 8 8f n 8 8f n 8 8f n S8 8f n 8 8f n baseb cvmGet b 0 0 break cstemp Empty cvmGet b 0 1 cvmGet b 0 2 0 cstemp cstemp_thresrcstemp_specklercstemp_ paint ppFile WriteString cstemp ppFile Close AfxMessageBox Calibration Done cvReleaseMat amp X cvReleaseMat amp Y cvReleaseMat amp b cvReleaseMat amp
56. resolution than other windows Harris 1978 a a typical speckle image b Complex spectrum of the speckle image Figure 4 3 A typical speckle image and its frequency spectrum The hanning window is defined as hann x 0 5 cos 27 5 0 lt x lt M 4 1 where M is the window size 53 A 2D hanning window is defined as hann x y 0 25 1 cos 2z FU cos 2z D 0 lt x lt M 0 lt y lt N 4 2 where M N are the 2d window size Figure 4 4 Hanning window Figure 4 4 shows the profile of the hanning window which is applied to the raw speckle image before it is transformed to the frequency domain The filtered speckle image whose boundary abruption is greatly tapered and the Fourier transform pattern of the filtered speckle image are shown in Figure 4 5 It is notable that the artifacts are not visible any more a Filtered speckle image b Complex spectrum of the filtered using hanning window speckle image Figure 4 5 Filtered speckle image and its frequency spectrum 54 4 2 Digital correlation procedure The theory of digital correlation has been described in Chapter 3 This section will focus on how it is implemented e Pyramid scheme The correlation procedure is very computationally expensive Since most of the calculation cost is on the Fourier Transform the software makes use of FFT W http www fftw org a free C subroutine library from MIT that is known for its speed and performance Howe
57. the reflected light wavelength which is an absolute quantity However there are several drawbacks associated with the Fiber Bragg Grating method e To measure the specimen strain the optical fiber can be either be mounted onto the surface or embedded in the body of the specimen When the fiber is embedded in the concrete mix the alkaline chemical environment starts to erode the thin coating that protects the fiber core For long term strain measurement the aging of the fiber might be a problem or even cause the loss of the measurement The protection of the leads of the fiber that exit from the concrete surface is also a concern in field applications of the method e tis questionable how faithfully the strain of the fiber follows the change of the strain of the specimen The loss of grip debond between the fiber and the concrete mix might happen causing a difference of the strain between them e Itis reported that when the FBG fiber is not aligned to the principal stress direction of the host material the strain detected by the FBG sensor will be much different than that of the host material Hong Nan Li 2007 1 2 5 Video extensometer A video extensometer measures the surface strain by tracking the coordinates of contrasting marks placed on the specimen The gauge marks can be in the form of grid of dots or lines as shown in Figure 1 6 with the dot diameter or line thickness ranging from half millimeter to a couple of millimeters T
58. them gives the relative displacement u n gt V Since there is a 2 1 parent child relations between the pixels at the level k 1 image and level K image in the image trees a pixel displacement in level k corresponds to a two pixel displacement in level k 1 Thus by multiplying the relative pixel displacement u v at level n by two the relative displacement between the image pairs at n the n 1 level can be estimated to be 2u 2v When conducting the correlation on the image pairs of level n 1 the two images are compared over the small area of M n by N n size with relative position displacement 2u 2v and produce the new relative displacement u v _ n The operation is repeated upward along the image tree till the root of the image tree is reached where the correlation between the two small areas of M n by N n size gives relative displacement H Vo Which is also the relative displacement between the original image pairs orrelation 24 2 5 2 4 S correlation i gt Image tree 1 Image tree 2 Figure 4 6 Pyramid scheme e Subimage scheme Given image pairs Il initial image and 2 post deformation image instead of correlating the two full images a small patch of I2 usually at the center of the image is used as a template The software slides the template from the top left to the bottom right of the image I1 56 trying to find the best match The reason that the correlat
59. used to circularize the beam and remove astigmatism Many laser heads in the market use acrylic lens because it is much easier to manufacture aspherical lens in plastic than in glass However since the optical sensor is supposed to be used in harsh environment where the ambient temperature can fluctuate tens of degree the thermal properties of the collimating lens have to be taken into consideration It is Known that the thermal expansion of acrylic 315 ppm C is several hundred times larger than that of the BK7 glass 0 98 ppm C Herzig 1997 To minimize the effect of the temperature change on the sensor performance a laser head with a glass collimating lens was preferred Another performance parameter of the laser module that is affected by the temperature is the beam pointing stability Laser pointing stability whose unit is typically urad 9C is a measure of how much the beam axis angle drifts over time as the temperature changes When the laser beam drifts the distance between the laser beams of the two modules will change too inadvertently affecting the accuracy of the strain sensor The laser heads used in the sensor are two high performance diode laser modules 1112A2 0001 from Diode Laser Concept company It is claimed by the company that the pointing stability is 10 urad 9C Using a laser profiler Beamstar V PCI PHIR company an attempt was made to verify the pointing stability rating of the laser module The laser profile
60. working unit suffered from the following drawbacks Assembly The single module design made use of one set of optical system component to capture the images of two areas on the object surface To achieve this three mirrors M1 M2 M3 in Figure 3 3 are used in the system The laser light had to be spilt and directed to the two surface areas to be measured using these three mirrors At the same time the three mirrors were also used to direct the reflected light from the two separate areas on the object surface onto the single camera whose size was only 6 4mm by 4 8mm The two folded adjustment of the three mirrors proved to be a very difficult task Gauge length With the single module design the gauge length was fixed after the sensor was assembled To change the gauge length the whole system would have to be redesigned and rebuilt Measurement range The single module design integrated two identical displacement measurement systems into one system by sharing the lens and camera To prevent the images of the two surface areas from overlapping each other on the single camera a mechanism was taken to separate the two speckle images One of the approaches involved manually blocking one of the two laser beam at a time This reguired additional manual operation by the user and significantly reduced the measurement pace The other approach was to block half of each of the laser beams which resulted in two side by side speckle patterns on the CCD came
61. 06 In the first step two complex spectrums of the image pair are obtained as follow F amp gt f a y e P70 gt dxdy 2 31 Fy 0 f amp x u y v e PO dydy 2 32 A basic property of Fourier Transform yields Bracewell 1978 F e 77 F 2 33 The normalized cross power spectrum of the two images is then calculated as Fi 2 0 F j2z ua rvo F 0 0 F 0 0 2 34 By applying a second step FFT to the resulting spectrum F e7 a d u v a pulse signal appears in the second FFT spectrum image at u v which represents the 26 displacement vector between the image pairs In this approach the FFT spectrums F 0 F consist of both magnitude information and the phase information After the normalization the magnitude information is removed and only the phase information is retained For verification purposes a pair of speckle images with relative shifting were cross correlated using both the normalized correlation algorithm and the phase correlation algorithm The resulting correlation images were very different In the correlation image produced by the normalized correlation algorithm as shown in Figure 2 6 the peak is suppressed by the high intensity of background noise While for the phase correlation as shown in Figure 2 7 the peak is very clear Figure 2 6 Normalized correlation results for a typical speckle image pairs Phase C
62. 2 start_x width 2 bs 2 blockfilter cvCreateMat bs bs CV_32FC1 CreateHanning bs bs blockfilter 1 bs float fvar CStdioFile ppFile if pVersionsArray 0 serialnumber 30052108 amp amp pVersionsArray 1 serialnumber 30052152 oldsensor 0 if ppFile Open C WProgram Files Strain parameter_new sensor txt CFile modeRead TRUE AfxMessageBox Failed to load parameter file if pVersionsArray 0 serialnumber 30052090 amp amp pVersionsArray 1 serialnumber 30052184 oldsensor 1 if ppFile Open C WProgram Files Strain parameter_old sensor txt CFile modeRead TRUE 131 AfxMessageBox Failed to load parameter file ppFile SeekToBegin int cc 0 CString tempString while ppFile ReadString tempString FALSE fvar float atof tempString Cort switch cc case 1 THRESHOLD fvar break case 2 thresholdhalf fvar break case 3 THRESHOLD_glass fvar break case 4 pl fvar break case 5 p2 fvar break case 6 p3 fvar break case 7 p4 fvar break case 8 gl fvar break case 9 g2 fvar break case 10 g3 fvar break case 11 g4 fvar break case 12 g5 fvar break ppFile Close imgl_fft fftwf complex fftwf malloc sizeof fftwf_ complex bs bs img2_fft fftwf_complex fftwf malloc sizeof fftwf complex bs bs res_fft fftwf_co
63. CO 32 Figure 3 4 Image splitting 22 3 v2 36 cede dedi vetindsicede desi decaedavedude dd decasd ive us a i eee asi i aiaia 32 Figure 3 5 Prototype based on single module design 0 00 eeescceeeessneeeeeessneeeeeeenaeeeesesseeeeeee 33 Figure 3 6 Interior view of the prototype based on the single module design 33 Fig re3 7 Stram Measurement ae AG a Gi aiden Fy A DC SEE ERES ASEN aout 34 Figure 3 8 Dual module Gesie m2 s c cnadgcede age i dn a gd od DYT LG FOD 0D Sd yD CD CCR 37 Figure 3 9 Schematic of the dual module design with dimensions labeled 9999 n 38 Figure 3 10 The third generation prototype based on dual module design 9 9 19199ennnnen 40 Figure 3 11 Interior view of the individual module of the third generation eee 40 Figure 3 12 The fourth generation prototype based on dual module design 9999 nn 42 Figure 3 13 Interior view of the fourth generation prototype eeeeeesceecceeeeeeeeeeetneeeeeeeeeeeeeees 43 Figure 3 14 Experiment to evaluate thermal expansion effect ee eeeeeeseeeeeeesneeeeeeenaeeeeeees 44 Figure 3 15 Saw tooth laser beam profiles aa ud WW BOG AR 45 Figure 3 16 Gaussian laser beam prolil amp i cei Gen day CF siete gdadevtamannesie 45 Figure 3 17 Laser pointing stability test 0 0 0 ee eseeeeessneeeceessneeeeeessaceecessneeecesssaeeesesssaeeesenes 47 Fig re 3 18 Multiple modul6es Setupau need yu o y Gwg gaa C4 sagas UF
64. DEVELOPMENT OF A PORTABLE OPTICAL STRAIN SENSOR WITH APPLICATIONS TO DIAGNOSTIC TESTING OF PRESTRESSED CONCRETE by WEIXIN ZHAO B S Huazhong University of Science and Technology 1998 M S Kansas State University 2006 AN ABSTRACT OF A DISSERTATION submitted in partial fulfillment of the reguirements for the degree DOCTOR OF PHILOSOPHY Department of Mechanical and Nuclear Engineering College of Engineering KANSAS STATE UNIVERSITY Manhattan Kansas 2011 Abstract The current experimental method to determine the transfer length in prestressed concrete members consists of measuring concrete surface strains before and after de tensioning with a mechanical strain gage The method is prone to significant human errors and inaccuracies In addition since it is a time consuming and tedious process transfer lengths are seldom if ever measured on a production basis A rapid non contact method for determining transfer lengths in prestressed concrete members has been developed The new method utilizes laser speckle patterns that are generated and digitally recorded at various points along the prestressed concrete member User friendly software incorporating robust and fast digital image processing algorithms was developed by the author to extract the surface strain information from the captured speckle patterns Based on the laser speckle measurement technigue four 4 successively improved generations of designs have been made A prot
65. FLOAT m_Exp 2 0f UpdateData TRUE switch m surfacetype 169 case 0 m_Lum 50 0f break case l m_Lum 70 0f break flags LUCAM_PROP_FLAG_AUTO BOOL AutoA 1 rt LucamSetProperty m_hCameraA LUCAM_PROP_EXPOSURE m_Exp flags if LucamSetProperty m_hCameraA LUCAM_PROP_AUTO_EXP_TARGET m_Lum flags MessageBox Failed to set exposure target AutoA 0 flags LUCAM PROP_FLAG AUTO rt LucamSetProperty m_hCameraB LUCAM_PROP_EXPOSURE m_Exp flags if ILucamSetProperty m_hCameraB LUCAM_PROP_AUTO_EXP_TARGET m_Lum flags MessageBox Failed to set exposure target Start the preview if LucamCreateDisplayWindow m_hCameraA Preview A WS_OVERLAPPEDWINDOWIWS_VISIBLE 0 200 640 480 this gt m_hWnd NULL amp amp LucamCreateDisplayWindow m_hCameraB Preview B WS_OVERLAPPEDWINDOWIWS_ VISIBLE 500 300 640 480 this gt m_hWnd NULL if LucamStream VideoControl m_hCameraA START_DISPLAY NULL amp amp LucamStreamVideoControl m_hCameraB START_DISPLAY NULL m_bPreviewing TRUE m_cbPreview EnableWindow TRUE m_cbPreview SetWindowText _T Stop m_cbMeasure EnableWindow FALSE else MessageBox Unable start previewing video Start Preview MB_OK LucamDestroyDisplayWindow m_hCameraA LucamDestroyDisplayWindow m_hCameraB 170 else MessageBox Unable create prev
66. Homography Projection from objection coordinate to camera coordinate 69 Figure 5 7 Setup of the first calibration Method uei eyd HU Gwy U OUT WED cya 71 Figure 5 8 Strain calculation with orientation difference of the two camera coordinate systems 74 Figure 5 9 Messphysk company s laser speckle extensometer ME53 33 919999eeneeeeeueeen 75 Figure 6 1 A two concrete block system sssessssseessssrerrsssrersssrressssrressseteesssrrersssrresesereeessereest 81 Figure 6 2 Comparison of laser speckle strain sensor and Digital dial gauge 0 0 eee 81 Figure 6 3 Difference between optical strain sensor and digital dial gauge measurements 82 Figure 6 4 A concrete beam under Compression ceeeseeceesssnceeeeesseeeetsesaeeeeeessaeeeeesstaeeeeeee 83 Figure 6 5 Surface deflection measurement obtained by Whittemore gauge and laser speckle Str n eic GWBNDEN FIT FFEFRYN FEN Muses cases loca datascedlausdccedia codes ca FERFAU EF FN FU NF YNN NT 83 Figure 6 6 Experiment setup of the comparison of laser speckle strain sensor and electrical resistance Strain gauge ESG Sensor scssc casas cayseascaenspndenssons gang anna ledeguasgaanvanntomsenteranntan senate 84 Figure 6 7 Measurement results of surface Strain ce eeesceceesssseeeeesesneeeeeessneeeessesaeeeeseesaeeeeeses 85 Figure 6 8 Difference of the measurements between optical strain sensor and electrical resistance SH IDDO Oe Y i Y Y A 3 85
67. L filterproduct refblockL blockfilter reffiltered filterproduct tplL backup blockfilter tplfiltered phase_correlation reffiltered tplfiltered poc_halfL cvMinMaxLoc poc_halfL amp minvalL amp maxvalL amp minlocL int xx yy if maxlocL x gt bs 2 xx oldstartL x bs maxlocL x else xx oldstartL x maxlocL x if maxlocL y gt bs 2 yy oldstartL y bs maxlocL y else yy oldstartL y maxlocL y int x maxlocL x int y maxlocL y CvMat real cvCreateMat bs bs CV_32FC1 CvMat im cvCreateMat bs bs CV_32FC1 for i 0 k 0 i lt bs it for j 0 3 lt bs jt t k cvmSet real i j float res_fft k 0 cvmSet im i j float res_fft k 1 CvMat in cvCreateMat bs bs CV_32FC2 int factor 25 CvMat out cvCreateMat 2 factor 2 factor CV_32FC2 161 amp maxlocL O upsampling in factor y l factor x l factor out 2 cvReleaseMat amp real cvReleaseMat amp im real cvCreateMat factor 2 factor 2 CV 32FCl im cvCreateMat factor 2 factor 2 CV _32FCl cvSplit out real im 0 O Compute the magnitude of the spectrum Mag sgrt Re 2 Im 2 cvPow real real 2 0 cvPow im im 2 0 cvAdd real im real NULL cvPow real real 0 5 cvMinMaxLoc real amp minvalL amp maxvalL amp minlocL amp maxlocL cvReleaseMat amp real cvReleaseMat
68. LONG numCameras LucamNumCameras LUCAM_VERSION pVersionsArray 20 ULONG tt LucamEnumCameras pVersionsArray numCameras UpdateData TRUE CvMat X cyCreateMat N 3 CV_32FC1 CvMat Y cyCreateMat N 1 CV_32FC1 CvMat b cvCreateMat 3 1 CV_32FC1 CvMat Xt cvCreateMat 3 N CV_32FC1 CvMat XtX cvCreateMat 3 3 CV_32FC1 CvMat tempM cvCreateMat 3 N CV_32FC1 b_keepgoing TRUE take initial reading pThread AfxBeginThread RUNTIME_CLASS CUIThread TakelnitialReading pThread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL Sleep 100 Sleep 2000 int i 0 179 while b_keepgoing 1 amp amp i lt N b_keepgoing TRUE pThread AfxBeginThread RUNTIME_CLASS CUIThread b_redo FALSE CalibrationReadingS peckle amp subxL amp subyL amp subxR amp subyR pThread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL Sleep 1000 cvmSet Y i 0 subxL baseb cvmSet X 1 0 subyL cvmSet X i 1 subxR cvmSet X i 2 subyR i cvTranspose X Xt cvMatMul Xt X XtX cvInvert XtX XtX cvMatMul XtX Xt tempM cvMatMul tempM Y b CvMat Xb cvCreateMat N 1 CV_32FC1 cvMatMul X b Xb cvSub Xb Y Xb cvAbs Xb Xb float residue 0 int maxindex l for i 0 i lt N it if residue lt cvmGet Xb i 0 residue cvmGet Xb i 0 maxindex i cvmSet Y maxindex 0 0 cvmSet X maxindex 0 0 cvmSet X maxindex 1 0 cvmSet X maxindex 2 0 cv
69. MatMul X b Xb 180 cvSub Xb Y Xb cvAbs Xb Xb residue 0 maxindex l for i 0 i lt N irr if residue cvmGet Xb i 0 residue lt cvmGet Xb i 0 maxindex i cvmSet Y maxindex 0 0 cvmSet X maxindex 0 0 cvmSet X maxindex 1 0 cvmSet X maxindex 2 0 cvTranspose X Xt cvMatMul Xt X XtX cvInvert XtX XtX cvMatMul XtX Xt tempM cvMatMul tempM Y b cvReleaseMat amp Xb CStdioFile DataFile cstemp Format S8 8f n 8 8f n 8 8f n 8 8f n cvmGet b 0 1 cvmGet b 0 2 if DataFile Open c para txt CFile modeWrite CFil DataFile SeekToEnd modeCreate TRUE DataFile WriteString cstemp DataFile WriteString An DataFile Close CStdioFile ppFile if pVersionsArray 0 serialnumber 30052108 baseb CFile amp amp pVersionsArray l serialnumber 30052152 181 cvmGet b 0 0 modeNoTruncate if ppFile Open C Program Files Strain parameter_new sensor txt CFile modeReadWrite TRUE AfxMessageBox Failed to load parameter file if pVersionsArray 0 serialnumber 30052090 amp amp pVersionsArray l serialnumber 30052184 if ppFile Open C Program Files Strain parameter_old sensor txt CFile modeReadWrite TRUE AfxMessageBox Failed to load parameter file float fvar
70. OLD AfxMessageBox might be fake peak measure again float deflection CStdioFile DataFile EN if DataFile Ope cstemp Forma 5 4f 5 4f 5 4f 5 4f subxL subyL subxR subyR n c data_fitting txt CFile modeWrite CFile modeNoTruncate CFile modeCreate TRUE DataFile SeekToEnd DataFile WriteString cstemp DataFile WriteString An DataFile Close 196 cvReleaseMat amp real cvReleaseMat amp im cvReleaseMat amp in cvReleaseMat amp o0ut for colindex 0 colindex lt colnum colindex t t for rowindex 0 rowindex lt rownum rowindex lt fftwf_free fft_halfL colindex rowindex fftwf_free fft_halfR colindex rowindex cvRe leasel mage amp poc cvRel cvRel asel asel mage amp refblockL mage amp refblockR cvRel asel cvRel cvRe cvRel cvRel cvRe cvRel cvRel easel leasel easel easel leasel easel asel mage amp tplRtemp mage amp tplLtemp mage amp tplR cvRel cvRel cvRel easel easel asel mage amp refR mage amp tplL cvRel if easel Lucam mage amp refL MessageBox Fail if pAllFr free pAllF ames rames m_cbPreview EnableWindow TRU m_cbPrevie
71. ReleaseImage amp reffiltered cvReleaseImage amp tplfiltered cvReleaseImage amp tplL backup cvReleaseIlmage amp tplR backup cvReleaseImage amp tplRtemp cvReleaseImage amp tplLtemp cvReleaseImage amp tplR cvReleaseImage amp refR cvReleaseImage amp poc_halfR cvReleaseImage amp tplL cvReleaseImage amp refL cvReleaseImage amp poc_halfL if LucamDisableSynchronousSnapshots hSynchronousSnapshots 165 MessageBox Failed to unsetup synchronous snapshots if pAllFrames free pAllFrames m_cbPreview EnableWindow TRUE m_cbPreview SetWindowText _T Preview Gl m_cbApply EnableWindow TRU return TRUE void CStrainDlg OnSelchangeMeasurementList int n CString str for int i 0 i lt m_measurementlist GetCount i if m_measurementlist GetSel i gt 0 n m_measurementlist GetTextLen i m_measurementlist GetText i str GetBuffer n str ReleaseBuffer measurementname str m_strPath defaultfoldert datasetname measurementname m_pointnumber 1 m_Spin SetPos m_pointnumber if m_measurementlist GetCurSel gt 0 CSpreadSheet SS defaultfolder tdatasetname xls Sheetl SS BeginTransaction SS AddCell measurementname l m measurementlist GetCurSel r l
72. Xt cvReleaseMat amp XtX cvReleaseMat amp tempM BOOL CStrainDlg TakeInitialReading long 1PixelSize 1 8 bits rnt ay Jy BOOL flag 0 CString cstemp UpdateData TRU CFileFind f r E 183 CString cfilenameL cfilenameR Dp F mpty cfilenameL defaultfolder L bmp cfilenameL cfilenameR Empty defaultfolder R bmp cfilename m_hCameraA hCameras 1 R R hCameras 0 1 m_hCameraB 0 for i i lt CAMNUM itt params i format height height params i format pixelFormat LUCAM PF_8 params i format subSampleX 1 params i format subSampleY 1 params i format width width params i format xOffset 0 params i format yOffset 0 params i exposure m_exposure 50 ms exposure params i gain 23 params i strobeDelay 0 0 unused params i timeout 3000 0 3000 ms params i useHwTrigger FALSE Set this to true for hardware triggered setup with daisy chaining params i useStrobe FALSE Set this to true if daisy chaining cameras params i exposureDelay 0 params i shutterType LUCAM SHUTTER_TYPE GLOBAL pParams i amp params i params 0 exposure m_exposureA params 1 exposure m_exposureB pAllFrames UCHAR malloc CAMNUM width height if pAl
