QUAD-G - Department of Geography
Contents
1. We seek transformations r A and s A that yield a predicted u v for any location By default QUAD G uses 2 degree polynomials for f and g That is i r A d atbAt cht dV ep fre D 0 5s A 0 g h i jA kP 11 2 where Y is the predicted image location If only the four corner control marks are available QUAD G uses a first degree polynomial in effect d e f j k l are zero A first degree polynomial can compensate for image rotation differential stretching in u and v and shearing A second degree polynomial is obviously even more flexible In particular it captures variations in projection scale that a linear function cannot The task is to choose the coefficients a through that are optimal in some sense As is standard QUAD G finds coefficients that reproduce the observed u v as close as possible in a least squares sense That is we solve the following optimization problem Find r A to minimize S Sir u F ii uf 3 n n Find s 4 to minimize S s 4 6 v X P n 9 i l i l The coefficients enter partial derivatives of S and S linearly so this is a problem in linear regression However because the image coordinates result from an automated procedure there is no guarantee they are accurate or even feasible Precautions are therefore essential to ensure the program does not fail catastrophically during the fitting step To this end QUAD G uses a technique known as2. colors of the k pixel of a mark Pixel locations within a mark are stored as offsets Au Av from the center of the mark Every feasible position i j within the search window evaluated with the misfit between the pattern and the image at i j given by N IAC Au V Av P k k l 1 u Au v Av P kK I u Au v Av P k P The location i j with smallest is taken as the mark location Note that this process always gives an optimum location whether or not the control mark actually appears in the search window After the search the root mean square misfit JE N is shown for each window Control Mark Adjustment optional manual mode only If the search results are unsatisfactory individual control marks including corners can be adjusted before going on to the next step Because final positions will be used as input data for the fitting procedure operators should position control marks as precisely as possible The adjustment procedure is the same as for corner adjustment Least squares Fitting and Error Analysis The input file gives a longitude latitude pair A for every control point The pattern search gives a image coordinates u v for the same marks in the scanned map This can be visualized in table form Control Mark Longitude Latitude u search v search No known known result result 1 1 1 uy V1 2 2 2 u2 v2 3 3 3 u3 V3 n i E Un Vn
3. a square map in lat long 5x9 5 rows in latitude 9 columns in longitude ControlMarkSpacing Control mark grid spacing Use only if mark spacing is the same in longitude and latitude see note below 00 10 2 5 0 041666 ControlMarkLonSpacing Longitude distance between control marks see note below 00 10 2 5 0 041666 ControlMarkLatSpacing Latitude distance between control marks see note below 00 10 2 5 0 041666 PerimeterMarksOnly Optional flag indicating interior marks are absent or should be ignored True T t False F f ControlMarkSize Size of control mark legs E g half the entire width of a mark Allowed values are IN inches pixels and f fraction Fraction means proportion of the image size Not case sensitive Inches assumed if no units are given 0 2In 0001f 25pixels 0 2 OutputGeographicFileName Optional name for output image file If omitted the input filename is used with the extension is removed and replaced by the suffix Geo tif mapl tif MadisonWest tif If omitted input file map tif becomes mapGeo tif Table 1 Input XML tags The XML input file can specify values for latitude longitude and grid spacing in either decimal degrees or degrees minutes seconds To obtain a value the program looks for a text string with one to three fields separated by blanks If there is only one field it is interpreted as
4. data via a dialog Diagnostic statistics are written to a log file for examination outside of QUAD G The program can read input images in uncompressed base TIFF format or binary ppm Other input formats are converted to ppm by a GDAL routine www gdal org Output images are stored as GeoTIFF rasters in a location determined by the user Please see http trac osgeo org geotiff for information about the GeoTIFF format Installation QUAD G is provided as Windows binary that can be executed from any directory There is no install procedure to run nor are any registry changes required The installation consists of copying the executable to any directory As with any application users can create desktop or program menu shortcuts pointing to the exe file The program requires the Microsoft NET framework version 4 or higher QUAD G uses the GDAL raster package as a helper for file conversion The required GDAL routines are distributed as Windows binaries by a variety of groups see downloads at www gdal org One of these packages must be installed for full capabilities of QUAD G The program looks for GDAL files in several standard install directories Program Files Program Files x86 and OSGEO4W If the package is not installed in one of those locations the user will need to modify the QUAD G source code accordingly Input Data File A single data file provides information for a suite or batch of input images Typic