73. _measurementlist GetText ref_index texttemp m_measurementlist DeleteString ref_index nIndex m_measurementlist InsertString ref_index texttemp m_measurementlist GetText selindex texttemp m_measurementlist DeleteString selindex nIndex m_measurementlist InsertString selindex texttemp RGB 255 O 0 ref_index selindex int n m_measurementlist GetCount m_measurementlist SetCurSel n 1 texttemp Reference is texttemp m_refname SetWindowText texttemp UpdateData FALSE void CStrainDlg Subpixel IplImage in double x double y int factor CvPoint minl maxl double minv maxv 173 int h 64 int w 64 CvRect rect IplImage peakarea cvCreateImage cvSize w h IPL_DEPTH_32F 1 cvMinMaxLoc in amp minv amp maxv amp minl amp maxl O int intx maxlI x int intyzmaxl y rect cvRect maxl x w 2 maxl y h 2 w h cvSetlmageROl in rect cvCopy in peakarea NULL cvResetImageROI in IplImage fftImg cvCreateImage cvSize w h IPL_DEPTH_32F 2 IplImage complexInput cvCreateImage cvSize w h IPL DEPTH 32F 2 IplImage imaginaryInput cvCreateImage cvSize w h IPL_DEPTH_32F 1 cvZero imaginaryInput cvDFT complexInput fftImg CV_DXT_FORWARD ICV_DXT_SCALE complexInput gt height cvMinMaxLoc peakarea amp minv amp maxv amp minl amp maxl 0 CvMat out cvCreateMat 2 factor 2 factor CV_32FC2 upsampling fftlmg factor maxl y 1 facto
74. _msg void OnConcrete afx_msg void OnSteel afx_msg void OnOpendatafile afx_msg void OnCalibrateIntensity afx_msg void OnSetreference afx_msg void OnCalibration I1 AFX_MSG DECLARE_MESSAGE_MAP private void UpdateMeasurementList CString defaultfolder CMessageDialog MessageDlg B endif defined AFX_STRAINDLG_H__235F11AA_4538_4BD7_88AA_66E784050B55__INCLUDED_ StrainDlg cpp include stdafx h include Strain h include StrainDlg h ifdef _DEBUG define new DEBUG_NEW 124 fundef THIS FIL Gl static char THIS_FILE endif a CStrainDlg CStrainDlg CWnd pParent NULL CDialog CStrainDlg IDD pParent AFX_DATA INIT CStrainDlg m_exposure 1 0f 23 0f m_gain m round ils m_DataSetList m_accuracy O m_ gl 0 0 m_delay m_duration 3 m_surfacetype m_ invert FALSE m configuration m_pointnumber uratio 0 m config m duration AFX_DATA_INIT m_hIco n T T AfxGetApp gt LoadIcon IDR MAINFRAM CW Gl void CStrainDlg DoDataExchange CDataExchange pDX CDialog DoDataEx AFX_DATA_MAP DDX_Control pDX DDX_Control pDX DDX_Control pDX DDX_Control pDX DDX_Control pDX DDX_Control pDX DDX_Control pDX DDX_Control pDX hange pDX StrainDlg DC_R EF NAM F m
75. _refname 17 DC_MeasurementList m measurementlist DC CalibrateIntensity m CalibrateIntensity DC _SPIN m Spin DC ROUND1 m_ cbRound DC PATH m path DC LOCK m cbLock DC APPLY m cbApply 125 DX_Text DX IDC_EXPOSURE m_exposure DX_Text pDX IDC_GAIN m_gain DX_Control pDX IDC CONNECT m cbConnect DX_Control pDX IDC PREVIEW m cbPreview I DX_Control pDX IDC_MEASURE m cbMeasure DX_Radio pDX IDC_ROUND1 m_ round DX_CBString pDX IDC DataSetList m DataSetList DX_Radio pDX IDC ACCURACY m accuracy DX_Text pDX IDC DELAY m delay DX_Text pDX IDC GL m gl DX_Radio pDX IDC Speckle m_surfacetype DX_Check pDX IDC INVERT m invert DX_Radio pDX IDC NORMCONFIG m configurationtype DX_Text pDX IDC POINTNUMBER m_pointnumber DX_Text pDX IDC Duration m duration AFX_DATA MAP EGIN MESSAGE MAP CStrainDlg CDialog AFX_MSG_MAP CStrainDlg y M_PAINT Z lt 10 G ERYDRAGICON Zz U A LICKED IDC_CONNECT OnButtonConnect i U Zz LICKED IDC_PREVIEW OnButtonPreview Z Z OSE IC z 2 06 O 2 EE EN zZ ED IDC APPLY OnButtonApply LICKED IDC_SETFOLDER OnButtonSetfolder zZ IC ED IDC MEASURE OnButtonMeasure Z LICKED
76. a is done by accumulation of photoelectrons on the CCD chip for a certain time period usually several milliseconds If the sensor moves during the integration period the captured images will suffer a blur effect The blurred images lose most of the high frequency information and this can cause the correlation algorithm to report a fake match of the speckle image pairs An algorithm is employed to detect blur in the captured image by transforming the image into the spatial frequency domain using a 2D Fourier transform and calculating the ratio of the low frequency power to the high frequency power If the ratio exceeds a predefined threshold it means the there is significant blur effect in the image The blurred image will then be discarded and the software will command the camera to capture another frame of the speckle pattern until a non blurred one is obtained e Histogram equalization The intensity of the image captured by the camera is usually unbalanced i e the grey level of the pixel concentrates to a limited range of the full grey range 0 255 For example the histogram of a typical raw speckle image is shown Figure 4 2 a in which the grey levels of the pixel concentrates to the range lower than 100 Histogram equalization increases the dynamic range of the grey level by making the histogram a uniform profile The histogram of the same speckle image after being equalized is shown in Figure 4 2 b
77. alue Decomposition SVD matrix A can be written in the form 4 A UXV ouv 4 33 i 0 The right singular vector of the matrix A corresponding to the smallest o is the solution h whose entries are the value of the four parameters f a P Now given the image displacements x y and x y that are detected by the two camera of the sensor we can calculate the distance change between the two observed points on the object surface using Eguation 4 28 and Eguation 4 25 Y Y 2L A X X By 272 2 72 171 171 qX By a x T And the surface strain can be then calculated by ar 78 The calibration method described above makes the sensor very flexible and easy to use The calibration procedure does not reguire any additional eguipment except a flat surface thus it can be done in the field or even when the sensor is mounted on the rail system The computation for solving the 4 parameters a B 2 B has been incorporated into the current software and most of the time that the procedure takes is to collect the N displacement vectors which needs less than 1 minutes This is significant less than the time reguired by the previous method which usually takes a couple of hours With the Auto Calibration method the adjustment of the gauge length becomes a trivial issue All the user needs to do is to unattach the two modules reattach the two modules with the desired gauge length and take l
78. ameraOpen iCamera m_hCameraB LucamCameraOpen iCamera 1 LONG numCameras LucamNumCameras LUCAM_VERSION pVersionsArray 20 ULONG tt LucamEnumCameras pVersionsArray numCameras if pVersionsArray 0 serialnumber gt pVersionsArray 1 serialnumber HANDLE swap swap m_hCameraA m_hCameraA m_hCameraB m_hCameraB swap if m_hCameraA NULL amp amp m_hCameraB NULL Now connected to the camera if ILucamSetProperty m_hCameraA LUCAM_PROP_EXPOSURE m_exposure 0 AfxMessageBox Unable to set exposure value if ILucamSetProperty m_hCameraA LUCAM_PROP_GAIN m_gain 0 AfxMessageBox Unable to set gain value if ILucamSetProperty m_hCameraB LUCAM_PROP_EXPOSURE m_exposure 0 AfxMessageBox Unable to set exposure value 130 if ILucamSetProperty m_hCameraB LUCAM_PROP_GAIN m_gain 0 AfxMessageBox Unable to set gain value m_bConnected TRUE m_cbConnect SetWindowText _T Disconnect m_cbPreview EnableWindow TRUE m_cbPreview SetWindowText _T Preview m_cbMeasure Enable Window TRUE m_cbApply EnableWindow TRUE m_CalibrateIntensity EnableWindow TRUE if LucamGetFormat m_hCameraA amp m_lffFormat amp m_fFrameRate MessageBox Unable to get camera video format Capture frames may not work properly Get Fromat MB_OK height m_IffFormat height width m_IffFormat width half_height height 2 half_width width 2 start_y height 2 bs
79. as been fabricated The portable design enables flexible adjustment of the gauge length and easy expansion to a rosette strain measurement configuration Extensive testing has been conducted in the laboratory environment to validate the sensor Furthermore the laser speckle strain sensor was applied to transfer length measurement of common prestressed concrete beams and prestressed concrete cross ties in the field The sensor yielded unprecedented measurements of transfer length in just a few minutes compared to the hours that are needed if using the current accepted method of measurement 13 Through this testing with different applications it has been shown that the newly developed portable laser speckle strain sensor can not only serve as an accurate instrument in the civil engineering laboratory where the deflection characteristics of a concrete member are needed but also can be readily used in the harsh environment of the prestress concrete industry with minimum surface preparation and staff training It also has the potential to rapidly process a large guantity of data points in an industrial setting It should be noted that as far as the author is aware the new developed sensor is the first device to successfully employ speckle method to determine transfer length of prestressed concrete Furthermore this development represents the first time that such a method has been demonstrated successfully in a harsh industrial environment with suf
80. asionally for research purposes This due to the fact that there does not exist a transfer length measurement method that is robust enough for the harsh environment of the plant and capable of keeping up with the working speed of the production line To evaluate the feasibility of the in plant transfer length measurements using the laser speckle strain sensor two trips have been made to the CXT concrete cross tie production plant in Grand Island NE One trip was on October 22nd in 2010 and the other trip was on February 8th in 2011 In order to facilitate the laser speckle measurements three small I4 inch diameter inserts were cast into each of the cross tie immediately after the pouring of the concrete mix The inserts allow an aluminum rail to sit on the top of the member surface conveniently The sensor was that was installed on the rail was able to traverse freely on it as shown in Figure 6 16 Figure 6 16 Laser speckle strain sensor mounted on a rail at CXT concrete cross tie plant 92 Before the detensioning of the cross tie initial laser speckle readings were taken every 0 5 inch for the first 10 points counting from the end of the tie and every 1 inch thereafter along the beam by traversing the laser speckle strain sensor on the rail manually A total of 70 data points were obtained for each tie with 35 data points for either side After the cross tie was detesioned i e the tensioned reinforcing strands were cut and prestress
81. ate In the homography projection model the homnogeneous coordinates are defined as x X 4 13 OF hg g y 4 13 1 1 where O is a point in the object coordinate and g is the projection point of O on the image plane Their relationship can be expressed as 69 X X Y Y M R i M R a R PRH 1 a 4 14 me x 1 1 where M is the magnification factor Since the camera plane is at the imaging plane the image shift is not sensitive to the object tilt rotation about the x axis and yaw rotation about the y axis the rotation angles about the x axis and the y axis of the object surface do not change the projection matrix between the object coordinate and the camera coordinate Therefore we can assume the B and 9 are zero and Equation 4 14 becomes x cos g sin 0 O t y M sin cos 6 0 ty 4 15 Nx amp Since the out of plane movement of the object surface is very small we can assume that the entry Z of the object position vector Q is zero Furthermore we are only interested in the relative movement of the objection surface point Q instead of its absolute projection position on the camera plane the translation entry t t can also be assumed to be zero Equation 4 15 can then be simplified to be i yw sin g gt 4 16 y sing cos Y or F lw PNP M sin p 4 17 y Msin M cos Y 70 5 3 Two calibration methods for the strain sens
82. b 0 i itcoff 2 pi nc factor Mc cvCreateMat nc noc CV_32FCl cvMatMul Ma Mb Mc realtemp cvCreateMat nc noc CV_32FC1 imaginarytemp cvCreateMat nc noc CV_32FC1 for i 0 i lt nc itt 145 for j 0 J lt noc j cvmSet realtemp i jJ cos cvmGet Mc i j cvmSet imaginarytemp i j sin cvmGet Mc i j temp cvCreateMat nor nc CV_32FC2 cvMatMul kr in temp cvMatMul temp kc out cvReleaseMat amp Ma cvReleaseMat amp Mb cvReleaseMat amp Mc cvReleaseMat amp realtemp cvReleaseMat amp imaginarytemp cvReleaseMat amp temp void CStrainDlg fftwcopy fftwf_complex source fftwf_complex target int size int i for i 0 i lt size i target i 0 source i 0 target i 1 source i 1 void CStrainDlg OnCreateDataSet CNameInputDialog Dlg Dlg m_DataPointNum 1 Dlg defaultfolder defaultfolder if Dlg DoModal IDOK datapointnumber Dlg m_DataPointNum CComboBox pCB CComboBox GetDlgItem IDC_DataSetList pCB gt ResetContent 146 CStringArray files CFileFind finder BOOL bWorking finder FindFile defaultfoldert while bWorking bWorking finder FindNextFile if finder IsDirectory amp amp finder IsDots files Add finder GetFilePath now cycle through the array for int i 0 i lt fil
83. both ends of the concrete beam to release the tension The stress transferred from the strands to the concrete is developed gradually from each end of the beam where the stress is zero and to the location far away from the end where the stress is at its full value The distance reguired to develop this stress is defined as The transfer length which is used to evaluate the guality and performance of concrete members To estimate the transfer length the surface strain profile of the prestressed concrete beam must be measured Many methods are available to measure strain either on the surface or in the body of the concrete structures The strain measurement under laboratory conditions is usually straightforward but it is much more difficult when applied in the field due to the fact that most of the strain measurements of structure materials must be done in a harsh environment or reguire long term monitoring In addition it is recognized that a measurement technigue that aimed to make its way to the diagnostic testing of large concrete structures must be easy to use Most civil engineers particularly field engineers are not experts in sophisticated sensor technology Therefore it is important to provide them with a practical solution instead of just a laboratory device with nanometer level resolution but could not be readily used in the field with minimum training To be incorporated into the diagnostic testing of modern concrete structures seaml
84. c Call this when linking to MFC statically endif CStrainDlg dlg m_pMainWnd amp dlg 118 int nResponse dlg DoModal if nResponse IDOK else if nResponse IDCANCEL Since the dialog has been closed return FALSE so that we exit the application rather than start the application s message pump return FALSE StrainDlg h if defined AFX_STRAINDLG_H__235F11AA_4538_4BD7_88AA_66E784050B55__INCLUDED_ define AFX_STRAINDLG_H__235F11AA_4538_4BD7_88AA_66E784050B55__ INCLUDED _ if _MSC_VER gt 1000 pragma once endif MSC_VER gt 1000 include lucamapi h include lucamsci h include SnapPreviewDlg h include lt stdio h gt include lt stdlib h gt include cv h include highgui h include fftw3 h include math h include lt iostream gt include lt windows h gt include lt mmsystem h gt include lt direct h gt include lt io h gt include lt afxtempl h gt include UIDialog h 119 pragma comment lib Winmm lib define CAMNUM 2 define bs 384 define cutoff 60 define MatchTempType CV_TM_CCOEFF_NORMED class CStrainDlg public CDialog Construction public BOOL CalibrationReadingSpeckle double subxL double subyL double subxR double subyR BOOL TakelInitialReading void MultipleMinLoc Ipllmage image CvPoint location void MultipleMaxLoc IplImage image CvPoint location void MinLoc Ipllmage image
85. ct cvRect colindex stepsize rowindex stepsize bs bs cvSetlmageROl refL rect cvCopy refL refblockL NULL filterproduct refblockL blockfilter reffiltered memcpy img1_half float reffiltered gt imageData bs bs sizeof float fftwf_execute fft_img1_halfsize fft_halfL colindex rowindex fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 2 1 fftwcopy img1_fft_half fft_halfL colindex rowindex bs bs 2 1 cvSetlmageROl refR rect cvCopy refR refblockR NULL filterproduct refblockR blockfilter reffiltered memcpy img1_half float reffiltered gt imageData bs bs sizeof float fftwf_execute fft_img1_halfsize fft_halfR colindex rowindex fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 2 1 fftwcopy img1_fft_half fft_halfR colindex rowindex bs bs 2 1 H CvPoint minlocL maxlocL minlocR maxlocR maxlocLhalf maxlocRhalf CvPoint minlocLtemp minlocRtemp maxlocLhalftemp maxlocRhalftemp double minvalL maxvalL minvalR maxvalR maxvalLhalf maxvalRhalf double minvalLtemp minvalRtemp maxvalLhalftemp maxvalRhalftemp int colL rowL colR rowR int colindex_store rowindex_store CvPoint oldstartL oldstartR startL startR oldstartL cvPoint 0 0 oldstartR cvPoint 0 0 startL cvPoint 0 0 startR cvPoint 0 0 BOOL peakflag workmode peakflag FALSE double oldLpeak oldRpeak oldLpeak 0 oldRpeak O0 max valLhalf 0 maxvalRhal
86. de A Tie 2 Side B ie in a y ddd sr a a Microstrain 75 76 77 78 79 80 Bi B2 83 84 85 8 87 88 89 90 91 S2 SG S4 55 56 S7 S8 99 10010140 Distance Along Beam in Figure 6 20 Cross tie surface strain measurement Tie 2 Side B 95 Tie3 Side A Microstrain 22345 677 8 S 130011 22 B 34 I5 146 17 18 39 20 23 22 23 24 25 BAB Distance Along Beam in Figure 6 21 Cross tie surface strain measurement Tie 3 Side A 96 Chapter 7 Conclusion This dissertation presented the development of a general non contact surface strain measurement technigue based on the laser speckle principle that is able to rapidly and accurately determine concrete surface strains The characteristics and behavior of the speckle were investigated The relationship between the multi degree of motion of the subject surface and the induced motion of the speckle pattern was also addressed Based on the laser speckle measurement technigue four 4 generations of designs have been made A prototype was fabricated for each design either on an optical breadboard for concept validation or in a portable form for field test operation For each generation design improvement was made based on the knowledge learned through the test of the previous generation prototype The fourth generation prototype incorporating a unigue modular design concept and unigue self calibration function exhibits several preferable features such as flexible adjustment
87. desired 5 axis five degree of freedom object movement was constructed during the early stage of the current laser speckle strain sensor development and resulted in the author s Master thesis Zhao 2006 The 5 axis motion measurement technigue employed in this optical breadboard layout was important to the strain sensor development in that the object surface during deformation is usually subjected to full 6 degrees of freedom movement The traditional optical speckle methods though designed to measure in plane movement are also sensitive to surface tilt and other rotational modes that are very likely to happen during the concrete surface strain measurement These rotation effects will introduce severe error to the surface displacement or strain measurement if properly taken into account The characteristics of the previously developed 5 axis motion measurement system are described below Zhao 2006 The system employed the concepts of translation only plane and tilt only plane to separate the surface in plane displacement and out of plane tilt thereby eliminating or greatly reducing rotation induced error The 5 axis movement in plane rotation was detected by using a polar correlation technique The 6 axis movement out of plane translation of the specimen surface is shown both theoretically and experimentally to have no contribution to the in plane surface strain and is therefore not measured by the sensor 29 A schematic
88. digital dial gauge of resolution of 0 001mm Shars 303 3506 The concrete block on the right was held stationary The system was used to create a relative linear displacement between the two concrete blocks by displacing the concrete block on the left while the concrete block on the right remained stationary The relative displacement between the two concrete blocks was increased from 0mm to 2mm with 0 1mm increments and was measured by both the digital dial gauge and the laser speckle strain sensor The results are shown in Figure 6 2 The readings by the two devices optical strain sensor and dial gauge have excellent agreement The differences between the two 80 sensor s readings are below 4 microns over the entire measurement range as shown in Figure 6 3 Figure 6 1 A two concrete block system Laser speckle strain sensor vs Digital dial gauge m E H o d G o o a dj Us d o et b4 o o A o o a a oH 1 Digital dial gauge mm Figure 6 2 Comparison of laser speckle strain sensor and Digital dial gauge 81 Difference between optical strain sensor and digital dial gauge 0 004 7 0 003 0 002 0 001 o e 0 001 9 1 1p 2 0 002 Difference mm 0 003 0 004 0 005 Displacement mm Figure 6 3 Difference between optical strain sensor and digital dial gauge measurements 6 2 Comparison wit
89. e rowindex histsize histsize height cvSetImageROI input rect cvEqualizeHist input input cvResetImageROI input cvEqualizeHist input output 175 BOOL CStrainDlg Blurcheck Ipllmage img1 Ipllmage img2 IplImage diff 0 diff cvCreateImage cvGetSize img1 IPL_DEPTH_8U 1 cvAbsDiff imgl img2 diff CvScalar dzcvAvg diff cvReleaselmage amp diff if d val 0 lt 16 return 0 else return 1 void CStrainDlg MaxLoc IplImage image double amp max CvPoint amp location float data int step CvSize size int x y cvGetRawData image uchar amp data amp step amp size step sizeof data 0 max 0 location x 0 location y 0 for y O y lt size height y data step for x 0 x lt size width x if data x gt max max double data x location x x location y y void CStrainDlg MinLoc IplImage image double amp min CvPoint amp location float data int step 176 CvSize size int x y cvGetRawData image uchar amp data amp step amp size step sizeof data 0 min double data 0 location x 0 location y 0 for y O y lt size height y data step for x 0 x lt size width x if data x lt min min double data x location x x location y y void CStrainDlg MultipleMaxLoc IplImage
90. e beam strain test 24 measurement period that might last for months To overcome the shortcomings associated to the traditional correlation method adapted digital correlation algorithms are more often used 2 4 1 Normalized correlation This method calculates the mean of the speckle image I1 and I2 then determines their normalized version Il and I2 The correlation coefficient is then obtained similar to Equation 2 27 The result is normalized again to obtain the normalized cross correlation coefficient N y mez L i DI x i y j Il UN a isl j R x y TI i 2 28 LG jy y L xsi ys j i l j l i l j l where N N DR I d Li j zy 2 29 N N YU LGat isyts I x i y Latiy p 2 30 N N A perfect match will give a peak equal to 1 and a complete no match will give a peak of 0 The normalization operation used by this method helps reduce effects of the image intensity variation to the matching of the image pairs 2 4 2 Phase correlation Although the normalized correlation algorithm works quite well on the regular images it was observed that it works poorly on the speckle image This is due to the unique characteristics of the speckle pattern which is a grainy pattern without any repeated feature If the speckle pattern is transformed to the frequency domain it can be observed that considerable information of the image is stored in the high spatial frequency domain contrary
91. e click on setup exe Follow the onscreen prompts to install the software drivers and user application After the software has been installed plug the power cable of the USB hub to the power supply and plug the unconnected USB cable into a free USB 2 0 High Speed port on the computer The Window s New Hardware Wizard will pop up two Lumenera Unconfigured Device dialogs Select Install the software automatically from the options that are presented to you and click Next A warning may appear notifying you that the drivers have not been digitally signed by Microsoft Click Continue Anyway to continue with the driver installation Then click Finish to install the drivers You must have administrator privileges to finish the above task Restart your computer Run the Strain application software from your Start All Programs Optical Strain Sensor menu to start the program 112 D 3 Software Operation and Making Measurement The Strain application is a program that incorporates a user friendly interface to help users make strain measurement with the sensor as well as manage the measurement data from multiple projects or subjects The software s main screen is divided into 3 sections Camera Control Measurement Settings and Measurement fc Strain Measurement Camera Control Measurement Settings Eee Dutput Unit micro strain Gain 1 23 23 Exposure ms 2 Reference Delay sec D Concrete Steel F Gauge
92. e concrete surface To evaluate the optical strain sensor s ability to measure transfer length several pretensioned concrete members were fabricated using different concrete mixtures and the strain 86 profile was measured on one side of each member using both the traditional Whittemore gage and the non contact laser speckle method 6 4 1 Surface strain measurement using the second generation prototype The laser speckle strain sensor used in this experiment was the second generation prototype based on single module design that has been discussed in Chapter 3 The mixtures used in this experiment corresponded to SCC mixtures that were part of another prestressed concrete study and previously reported Peterman 2007 The pretensioned members were each 9 6 long with a trapezoidal cross section as shown in Figure 6 10 Surface strain measurements for the trapezoidal specimens were obtained using both the standard Whittemore technigue and the laser speckle optical technigue In Jin 24 in 2 1 Figure 6 10 Cross section of the pretensioned concrete member In order to facilitate the laser speckle measurements an aluminum rail was mounted to the side of the member as shown in Figure 6 11 The rail was attached to the members using small 4 inch diameter inserts that were cast into the sides of the pre tensioned concrete members Because of the insensitivity of the sensor to undesirable degrees of freedom no high precis
93. e subimage Defblock produce n peaks and n displacement vector which are denoted as p and u v The sub window pair with the maximum peak indicates the best match and is used further to extract the subpixel displacement information Figure 4 7 Sub image scheme 57 4 3 Sub pixel interpolation The displacement vector of the correlation has an integer pixel resolution with an uncertainty of 1 2 pixel Two approaches have been identified to achieve sub pixel resolution for image correlation process One possible approach is to upsample both images then conduct the correlation on the upsampled image pairs For example to achieve 1 20 pixel subpixel resolution for the correlation on two sub images of 256x256 pixels both sub images have to be upsampled to 20 times larger images with the size of 5120x5120 pixel The correlation computation reguired on the image pairs of 5120x5120 is enormous and very time consuming The other approach is to compute the correlation on the image pairs without being upsampled and then interpolate in the region near the peak location to extract the peak location in sub pixel resolution At first it may look like try to extract unmeasured information However the interpolation procedure can be justified by the reasoning that the correlation image is generated by matching thousands of individual speckles in the image pairs Each pair of speckle image element gives an integer pixel displacement The actual peak
94. ection E between two surface points 8 apart is less than 0 2mm Substituting E 0 2mm L 203 2mm and assuming the uncertainty of measuring the gauge length is u mm 2 2 u ae 1 ue z 20microstrain 203 2 203 2 Thus the estimated uncertainty of the optical strain sensor is about 20 microstrain 107 Appendix D Strain Sensor User s Manual D 1 Introduction D 1 1 Background Information of the Optical Strain Sensor The Optical Strain Sensor utilizes the principle of speckle correlation to measure the subject surface strain The major components of the system are two identical modules rigidly attached side by side by steel channels Each of the modules emits a laser beam to the subject surface and the reflected speckle images are captured by the cameras in the sensor The system automatically analyzes the speckle images that are taken before and after the surface deformation to extract the strain information and present the results to the user in real time The sensor can operates either in hand held or stationary state depending on the application D 1 2 Laser Safety The Laser Head emits visible red light beams The laser intensity is low and can not damage human skin However looking directly into the laser beam can cause injury Laser medium Diode Radiant Power lt 5 milliwatt Wavelength 632 8 nanometers D 1 3 System and Power Requirement 120V 60HZ AC or 5V DC Windows 2000 o
95. ed correlation image with 20 times higher resolution than the original correlation image However the inverse Fourier transform of the full zero padded pattern calculates the interpolation value at every location while we are only interested in the upsampled information in the proximity of the peak Furthermore computation of the inverse Fourier transform of the full zero padded pattern is not practical due to the extremely large size of the padded pattern Alternatively the software calculates the upsampled information around the peak area only by employing a matrix implementation of the inverted Digital Fourier Transform 256 20 256 20 Figure 4 8 Zero padding interpolation The Digital Fourier Transform is defined as Deng 2008 N 1N 1 x k k X i i wM kk 0 1 N 1 4 4 i 0 i 0 where w e If X i i is the zero padded spectrum pattern then x k k represent the interpolated correlation image matrix that is being calculated The matrix form of Equation 4 4 can be written as Xyxn WyxnX nxnWyxn 4 5 59 where X x is the matrix form of the zero padded spectrum pattern x is matrix form of the interpolated correlation image matrix that we are trying to obtain and W is a N by N matrix whose k i entry is w e i e 1 1 1 1 1 2 N 1 w w w 2 4 2 N 1 Wry 1 w We ae MW 4 6 N 1 2 N 1 N 1 N 1 l w WP okt WAAR Using the matrix form of the Fouri
96. eeeeeseeennaeeeeeeees 80 6 2 Comparison with a Whittemore gauge during compressed concrete beam strain measure ment oe Ge i Ew Fy Fd FF BCT EF FU ties 82 6 3 Comparison with an electrical resistance strain gauge during compressed concrete beam strain mMeds re mie nyf a au O GRON Cd O dn SA ee cued Y NO GN ONN 84 6 4 Application of the optical strain sensor to a prestressed concrete member 85 6 4 1 Surface strain measurement using the second generation prototyp 87 6 4 2 Surface strain measurement using the fourth generation prot0typ e 90 6 5 Transfer length measurement of prestressed railroad tie eee eeeesseeceeeeeeeseeeenneeeeeeeees 91 Chapter 716 One LU S100 au ee haus ae Y ee aves ee an YN a 97 Relenences a A O GG lS a FA 99 Appendix A Hardware Components List the fourth generation prototype 99 103 Appendix B Specifications and SolidWork model of the laser speckle strain sensor the fourth PENETALION PIOLOLYPE TH m m REN SS S a a ai aaisan 104 Appendix C Uncertainty Analysis essseeeeeseeessesrersssreessssreessserersssrretsssrteseseteesssereesssrreesseret 105 Appendix D Strain Sensor User s Manual eesssseseesseesesseessserersssrressssrresrsereessserersssrressssree 108 Appe dix E Source coden au a CN CL Y YD GO 118 List of Figures Faer Whittemore aauc ee RG GO aay ON FE y
97. een developed The new method utilizes laser speckle patterns that are generated and digitally recorded at various points along the prestressed concrete member User friendly software incorporating robust and fast digital image processing algorithms was developed by the author to extract the surface strain information from the captured speckle patterns Based on the laser speckle measurement technigue four 4 successively improved generations of designs have been made A prototype was fabricated for each design either on an optical breadboard for concept validation or in a portable self contained unit for field testing For each design improvements were made based on the knowledge learned through the testing of the previous version prototype The most recent generation prototype incorporating a unigue modular design concept and self calibration function has several preferable features These include flexible adjustment of the gauge length easy expansion to two axis strain measurement robustness and higher accuracy Extensive testing has been conducted in the laboratory environment for validation of the sensor s capability in concrete surface strain measurement The experimental results from the laboratory testing have shown that the measurement precision of this new laser speckle strain measurement technigue can easily achieve 20 microstrain Comparison of the new sensor measurement results with those obtained using traditional strain gauges Whittemo
98. ement Tie 2 Side B eeeeeesrereesrrerererrerrsere 95 Figure 6 21 Cross tie surface strain measurement Tie 3 Side A eeeeereeesrerrsrrerererrrrrseree 96 xi List of Tables Table 5 1 Data of the camera distortion experiMeNt eee eeeeseeeececceeeeseeesnneeeeeeeeeeesesnneeeeeeeees 64 Table 5 2 Error caused by the sensor misalignment eescceeeeseeceeeessneeeeeesseeeceessaeeeeesenaeeees 66 Table 5 3 Calibration data from camera A and camera B c eee eeceeeeesneeeeeeesseeeeeeenaeeeeensnaeeees 12 Table 5 4 Experimental data of the Auto Calibration method demonstration 999 77 Table C 1 Experiment data for uncertainty amalysis eeecesssccccccceeeeeeeenneeceeeeeeeeeeeennaeeeeeeees 106 xii Acknowledgements I wish to express appreciation to my advisor Dr B Terry Beck for his friendship and guidance I would also like to thank Dr Robert J Peterman and Dr Chih Hang John Wu with whom I have worked throughout my graduate study for their valuable advice and insight into my research Thank Dr Yougi Wang and Dr Ruth Douglas Miller for reviewing my thesis There are many collaborators to whom I owe my appreciation These include Rob Murphy Steven Hammerschmidt and Ed Volkmer with whom I conducted extensive testing and many experiments using the laser speckle strain sensor Trevor Heitman and Ryan Benteman also provided excellent technician support for the fabrication of the sens
99. emental displacements of the speckle image corresponding to 0 1 mm surface displacement are calculated 71 Table 5 3 Calibration data from camera A and camera B a Cumulative x1 Cumulative yl Incremental x1 Incremental yl displacement displacement displacement displacement Surface pixel pixel pixel movement mm 2308 on mba om nuj ou sef o nn ore 201 onaj e o da maf ou mnu ew WED ON Cumulative x2 Cumulative x2 Incremental x2 Incremental x2 displacement displacement displacement displacement Surface pixel pixel pixel movement mm 23 12 22 80 P0986 12 8 sae 21 96 22 40 12 92 i iw 11 96 0 12 0 84 16 48 2 04 1 06 Average 224 56 72 By Eguation 4 17 we have x _ M cos g M sin 4 X J E M sin M cos Y vaj M cos g M sin X y G M sin 0 M cos Y Alternatively Eguation 4 18 and Eguation 4 19 can be written as and and Using the data in Table 5 3 fls uk and p l 22 324 a 0 204 B 22 456 a 0 206 4 18 4 19 4 20 4 21 4 22 4 23 solving Equation 4 22 and Equation 4 23 we obtain the value of parameters f a 8 Thus the displacement vectors x y and x y detected by the two cameras can be converted to the displacement vectors X Y and X Y on the object plane by using Equation 4 20 and Equation 4 21 As
100. er NULL browse pszDisplayName m_strPath GetBuffer MAX_PATH browse lpszTitle Please select a folder LPITEMIDLIST lpItem SHBrowseForFolder amp browse if lpItem NULL return A 138 m_ strPath ReleaseBuffer if SHGetPathFromIDList lpItem m_ strPath GetBuffer MAX PATH false return m_path SetWindowText m_strPath m_ strPath ReleaseBuffer void CStrainDlg OnButtonMeasure UpdateData TRUE b_keepgoing TRUE if m_measurementlist GetSel 0 gt 0 amp amp m_measurementlist GetCount gt 1 if IDNO AfxMessageBox You are going to resume or retake the referenc image shots Proceed MB _YESNO return pThread AfxBeginThread RUNTIME CLASS CUIThread T while b_keepgoing TRUE b_redo FALSE Gl m_Spin SetPos m_pointnumber UpdateData TRUE Measure pThread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL Sleep 100 if m pointnumber datapointnumber amp amp b_redo FALS break if b_redo FALSE amp amp b_keepgoing TRUI Gl Gl m_pointnumber void CStrainDlg OnButtonLock UpdateData TRUE if m_bLocked 1 139 m_bLocked 0 m_cbLoc GetDlgI GetDlgI GetDlgI GetDlgI GetDlgI GetDlgI else if Ce Ce Ce Le Ue Ue m I m I m I m I m I m I DC_ROUND1 gt E
101. er transform implementation the value of any single point on the interpolation image can be calculated For example the value of the entry i k in the interpolation image is calculated by x i k Win row amp xx Wi kth column 4 7 Thus the area near the peak location can be interpolated using Eguation 4 7 to yield the desired peak location to sub pixel resolution It is notable that the zero padding interpolation method makes use of the information from all the pixels value in the correlation image thus it is more accurate and suffers no peak locking error problem 4 4 Refreshing reference It is possible to have the camera take multiple speckle images as the object continuously undergoes displacement and then use an incremental method to increase the measurement range Initially the first speckle image would serve as a reference baseline image and every newly taken image would be correlated with it to extract the object surface motion information As the object surface displaces further from the initial position the correlation coefficient decreases due to the de correlation effect that happens when the two images share less similarity Before the correlation coefficient drops to a predefined threshold indicating significant de correlation the reference image would be replaced by the newest speckle image Following this procedure it should be possible to recover a well defined correlation peak Further speckle displaceme
102. es GetSize i pCB gt InsertString 1 files i Right files i GetLength defaultfolder GetLength 1 int nIndex 0 if nIndex pCB gt FindString nIndex Dlg m_DataSetName CB_ERR pCB gt SetCurSel nIndex if File m_hFile CFile hFileNull File Close datapointnumber Dlg m_DataPointNum File Open defaultfolder Dlg m_DataSetName index txt CFile modeCreate CFile modeWrite CString temp temp Format d datapointnumber File WriteString temp File WriteString n File Close OnSelchangeDataSetList void CStrainDlg UpdateMeasurementList CString texttemp CString ReadMeasurementString while m_measurementlist GetCount gt 0 147 m measurementlist DeleteString O int i 0 if File m hFile CFile hFileNull File Close if File Open defaultfolderr AYWW rdatasetnamer AAindex txt TRUE File SeekToBegin File ReadString ReadMeasurementString while File ReadString ReadMeasurementString FALSE if i ref_index CFile modeRead m_measurementlist InsertString i ReadMeasurementString RGB 255 0 0 texttemp Reference is ReadMeasurementString m_refname SetWindowText texttemp else m_measurementlist InsertString i ReadMeasurementString i measu
103. esley Longman Inc Helena Huiqing Jin W Y L 2006 Strain Measurement of Microsamples Using Laser Interferometry International Mechanical Engineering Congress and Exposition pp 563 567 Chicago Illinois USA ASME Herzig H P 1997 Micro optics Elements Systems and Applications Taylor Francis Books Ltd Hong Nan Li G D Z S 2007 Strain transfer analysis of embedded fiber Bragg grating sensor under nonaxial stress Optical Engineering http mathworld wolfram com BesselFunctionoftheFirstKind html n d L X Yang R W 1999 Strain and stress analysis by means of a novel sensor MicroStar International Workshop on Video Controlled Materials Testing Nancy France Malo K B 2008 Planar Strain Measurements on Wood Specimens Experimental Mechanics vol 49 575 586 100 ME 53 Laser speckle extensometer manual n d Retrieved from http www messphysik com index php id 16 amp L 1 Merzbacher A D 1996 Fiber optic sensors in concrete structures a review Smart Materials and Structures MTS LX laser extensometer 2009 Retrieved from http www mts com ucm groups public documents library dev_003699 pdf Mufti A A 2003 Integrated sensing of civil and innovative FRP structures Progress in Structural Engineering and Materials vol5 115 126 MUSPRATT M A 1969 STRAIN MEASUREMENT IN REINFORCED CONCRETE SLABS Strain 152 156 Naaman A E 1982 Prestressed concrete analysis a
104. ess than 1 minute to calibrate the parameters 79 Chapter 6 Validation and application of the laser speckle strain Sensor The purpose this chapter is to demonstrate that the new developed optical strain sensor is a general strain measurement device that can be readily used not only in the laboratory but also in the harsh environment of the prestressed concrete industry Furthermore it reguires with minimum surface preparation A series of laboratory setups were fabricated and used to conduct direct comparisons with various conventional measurement technigues including Whittemore gauge and the electrical resistance strain sensor These experiments were conducted both indoors laboratory and outdoors field in order to validate the ability of the optical strain sensor to measure concrete surface strain with high resolution and consistency After the validation the sensor was further applied to the real field measurement for the diagnostic testing of prestressed concrete members and in particular prestressed railroad cross ties 6 1 Validation using a two concrete block system Shown in Figure 6 1 is the manual motion system used for verifying the performance of the optical strain sensor in a concrete surface measurement Two small pieces of concrete block were positioned side by side with approximately 8 inches apart The concrete block shown on the left was attached to a manual traverse system whose displacement was measured by a
105. essly the sensor must be able to provide rapid working speed and not reguire any special training of the workers The characteristics desired for a strain sensor suitable for diagnostic testing of prestressed concrete members in the field include e Robustness e Portability e Adequate sensitivity and dynamic range e No contact to the surface e Insensitivity to out of plane motion of the surface e Insensitivity to environmental temperature fluctuation e Removable from the surface during downtime 1 2 Literature review of strain measurement technigues In this section several available strain measurement technigues are discussed 1 2 1 The Whittemore gauge Gauge Length Figure 1 1 Whittemore gauge The Whittemore gauge as shown in Figure 1 1 is a mechanical strain gauge that has been widely used for measuring surface strain of concrete structures for decades Before a strain reading can be made small steel circular buttons with a precision pinhole at the center called points are bonded on the concrete surface by using epoxy as shown in Figure 1 2 The Whittemore gauge measures the distance between the pinholes of successive pairs of points Prior to the surface deformation a set of reference length measurement are made representing the unstrained positions of the points Then a second measurement is taken after the surface deformation The difference between the second measurement and the reference lengt
106. f 0 colL 0 rowL 0 colR 0 rowR 0 CButton pCBmode CButton GetDlgItem IDC HANDHOLD pCBmode gt UpdateData TRUE 157 if pCBmode gt GetCheck BST_CHECKED workmode 0 else workmode 1 m_tStartTime GetTickCount clock while b_keepgoing 1 amp amp b_redo FALSE m_tEndTime GetTickCount clockQ dElapsed m_tEndTime m_tStartTime if peakflag TRUE amp amp dElapsed gt 0 Il peakflag TRUE amp amp maxvalLhalf gt thresholdhalf amp amp maxvalRhalf gt thresholdhalf pThread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL PlaySound MAKEINTRESOURCE UIDR_WAVE1 AfxGetResourceHandle SND_AS YNCISND_RESOURCEIS ND_NODEFAULT break LucamTakeS ynchronousSnapshots hS ynchronousSnapshots ppFrames tplL gt imageData char ppFrames 0 rect cvRect width 2 bs 2 height 2 bs 2 bs bs cvSetlmageROI tplL rect cvCopy tplL tplLtemp NULL filterproduct tplLtemp blockfilter tplfiltered memcpy img1_half float tplfiltered gt imageData bs bs sizeof float fftwf_execute fft_img1_halfsize maxvalLhalf O colindex_store 0 rowindex_store 0 for colindex 0 colindex lt colnum colindex for rowindex 0 rowindex lt rownum rowindex fftwcopy img1_fft_halfL img1_fft_half bs bs 2 1 phase_correlation_block fft_halfL colindex rowindex img1_fft_half poc_halfL cvMinMaxLoc poc_halfL amp minvalLtemp amp maxvalLhalftemp