5. degrees If the second field is present it is taken as minutes If there is a third field it is interpreted as seconds Any of the three can have a decimal point A minus sign means the entire value is negative Table 2 provides a few examples Input String Interpretation 44 5 44 5 or 44 30 44 30 44 5 44 30 44 30 30 44 50833 44 30 30 44 50833 44 50833 44 30 30 00 2 5 2 5 81 30 81 5 west longitude Table 2 Examples of input for latitude and longitude The control mark grid can have rectangular cells with different spacings in latitude and longitude In that case use the xml tags ControlMarkLatSpacing and ControlMarkLonSpacing to specify respective values If the grid is uniform with square cells the same value can be supplied for each or the tag ControlMarkSpacing can be used Some map series have control marks only on the quad edges and thus lack the symbols seen in Figure 1 In such cases the PerimeterMarksOnly tag should be set to True There is an important distinction between the input XML data values described here and user settings or preferences discussed below Obviously data values can vary from scan to scan One image within a batch could have a scale of 1 24000 whereas another might be at 1 62500 By contrast user preferences apply to the entire batch Preferences can be changed in a QUAD G session but not while a single im
6. singular value decomposition SVD SVD is certain to return a solution even in pathological situations such as collinear control points Numerical accuracy of the fit is improved by scaling all values A u v to the unit square before optimization Solutions to 3 and 4 give the best fitting polynomial transformation based on all control points Cross validation is used to assess the ability of the polynomial to capture the pattern of the control points If the transformation is a good one it ought to be able to successfully predict the location of a new control point not part of the fitting procedure Cross validation implements this idea by excluding a control point from 3 and 4 and using the resulting functions to predict that excluded point Within QUAD G the cross validation error is reported as the distance in pixels between the predicted and search window locations of the control mark Ei i 4 by v 6 The predicted values u in 6 are those generated by a model with the oo data point omitted Each control point is dropped in turn and the resulting error is computed Note there are two predictions and two errors for each point First there are the model predictions and corresponding residuals In addition we have the cross validation predictions and errors By default he program displays the model predictions and errors in the thumbnail windows Selecting CV predictions changes
7. to cross validation values Either method provides visual feedback as to whether or not the polynomial successfully models various parts of the map In addition a global error measure is found by summing over all excluded points Cross validation errors are larger than S and S because the errors are measured at points not used during optimization Cross validation errors are meant to give the error expected with independent data and are mostly useful for identifying points with a failed search The model errors give a better estimate overall success of the transformation thus they are written to an output error XML file so they can be examined later Only images whose model errors are smaller than a user specified threshold will be georeferenced In manual mode there is once again an opportunity for user adjustment of the pixel locations used for fitting After adjustment step 7 must be repeated for adjustments to take effect 10 Creation of Output Georeferenced Image optional The output from QUAD G is a new image whose pixels are longitude latitude A squares These pixels lie on a regular grid oriented north south and east west The goal is to match the input image resolution That is we desire at least as many output pixels in each direction as were present on the original image Because input pixels are not squares in A this means there will be more pixels in one dimension in the output than were found on the input im
8. Files G _QuadG_Example Qutput lt ML Transformed Image Files G _QuadG_Example OutputT if E rror XML Files i G 4_QuadG_Example Output lt ML Linear Fit Only Never Rotate _ No Georeferenced Image Apply Save Figure 6 Preferences Dialog Process Buttons This group invokes the processing steps discussed above Automatic Mode In automatic mode the user clicks Run Batch and the full sequence is of steps executed for the batch starting with the selected file and continuing to the end of the file list After a batch starts the Stop button becomes enabled and can be used to interrupt the process Figure 7 17 Figure 7 Process buttons for automatic mode Manual Mode In manual mode a user selects an input image and then progresses through the sequence of processing steps by clicking the appropriate button for that step Buttons are enabled when the corresponding processing step is appropriate Figure 8 Figure 8 Process buttons for manual mode as they appear immediately following selection of an input image file The buttons exposed in manual mode are Find Corners XML scan information is read for a single file and corners are located by successive evaluation of candidate quadrilaterals The main panel is used to display assembly of quadrilaterals during the search The user may adjust corners after this button executes and such adjustment is highly recommended if one or more corners are n
9. USGS Historical Quadrangles Scanning Project QUAD G Automated Georeferencing of Scanned Map Images User Manual Version 2 10 9 3 2014 J E Burt J White G J Allord Overview This document describes software for georeferencing scanned topographic quadrangles and other map images In other words the software converts an image from a scanner s coordinate system to a known spatial reference system In this case the output reference system is the geographic latitude longitude coordinate system implicit in the original map The software known as QUAD G operates in much the same way as other georeferencing tools a small number of points are located in both image coordinates and geographic coordinates These so called control points are used to establish a relationship between the image and world This relationship defines a mapping between the two systems that is applied to the scanned image The result is a new image whose pixels comprise a grid of squares in latitude and longitude aligned with the cardinal directions QUAD G differs from standard tools in that the control points are found automatically rather than provided by a user through on screen digitizing Thus the software can process an arbitrarily large batch of scanned images without operator supervision Generally speaking QUAD G has advantages over standard tools whenever one has more than a few maps in a series that require georeferencing QUAD G was developed with f
10. WY Offset O Figure 10 Color left and Grayscale right visualizations The visualization setting has no effect on the output file It is intended primarily for use in manual mode as an aid when adjusting control mark coordinates Comment Box Any text entered in this control is copied to the output XML files This provides a way to include an operator name short notes or other information in the output stream 20
11. age For example consider a 7 5 minute map The input image is rectangular there might be 15 000 pixels in the vertical but only 10 000 in the horizontal In this case QUAD G would yield an output image that is approximately 15 000 x 15 000 Obviously to do otherwise would require non square output pixels or else a loss of information in the north south direction Figure 2 provides an example from a 7 5 minute scan 10 DIVISION DIVISION GEOLOGICAL AND N7 GEOLOGICAL AND N2 STURGEON BAY 18 ME 65 FO STURGEON BAY 18 Mi 3 a EE E FORESTVILLE Mi 65 FO Figure 2 Sections of input left and georeferenced right images On the original the sub area is rectangular in terms of both land surface area and image extent with more pixels in the vertical Figure 2 left However the area shown is roughly square in longitude and latitude Obviously with fewer pixels in the horizontal pixels on the original are wider in longitude than in latitude Because the area shown is nearly square in A the georeferenced image of the same area is nearly square In order to preserve detail the output image has as many pixels in the vertical as on the original This results in more output pixels in the horizontal than on the original Construction of the georeferenced image proceeds as follows i First the output grid is established The spacing of the grid is determined by the input resolution as discussed above The grid extent is the mapp
12. age or batch of images is being processed For example the pixel error threshold used to flag problem images is constant within a batch Thus if some images are scanned at 300DPI and others are at 600DPI it would likely be a mistake to process them in the same batch An acceptable error at 600DPI might be too large for use at 300DPI Georeferencing Steps The complete sequence is as follows 1 Preference File Processing When QUAD G is loaded into memory it searches the installation directory for an XML file saved from a previous session User option values are read from the file if it exists otherwise they are set to default values Preference files are not maintained directly by the user thus its format is not described here Preference values are changed via a dialog box described in the next section 2 Input File Processing The XML lt scan gt block for an image is processed to obtain the latitude and longitude coordinates for all control points That is from the corner location the grid spacing and the number of rows and columns the program finds true latitude and longitude for every control point The image orientation normal or transverse is also determined in this step 3 Find Corners The input image is searched for the bounding neatline The search is informed by the quad aspect ratio which is known given the control mark layout If the map scale is provided the approximate pixel dimensions of the quad are known and can