107. f the wire The vibrating wire gauge is very simple in design Two anchors are installed on the specimen surface and the two ends of the wire are attached to the anchors Once the stress o of the wire is Known using Eguation 1 1 the strain of the surface can be found too assuming the wire deformation faithfully follows the surface deformation The advantage of the wire vibration gauge is that the gauge wire can be removed from the specimen which makes it a suitable tools for long term monitoring of strain change in the concrete structure The major drawback of the wire vibration gauge is its sensitivity to ambient temperature It is reported that a 1 degree temperature change causes a 20 change in the strain measurement Neild 2005 In some situations the specimen temperature changes rapidly either due to ambient temperature change or due to active heating to the concrete mix to expedite the cure process Unless the thermal expansion factor of the specimen is the same of that of the wire measurement error is introduced It is possible to compensate the error caused by the temperature change but doing so greatly complicates the system 1 2 4 Fiber optics strain sensor Fiber optics based measurement technigues are very versatile As many as 60 different guantities including temperature pressure and strain can be measured by fiber optics Fuhr 2000 For strain measurement Fiber Bragg Grating is one of the most popular methods i
108. ficient resolution and accuracy to make automated transfer length measurement possible in the concrete railroad cross tie manufactory industry The chapters in this dissertation are arranged as follows e Chapter 2 Theoretical background of the laser speckle strain measurement Theoretical modeling of the speckle will be presented using Fourier optics The principle behind the 5 axis motion measurement which serves as the foundation of the design of the laser speckle strain sensor will be discussed The digital image correlation technigue which is the crucial part of the data analysis will also be presented e Chapter 3 Hardware design of the optical strain sensor Various different hardware designs of the laser speckle strain sensor will be described in detail Their advantages and drawbacks will be compared Technical detail of the individual components including laser head CCD camera lens and the alignment mechanism used by the sensor will be presented e Chapter 4 Software development of the optical strain sensor The preprocessing of the captured speckle images will be presented A detailed explanation of how the digital correlation procedure is conducted to extract the relative shifting of the speckle patterns in sub pixel resolution will be presented along with the various technigues that are implemented to speed up the correlation computation 14 e Chapter 5 Calibration The general procedure for the calibration of the laser
109. ftwf_free imgl_fft_halfL fftwf_free imgl_fft_halfR free imgl_ half free img2_ half free res_half fftwf_free imgl_fft_half fftwf_free img2_fft_half fftwf_free res_fft_half if File m_hFile CFile hFileNull File Close void CStrainDlg OnClose CDialog OnClose void CStrainDlg OnButtonApply 137 UpdateData TRUE if m_gain gt 23 m_gain lt 0 1 l m_exposure lt 0 001 return if m_bPreviewing if LucamStreamVideoControl m hCameraA STOP_STREAMING NULL amp amp LucamStreamVideoControl m hCameraB STOP STREAMING NULL MessageBox Unable to Apply Update MB OK return LucamDestroyDisplayWindow m_hCameraA LucamDestroyDisplayWindow m_hCameraB if LucamSetProperty m_hCameraA LUCAM_PROP_EXPOSURE m_exposure 0 AfxMessageBox Unable to set exposure value if LucamSetProperty m_hCameraA LUCAM_PROP_GAIN m gain 0 AfxMessageBox Unable to set gain value if LucamSetProperty m_hCameraB LUCAM_PROP_EXPOSURE m_exposure 0 AfxMessageBox Unable to set exposure value if LucamSetProperty m_hCameraB LUCAM_PROP_GAIN m gain 0 AfxMessageBox Unable to set gain value void CStrainDlg OnButtonSet folder BROWSEINFO browse ZeroMemory amp browse sizeof browse browse hwndOwn
110. h a Whittemore gauge during compressed concrete beam strain measurement An experiment was conducted to measure the surface strain of a small concrete beam under different compressional loads by using both the optical strain sensor the second generation prototype and the Whittemore gauge for direct comparison A concrete beam of length 300mm and 90mm by 90mm sguare cross section was mounted on the compression test setup as shown in Figure 6 4 The concrete beam was loaded with two different compressional loads at 2 000 bs and 4 000 bs At each load level the deflections between two fixed points mounted on the beam surface was measured by the Whittemore gauge The laser speckle strain sensor was also used in handheld mode to measure the surface deformation at the same time for comparison purposes The results from the optical strain sensor and the Whittemore gauge are both shown in Figure 6 5 It can be seen that the readings by the two sensors have excellent agreement with differences between the results below 6 microns The repeatability of the 5 measurements shown using the laser speckle strain sensor at the same load level are excellent too 82 Figure 6 4 A concrete beam under compression Optical sensor Vs Whittemore 0 080 0 070 9 id E Whittemore 0 060 Optical Handheld 1 0 050 oN r Optical Handheld 2 0 040 a Optical Handheld 3 o a nen e Optical Handheld 4
111. h is divided by the gauge length 203 2mm 8 giving the strains on the concrete surface When a reasonable strain profile is reguired tens of points must be bonded onto the concrete surface which is very time consuming and labor intensive Furthermore the measurement results are heavily influenced by the users habits and skills Experience shows that different users can produce readings that are greatly different It requires a considerable amount of training and experience to achieve consistent and repeatable results Figure 1 2 Prestressed concrete with metal points mounted on the surface 1 2 2 Electrical resistance strain gauge Another traditional gauge used to measurement concrete surface strain is the electrical resistance strain gauge MUSPRATT 1969 It employs the principle that metallic conductors subjected to mechanical strain exhibit a change in their electrical resistance By converting mechanical strain into an electronic signal the electrical resistance strain gauge can measure strain to guite high resolution Figure 1 3 Electrical resistance strain gauge In general the electrical resistance strain gauge is precise reliable and easy to use However the technigue has several disadvantages The technigue reguires gluing the gauge on the specimen surface Since the gauge has contact with the specimen surface it may influence the surface strain and cause measurement error It is difficult to bond the gauge
112. half of the areas around point A and point B are illuminated This results in two simultaneous side by side speckle patterns Figure 3 4 on the CCD camera generated by point A and point B respectively 31 AM AB Concrete Surface NY MS Mirror 1 M2 Stop 1 M1 7 0 Diode Laser Stop 2 Mirror 3 M3 Beam Expander Stop 3 Lens L3 B2 Miror 4 CCD Camera M4 Speckle Pattern images Figure 3 3 Single module design A2 B2 Figure 3 4 Image splitting The prototype of single module design and its interior view dust cover removed are shown in Figure 3 5 and Figure 3 6 respectively The optical system was built using the Cage System components from Thorlabs company for rapid fabrication The camera is a Marlin 32 F145B CCD camera with a resolution of 1392 by 1040 pixels All the optics components the camera and the laser head are mounted on a single piece of 8 x2 steel base The sensor were enclosured in a metal box with the dimension of 11 x7 x4 The mass of the sensor is 7 lbs i ty O AAN Se ee o ec o es eee cs a Figure 3 6 Interior view of the prototype based on the single module design During the measurement the optical strain sensor is first positioned onto the concrete surface before the detensioning The CCD camera then captures a speckle image of the two side 33 by side speckle patterns as shown in Figure 3 4 which are generated by po
113. hanism 50 Chapter 4 Software development for the optical strain sensor The software of the optical strain sensor was coded using Visual C for flexible hardware control and high speed computation A user friendly interface was developed that is described in detail in the software manual in Appendix D This chapter mainly discusses the data processing of the raw speckle image pairs captured by the cameras of the optical sensor to extract the strain information from the object surface The data processing for each separate camera module consists of several steps as shown in the Figure 4 1 Image acquisition Blur detection No blur blurred 7 lt r Peak lt threshold Histogram equalization l Apply hanning window Image correlation with Pyramid scheme __ Peak gt threshold Subpixel interpolation Strain calculation and report the result Figure 4 1 Image processing diagram 4 1 Preprocessing The raw speckle image usually suffers blur noise and imbalance of intensity The direct application of a digital correlation algorithm at this early stage could result in loss of correlation and broad correlation peak that gives low accuracy in the prediction of image shifting from surface strain The goal of preprocessing is to make the raw images more suitable for analysis by a digital correlation algorithm 51 e Blur detection The acquisition of an image by the CCD camer
114. he video image captured by the digital camera is analyzed by the image processing algorithms to locate the centers of the dots or the edges of the lines During the test the centers of the dots or the edges of the lines are followed automatically by the software Their coordinate changes are used to extract the specimen strain information Malo 2008 F COMPUTER Ne eg cry LOADING UNIT CAMERA coomwwms Figure 1 6 Video extensometer configuration Malo 2008 Since the surface strain is measured by tracking a center of the mark fine marks must be applied to the surface such as a 7x7 grid of 0 5 mm diameter dots as described in wood surface strain measurement Malo 2008 For a material with irregular or soft surface the application of marks may not be practical Some other Video extensometers such as the MTS LX Laser Extensometer MTS LX laser extensometer 2009 use tapes instead of marks to tag the surface displacement The tape that attaches to the specimen surface has strip spacing on it The extensometer determines the surface strain by measuring the extension of the strip spacing The technique is not a real non contact measurement method since the tape contacts the specimen surface It is possible that the strip and the specimen extend or shrink by different amounts due to a creep effect so that the strain measured from the tape does not faithfully represent the actual specimen strain The resolution is limited
115. hough it is possible to compensate for the error by recording an appropriate temperature change of the connection bar during the operation of the strain sensor the compensation may not be accurate and may make the system far more complicated e Change of the optical imaging system In the optical imaging system the relative positions of various components are supposed to be fixed However since the optics were mounted on one edge of the L shaped steel bracket and the camera were on the other edge the deformation of the bracket could change the relative positions of the optics and the camera and in turn cause measurement error by the sensor More importantly it is difficult to predict how much error will be introduced by the deformation of the bracket since this will depend on the temperature distribution of the steel bracket thus the compensation of this error could be difficult to implement in practice 41 3 3 2 The fourth generation prototype The fourth generation prototype was also based on dual module design and was developed with multiple measures to minimize the effects of thermal expansion on the sensor strain measurement Figure 3 12 The fourth generation prototype based on dual module design Figure 3 12 shows the fourth generation prototype For this design the two modules were attached rigidly to each other using two carbon fiber rods whose thermal expansion coefficient is 1 10 of that of steel This significant
116. i for j 0 j lt width jt k if itshifty gt height 1 itshifty lt 0O j shiftx gt width 1 j shiftx lt 0 tmp 0 else tmp 1 0 productimg_data k tmp float img data k void CStrainDlg OnConcrete m_exposure 4 CEdit pEd CEdit GetDlgItem IDC_EXPOSUR LJ pEd gt SetWindowText 4 void CStrainDlg OnSteel m_exposure 18 CEdit pEd CEdit GetDlgItem IDC_EXPOSUR Gl lt pEd gt SetWindowText 18 168 void CStrainDlg INTfilter IplImage img CvMat filter IplImage productimg int shiftx int shifty int Tsp Gy KF float tmp get image properties int width img gt width int height img gt height uchar img_data uchar img gt imageData uchar productimg_data uchar productimg gt imageData for i 0 k 0 i lt height i for j 0 j lt width j k if itshifty gt height 1 it shifty lt 0 j shiftx gt width 1 j shiftx lt 0 tmp 0 else tmp cvmGet filter it tshifty j shiftx productimg_data k uchar tmp float img_data k 0 5 void CStrainDlg OnOpendatafile ShellExecute NULL open defaultfolder datasetnamet xls NULL NULL SW_S HOWNORMAL void CStrainDlg OnCalibrateIntensity if m_bPreviewing FLOAT m Lum LONG flags BOOL rt
117. iew window Start Preview MB_OK CDelayMessageBox mbox this mbox MessageBox Calibating Intensity 5 TRUE CDelayMessageBox MBIcon MBICONNONE long dumb 0 LucamGetProperty m_hCameraA LUCAM_PROP_EXPOSURE amp m_exposureA amp dumb LucamGetProperty m_hCameraB LUCAM_PROP_EXPOSURE amp m_exposureB amp dumb if AutoA 0 m_exposureA m_exposureB if LucamStream VideoControl m_hCameraA STOP_STREAMING NULL amp amp LucamStream VideoControl m_hCameraB STOP_STREAMING NULL m_cbPreview Enable Window TRUE m_cbPreview SetWindowText _T Preview m_cbMeasure Enable Window TRUE else MessageBox Unable STOP previewing video Stop Preview MB_OK LucamDestroyDisplayWindow m_hCameraA LucamDestroyDisplayWindow m_hCameraB m_bPreviewing FALSE void CStrainDlg phase_correlation_block fftwf_complex fftl fftwf_complex fft2 IplImage poc float tmp get image properties float fft_size float bs bs setup pointers to images float poc_data float poc gt imageData for int i 0 i lt bs bs 2 1 i res_fft_half i O fft2 i O fft1 i 0 fft2 i 1 C fft1 iJ 1 res_fft_half i 1 fft2 i O fft1 iJ 1 CfFE2 i 1 fft1 i O tmp sgrt pow res_fft_half i 0 2 pow res_fft_half i 1 2 res_fft_half i 0 tmp 171 res_fft_half i 1 tmp obtain the phase correlation array fftwf_execute
118. if pAllFrames NULL MessageBox No memory for frames for i 0 i lt CAMNUM i 154 ppFrames i pAllFrames i width height hSynchronousSnapshots LucamEnableSynchronousSnapshots CAMNUM hCameras pParams m_cbPreview EnableWindow FALSE m_cbPreview SetWindowText _T Preview m_cbApply EnableWindow FALSE CString refmeasurementname int n n m_measurementlist GetTextLen ref index m_measurementlist GetText ref_index refmeasurementname GetBuffer n refmeasurementname ReleaseBuffer cRefFilenameL Empty cRefFilenameL Format s_ dL bmp defaultfolder datasetname refmeasurementname m_pointnumbe r cRefFilenameR Empty cRefFilenameR Format s_ dR bmp defaultfolder datasetname refmeasurementname m_pointnumbe r if f FindFile cRefFilenameL AfxMessageBox Left Reference image doesn t exist Can t do correlation LucamDisableSynchronousSnapshots hSynchronousSnapshots free pAllFrames b_keepgoing FALSE return FALSE if f FindFile cRefFilenameR AfxMessageBox Right Reference image doesn t exist Can t do correlation LucamDisableSynchronousSnapshots hS ynchronousSnapshots free pAllFrames b_keepgoing FALSE return FALSE int colindex rowindex IplImage poc_halfL 0 155 IplImage tplL 0 IplImage refL 0 IplImage poc_halfR 0 IplImage tplR 0 IplImage refR 0 IplImage poc 0
119. ifft_res_halfsize normalize and copy to result image for i 0 1 lt bs bs i poc_data 1 res_half i fft_size deallocate FFTW arrays and plans void CStrainDlg ShowImage IplImage img int width img gt width int height img gt height IplImage img_temp img_temp cvCreateImage cvSize width height IPL_DEPTH_32F 1 double minv maxv CvPoint minl max cvMinMaxLoc img amp minv amp maxv amp minl amp maxl 0 cvScale img img_temp 1 0 maxv minv 1 0 minv maxv minv cvNamedWindow image 0 cvShowImage image img_temp cvResize Window image width height cvWaitKey 0 cvDestroyWindow image cvReleaselmage amp img_temp void CStrainDlg CreateLargeHanning int M int N CvMat filter double pi 3 1415926 float temp CvMat Ma cvCreateMat M 1 CV_32FC1 CvMat Mb cvCreateMat 1 N CV_32FC1 for int row 0 row lt M 2 row temp float 0 5 1 cos pi float row float M 2 172 cvmSet Ma row 1 temp cvmSet Ma M row 1 1 temp for int col 0 col lt N 2 col temp float 0 5 1 cos pi float col float N 2 cvmSet Mb col temp cvmSet Mb 1 N col 1 temp cvMatMul Ma Mb filter cvReleaseMat amp Ma cvReleaseMat amp Mb return TRUE void CStrainDlg OnSetreference int selindex nIndex selindex m_measurementlist GetCurSel CString texttemp m
120. inDlg OnButtonStop m_bStop 1 void CStrainDlg upsampling CvArr in int factor int roff int coff CvArr out int N in input FT complex spetrum 2 channels factor upsampling factor int i j CvMat Ma Mb Mc realtemp imaginarytemp temp float pi 3 14159f int nor factor N noc factor N upsampled output matrix size CvSize size size cvGetSize in int nr size height nc size width input matrix size int nr 4 nc 4 144 CvMat kr cvCreateMat nor nr CV_ 32FC2 CvMat kc cvCreateMat nc noc CV_32FC2 cvZero kc cvZero kr Ma cvCreateMat nor 1 CV_32FC1 for i 0 i lt nor i cvmSet Ma i 0 itroff 2 pi nr factor Mb cvCreateMat l nr CV_ 32FCl for i 0 i lt nr i cvmSet Ma itnr 2 0 i nr nr nr 2 Mc cvCreateMat nor nr CV_32FC1 cvMatMul Ma Mb Mc realtemp cvCreateMat nor nr CV_32FC1 imaginarytemp cvCreateMat nor nr CV_32FC1 for i 0 i lt nor itt for j 0 J lt nr j cvmSet realtemp i jJ cos cvmGet Mc i j cvmSet imaginarytemp i jJ sin cvmGet Mc i j cvReleaseMat amp Ma cvReleaseMat amp Mb cvReleaseMat amp Mc cvReleaseMat amp realtemp cvReleaseMat amp imaginarytemp Ma cvCreateMat nc 1 CV_32FC1 for 1 0 i lt nr i cvmSet Ma itnr 2 0 itnr nr nr 2 Mb cvCreateMat 1 noc CV_32FC1 for i 0 i lt noc i cvmSet M
121. index stepsize cvSetImageROI refL rect cvCopy refL refblockL NULL cvResetImageROI refL rowindex stepsize bs bs i i i filterproduct refblock blockfilter reffiltered memcpy imgl_half fftwf_execute float fft_imgl_halfsize 189 reffiltered gt imageData bs bs sizeof float Ye fft_halfL colindex rowindex bs 241 fftwf_complex fftwf complex bs fftwcopy imgl_fft_half fft_halfL colindex rowindex bs cvSetlmageROI refR rect cvCopy refR refblockR NULL cvResetImageROI refR filterproduct refblockR blockfilter reffiltered memcpy imgl_half float fftwf_execute fft_imgl_halfsize fft_halfR colindex rowindex fftwf_complex fftwf complex bs bs 2 1 fftwcopy imgl_fft_half fft_halfR colindex rowindex bs fftwf malloc reffiltered gt imageData bs bs sizeof fftwf malloc sizeof bs 2 1 float sizeof bs 2 1 CvPoint minlocL maxlocL minlocR maxlocR maxlocLhalf maxlocRhalf CvPoint minlocLtemp minlocRtemp maxlocLhalftemp maxlocRhalftemp double minvalL maxvalL minvalR maxvalR maxvalLhalf maxvalRhalf double minvalLtemp minvalRtemp maxvalLhalftemp maxvalRhalftemp int colL rowL colR rowR int colindex_store
122. ing force was transferred to the concrete member post dententioning readings were taken The two sets of readings were compared correspondingly to extract the strain information at each location which in turn was used to plot the strain profile of the cross tie for the transfer length determination The application of the laser speckle strain sensor to the cross tie transfer length measurement was not successful in our first trip to CXT concrete cross tie production plant The software failed to find correlation between the corresponding speckle image pairs Thus it could not extract the surface strain information along the cross ties This was due to a de correlation effect that was caused by the dramatic physical change of the cross tie surface when the ties went through the saw cutting machine Here the cross tie surface undergoes severe abrasions including washing scrubbing wiping and vacuuming as shown in Figure 6 17 The change of the concrete surface s microscopic profile causes significant de correlation to the speckle image pairs To reduce the de correlation effect microscopic reflective particles were bonded to the cross ties to serve as artificial speckle before the initial readings were taken as shown in Figure 6 18 The particles were much less vulnerable to the severe abrasions than the concrete surface itself and helped keep the correlation that was critical for the laser speckle strain sensor to be functional in the this e
123. int A and point B respectively These two patterns are denoted as A1 and B1 The sensor is then removed from the concrete surfaces After the detensioning the optical sensor is positioned mounted back onto the surface The camera captures another speckle image with two speckle patterns which are denoted as A2 and B2 By applying a cross correlation technigue to the pair of speckle patterns A1 and A2 before and after the detensioning processes the displacement AA can be extracted The displacement AB can be extracted from pattern B1 and pattern B2 in a similar fashion As shown in Figure 3 7 the axial surface strain between point A and point B can thus 5 AB AA be determined from FU where L is the gauge length 203 2 mm 8 inches for the current setup Prestressed Concrete Beam Beforce Detensioning L 8 in 203 2mm B gt After Detensioning Figure 3 7 Strain Measurement The single module strain sensor was extensively tested in a laboratory environment as well as on actual prestressed concrete beams in the field The measurement results were compared with those obtained using traditional gauges Whittemore gauge and the electrical resistance strain gauge and showed good consistency The experiments and applications are discussed in detail in Chapter 6 A provisional patent was also granted for the design 34 The laser speckle strain sensor based on the single module design although proved to be a