13. ally all of the images in a batch come from one map series but the definition of a batch is left completely to the user The only requirement is that all images within a batch reside in the same folder Likewise all output images for a batch are placed in a single folder Input and output folders are specified by the user The data file must be provided as an Extensible Markup Language XML file see www w3 org TR REC xml The XML file consists of a standard preamble and a list of scans Within the scan list each image is described in an XML block delimited by lt scan gt and lt scan gt tags For example the following listing shows a data file for georeferencing two images lt xml version 1 0 gt lt ScanList gt lt Scan gt lt FileName gt 500010 tif lt FileName gt lt MapName gt Algoma Quadrangle lt MapName gt lt Datum gt NAD27 lt Datum gt lt ControlMarkSpacing gt 0 04 1 6666667 lt ControlMarkSpacing gt lt Longitude gt 87 375 lt Longitude gt lt Latitude gt 44 5 lt Latitude gt lt Resolution gt 600 lt Resolution gt lt ControlMarkLayout gt 4x4 lt ControlMarkLayout gt lt ControlMarkSize gt 0 2in lt ControlMarkSize gt lt OutputGeographicFileName gt Algoma_24000 GTIFF lt OutputGeographicFileName gt lt Scan gt lt Scan gt lt FileName gt AK_ Ambler River _A 2 1985 63360 tif lt FileName gt lt MapName gt Ambler River A 2 lt MapName gt lt Datum gt NAD27 lt Datum gt lt Scale gt 63360 lt Scale gt
14. also be used in the search The search first finds candidate edges nearly vertical and horizontal lines that might be part of the quadrangle boundary All combinations of candidate edges are projected to intersections that define a set of candidate quadrangles The candidate quadrangle whose aspect ratio or size is closest to the known quadrangle is taken as the best guess This is used to guess at image coordinates of each corner Windows are placed around each guess to show the results 4 Adjustment of Corners optional manual mode only If an actual corner is not within its window it and other control marks will not be successfully located during the pattern search An option is therefore provided for user adjustment of corner locations The operator clicks within a window and drags until the actual corner is reasonably close to the center of the window The arrow keys can be used to move the image one pixel at a time A shift arrow combination accelerates movement to 10 pixels There is no need for fine adjustment of corners approximate image locations are sufficient 5 Identify Search Windows The program uses the grid layout and the image corner locations to guess at control mark locations That is a control mark grid is established based on the presumed corner locations These become the center of search windows for all the control points including the corner marks If a control mark is larger than 2 3 of the search window the mark
15. e for each transformation Error XML File Directory location of XML information for images whose error threshold was exceed Information for all scans is put in a single file This file can be opened later as an input file for individual manual processing of problematic scans Linear Fit Only Button use linear rather than quadratic polynomial regardless of control mark count This can prevent unrealistic extrapolation in areas far from control marks Never Rotate Button disable automatic detection and rotation of transverse scans No Georeferenced Image Button Control marks and least squares fit written to output xml file but the scanned image is not transformed No QA Button No quality analysis is performed on the georeferenced image Apply Button Apply changes to the current QUAD G session but do not save settings Dialog remains open Save Button Apply changes and save settings to the current preference file Close Button Apply and close the dialog without saving changes Any changes made are used in the QUAD G session but they are not stored externally for use later 16 Preferences File Preferences xml File Window Dimensions Error Thresholds Thumb Width x Height Level 1 160 160 2 Search Window Size Level 2 200 3 Level 3 Crosshair Colors 5 3 Mark Location write File J gt Prediction Directories Source Images G 4_QuadG_Example Transformation XML
16. ed area plus an optional map collar The grid spacing and extent determine the position and size of the grid At the end of this step the latitude and longitude is computable for every pixel in the output image ii The next step assigns a color to each output pixel using the input image For a given pixel center 4 equations 1 and 2 are used to find the four surrounding image pixels Red green and blue values at those locations are linearly interpolated onto A and written to a temporary tiff file iii After all pixels are processed the temporary file is converted to geoTiff format using GDAL That is the input datum and other geographic metadata are composed and written to the final output file along with the pixel data 11 Quality Analysis of Georeferenced File optional The output image has pixels that are square in latitude and longitude The latitude and longitude of every control point is known Therefore the expected the pixel coordinates of a control point A are simply A Asin W 1 u A Prax P H 1 i A 11 where W and H are the output image width and height respectively and Anin Pin ANd Anax gt Pax are the corresponding image longitude latitude corners The quality assessment proceeds by extracting a window surrounding u v and searching that window for a control mark This gives a found position The difference between A and A is converted to ground distance erro
17. entire image By sorting on any column one could use this file to examine a large number of scans for potential problems Errors for individual marks are available in the xml file mentioned above Using QUAD G QUAD G is a standard Windows application All of its functionality is accessed though menus and buttons displayed on the main screen Figure 4 The main screen is also used to display diagnostic text and graphical elements enabling a user to monitor progress on a batch of images This section describes how to use and interpret QUAD G features Fela Smaa Oma TX 13 N Mak Uyat Lat Mak Spacing 0041605670 Lon Mak Spong 00416868670 Perimeter Maks Oriy Fake Hak Soe 0n Dapa Fie LES ke Li u 0 0 9 233 13 ES f AT 7 z rx l x 0 0 e 0 183 uy 0 0 620 212 uj 0 0 620 198 u 0 0 e0 108 uein u y 0 0 050 150 PAASA 0 0 620 175 7 a00 00 0 0 e 0 236 u 0 0 e 0 262 Fie Lint Lando TX 1950_2500004 a na Soo a sone 70117100 F er DX 18 12s 104958 24000 tf Angruas_ TX 1963 1000009 Battin ay_T _1904_ 100000 u Baher Goreng 1X 197562900 1 Daksdse_ TX 1544 iF tner Inchide Colu O Lore Coty Boh 2 thes O OV Predictions Figure 4 Main Program Screen Top image shows scan with interior control points Bottom image shows a case where only edge perimeter marks are used File Menu Figure 5 This menu is used to open an input XML file input parameters
18. for a single scan file and to exit the program Typically a user will load an XML file for a batch of scans which has the effect of populating the File List box seen in the upper right corner of Figure 4 The list can then be processed with no operator intervention in automatic mode or individual files can be highlighted and processed in manual mode Typically a user will first employ automatic mode for a batch and then return to any problem files in manual mode The single scan option provides an alternative to input via an XML input file In this case a dialog opens allowing for direct entry of scan file parameters see Fig 5b QUAD G v 2 05 Open Batch File Single Scan Processing Figure 5 File Menu 14 EE Single File Input File Names Scan File Lower Right Corner Longitude Anahuac_TX_1983_100000 ti 34 Quad Name optional Latitude 29 30 Anahuac Texas Metric Output File Name optional Control Mark Info Rows Columns Other B 5 Datum Lat Spacing Lon Spacing NAD27 0015 0015 Map Scale optional Size 1 100000 2p Resolution DPI 600 Perimeter Only Figure 5b Direct Input Dialog Preferences Menu and Dialog Opening the Preferences menu exposes the dialog shown in Figure 6 The dialog s File submenu is used to read an existing preferences XML file This allows settings saved previously to be easily re established The File submenu a
19. itude spacing The four corners of the map area comprise the minimum required set of control points but use of edge and interior control points is strongly recommended As seen in Figure 1 a standard set of shapes is assumed for the control marks Note that interior corner and edge marks each have a distinct shape These shapes are built into QUAD G in the orientations shown Thus an interior mark is always a shape and a top edge mark is always a shape No assumptions are made ian the number or arrangement of control ade other than that they comprise a regular grid and that all are black in color The distribution version of QUAD G assumes the quadrangle is bounded by a black neatline and that the area outside the neatline is white The program can be easily modified if these assumptions are not met T 29 Madi 207 17 30 33 Os SS coe ae L 7 302 lp Figure 1 Standard control marks in the 4x4 layout of 7 5 minute quadrangles 2 As is true for standard georeferencing tools we assume a low degree polynomial can adequately represent the transformation from image to geographic coordinates As a practical matter this amounts to assuming the map scale is so large that the map projection is essentially undetectable e g great circles appear straight on the map Small scale maps with geodesics that depart drastically from straight lines are outside the design parameters of this project Thus while QUAD G is pe
20. lso provides for saving a modified suite of preferences as a new file Meanings of the settings are as follows Thumb Dimensions controls the display size of search windows If the thumbnail size is too large for all search windows to fit in the panel the display size is adjusted downward However the search window size is unaffected by this It remains at the value shown Search Window Size determines size of image subareas searched for control marks The default search window size is 250 by 250 pixels Large high resolution scans sometimes benefit from a larger window size in order to find a control mark Smaller sizes result in somewhat faster searches 15 Crosshair Colors Sets color of cross hairs used to display search window locations and model predictions Error Thresholds cross validation errors equation 6 are flagged with colors indicating increasing levels of severity Levels 1 3 are simply visual cues By contrast the Error Level determines whether or not an output image is produced A georeferenced image will be generated only if all errors are below the threshold Source Image Directory location of input images Transformed Image Directory location of output georeferenced images and quality analysis images Transformation Information Directory location of XML files for successfully transformed images errors below threshold Control mark locations transformation errors etc are placed in a separate fil