124. ion is conducted using multiple subimage pairs instead of the full image pairs is due to the fact that correlation on small size images are more robust to image rotation Correlation inherently is suitable for tracking the translation of the image but not the rotation Weixin Zhao 2004 For a large image the disruption of the image rotation on the correlation is more severe because a small rotation can cause large translations on the image boundary For a small size image the effect of image rotation is less severe However too small template will consist of limited size of speckle pattern and this makes it hard for the searching process to identify a match In the software a template of 256x256 size was used which provided robustness against the image rotation and enough resolution The detailed procedure is described below The captured reference baseline image I1 is split into many sub window areas of 256 pixels by 256 pixels size denoted as Refblock O lt i lt N 1 where i is the block index and N is the total subimage number A sub window of 256 pixels by 256 pixels size is also extracted from the center area in the after deformation image I2 denoted as Defblock Since the sizes of the subimages are small it can be assumed that there is no change within the sub image and the shift between the sub image pairs is uniform at every point Thus using the phase correlation procedure described in Chapter 3 The N subimages Refblock and th
125. ion of the waiting time is defined in the Reference Delay in Measurement Settings section During this period the user can align the sensor to the marks that the user 116 makes in Step 3 When the delay time is over the initial reading will be made and the sound of ding is played For the readings other than the initial readings try to align the sensor to the same marks while the program continuously scans the subject surface Wiggle the sensor to facilitate the scanning process When the program detects a correlation the measurement will be made automatically A ding sound will be played and the measurement result will be displayed on the screen After the measurement on the current point is made the spin button Point ff will increment automatically proceeding to the measurement of the next point Repeat step 4 117 Appendix E Source code Strain cpp include stdafx h include Strain h include StrainDlg h ifdef DEBUG define new DEBUG_NEW undef THIS_FILE static char THIS_FILE __ FILE _ endif BEGIN_MESSAGE_MAP CStrainApp CWinApp IH AFX_MSG_MAP CStrainApp I AFX_MSG ON_COMMAND ID HELP CWinApp OnHelp END_MESSAGE_ MAP CStrainApp CStrainAppO CStrainApp theApp BOOL CStrainApp InitInstance AfxEnableControlContainer Standard initialization ifdef _AFXDLL Enable3dControls Call this when using MFC in a shared DLL else Enable3dControlsStati
126. ion traverse setup was reguired and simple visual manual positioning was adeguate This is very important for the measurement of prestress concrete strain in the field Since the laser speckle method utilizes the concrete surface characteristics to measure displacements the surface strains can be guickly determined without the time consuming process of adhering gage points to the concrete surface 87 In addition to using the laser speckle strain sensor gage points were also bonded to the concrete beam surface in order to provide a direct comparison between the two methods of strain measurement Figure 6 11 Experiment setup for transfer length measurement of prestressed concrete member using the second generation prototype c T 800 O 700 Ss E 600 500 s B 400 o E 300 o o 200 2 2 100 Whittemore z Optical LSI c o 0 o 0 20 40 60 80 100 Figure 6 12 Comparison of strain measurements immediately after de tensioning of a pre tensioned specimen 88 The results from this beam test are shown in Figure 6 12 It can be seen that the laser speckle strain sensor results in much smoother data with less scatter than that generated from the existing surface strain measurement technigue using the Wittemore gauge The laser speckle technigue has been validated on members cast in both indoor and outdoor operations
127. is assembled The analysis in Chapter 5 shows that even a 0 4 degree orientation difference can cause an error comparable to the nominal resolution of the sensor if not taken into account To deal with the problem caused by the orientation difference of the two cameras a more complicated calibration algorithm has to be implemented Overall the traits of the dual module design overweigh its disadvantages Efforts were subsequently expended on its implementation and perfection of the dual module design Both the third and the fourth generation prototypes are fabricated based on the dual module deign They are discussed below 39 3 3 1 The third generation prototype Figure 3 10 shows the third generation prototype that is based on dual module design The two modules left and right handed shown are attached rigidly to each other using two L shaped cross section steel channels 1 flange length 7 long and eight 4 20 mounting screws Figure 3 11 shows the interior view of an individual module A L shaped bracket was used as the base for the module The optics are mounted on one edge of the bracket and the camera are on the other edge The module was covered with a plastic enclosure made from rapid prototyped ABS Plastics The dimensions of the individual module were 5 x4 x2 and the total weight of the prototype was 5 2 Ibs a a aA 2 ae See Figure 3 10 The third generation prototype based on dual module design
128. ity distribution function of subjective speckle 08 06 J QaDri id Dr Ad 04 4 0 2 7 Figure 2 2 Plot of speckle intensity distribution function 2 2 Review of the 5 Axis motion measurement system Axial strain measurement is accomplished by taking the differential of the relative displacements between two points on the object surface A typical speckle measurement is fulfilled by first illuminating the associated specimen surface with coherent light laser The random reflections from the surface features roughness generate a grainy speckle pattern image at the camera plane This speckle pattern could be thought of as a fingerprint of the illuminated area in the sense that the speckle pattern produced by a given surface region is unique Furthermore when the surface undergoes movement or deformation the speckle pattern in the image plane will also move or deform accordingly This tracking feature is the basis of the displacement measurement of the laser speckle technology However the speckle displacement at the camera plane usually is not only sensitive to the object surface displacement but also to other axis movements of the object surface especially the out of plane rotations tilt and yaw which result in error in the strain measurement To extract the displacements accurately without being affected by other axis movements a 5 axis motion measurement technique was developed that is able to
129. l speckle photography DSP They relate to different methods of producing and processing the speckle image The ESPI technigue measures the object surface displacement or deformation by detecting the corresponding phase change of the light wavefronts reflected from the surface just as a conventional Michelson interferometer does The image taken in an ESPI system called a speckle interferogram is produced by interfering the speckle radiation reflected from an object surface with a reference 10 light field either a uniform coherent light beam or another speckle field Dainty 1975 In practice the speckle interferograms are taken both before and after the object displacement or deformation A characteristic fringe pattern can be obtained by subtracting the two speckle interferograms The fringe spacing corresponds to a 27 phase change of the wavefronts resulted from the object surface deformation as is the case with a Michelson interferometer with a mirror displacement Thus the surface deformation and displacement can be readily determined by counting the number of fringe changes As an interferometry method the ESPI technique has high resolution on the order of a fraction of a light wavelength and the resolution is not limited by the resolving power of the imaging system Samala 2005 Helena Huiqing Jin 2006 As fringe counting is involved this method has a 27 ambiguity limitation that is the periodical fringe pattern resembles
130. lFrames NULL MessageBox No memory for frames for i 0 i lt CAMNUM i ppFrames i pAllFrames i width height 184 pParams hSynchronousSnapshots m_cbPreview UpdateWindow EnableWindow FA m_cbApply EnableWindow FALSI QO 1S LucamEnableSynchronousSnapshots CAMNUM Dp F Gl m_cbPreview SetWindowText T Preview i pThread gt PostThreadMessage WM_SHOWDIALOG NULL NULL in E Edit pEB t delay C Edit p if m delay lt EB gt UpdateData 0 TRUE m_delay 0 for int k 0 k lt m_delay k Sleep 1000 if b_keepgoing FALSE if b_keepgoing TRU IplImage rL IplImage rR rL cvCrea Gl 0 0 Imag cvSize rR cvCrea IplImage rL Imag _bI cvSize urcheck 0 IplImage r rL_blurchec R_bl k cvCreat lurcheck 0 b_redo TRU amp amp b_redo FALSI k cvCr Imag at rR_blurchec cvZero rL blurcheck cvZero rR blurcheck Imag i BOOL no_blur 0 while if b_ break m tStartTime LucamTa keepgoing no_blur 0 b_redo TRU GetDlgItem IDC DELAY Gl break Gl lt width height IPL_DEPTH_8U 1 width height IPL_DEPTH_8U 1 cvSize width height
131. lImage poc void upsampling CvArr in int factor int roff int coff CvArr out int N CString m_strPath LUCAM_SNAPSHOT IsSettings BOOL m_bLocked void CleanUp HICON m hlIcon BOOL CreateHanning int M int N CvMat filter BOOL flag int d CvMat hanningfilter CvMat blockfilter HANDLE hCameras CAMNUM LUCAM_SNAPSHOT pParams CAMNUM Array of ptrs to the param struct LUCAM_SNAPSHOT params CAMNUM UCHAR ppFrames CAMNUM Array of pointers to frames UCHAR pAllFrames BYTE pBmpBuffer HANDLE hSynchronousSnapshots HANDLE m_hCameraA m_hCameraB BOOL m_bConnected BOOL m_bPreviewing BOOL m_bSnapping LUCAM FRAME _FORMAT m_lffFormat CFileFind f CString cRefFilenameL cRefFilenameR float m_fFrameRate clock_t m_tStartTime clock_t m_tEndTime clock_t dElapsed Generated message map functions IH AFX_MSG CStrainDlg virtual BOOL OnInitDialog afx_msg void OnPaint afx_msg HCURSOR OnQueryDragIcon afx_msg void OnButtonConnect afx_msg void OnButtonPreview afx_msg void OnClose 123 afx_msg void OnButtonApply0O afx_msg void OnButtonSetfolder afx_msg void OnButtonMeasure afx_msg void OnButtonLock afx_msg void OnButtonStop afx_msg void OnCreateDataSet afx_msg void OnEditchangeDataSetList afx_msg void OnSelchangeDataSetList afx_msg void OnNewMeasurement afx_msg void OnDeleteMeasurement afx_msg void OnSelchangeMeasurementList afx
132. lane displacements AA and AB of the two nearby surface points A and B as discussed in Section 3 1 E and the surface strain between point A and point B is determined by the equation T where E AB AA is the relative deflection between point A and point B and L is the gauge length of 203 2 mm 8 inches for the current setup The uncertainty comes from three sources 1 uncertainty of surface displacement measurement 2 uncertainty of digital dial gauge used to calibrate the optical sensor and 3 uncertainty of gauge length L measurement The three uncertainties are estimated individually below and the total or combined uncertainty is obtained by using error propagation methods The manual motion system as shown in Figure 6 1 was used to estimate the uncertainty of the surface displacement measurement of the optical sensor The concrete block on the right was stationary The concrete block on the left was attached to a manual traverse system whose displacement was indicated by a digital dial gauge of resolution of 0 001mm Shars 303 3506 The displacement of the concrete block was also measured by the laser speckle strain sensor The experiment was conducted by displacing the concrete block from 0mm to 1 000mm with increments of 0 100mm The data is listed in Table C 1 and plotted in Figure C 1 The residual plot is shown in Figure B 2 which shows the deviation of the measured displacement from the laser speckle strain senso
133. lightly different place than the light rays close to the axis When using a spherical lens for imaging an object surface to the camera plane the images obtained exhibit radial distortion as shown in Figure 5 1 The light beams farther from the lens axis bend more than the beams close to the lens axis Thus the distortion increases from 0 at the center of the image to higher level at the edge of the image 62 Figure 5 1 Camera image distortion Obviously the image distortion introduces measurement error to the strain sensor since the image shift is not linear to the surface movement if the image is distorted To reduce this effect and keep the cost low at the same time triplet lenses were chosen as the imaging lenses of the sensor The triplets lenses consists of one concave lens and two convex lenses whose aberration effect cancels out producing an almost aberration free image To evaluate the scale of the error that is caused by the image distortion an experiment was conducted which is described Image pairs initial baseline reading and post movement reading were taken by the camera using the triplet lens as the imaging lens The object surface under observation undergoes primarily only linear motion Thus the pixel shift of the image pairs at any location of the images should be identical if without image distortion The pixel shifts were calculated at 3 regions in each of 5 image pairs One of the image pairs are shown in Figure 5 2
134. lly designed with a large margin of safety with the nominal load 2 or 3 times of the actual load A A Mufti 2008 However the aging of the structure and excessive usage always lead to a reduced factor of safety For instance it is reported that more than 200 000 bridges in United States and 30 000 bridges in Canada are operating at a deficient condition due to the inadeguate maintenance and excessive loading Mufti 2003 It is risky to keep the aged civil infrastructures in service without reliable information of them Either to design and build civil infrastructures of extended lifetime without compromising the safety or increased cost or to effectively qualify their performance in the term of safety it is important to find a convenient way to collect the information about the structure performance either at time of the construction or at the time of service One of the factors that are always used to evaluate the performance of concrete member is the stress or strain information in the member For example the evaluation of the bridge health is usually done by measuring the in situ strain responding to traffic flow Ceravolo 2005 Of particular importance to civil infrastructure is prestressed concrete which is usually fabricated by casting concrete mix around already tensioned steel strands After the casting process is complete and the concrete has hardened a detensioning procedure is undertaken by cutting the reinforcing strands at
135. location is determined by averaging the large guantity of displacement vectors provided by these speckle pair matching Thus the information that is used to extract the sub pixel location of the peak is already embedded in the correlation image There are many interpolation methods available to extract sub pixel peak location including parabolic peak fitting Gaussian peak fitting etc These peak fitting methods only use the neighboring points information for the interpolation and suffer from the peak locking effect i e the sub pixel peak detected tends to be close to the integer pixel location Sung 2004 A more advanced interpolation called zero padding fitting is used in the image analysis for the optical strain sensor software First the correlation image is transformed to the freguency domain by means of a Fourier transform Suppose the Fourier transform pattern of the correlation image has a size of 256 by 256 pixels To achieve 1 20 pixel peak location resolution zero valued pixels are appended to the high freguency end so that we have a Fourier pattern of 20 256 by 20 256 pixel as shown in Figure 4 8 After padding zeroes in the high freguency area there is no change in the real and imaginary parts of the Fourier spectrum nor in the phase spectrum The only change is in the densified spatial sampling freguency If an inverse Fourier 58 transform is applied to the zero padded spectrum pattern the result will be an interpolat
136. ly reduces the sensor gauge length change caused by temperature change Figure 3 13 shows the interior view of an individual module The optics and the camera s CCD chip were attached together as one piece The whole imaging system of each individual module is mounted rigidly on an aluminum base This prevented the relative positions of the imaging system components from change In addition the main heat generation component which is the camera circuit board was positioned far away from the rest of the imaging system Air vent slits were made at the top of the enclosure to facility heat dissipation to outside by free convection 42 The module was enclosured in a box fabricated from ABS plastic using a Rapid prototyping system The dimension of an individual module is 4 x3 x2 and the total weight of the prototype is 2 6 Ibs Vent slits Camera circuit Camera Optics Figure 3 13 Interior view of the fourth generation prototype An experiment was conducted to quantitatively evaluate the effect of thermal expansion on the strain measurement This was accomplished by measuring strain on a non deforming surface using the third generation prototype and the fourth generation prototype Tests started from a state initially in equilibrium with the ambient temperature Ideally both sensor prototypes should report zero since the specimen surface undergoes no deformation However as the camera dissipates heat and the temperature of the sen
137. maxlocL x x maxlocL x y gt bs 2 yftbs maxlocL y y maxlocL y locL x cL y cvCreateMat bs bs CV_32FC1 CreateMat bs bs CV_32FC1 k O i lt bs it J lt bs jtt k i j float res_fft k 0 i j float res_fft k 1 194 CvMat in cvCreateMat bs bs CV_ _32FC2 int factor 25 CvMat out cvCreateMat 2 factor 2 factor CV_32FC2 upsampling in factor y 1 factor x 1 factor out 2 cvRel cvRel leaseMa leaseMa real cvCrea im cvCrea cvSplit ou Compute cvPow real cvPow im cvAdd real cvPow real t amp real t amp im teMat factor 2 factor 2 CV 32FCl1 teMat factor 2 factor 2 CV_32FC1 t real dm 0 0 y the magnitude of the spectrum Mag sgrt Re 2 r Im 2 real 2 0 im 2 0 im real NULL pL ead OS cvMinMaxLoc real amp minvalL amp maxvalL amp minlocL amp maxlocL 0 cvRel cvRel leaseMa leaseMa subxL widt t amp real t amp im h 2 bs 2 xx 1 float maxlocL x factor subyL yy rect cvRec height 2 bs 2 1 float maxlocL y factor t oldstartR x oldstartR y bs bs cvSetlmag cvCopy refR cvResetImageROI refL ROI refR rect vefblockR NULL filterprodu filterprodu ct refblockR blockfilter reffiltered ct tplR_backup blockfilter tplfiltered
138. me situations the sensor does not work in a stationary manner For example for the prestressed concrete beam strain measurement the sensor is reguired to be removed from the concrete beam surface before the 22 detensioning and then be put back on the surface after the detensioning due to the fact that the detensioning process involves a release of a 30000 bs 13607 kg traction force which could damage the sensor if left on the concrete surface Thus the repositioning of the laser speckle sensor will cause the distance between the lens of the optical system to the concrete surface to vary inevitably This arises a problem that the change of the object distance will affect the system sensitivity Figure 2 4 shows the measured systematic sensitivity to object translation at different object sensor distances It shows that the system sensitivity varies significantly when the sensor is at different depth positions relative to the object plane This is obviously an undesired characteristic for a sensor 70 68 F 66 F 64 F 62 7 60 F Sensitivity pixels mm 97 98 99 100 101 102 103 Object distance mm Figure 2 4 System sensitivity changes due to the object distance change Zhao 2006 However it is proved that under the condition of collimated and normal illumination the system sensitivity becomes insensitive to the out of plane object movement Zhao 2006 Experiments as shown in Figure 2 5
139. mp UpdateData TRUE CFileFind f uble subxR double subyR 8 bits 186 GI n O SND_ASYNC SND_R cfilenameL cfilenameR eSynchronousSnapshots hSynchronousSnapshots double subyL hCameras 0 m_hCameraA hCameras 1 m_hCameraB for i 0 i lt CAMNUM i params i format pixelFormat LUCAM PF_8 params i format subSamplexX 1 params i format subSampleY 1 params i format height height params i format width width params i format xOffset 0 params i format yOffset 0 params i exposure m_exposure 50 ms exposure params i gain 23 params i strobeDelay 0 0 unused params i timeout 3000 0 3000 ms params i useHwTrigger FALSE Set this to true for hardware triggered setup with daisy chaining params i useStrobe FALSE Set this to true if daisy chaining cameras params i exposureDelay 0 params i shutterType LUCAM SHUTTER_TYPE GLOBAL pParams i amp params i params 0 exposure m_exposureA params 1 exposure m_exposureB pAllFrames UCHAR malloc CAMNUM width height if pAllFrames NULL MessageBox No memory for frames for i 0 i lt CAMNUM i ppFrames i pAllFrames i width height hSynchronousSnapshots LucamEnableSynchronousSnapshots CAMNUM pParams m_cbPreview EnableWindo
140. mplex fftwf_malloc sizeof fftwf_complex bs bs imgl fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs img2 fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs res fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 132 imgl_half float malloc sizeof float bs bs imgl_fft_half fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 2 1 imgl_fft_halfL fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 2 1 imgl_fft_halfR fftwf_complex fftwf_malloc sizeof fftwf_complex bs bs 2 1 img2_half float malloc sizeof float bs bs img2_fft_half fftwf_complex fftwf_malloc sizeof fftwf complex bs bs 2 1 res_half float malloc sizeof float bs bs res_fft_half fftwf_complex fftwf_malloc sizeof fftwf_complex bs 2 1 FILE planfile planfile fopen plan wisdom r if planfile NULL fftwf import_wisdom from file planfile fclose planfile fft imgl halfsize fftwf_plan dft_r2c_2d bs bs imgl_half imgl_fft_ half FFTW ESTIMATE fft img2 halfsize fftwf_plan dft_r2c_2d bs bs img2_half img2_ fft_ half FFTW ESTIMATE ifft_res_halfsize fftwf_plan dft_c2r_2d bs bs res_fft_half res_half FFTW ESTIMATE fft_imgl fftwf_ plan dft_2d bs bs
141. n recent years The optical fiber used in this method is fabricated in a way that there is a periodic variation of the refractive index in the fiber core called a Bragg grating Suppose the grating interval is G as shown in Figure 1 4 The Bragg wavelength is calculated by A 2n G 1 2 where n is the average refractive index When incident light passes through the Bragg grating only the light of the wavelength egual to the Bragg wavelength will be reflected and the light of the wavelength other than the Bragg wavelength transfers through Since the Bragg wavelength A is dependent on the Bragg grating interval G which is in turn directly related to the applied strain the applied strain can be determined by measuring the Bragg wavelength i e the wavelength of the reflected light Grating Interval G Incident Light Transferred light Reflected light Bragg wavelength A Figure 1 4 Fiber Bragg Gratings Merzbacher 1996 input te transmitted r reflected strain induced eta JK signal shift Sg AN y Xe A Figure 1 5 Transmission and reflection of Fiber Bragg Grating Merzbacher 1996 Fiber Bragg Grating method has several advantages over the traditional eletrical resistance strain sensor First it is immune to electromagnetic interference from the industrial enviroment Merzbacher 1996 In addition it does not suffer from any light intensity fluctuation in that it measures the strain based on the change of
142. nFile Open defaultfoldert tdatasetnamet description txt CFile modeRead while DescriptionFile ReadString temp2 FALSI Description Descriptiont temp2 r n pED gt SetWindowText Description Gl 149 DescriptionFile Close CSpreadSheet SS defaultfolderr AW rdatasetnamer xls SS BeginTransaction CStringArray sampleArray sampleArray RemoveAll CString ctemp sampleArray Add Point for int i l i lt datapointnumbertrl irr ctemp Format Sd 1 sampleArray Add ctemp SS AddHeaders sampleArray SS Commit void CStrainDlg OnNewMeasurement CMeasurementNameDialog Dlg if Dlg DoModal IDOK measurementname Dlg m_MeasurementName if File m_hFile CFile hFileNull File Close Sheetl if File Open defaultfoldert datasetnamet index txt CFile modeWrite TRUE File SeekToEnd File WriteString measurementname File WriteString n File Close m_pointnumber 1 m_Spin SetPos m_pointnumber UpdateMeasurementList void CStrainDlg OnDeleteMeasurement 150 CString ReadMeasurementString Ef File m hFile CFile hFileNull File Close CStdioFile temp if File Open defaultfoldert datasetnamet index txt CFile modeReadWrite TRUE amp amp
143. nableWi DC_ROU D2 gt EnableWi DC_ROU D3 gt E T i nableWi T i nableWi Eal nableWi DC_ROUND5 gt DC_ROU N N DC_ROUND4 gt E N N D6 gt E i nableWi k SetWindowText T Lock ndow 1 ndow 1 ndow 1 ndow 1 ndow 1 ndow 1 m_round gt 0 amp amp m_round lt 5 m_bLocked 1 m_cbRou m_cbLoc GetDlgI GetDlgI GetDlgI GetDlgI GetDlgI GetDlgI switch m_rou case 0 Ce Ue Ce Ce Ue Ue nd m I m I m I m I m I m I EnableWindow DC_ROUND1 gt E FALSE k SetWindowText _T Unloc T i nableWi DC_ROUND2 gt l Eal nableWi DC_ROU D3 gt E T i nableWi Eal nableWi DC_ROU D5 gt E T i nableWi N N DC_ROUND4 gt N N T i nableWi DC_ROU nd GetDlgItem IDC_ ROUND1 gt break case l GetDlgItem IDC_ ROUND2 gt l break case 2 GetDlgItem IDC_ ROUND3 gt l break case 3 GetDlgItem IDC_ ROUND4 gt 1 break case 4 GetDlgItem IDC_ ROUND5 gt break D6 gt E k ndow FAL ndow FAL EnableWindow TRU MAU N FH FH FH DW ndow FAL ndow FAL ndow FAL ndow FAL U U NU NU UU
144. nce strain gauge ESG sensor 84 The results from the optical strain sensor and the electrical resistance strain gauge are shown in Figure 6 7 and the differences between these two different sensor measurements are shown in Figure 6 8 It can be seen that the readings by the two sensors have excellent agreement with differences between the results falling below 6 microstrain 500 450 FSG 400 350 Optical Sensor 300 250 200 Microstrain 150 100 50 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 Load Ibs Figure 6 7 Measurement results of surface strain Differences Between Optical and ESG 7 6 5 4 3 amp B2 nl YZ 2 0 a 2000 4000 10000 12000 14000 18900 20000 2 3 4 Load Ibs Figure 6 8 Difference of the measurements between optical strain sensor and electrical resistance strain sensor 6 4 Application of the optical strain sensor to a prestressed concrete member Currently prestressed concrete is widely used in civil engineering infrastructure especially in applications that require extra length like bridges skyscrapers foundations pipes and piles pavements and water tanks due to its great advantage over traditional concrete in 85 sustaining tension Naaman 1982 Prestressed concrete is manufactured by casting around
145. nd design fundamentals New York McGraw Hill Neild S A 2005 Development of a Vibrating Wire Strain Gauge for Measuring Small Strains in Concrete Beams Strain 3 9 Peterman R 2007 The Effects of As Cast Depth and Concrete Fluidity on Strand Bond PCI Journal 72 101 R Wegner A 1999 The miniaturization of speckle interferometry for rapid strain analysis Proceedings of SPIE v 3824 pp 30 36 Munich Germany S C Liu K P 1995 Civil infrastructure systems research hazard mitigation and intelligent material systems Smart material structure 169 174 Samala P R Su M amp etc 2005 Strain measurement of a mouse bone by 3D electronic speckle pattern interferometry Annual SPIE Optics and Phootonics Symposium Proceeding of SPIE pp 58800C1 9 San Diego CA SPIE 101 Shapiro L G 2001 Computer Vision Prentence Hall Sung B J 2004 Sub pixel image interpolations for PIV Proceedings of 2004 Korea Japan Joint Seminar POSTECH Korea Weixin Zhao B T 2004 A novel optical technigue for measuring 5 axis surface movement Two and Three Dimensional Vision Systems for Inspection Control and Metrology Il Proceedings Vol 5606 pp 66 73 SPIE Zhao W 2006 A novel optical technigue for measuring 5 axis surface movement M S thesis Kansas State University Kansas 102 Appendix A Hardware Components List the fourth generation
146. neration protofyDB Gn og Ru Yg O grd Ydd og yd Dydd yd 42 3 4 Components of the optics GV SECT ii RG Gu AW oom y o SR FF eRoees 44 al aser Head ceu ed dow EU aU CC UL CG GU OF 44 342 CCD Camerina ud dad dd add dd yddu ddd de edie a 47 5 AS Ais MIMENE mechanisering esgere dd Cn gn GOF a Og penseetypegsasbeneen 49 Chapter 4 Software development for the optical strain SCNSOT essceeeeeesseeeeeesneeeeeeenaeeees 51 AN Preprocessing 5 i teaser dn debe UN AG CU y ONN Tn idee escent 51 4 2 Digital correlation procedure vc cccsssisieasedescieadessusanntcaaccavess COD YF YRS Ed SS NOD ANNE Ed O YF Yd 55 4 3 S b pixelinterpolatiOi is ueu ee edyn dug dean ynddo dya yd wd TN y URL o Gn ON 58 4 4 Refreshing referente eii Ydi Fi i YNYR YL YNG ecstssaanccchecdevedsvsanvaiesedevsioeessyadnvedacatesesnvaabeeddees 60 Chapter a CalibraHon se GG GR 62 5 1 Measurement error sources of the laser speckle strain sensor 9 1 iiiiiiei nn 62 a Distortion due tothe enau ee A FD I DEAL 62 5 1 2 Error due to misalignment cu WR A Ae as aA 65 5 2 Homography projection siiscesissiucsssccivcenaveduesiaivissacsceavssasatpneaacaioessuvvaneanssaaneesgvaisennsadbectayes 67 5 3 Two calibration methods for the strain SCNSOL eeeseeeeessneceeeeeseeeeesssaeeeeeseseeeeesenaes 71 Chapter 6 Validation and application of the laser speckle strain sensor 80 6 1 Validation using a two concrete block SYStOM eee eeeeeseeeceeeeeeeeseetnceeee
147. nsor Sensor module module module module Figure 3 18 Multiple modules setup 48 Figure 3 19 Rosette setup for two dimensional strain measurement 3 4 3 Alignment mechanism The measurement procedure generally consists of two steps 1 Take speckle images from the undeformed surface as an initial or baseline reading 2 Take speckle images from the deformed object surface Since the dynamic range of the optical strain sensor is 2mm to conduct step 2 of the measurement procedure a mechanism must be introduced to help the user align the sensor onto the position on the object surface within 2mm of where the initial reading takes place If the sensor is mounted on a track the alignment would be ensured by the track itself However in the case of manual operation there have to be some markings left on the object surface to guide the alignment This is needed to ensure correlation between the displaced image and the baseline image Such kind of markings must satisfy the following requirements e Inexpensive e Good use on concrete and metals e Must not fade or rub off Several possible marking mechanisms were evaluated for practicality e Fluorescent marking The idea is to use a fluorescent material to serve as the marker When the marker is illuminated by an ultraviolet UV light emitter that is mounted on the sensor it would 49 absorbs ultraviolet light and emits visible light signaling the user that the sensor is in alignmen
148. nt 60 could then be determined by adding up the speckle displacements before the replacement of the reference pattern Theoretically the measurement range is indefinite as long as every pair of consecutive speckle images does not exceed the threshold of significant de correlation 61 Chapter 5 Calibration This chapter discusses measurement error sources associated with the laser speckle strain sensor including distortion of the lens misalignment between the initial baseline reading and the second reading and misalignment between the two modules of the sensor Their effects on the strain measurement accuracy of the sensor are investigated Two calibration methods are proposed to correct these errors using the homography projection technigue the mathematical instrument that describes the relationship between the physical scene and the image captured by the camera 5 1 Measurement error sources of the laser speckle strain sensor 5 1 1 Distortion due to the lens No lens is perfect There are several kinds of aberrations associated with lens The most common one is spherical aberration Many simple lenses are made into the spherical profile for lower manufacturing cost There do exist parabolic lenses which have a more mathematically ideal profile with minimum aberration but they are is expensive and the lens selection is limited Spherical aberration causes the parallel light rays distant from the lens axis to be focused in a s
149. on of a vector is eguivalent to giving the vector a new description in a different coordinate system which is implemented by multiplying the vector by a sguare matrix of the appropriate size A two dimensional rotation of angle a is represented as a multiplication of the vector by a 3x3 matrix as shown in Eguation 4 8 Rotation in the three dimensional space is eguivalent to three two dimensional rotations on the X Y Z axes respectively Rotating on the X Y Z axis respectively with angles a 6 is given by the product of the three matrices R a R B R 6 1 0 0 R a 0 cos a sin a 4 8 O sin a cos a cos B 0 sin B R B O 1 0 4 9 sin B O cos P cos 0 sin 0 O R sin g cos 0 O 4 10 0 0 1 And 68 R R a R B R 6 4 11 The translation vector is the offset from the origin of the object coordinate to the origin of the camera coordinate system that is t O svjet Qcamera 4 12 Given a point Q X Y Z in the object coordinate system its coordinate g x y z in the camera coordinate can be expressed as q x y z RO X Y Z t Homography is a special case of the coordinates transformation Faugeras 1993 It is used to describe the projection from a two dimentional surface object surface to another two dimentional surface Camera surface as shown in Figure 5 6 gt 1 Camera plane Object plane Figure 5 6 Homography Projection from objection coordinate to camera coordin
150. on the other hand is based on intensity correlation By comparing two speckle images taken before and after surface deformation the in plane displacement vector resulting from the loading can be determined Once the complete displacement field is obtained it can be differentiated to obtain an in plane strain map DSP generally has lower resolution than ESPI but larger dynamic range The resolution is limited by the speckle size which typically ranges at the micrometer level and the resolving power of the imaging system There exists some strain measurement devices in the market utilizing DSP technigue One of them is ME 53 extensometer ME 53 Laser speckle extensometer manual Eduard Schenuit 2008 from Messphysik company It has two variations One version consists of two cameras and a servo drive that controls the motion of the cameras for a large gauge length setup as shown in Figure 1 10 the other version consists of a single camera for small area surface strain measurement The ME 53 laser speckle extensometer makes non contact strain measurement based on the DSP technique and does not require any surface marking However it is mainly designed for laboratory use The sensors must be mounted on a vertical track and the specimen must be installed on a laboratory bench This is to prevent the relative rigid motion between the sensor and the specimen Although the DSP technique is designed to measure in plane movement it is also commonly sensitive
151. onal calibration equipment 38 e Measurement range Unlike the single module design that requires mechanisms to prevent the speckle images of the two inspected area from being overlapped with each other the dual module design is free of this problem since each of the speckle images is captured by a separate CCD camera The measurement range is also doubled compared to that of the single module design that adopts the splitting image approach because the entire CCD array is devoted to measure the separate end point displacement The dual module design also has its drawbacks e Since the dual module design consists of two sets of nearly identical optical systems the total cost is doubled compared to the single module design e As an electronic component the image captured by the CCD camera is subject to drift due to various noises In the dual module design the two cameras each suffer their own image drift The sum of the independent drifts of the two speckle images from left and right modules is reflected as a distance change in the measurement causing error in the strain reading The single module design does not suffer from this problem due to the fact that both speckle images are captured by the same camera such that the camera drift affects both speckle images which in turn is canceled out in the measurement results e In practice the two cameras in the dual design will have a minor orientation difference no matter how well the sensor
152. or In this section two calibration methods are proposed to correct the misalignment errors that are discussed in Section 5 1 The first method translates the displacement vectors from the two different camera coordinates to the same coordinate systems by using the rotation matrix described in section 5 3 The calibration setup is shown in Figure 5 7 The sensor is mounted on a traverse stage that does horizontal movement The displacement is accurately measured by a dial gauge A piece of specimen for example a piece of concrete block is kept stationary in front of the sensor so that both observation points of the two cameras are on the specimen surface Furthermore throughout the experiment the specimen is not subjected to any load This ensures that the surfaces seen by the two cameras always displace at the same direction by an egual distance The next step is to displace the traverse stage from 0 to 1mm with increments of 0 1mm At every increment the images are recorded by both camera and are then analyzed by the cross correlation algorithm described in Chapter 4 The displacement of the traverse stage and the corresponding image shift of the cameras in the unit of pixel are shown in the Table 5 3 Since there is always an angle between the x axis of the camera coordinate system and the transverse stage displacement direction the speckle image taken by the camera has both x and y component At the last row of the tables the average incr
153. or The sensor has been greatly improved due to their work I am also grateful of Kansas Depart of Transportation KDOT University Transportation Center UTC Advanced Manufacturing Institute AMI and Precast Prestressed Concrete Institute PCI for their kind financial support xiii Dedication To my wife Grace xiv Chapter 1 Introduction 1 1 Background of strain measurement for civil infrastructure Civil engineering infrastructure comprises some of the most massively built assets in the world For centuries engineers have been trying to build more reliable and long lasting infrastructure from the Great Wall in China to the modern Interstate Highway System in the USA Many new materials and design concepts have been proposed aimed at reducing weight increasing spans achieving longer infrastructure life and lower cost However the civil engineering field has been conservative in adopting the new materials and technologies due to concern for compromising safety standards S C Liu 1995 Faber amp Stewartb 2003 This resulted in a relatively slow evolution of technologies in the field of civil engineering compared to those of other disciplines such as computer science and electronic engineering The barrier in adopting the new materials and technologies lies partly in the lack of convenient and reliable methods to evaluate their performance and implement safety control S C Liu 1995 In addition civil infrastructure is usua
154. orrelation 100 200 Peak 300 500 600 Figure 2 7 Phase correlation results for a typical speckle image pairs 27 Chapter 3 Hardware design for the optical strain sensor This chapter discusses the hardware design of the strain sensor Multiple factors must be brought into the consideration during this design stage The main objectives of the design are listed in the following e To develop a portable surface strain sensor capable of measuring surface strain The sensor dimension should be as small as possible and the weight should be light for the portability of the sensor e Measurement uncertainty to be on the order of 25 50 microstrain in order to have capability similar to the Whittemore gauge whose uncertainty is about 50 microstrain e Nominal range of measurement should be large enough to facility easy positioning and alignment of the sensor in handheld work mode e The sensor is aimed for a commercial product to replace the industrial standard Whittemore gauge Thus it should be easy to manufacture and assemble e The sensor is to be for use in harsh environment where various condition including extreme temperature humid vibration and dust pose challenge to the sensor s function For example one of the field application that the sensor has been applied to is the transfer length measurement of railroad cross tie in a manufacturing plant The temperature in the plant varies as much as 60 F through the year
155. otype was fabricated for each design either on an optical breadboard for concept validation or in a portable self contained unit for field testing For each design improvements were made based on the knowledge learned through the testing of the previous version prototype The most recent generation prototype incorporating a unigue modular design concept and self calibration function has several preferable features These include flexible adjustment of the gauge length easy expansion to two axis strain measurement robustness and higher accuracy Extensive testing has been conducted in the laboratory environment for validation of the sensor s capability in concrete surface strain measurement The experimental results from the laboratory testing have shown that the measurement precision of this new laser speckle strain measurement technigue can easily achieve 20 microstrain Comparison of the new sensor measurement results with those obtained using traditional strain gauges Whittemore gauge and the electrical resistance strain gauge showed excellent agreement Furthermore the laser speckle strain sensor was applied to transfer length measurement of typical prestressed concrete beams for both short term and long term monitoring The measurement of transfer length by the sensor was unprecedented since it appears that it was the first time that laser speckle technique was applied to prestressed concrete inspection and particularly for use in transfer leng
156. oyDisplayWindow m_hCameraB else Stop the preview reate pr view window Start Preview MB OK if LucamStreamVideoControl m hCameraA STOP STREAMING NULL amp amp LucamStreamVideoCont m_cbPreview EnableWindow TRU rol m_ hc ameraB STOP_STREAMING NULL m_ cbPreview SetWindowText _T m cbMeasure EnableWindow TRU else E Preview E MessageBox Unable STOP previewing video Stop Preview MB_ OK m_ bPreviewing FALSE Gl f void CStrainDlg CleanUp cvReleaseMat amp blockfilter if m bPreviewing LucamDestroyDisplayWindow m_hCameraA LucamDestroyDisplayWindow m_hCameraB if IDYES AfxMessageBox Currently previewing Do you wish to stop MB_YESNOCANCEL OnButtonPreview else return 136 if m_hCameraA NULL LucamCameraClose m_hCameraA if m_hCameraB NULL LucamCameraClose m_hCameraB if imgl_half NULL fffwf_destroy_plan fft_imgl fftwf_destroy_plan fft_img2 fftwf_destroy_plan ifft_res fftwf_destroy_plan fft_imgl_halfsize fftwf_destroy_plan fft_img2_halfsize fftwf_destroy_plan ifft_res_halfsize fftwf_free imgl fftwf_free img2 fftwf_free res fftwf_free imgl fft fftwf_free img2_fft fftwf_free res_fft f