21. lt ControlMarkLonSpacing gt 00 15 lt ControlMarkLonSpacing gt lt ControlMarkLatSpacing gt 00 10 lt ControlMarkLatSpacing gt lt Longitude gt 156 5 lt Longitude gt lt Latitude gt 67 lt Latitude gt lt Resolution gt 600 lt Resolution gt lt ControlMarkLayout gt 4x4 lt ControlMarkLayout gt lt ControlMarkSize gt 0 1 lt ControlMarkSize gt lt OutputGeographicFileName gt Ambler_River_A 2geo tif lt OutputGeographicFileName gt lt Scan gt lt lt ScanList gt The meaning of each tag is given Table below Tag Name Meaning Examples FileName Input image file relative to Jacoma tif 167001 tif source image directory MapName Optional text name of map West Madison Quadrangle Datum Optional datum of original map NAD27 NAD83 WGS72 needed for geoTiff header Must WGS84 be one of the examples shown at right Defaults to NAD27 Longitude Longitude of lower right map 88 5 88 30 00 see note below 81 375 81 22 30 Latitude Latitude of lower right map 42 125 42 7 5 00 corner see note below 15 5 15 30 00 Resolution Scan resolution in dots per inch 600 Scale Optional map scale denominator 24000 62500 100000 250000 in representative fraction Providing this tag will result in a better edge search ControlMarkLayout Number of rows and columns of control marks on map when viewed in normal orientation Do not adjust for transverse orientation Note rows first then colums 4x4
22. nels of radio buttons control basic attributes of execution a Mode Extent Manual Include Collar Automatic Lose Collar Visualization O Both 2 files Er Color Fit O Grayscale C C Predictions Figure 9 Mode Buttons Mode In manual mode only a single file is processed namely that selected in the file list box In automatic mode the entire list is processed starting with the selected file The current file will be indicated and the screen will be updated to show progress through the batch Extent This governs the area of the scan that appears in the output file The Include Collar option causes the entire scan to be processed which includes text and all other material surrounding the map rectangle If Lose Collar is chosen only the map area within the quad neatline appears in the output georeferenced file If Both is selected two files are output one with collar one without The default file names are xxxxxGeo tif and xxxxxNoCollarGeo tif where xxxxx is the input file name If the output file name has been specified in the preferences NoCollar will be inserted before the file extension CV Predictions Selecting this means cross validation predictions are plotted and cross validation errors are shown for each window Otherwise model values are plotted and printed 19 Visualization This controls search window background color Figure 10 U Offset O Y Offset 0 U Offset 0
23. ot found Corner images should be adjusted so that corners are near the center of the search window As mentioned earlier exact placement is not essential This button is enabled whenever an input image file is selected from the file list box seen in Figure 4 Find Windows Corner information is used to establish search windows presumed to contain control marks The user may adjust search windows after this step Once again exact placement is not required Search Windows Each window is searched for its mark After this the user may manually identify its location by positioning the image so that the appropriate edge mark is directly beneath the cross hair The adjusted pixel values will be taken as the exact cross hair location in the least squares fit thus it is essential 18 that the user be as precise as possible Recall that the arrow keys can be used for small movements of control marks Fit Performs the least squares fit and generates error values As seen in Figure 4 prediction errors are shown for each control mark with color coding determined by the error thresholds Export Writes control mark coordinates and least squares fit to output xml file If requested create the output image and performs quality analysis on the same Stop Interrupts georeferencing at the first available opportunity Processing will not stop within a swath of pixels or while the GDAL helper is running Mode Button Group As seen in Figure 9 three pa
24. rfectly adequate for scales of 1 100 000 and above it is not recommended for maps of the full globe Error statistics computed by the program will alert the user to situations where the polynomial model is inadequate 3 We have observed that scan operators typical feed map sheets in a normal orientation so that the northern border is toward the top of the image However sheets that are much wider than they are tall are sometimes fed transversely with the northern border on the right As a working hypothesis QUAD G assumes that the map has been scanned in normal orientation For quads with more columns than rows of control marks the program compares the aspect ratio of the image to the aspect ratio of the latitude longitude quadrangle If the image is inconsistent with the quadrangle the program assumes the map has been scanned transversely and image coordinates are rotated accordingly Please note that regardless of input orientation the output image file will always have north toward the top Use of QUAD G is straightforward a data file is provided that describes the images to be georeferenced The required information for each image includes the image file name map extent control mark layout and several other parameters described below QUAD G reads the data file and processes each image in turn Alternatively there are options for manual selection and processing of individual images listed in the input file and for direct input of image
25. rs in meters A A p oa where p and g are respectively the length of a degree longitude and latitude at latitude The errors are stored in the output XML file as GeoTiffErrors and they are shown on screen as seen in Figure 3 below The QA image is also stored in portable network graphics format as a file named xxxxGeoVerify png where xxxx is the filename prefix used for the georeferenced file ay ae Error 1 67m Poss Error 1 23m coo Error 0 83m Roes Error 0 E Figure 3 GeoTiff quality analysis screen Please note that these error measures could be wrong for either of the following reasons a the map ellipsoid might be incorrectly specified leading to the wrong values of p and q or b the pattern search might fail giving the wrong No failed pattern searches have been observed in testing with thousands of images 12 Numeric error measures are also saved in a comma separated text file named verify csv This file is placed in the output image directory and contains summary errors for each georeferenced image See below for an example RMS Error m Max Error m Min Error m _ Mean Error m n G Verify csv F 1 2 1 681 2 278 3 1 938 3 893 4 1 05 2 036 5 6 i Georeferenced File 510745Geo tif 510753Geo tif 510768Geo tif The statistics shown are accumulated over all n available control marks in the image thus the values provide an aggregate measure of error for the
26. size is reduced accordingly to avoid the possibility of the mark extending beyond a window boundary If the PerimeterMarksOnly tag contains T or t no interior windows are found 6 Adjust Search Windows optional manual mode only If a control mark is not within its window it will not be successfully located during the pattern search In this case the user should move the window until it is roughly centered over the correct control mark The adjustment procedure is the same as for corner adjustment Only approximate placement is required 7 Search Windows Each window is searched for its control mark The proper control mark template is placed over every possible pixel in the search window and deviations between the template and the underlying image are noted The pixel location with smallest deviation is taken as the control mark location In our experience scanned images are typically rotated by a few tenths of degree from vertical In addition convergence of meridians guarantees that the quadrangle is not a true rectangle Thus the edges are only assumed to be within 2 of horizontal or vertical A trivial change in the program would accommodate other tolerance values 8 9 In particular let u v 1 u v Ip u v be the red green blue values of the image search window at coordinates u v A control mark pattern consists of a set of N pixels with red green blue components Let P k P k P k be the
27. unding from the United States Geological Survey to support its Historical Quadrangle Scanning Project Therefore the default QUAD G setup is for topographic quadrangles and similar USGS map sheets but it can be used with scans of large scale maps from any source The program is distributed as a MS Windows executable and in source code format as a C program under the GNU General Public License The program is available at www geography wisc edu Quad G as are sample input datasets and output files QUAD G was developed under a cooperative agreement between the United States Geological Survey and the University of Wisconsin Madison Geography Department The project was a collaborative effort between Gregory J Allord USGS James E Burt UW and Jeremy White UW with assistance from A Xing Zhu UW Jim Burt and Jeremy White coded the program This manual was drafted by Jim Burt Use of QUAD G rests on the following assumptions 1 The mapped area comprises a rectangle in latitude and longitude The four corners of the rectangle serve as control marks Thus the extent of this rectangle is assumed known as well as the layout of any additional control marks of likewise known latitude and longitude For example a 7 5 minute USGS topographic quadrangle has 16 control marks arranged in a 4 x 4 grid with 2 5 minute spacing Figure 1 The grid need not have the same number of rows as columns and the spacing in longitude need not match the lat
Download Pdf Manuals
Related Search