157. p 4 a ind 4 em e 6 a a e 9 4 e e a r Figure 3 11 Interior view of the individual module of the third generation 40 This third generation prototype based on dual module design was extensively tested in laboratory environment Severe errors were observed in the measurement of surface displacement and strain Through investigation it was found that the significant errors were from a thermal expansion effect primarily related to the camera heating As an electrical device the speckle strain sensor has an inherent thermal expansion effect which resulted in an enormous displacement or strain The moment the sensor is turned on the electrical components mainly the camera running at 4 Watts starts to dissipate heat and causes the continuous thermal expansion on the hardware The effect of the thermal expansion on the strain sensor is two fold First there is a possible change of the sensor gauge length and second there is possible an accompanying change in the optical imaging system configuration e Change of the sensor gauge length Since the channel bars used in this third generation prototype that connect the two modules are steel whose thermal expansion coefficient nominally is 7 3 10 in in F every degree Fahrenheit increase in temperature will cause the steel bar or the sensor gauge length to expand as much as 7 3 microstrain which will in turn be falsely recorded by the sensor as the specimen deformation Alt
158. ped at the early stage of the optical strain sensor development It is important to the strain sensor development in that the object surface during deformation is usually subjected to 6 degrees of freedom movement However only the in plane displacement information is used to calculate the strain The motion of other axis especially the out of plane rotations act as error source to the strain measurement The 5 axis motion measurement system is able to separate the 5 axis movement so that the movement of each axis can be measured independently thus the effect of the out of plane rotations can be eliminated for the strain measurement To the end the digital image correlation technigue is discussed It is an image process technigue that estimates the relative shift of the speckle image pairs taken before and after the surface deformation such that the displacement or strain information can be extracted 2 1 Mathematical description of the speckle phenomenon Speckle is generated by illuminating a rough surface with coherent light as shown in Figure 1 7 The random reflected waves interfere with each other resulting in a grainy image as shown Figure 1 8 Figure 2 1 shows an imaging system of recording the subjective speckle filed by a CCD camera The object plane XY is illuminated by a laser light beam The reflected laser lights from the rough object surface is collected by the lens and then imaged onto the camera plane xy from the object The
159. r maxl x 1 factor out 2 Ipllmage real cvCreateImage cvSize factor 2 factor 2 IPL_DEPTH_32F 1 IplImage im cvCreateImage cvSize factor 2 factor 2 IPL_DEPTH_32F 1 cvSplit out real im 0 0 Compute the magnitude of the spectrum Mag sgrt Re 2 Im 2 cvPow real real 2 0 cvPow im im 2 0 cvAdd real im real NULL cvPow real real 0 5 cvMinMaxLoc real amp minv amp maxv amp minl amp maxl 0 x intx l double maxl x double factor y inty 1 double maxl y double factor cvReleaselmage amp real cvReleaseImage amp im cvReleaselmage amp peakarea cvReleaselmage amp fftlmg cvReleaselmage amp complexInput 174 cvReleaselmage amp imaginarylnput cvReleaseMat amp out void CStrainDlg PreProcess IplImage input Ipllmage output cvSmooth input output CV_MEDIAN 3 3 AdaptiveHist output output UpdateData TRUE if m_invert TRUE cvXorS output cvScalar 255 output cvThreshold output output 120 0 CV_THRESH_TOZERO cvEqualizeHist output output void CStrainDlg AdaptiveHist IplImage input Ipllmage output int histsize 4 CvRect rect int width input gt width int height input gt height int colnum floor width histsize int rownum 1 for int colindex 0 colindex lt colnum colindex for int rowindex 0 rowindex lt rownum rowindex rect cvRect colindex histsiz
160. r Windows XP 450 MHz Pentium III or higher 108 512 MB RAM USB 2 0 Port D 1 4 System Limitation Dynamic range 2mm Scanning rate 5hz Maximum ambient temperature 130 F D 2 Installation D 2 1 Components Checklist Sensor body USB Hub and power adapter USB cables 109 Figure D 1 Sensor body D 2 2 Hardware Assembly 1 Cable connection Connect the two USB cables to the two USB sockets of the sensor The match of the USB cable and the socket is random The connection diagram is shown in Figure 2 Important Do NOT plug the unconnected USB cable to the computer before installing the software 110 i o gt Figure D 2 Cable Connection Diagram Power Supply 2 Gauge length adjustment The gauge length of the strain sensor is egual to the distance of the two illuminated points The default gauge length of the strain sensor is 8 inches The user can adjust the gauge length to meet the reguirement of various applications To change the gauge length unscrew the 8 set screws at the bottom of the sensor slide the sensor modules along the carbon fiber rods till the desired gauge length is achieved tighten the 8 set screws to lock the gauge length Set Screws A a Figure D 3 Adjust gauge length D 2 3 Installing the Software wr D Close all application software that is running and then insert the Installation CD into your CD DVD ROM drive Doubl
161. r as a function of the displacement mm measured by the digital dial gauge 105 Table C 1 Experiment data for uncertainty analysis Dial gauge mm Laser speckle sensor mm Deviation mm 0 000 0 000 0 000 0 100 0 100 0 000 0 200 0 202 0 002 0 300 0 300 0 000 0 400 0 400 0 000 0 500 0 499 0 001 0 600 0 598 0 002 0 700 0 700 0 000 0 800 0 797 0 003 0 900 0 902 0 002 1 000 0 999 0 001 LSI sensor Vs Digital dial gauge 0 4 0 6 Digital dial gauge mm Figure C 1 Uncertainty analysis experiment data 106 Deviation 0 004 Uncertainty Interval 0 003 0 002 Deviation mm o o o o o o r o gt 0 002 0 003 0 004 Dial gauge mm Figure C 2 Deviation between optical sensor and digital dial gauge Figure C 2 also shows the uncertainty band covering all the residual points to within about 0 003 mm which is denoted as uu 9 003mm The uncertainty of the digital dial gauge iS u 0 001mm Using the propagation of errors method assuming these errors are independent the total uncertainty of the deflection measurement is given as Uy sym Uptaconene Uue 0 003 0 003 0 001 0 004mm gauge Since the strain is calculated by 7 the uncertainty of the strain measurement is Mei led Since the surface strain of the prestressed concrete beam is usually less than 1000 microstrain This means that the defl
162. r was positioned 0 7m distance from the sensor After the sensor was turned on the temperature inside the sensor and the laser pointing location on the laser profiler were monitored for 80 minutes Figure 3 17 shows the laser pointing drift to be as high as 40 microns as the temperature increased by 5 C due to the heat dissipated from the camera Thus the pointing stability can be calculated as 40um 0 7m 5 C z11 4urad C The result is guite close to the claimed rating of the laser head 46 Laser poitning drift at 0 7m distance drift um 0 T T T T 1 0 20 40 60 80 100 minutes after cold start Figure 3 17 Laser pointing stability test 3 4 2 CCD Camera Besides the compactness low power usage and low noise there are other critical features that must be taken into consideration with the camera selection for the optical strain sensor e Synchronous acquisition and global shutter The two cameras of the optical strain sensor must be able to take synchronous shots otherwise any rigid motion between the sensor and the object within the time of the two shots will be reflected in the strain measurement and cause significant error It is the same reason that a global shutter feature is necessary A CCD has millions of opto detectors with each one corresponding to a pixel With a global shutter all the opte detectors start and stop exposing at the same time eliminating the potential error
163. ra Each speckle pattern used only half area of the CCD chip as shown in Figure 3 4 However this approach causes the measurement range of the sensor be halved Weight The prototype sensor weighed 7lbs It makes the user fatigue quickly when the sensor was operated in manual mode and was handheld by the user 35 3 3 Dual module design In observing the drawbacks associated with the single module design the next generation of the sensor took the form of modular design which consisted of two identical modules that were attached to each other rigidly side by side The new sensor detects the displacement at two points usually 8 inches apart but adjustable on the specimen surface which is converted into the surface strain by dividing the net displacement by the gauge length The displacement measurement is based on the Digital Speckle Photography DSP technigue The principle of the measurement technigue is as follows A typical measurement is fulfilled by recording the speckle images before and after the object displacement or deformation The two images are then cross correlated and the peak position of the correlation image indicates the displacement of each ends A schematic of the dual module design is shown in Figure 3 8 The two individual modules are identical except that one is a right handed module and the other is a mirror arranged or left handed copy of the right handed module Each of the modules consists of a 5mW diode
164. re gauge and the electrical resistance strain gauge showed excellent agreement Furthermore the laser speckle strain sensor was applied to transfer length measurement of typical prestressed concrete beams for both short term and long term monitoring The measurement of transfer length by the sensor was unprecedented since it appears that it was the first time that laser speckle technique was applied to prestressed concrete inspection and particularly for use in transfer length measurement In the subsequent field application of the laser speckle strain sensor in a CXT railroad cross tie plant the technigue reached 50 microstrain resolution comparable to what could be obtained using mechanical gauge technology It was also demonstrated that the technigue was able to withstand extremely harsh manufacturing environments making possible transfer length measurement on a production basis for the first time Table of Contents Wu NF Y Y O O NN HR FA A FAR ix List of Tables i DL aes Ree ate A DN AU xii Acknowledgements id GYD GO Rd a sutra NY xiii Dedicatio Maii a epee O ee ee na xiv Chapter nroddu NO a A A AR ERE R 1 1 1 Background of strain measurement for civil infrastructure eeeeeeeeeeeeeeeerereeserrresererseseee 1 1 2 Literature review of strain measurement teChniques ccccccccceeeeseeeenneeeeeeeeeeeeeesnneeeeeeees 3 1 2 1 The Whittemore gaugei esineeseen iieii tiiir FARUS Cd NSF Fd 3 1 2 2 Electrical resistance strain Sa
165. rementname ReadMeasurementString int n m_measurementlist GetCount m_measurementlist SetCurSel n 1 OnSelchangeMeasurementList void CStrainDlg OnEditchangeDataSetList void CStrainDlg OnSelchangeDataSetList if File m_hFile CFile hFileNull File Close CEdit pED CEdit GetDlgItem IDC Description 148 m_strPath default foldert datasetname measurementname pED gt Clear CComboBox pCB CComboBox GetDlgItem IDC DataSetList pCB gt UpdateData TRUE int nIndex pCB gt GetCurSel pCB gt GetLBText nIndex datasetname if _access defaultfolder datasetname index txt 0 1 if File Open defaultfoldert tdatasetname index txt CFile modeRead TRUE File SeekToBegin CString temp File ReadString temp datapointnumber _ttoi temp gal c dit pED CEdit GetDlgItem IDC DataPointNumber CString pointnumber pointnumber Format d datapointnumber EA pED gt SetWindowText pointnumber m_Spin SetRange 1 datapointnumber m_Spin SetPos 1 ref_index 0 UpdateMeasurementList if _access defaultfolder datasetname description txt 0 1 CString temp2 CString Description Description Empty CStdioFile DescriptionFile Descriptio
166. reviewing FLOAT m_Lum LONG flags BOOL rt FLOAT m_Exp 2 0f m_Lum 60 0f flags LUCAM PROP_FLAG AUTO rt LucamSetProperty m_hCameraA LUCAM_PROP_EXPOSURE m_Exp flags if LucamSetProperty m_hCameraA LUCAM_PROP_AUTO_EXP_TARGET m_Lum flags MessageBox Failed to set exposure target flags LUCAM PROP_FLAG AUTO rt LucamSetProperty m_hCameraB LUCAM_PROP_EXPOSURE m_Exp flags if LucamSetProperty m_hCameraB LUCAM_PROP_AUTO_EXP_TARGET m_Lum flags MessageBox Failed to set exposure target if LucamCreateDisplayWindow m_hCameraA 200 WS_OVERLAPPEDWINDOW WS_VISIBLE 0 640 LucamCreateDisplayWindow m_hCameraB 500 WS_OVERLAPPEDWINDOW WS_VISIBLE if LucamStreamVideoControl m hCameraA 300 640 LucamStreamVideoControl m_hCameraB m_bPreviewing TRUE GI m cbPreview EnableWindow TRU 135 480 Preview B 480 Preview A START DISPLAY NULL NULL NULL amp amp NULL NULL START_DISPLAY NULL amp amp m_cbPreview SetWindowText T Stop m cbMeasure EnableWindow FALSE else MessageBox Unable start previewing video Start Preview MB OK else MessageBox Unable c LucamDestroyDisplayWindow m_hCameraA LucamDestr
167. ring large object deformation Applied Optics Applied Optics Ceravolo A D 2005 Monitoring and Response of CFRP Prestressed Concrete Bridge Sensing Issues in Civil Structural Health Monitoring 85 94 D A Gregory 1976 Basic physical principles of defocused speckle photography a tilt topology inspection technique Optics and Laser Technology Vol 8 Dainty J C 1975 Laser speckle and related phenomena New York Springer Verlag Berlin Heidelberg Deng L e 2008 Efficient image reconstruction using partial 2D Fourier transform Proceedings of the IEEE Workshop on Signal Processing Systems pp 49 54 Washington D C Metro Area USA SiPS 99 Eduard Schenuit R B 2008 Optical strain measurement on small specimens based on laser speckles Materials Science Forum 231 242 Faber M H amp Stewartb M G 2003 Risk assessment for civil engineering facilities critical overview and discussion Reliability Engineering amp System Safety 173 184 Faugeras O 1993 Three Dimensional Computer Vision The MIT Press Fuhr P 2000 Measuring with Light Part 2 Fiber Optic Sensing From Theory to Practice Sensors Goodman J W 1996 Introduction to Fourier Optics New York McGraw Hill Science Harris F J 1978 On the use of windows for harmonic analysis with the discrete Fourier transform Proceedings of the IEEE Vol 66 No 1 51 83 Hecht E 1998 In Optics Third Edition Addison W
168. rowindex_store CvPoint oldstartL oldstartR startL startR oldstartL cvPoint 0 0 oldstartR cvPoint 0 0 startL cvPoint 0 0 startR cvPoint 0 0 BOOL peakflag workmode peakflag FALSE double oldLpeak oldRpeak oldLpeak 0 oldRpeak 0 maxvalLhalf 0 maxvalRhalf 0 colL 0 rowL 0 colR 0 rowR 0 CButton pCBmode CButton GetDlgItem IDC_HANDHOLD pCBmode gt UpdateData TRUE if pCBmode gt GetCheck BST_CHECKED workmode 0 else workmode 1 m_tStartTime GetTickCount clock 190 while b_keepgoing 1 amp amp b_redo FALS m_tEndTime GetTickCount cl dElapsed m_tEndTime m tS if peakflag TRUE amp amp dEl maxvalLhalf gt thresholdhalf amp amp pT PlaySound MAKEINTRESOURC OURCE SND_NODEFAULT break Luca Gl lock tartTime apsed gt 0 peakflag TRU Gl lt amp amp maxvalRhalf gt thresholdhalf hread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL mTakeSynchronousSnapshots hSynchronousSnapshots E IDR WAVE1 AfxGetResourceHandl GI SND_ASYNC SND_R ppFrames tplL gt imageData char ppFrames 0 rect cvRect width 2 bs 2 cvSetlmageROI tplL cvCopy tplL height 2 bs 2 bs bs r tplLtemp cvResetImageROI tplL filterproduct tplLtemp memcp
169. separate the 5 21 axis movement so that the movement of each axis can be measured independently thus the effect of the out of plane rotations are mostly eliminated The 5 axis measurement principle is based on the fact that for subjective speckle there exist two characteristic planes behind the lens a tilt only plane and a translation only plane such that the speckle image at the tilt only plane is only sensitive to the tilt of the specimen and the speckle image at the translation only plane is only sensitive to the translation of the specimen D A Gregory 1976 These planes are shown in Figure 2 3 The 5 axis motion measurement technigue was developed during the early stage of the laser speckle strain sensor development Zhao 2006 The principle of the 5 axis motion measurement was later applied to the design of the optical strain sensor in which the camera is positioned at the translation only plane of the optical system to eliminate the effect of the surface motion other than the in plane displacement A detailed discussion of the 5 axis motion measurement technigue can be found in author s M S thesis Zhao 2006 Object surface plane tilt roll Ree P Bree Tilt_only f MR lens pime Translation only pitch Figure 2 3 Tilt Only Plane and Translation Only Plane 2 3 Insensitivity of system sensitivity to out of plane movement During the operation of the optical strain sensor in the field in so
170. shown in Figure 5 8 since the vectors X Y and X Y are with respect to the same coordinate the surface strain can be readily calculated by 13 0 0 L 4 24 where 0 0 Lis the distance between the two observation points on the object surface which is also defined as the gauge length of the sensor Y pr X1 Y Ob Oa L q X2 Y2 Figure 5 8 Strain calculation with orientation difference of the two camera coordinate systems The drawback of this method is that it is time consuming and requires addition equipments a traverse system and the dial gauge The calibration can usually be only done in the lab Considering the fact that every time when the user needs to adjust the gauge length it is necessary unattach and reattach the two modules causing the orientation difference of the two cameras change Therefore the sensor must be recalibrated every time when the gauge length is adjusted which is not convenient for the user In fact most commercial strain sensors require specific equipment for calibration For example the Messphysk company s laser speckle extensometer ME53 33 incorporates a calibration system into the sensor system as shown in Figure 5 9 which enables the end users to calibrate the sensor by themselves ME 53 Laser speckle extensometer manual However it makes the system bulky and the calibration procedure is still time consuming 14 Figure 5 9 Messphysk company s laser speckle
171. sioning of a pre tensioned specimen using laser speckle strain sensor 6 5 Transfer length measurement of prestressed railroad cross tie Prestressed concrete railroad cross ties are becoming increasingly popular in the United States and are an essential component for high speed railway lines Currently the production of the prestress concrete rail cross ties is a highly automatic process in the USA For instance in the CXT concrete cross tie production plant in Grand Island NE the production of the cross tie involves pouring concrete mix on a grid of strands of 386 feet length that are tensioned at 300 thousand lbs force in a heated bed The headed bed expedites the curing process of the concrete mix that typically takes days to a period as short as 8 hours After the concrete mix is cured the 91 386 feet long prestressed beam goes through a saw cut machine continuously and is cut into 8 6 long cross ties The plant is capable of producing several thousand cross ties each day In order for these prestress concrete ties to function adeguately over their expected service life the prestressing force must be fully introduced into the railroad tie at a location well before the rail load is applied Once again the length required to transfer the prestressing force into the concrete cross tie member is the Transfer Length However currently the concrete rail cross tie industry does not conduct transfer length measurements except occ
172. sor starts to climb the effect of the thermal expansion will be reflected by the non zero readings of both prototypes Figure 3 14 shows the measured deflection reading from the third generation prototype in pink curve and that from the fourth generation prototype in blue curve It can be seen that the deflection reported by the third generation prototype keeps increasing until it stabilizes at 12 microns Considering the system s nominal resolution of 20 microstrain see Appendix C with an 8 203 2mm gauge length which corresponds to 4 microns the error from thermal expansion is three times of the nominal 43 resolution of the sensor and must be reduced to make the sensor usable in real application In contrast the fourth generation prototype reported less than 2 microns throughout the experiment showing that the measures that were taken in the design of the fourth generation prototype successfully reduced the effect of the thermal expansion on the strain sensor performance Thermal drift test 0 014 0 012 0 01 0 008 generation z prototype 0 006 1 eThe fourth 0 004 S 0 002 0 won 100 200 300 minute Figure 3 14 Experiment to evaluate thermal expansion effect 3 4 Components of the optics system 3 4 1 Laser Head The laser head used in the optical strain sensor is a compact laser diode module It consists of a laser diode circuit collimating lens and the drive circuit packaged in
173. speckle strain sensor will be presented Then the effect of the sensor misalignment will be investigated An improved calibration method that is able to correct the error caused by the orientation difference of the two camera coordinates will be presented e Chapter 6 Conclusion The work that has been done for the development of the laser speckle strain sensor will be summarized The recommendation of future work based on the efforts presented here will be discussed The appendices at the end of the dissertation contain the references cited in the text The hardware components list and the specifications for the sensor along with the uncertainty analysis are also included An operation manual for the laser speckle strain sensor and the software source code are attached to the end of the dissertation 15 Chapter 2 Theoretical background of laser speckle strain measurement technigue Speckle is generated by illuminating a rough surface by coherent light The reflected light interferes constructively and destructively creating a grainy pattern at the observation plane In this chapter Fourier optics statistical tools and imaging theory are used to explore the various characteristics of the speckle The characteristics of the speckle are described mathematically and the optimal spatial sampling resolution is obtained Furthermore the theory of 5 axis motion measurement technigue will be described The 5 axis motion measurement was develo
174. src arr CvSize dst_size cvGetSize dst_ arr Int cx cy if dst_size width size width dst_size height size height cvError CV_StsUnmatchedSizes cvShiftDFT Source and Destination arrays must have equal sizes FILE LINE if src_arr dst_arr tmp cvCreateMat size height 2 size width 2 cvGetElemType src_arr cx size width 2 cy size height 2 image center gl cvGetSubRect src_arr amp glstub cvRect 0 0 cx cy q2 cvGetSubRect src_arr amp g2stub cvRect cx 0 cx cy q3 cvGetSubRect src_arr amp q3stub cvRect cx cy cx cy q4 cvGetSubRect src_arr amp g4stub cvRect 0 cy cx cy dl cvGetSubRect src_arr amp dlstub cvRect 0 0 cx cy d2 cvGetSubRect src_arr amp d2stub cvRect cx 0 cx cy d3 cvGetSubRect src_arr amp d3stub cvRect cx cy cx cy d4 cvGetSubRect src_arr amp d4stub cvRect 0 cy cx cy 143 if src_arr dst_arr if CV_ARE_TYPES_EQ ql dl cvError CV_StsUnmatchedFormats cvShiftDFT Source and Destination arrays LI must have the same format FIL cvCopy q3 dl 0 LINE__ cvCopy q4 d2 0 cvCopy q1 d3 0 cvCopy q2 d4 0 else cvCopy q3 tmp 0 cvCopy q1 q3 0 cvCopy tmp ql 0 cvCopy q4 tmp 0 cvCopy q2 q4 0 cvCopy tmp q2 0 void CStra
175. surface such as steel aluminum and fiber glass 97 The sensor has shown its capability of rapidly obtaining the surface strain profile for the transfer length determination The procedure has the potential to be further expedited by mounting the sensor on an automatic traverse instead of manually moving it In contrast to the conventional transfer length measurement methods that interrupt the daily casting and de tensioning seguence the automation of the transfer length measurement could be incorporated into the manufacturing production line for prestressed concrete member such as the railroad cross tie application thus bringing the real time online diagnostics and monitoring of the prestressed concrete production into reality 98 References A L Window G S 1982 Strain gauge technology London New York Applied Science A A Mufti J N 2008 Health monitoring of structures amp related education and training needs of civil engineers Bridge Maintenance Safety Management Health Monitoring and Informatics IABMAS 08 Proceedings of the Fourth International IABMAS Conference pp 33 39 Seoul Korea Bracewell R N 1978 The Fourier Transform and Its Applications New York McGraw Hill Brinson M E 1984 Resistance foil strain gage technology as applied to composite materials Experimental Mechanics Vol 26 153 154 C Joenathan B F 1998 Speckle interferometry with temporal phase evaluation for measu
176. system 4 25 will result in be a very complicated expression and the unknown variables will be difficult to determine Observing that Y2 Y1 X X and Y Y are of the same scale and Y Y L X X yield AX 4 29 Therefore Which is the second order term in Eguation 4 28 is temporarily Y X 2L ignored and will be calculated later Thus by eguation system 4 25 A x By a Byy 4 30 To determine the distance change by the image motion detected by the two cameras the four unknown parameters Pi gt Pa must be determined This can be done by the following procedure 76 Position the sensor on a flat specimen surface such that each camera observes a point on the surface The specimen surface should not have any deformation throughout the experiment Take the reference or baseline images Remove the sensor and then put it back Try to align the sensor to the position where the reference images are taken Take the images and calculate their relative displacement with respect to the reference image using the cross correlation method described in Chapter 4 The results are two displacement vectors x y and x y in units of pixels Repeat step 3 and step 4 for N times N is recommended to be larger than 10 for enough a sample size The demo experiment results are shown in Table 5 4 Table 5 4 Experimental data of the Auto Calibration
177. t This kind of marking mechanism does not leave visible markings on the object surface which might be appealing to some applications However the UV lasers are expensive Too expensive to the point where only making an alignment point is not worth it In addition since there is no visible light be emitted when the alignment is not close the user might lack the guidance to show where the target location is or what direction to move for alignment e Visible marking tracked by crosshair laser light With this solution a crosshair laser is attached to the optical strain sensor The object surface is marked with a cross sign using a paint marker The user positions the sensor by aligning the crosshair beam of the laser to the cross sign on the object surface This solution has the advantage of easy application of the marking However one drawback is that the user s eyes become fatigue after several minutes of watching the laser light In addition the crosshair laser light becomes less visible under the sunlight e Visible marking and supporting legs In this approach the sensor is equipped with three or four supporting legs as shown in Figure 3 20 Two dots are marked on the object surface The user positions the sensor by aligning two tips of the supporting legs with the two dots In practice this solution was found to work quite well It is also quite simple to implement Figure 3 20 Visible markings and supporting legs used as the alignment mec
178. t measure the displacements at point A and point B respectively A laser is collimated by lens L1 L2 and then directed to the specimen surface at point A and point B respectively using polarization beam splitter B1 The reflected waves from the diffusive surface are directed through the beam splitter B1 and the lens L3 Right in front of lens L3 is a non polarizing beam splitter B2 that sends the reflected laser beams to Mirror M4 The light beams then go back through the beam splitter B2 and are finally captured by the CCD camera Mirror M4 is used to make the sensor more compact by folding doubling the optical path length Since there are two laser beams reflected from point A and point B on object surface and sent to a CCD camera the CCD camera actually captures two speckle patterns at the same time The analysis of the speckle images would be more complicated if the two speckle patterns overlapped each other There are two approaches to prevent this from happening One approach involves manually blocking one laser beam at a time and taking the speckle image twice When the laser beam that illuminates point A is blocked the speckle image captured by the camera is generated by point B only Likewise when the laser beam that illuminates point B is blocked the speckle image captured by the camera is generated by point A only The other approach is to block half of each of the laser beams with Stop 1 and Stop 2 as shown in Figure 3 3 such that only
179. tL x 0 if oldstartL y lt 0 oldstartL y 0 if oldstartR x lt 0 holdhalf amp amp tplR backup NULL tplL backup NULL maxvalRhalf gt oldRpeak i Gl tbs maxlocRhalf tore stepsize maxlocRhalf x tbs maxlocRhalf tore stepsize maxlocRhalf y 193 Xi y maxvalRhalf gt thresholdhalf oldstartR x 0 if oldstart R y lt 0 oldstartR y 0 if oldstart oldstartL x width bs 1 if oldstart oldstartL y if oldstart oldstartR x if oldstart oldstartR y rect cvRect cvSet ImageRO L xtbs gt width L ytbs gt height height bs 1 R xtbs gt width width bs 1 R ytbs gt height height bs l1 oldstartL x oldstartL y bs bs cvCopy refL I refL rect refblockL NULL cvResetlmag filterproduc filterproduc phase_correl cvMinMaxLoc int xx yy if maxlocL xx oldstartL else xx oldstartL if maxlocL yy oldstartL else yy oldstartL int x max int y maxlo CvMat real CvMat im cv for i O for j O cvmSet real cvmSet im ROI refL t refblockL blockfilter reffiltered t tplL backup blockfilter tplfiltered ation reffiltered tplfiltered poc_halfL poc_halfL amp minvalL amp maxvalL amp minlocL amp maxlocL x gt bs 2 xtbs
180. th measurement In the subsequent field application of the laser speckle strain sensor in a CXT railroad cross tie plant the technigue reached 50 microstrain resolution comparable to what could be obtained using mechanical gauge technology It was also demonstrated that the technigue was able to withstand extremely harsh manufacturing environments making possible transfer length measurement on a production basis for the first time DEVELOPMENT OF A PORTABLE OPTICAL STRAIN SENSOR WITH APPLICATIONS TO DIAGNOSTIC TESTING OF PRESTRESSED CONCRETE by WEIXIN ZHAO B S Huazhong University of Science and Technology 1998 M S Kansas State University 2006 A DISSERTATION submitted in partial fulfillment of the reguirements for the degree DOCTOR OF PHILOSOPHY Department of Mechanical and Nuclear Engineering College of Engineering KANSAS STATE UNIVERSITY Manhattan Kansas 2011 Approved by Major Professor B Terry Beck Abstract The current experimental method to determine the transfer length in prestressed concrete members consists of measuring concrete surface strains before and after de tensioning with a mechanical strain gage The method is prone to significant human errors and inaccuracies In addition since it is a time consuming and tedious process transfer lengths are seldom if ever measured on a production basis A rapid non contact method for determining transfer lengths in prestressed concrete members has b
181. to surface tilt yaw and pitch which brings error into the strain measurement The bulky size of the system also makes the system impractical for use in the field Furthermore a calibration procedure involving displacement of the camera by a certain known distance using the servo system that comes with the system must be conducted by the end user prior to the measurement In addition whenever the distance between the sensor and the specimen surface changes the system must be re calibrated 12 Figure 1 10 ME 53 laser speckle extensometer 1 3 Overview of dissertation The importance of having a reliable and robust strain measurement technigue for either factory monitoring or field testing of concrete structural members has been described above The current available methods are either more opted for laboratory testing or are too slow to allow online monitoring The work presented in this dissertation illustrates the development of a general strain measurement technigue based on the laser speckle principle that is able to rapidly and accurately determine concrete surface strains An understanding of the relationship between the multi degree motion of the subject surface and the induced motion of the speckle pattern is reguired in order to make the laser speckle measurement technigue applicable to the typically harsh industrial environment A portable prototype incorporating unigue modular design concept and self calibration feature h
182. to the regular image for which most information is stored in the low spatial frequency domain The intensity of every speckle element in the pattern does not carry much information due to the fact that its intensity tends to fluctuate continuously and randomly when the object surface moves The fluctuation of 25 the intensity of individual speckle is in fact noise and should not be taken into account when matching the speckle image pairs Instead it is the relative location of the speckles between the speckle image pairs that determines if the image pairs are correlated and the amount of relative shifting This explains why the normalized correlation algorithm functioning well on the regular images by matching the image pairs pixel by pixel according to their intensity level does not work well on the speckle image pairs Alternatively a phase correlation algorithm based on the Fourier Transform is able to discard the intensity information and reply primarily on the phase information for matching the image pairs The overall procedure using a phase correlation technigue for speckle image shift detection is described below Suppose a pair of speckle patterns is given corresponding to deformed and un deformed states The two speckle images can be represented as f x y for the un deformed one and f x u y v for the deformed one where u v denotes the components of the local displacement vector which are regarded as constants here Zhao 20
183. to the rough surface such as concrete Temperature material properties and the adhesive that bonds the metallic conductors to the surface all affect the detected resistance and hence can interfere with the accuracy of the strain measurement A L Window 1982 The electrical resistance strain gauge is sensitive to electromagnetic interference EMI which could cause measurement error when used in the industrial environment where many types of EMI inducing eguipment are present such as motors or electrical heaters In a harsh environment the glue may debond and the gauge may break off from the specimen surface making the measurement impossible In the case of large and suddenly changing surface strain the gauge may suffer from creep effect Experiments have shown that the reading of the electrical resistance strain gauge tends to decrease from an initial value if the specimen surface is subjected to a suddenly large load The creep effect is caused by the partial debonding of the glue that bonds the gauge to the surface which results in measurement error Brinson 1984 1 2 3 Vibrating wire strain gauge The vibrating wire strain gauge operates on the principle that the natural freguency of a pretensioned wire is affected by the stress applied to it The relationship between the natural frequency f and the stress o is described by A L Window 1982 1 jo 1 1 21 p where p is the wire material density and the length o
184. ug Gu Re gi Y yy Gn ng Y yR 4 1 2 3 Vibrating wire strain gauedig yd YND FYN ROL Y coadsaieseasvastcondsedeedousesseaaaceurs 6 12 4 Fiber opHOs StTaim SGRSOL ci a nre FANNA a NT nU dL AO 6 1 2 5 Video extenso Meter LO i Gw gn aca Ge dyd O ee Ahad Fd dd O 8 1 2 6 Laser speckle strain measurement 550 ddo ae ceed teense eee 9 I Overview oFdissertatloll AO OO ee GOR 13 Chapter 2 Theoretical background of laser speckle strain measurement technigUu 16 2 1 Mathematical description of the speckle phenomenon 1 0 0 0 cceeeeeeeeneeeeeceeeeeeeeetnneeeeeeees 16 2 2 Review of the 5 Axis motion measurement SYStEM cccccccceeeeseeeenneeeeeeeeeeseeeenneeeeeeeees 21 2 3 Insensitivity of system sensitivity to out of plane movement sssseeeserereeererreerrerrsereee 22 2 4 Digital correlation technique sseeeeseseeessereessssrerssserersssrrerrsereesssereesssrreessseressssrereesssree 24 2 41 Normalized eorrelatioll uei daa YL e HC EE dU E EES 25 2 4 2 Phase correlato ceninin naia ar OY Ed gd DD aise ai 25 Chapter 3 Hardware design for the optical strain sensor essssseesessserssereeesssrerrrserersrsrreessseee 28 3 1 5 axis motion TIC SUT CDC iS y Ste a dd WR Y 29 3 2 Single moduledesign Y HF ET FR FFCD 31 3 Pia moduledesitTlu si dd AD CN e UR NC SN 36 3 3 1 The third generation prototype wa siaisdesiiaivaisciaandes ccaisgas ddd dd ydd d y SG Dydd YY d y Gyd dd 40 3 3 2 The fourth ge
185. ver even with FFTW it takes about 2 seconds to compute the correlation on an image pairs of 1392x1040 pixels using a P4 computer To speed up the correlation procedure between the image pairs a pyramid scheme is employed in the optical strain software The two images that are to be correlated are both scaled down to half size n time creating two image trees with each consisting of n 1 images where n is the downsampling depth as shown in Figure 4 6 The downsampling technigue is implemented by convolving the image with a two dimensional Gaussian filter defined as o 1 f x y Ee 4 3 21c because it provides better performance for the downsampling task compared to other types of filters Shapiro 2001 The images of the same image tree are similar to each other but with different size and resolution The parent image is twice the size and higher resolution of the child image The idea of the pyramid is that the lower resolution image pairs can be correlated first in order to estimate the possible start positions for the correlation computation for their parent image pairs The parent image pairs can then be correlated in a small area around the estimated start position to find the actual displacement The procedure is described as follows Suppose the original image size is M by N and the image trees have n levels The image pairs at the bottom of the tree have the size of M n by N n pixels and the correlation between 55
186. w FALSE m_cbPreview SetWindowText _T Preview Gl m_cbApply EnableWindow FALSI 187 hCameras int n CString refmeasurementname cRefFilenameL Empty cRefFilenameL defaultfoldert L bmp cRefFilenameR Empty cRefFilenameR defaultfoldert R bmp if f FindFile cRefFilenameL AfxMessageBox Left Reference image doesn t exist Can t do correlation LucamDisableSynchronousSnapshots hSynchronousSnapshots free pAL1Frames b_keepgoing FALSE return FALSE if f FindFile cRefFilenameR AfxMessageBox Right Reference image doesn t exist Can t do correlation LucamDisableSynchronousSnapshots hSynchronousSnapshots free pAllFrames b_keepgoing FALSE return FALSE int colindex rowindex IplImage poc_halfL 0 IplImage tplL 0 IplImage refL 0 IplImage poc_halfR 0 IplImage tplR 0 IplImage refR 0 IplImage poc 0 IplImage refblockL 0 IplImage refblockR 0 IplImage reffiltered 0 IplImage tplfiltered 0 IplImage tplLtemp 0 IplImage tplRtemp 0 IplImage tplL backup 0 188 IplImage tplR backup 0 tplLtemp cvCreateImage cvSize bs
187. w SetWindowText _T mage amp reffiltered mage amp tplfiltered mage amp tplL_backup mage amp tplR_backup mage amp poc_halfR mage amp poc_halfL DisableSynchronousSnapshots hSynchronousSnapshots led to unsetup synchronous snapshots Gl Preview 197 m_cbApply EnableWindow TRU return true Gl 198
188. xGetResourceHandle SND_AS YNCISND_RESOURCEIS ND_NODEFAULT pThread gt PostThreadMessage WM_CLOSEDIALOG NULL NULL LucamSaveImage width height LUCAM_PF_8 ppFrames 0 cfilenameL LucamSavelmage width height LUCAM_PF_8 ppFrames 1 cfilenameR Sleep 100 if ILucamDisableSynchronousSnapshots hSynchronousSnapshots MessageBox Failed to unsetup synchronous snapshots if pAllFrames free pAllFrames 153 else for G 0 i lt CAMNUM i params i format pixelFormat LUCAM PF 8 params i format subSampleX 1 i i params i format subSampleY 1 params i format height height i params i format width width params i format xOffset O params i format yOffset 0 params i exposure m_exposure 50 ms exposure params i gain 23 nl params i gainBlue 1 0 n params i gainGrn1 1 0 nl params i gainGrn2 1 0 n params i gainRed 1 0 params i strobeDelay 0 0 unused params i timeout 3000 0 3000 ms params i useHwTrigger FALSE Set this to true for hardware triggered setup with daisy chaining params i useStrobe FALSE Set this to true if daisy chaining cameras params i exposureDelay 0 params i shutterType LUCAM_SHUTTER_TYPE_GLOBAL pParams i amp params i params 0 exposure m_exposureA params 1 exposure m_exposureB pAllFrames UCHAR malloc CAMNUM width height
189. xtreme situation 93 a Cutting b Washing scrubbing wiping and vacuuming on the tie surface Figure 6 17 Severe abrasions to the cross tie surface at the saw cutting machine Figure 6 18 Cross tie surface bonded with microscopic reflective particles With the microscopic particles applied to the surface of the cross tie the laser speckle strain sensor was able to find the correlation between the corresponding speckle image pairs and extract the surface strain information of the cross ties The total time for measuring each tie was about 3 minutes This was made possible because no high precision traverse setup was reguired and simple visual manual positioning was adeguate The results from the in plant cross tie measurement are shown in Figure 6 19 6 20 6 21 To evaluate the capability of the laser speckle strain sensor to monitor the long term surface strain trend another set of readings were taken for 94 each cross tie ten days after the pouring It can be seen in that the laser speckle strain sensor has successful extracted the surface strain information both for short term and long term effect Tie 1 Side A 1100 1000 al aua a ee l 700 gr cine fs Microstrain After Cut Poured 2 7 11 2 17 11 10 Days After Pour Jy 23 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 2B 0 Distance Along Beam in Figure 6 19 Cross tie surface strain measurement Tie 1 Si
190. y fftwf_ execute imgl half fl maxvalLhalf 0 colindex_store 0 rowindex_store 0 for colindex 0 colindex lt col for rowindex 0 rowindex lt row phase_correlation_bloc poc_halfL cvMinMax amp maxlocL if maxval maxval ihal Loc maxloc hal halftemp f maxloc poc_halfL ect NULL rblockfilter tplfiltered oat nu nu k fft amp minvalI tplfiltered gt imageData bs bs sizeof m colindexrr m rowindexrr _halfL colindex rowindex float fft imgl halfsize imgl_fft_half temp amp maxvalLhalftemp amp minlocLtemp O lLhalftemp gt maxvalL f maxvalLhal hal half ftemp ftemp 191 colindex_store colindex rowindex_store rowindex if maxlocLhalf x gt bs 2 startL x c else startL X c startL y r startL y r tbs maxlocLhalf olindex_store stepsiz olindex_store stepsize maxlocLhalf x if maxlocLhalf y gt bs 2 owindex_store stepsiz tbs maxlocLhalf owindex_store stepsize maxlocLhalf y if maxvalLhalf gt THRESHOLD tplR gt imageData char ppFrames 1 rect cvRe cvSetlmage cvCopy tpl ct widt ROI tpl R tplR R rect temp NULL cvResetImageROI tplR filterprod uct tpl memcpy imgl_half fftwf_exec maxvalRhal

DEVELOPMENT OF A PORTABLE OPTICAL STRAIN - K-REx

Contents

Download Pdf Manuals

Related Search

Related Contents