Home

the aerospace technology group, ardita aeronautica division

1. We have taken the Futaba 9202 servo as our sample servo model with parameters given by 38 2261 ang C 0 5118 However the digital Futaba 9253 servo is much faster that Futaba 9202 servo The parameters are given by n 32 2261 and 0 5118 The propagation state is done using RK4 routine Atmosphere Earth Model In this section we define an atmospheric model which yields the density air air pressure air temperature local speed of sound as a function of current density altitude The table 3 VI 6 shows the atmospheric model parameters p Density of air slug ft pressure Pa Air pressure Ib ft temperature Tai Air temperature R sp sound c Local speed of sound ft s densAlt altitude H Current density altitude ft MSL up p 0 0023769 1 0 68753 0 003264 H x 10 H x 10 7 IL ois 518 6G9 1 0 687532 H x 10 b 0 003298 H x 10 Pais 1716 5 P Thi Ca 1116 45 y Tair 518 69 Table 3 VI 6 Atmosphere Earth model 3 VI PHOTOGRAMMETRY MODULE This is a complete photogrammetry tool consisting of tools such as normalization triangulation stereo plotter rectification interior orientation exterior orientation and DEM modelling Unlike other similar tools though it also has a 6 parameter iterative algorithm capable of correcting imaging system distortions so mosaic generation and image registration improves particularly where images are sourced from multiple disparate platforms a
2. A Peregd gt Contribubons T Wutztel Ge eostatistics Grids b Geostatistics Knging b Geostatistics Points gt Grid Analysis gt Grid Calculus gt Grid Discretisation gt Grid Fite b Gnd Grid g gt Grid Spine interpolatiol b gt Gnd Toots gt Grid Visualisation gt image Process ng gt imagery Opency t e nagery Segmentat ol gt trwort GPS Data P moort Expor DKF enportExpOrt ES EDO TportExport GDALIOG imoort Export GRE Fue oort Export Gids 009 imoart Expart images p in Dort Export Shapes Gee ee Load Save 7er ready Module Libraries Figure 3 1 1 The Raphael GCS main window It shows the graphical user interface The whole functionality of Raphael GCS can be found from this main window In the following sections we present the various modules in more detail 3 11 DYNAMIC MISSION PLANNING AND TASKING Todays high altitude endurance HAE reconnaissance unmanned aerial vehicles UAVs are extremely complex and capable systems They are only as good as the quality of their implementation however Mission planning is rapidly increasing in complexity to accommodate the requirements of increasing aircraft and information control capabilities Effective mission planning is the key to effective use of airborne reconnaissance assets which demand extremely intensive and detailed mission planning The mission pl
3. COS YSIN K rig COS w SIN COS sinc sim amp T3 siny lag COS w SIN i SIN amp SINW cos amp T32 SIN amp cosy r33 COS Ww COS f The intrinsic camera parameters usually include the effective focal length f scale factor s and the image center 4b Y L also called the principal point Here as usual in computer vision literature the origin of the image coordinate system is in the upper left corner of the image array The unit of the image coordinates is pixel and therefore coefficients Du and Dv are needed to change the metric units to pixels In fact their precise values are not necessary because they are linearly dependent on the focal length f and the scale factor s By using the pinhole model the projection of the point xi yi zi to the image plane is expressed as The corresponding image coordinates ui vi in pixels are obtained from the projection i V i by applying the following transformation D uSu u i Uy D vi Up v Usually the pinhole model is a basis that is extended with some corrections for the systematically distorted image coordinates The most commonly used correction is for the radial lens distortion that causes the actual image point to be displaced radially in the image plane The radial distortion can be approximated using the following expression radial 4 Aa A kir ker cC ae i j 2 A Aa 0994 Vv kir kor i
4. and expressed in the reference frame 2 the desired body frame Xa amp P4 with respect to the world frame X eb And where and Sa are approach Figure 3 V 1 It shows relationship between inertial and body fixed coordinates frames for a UAV on a landing 1 given by P Note There are 2 kinds of projection used in vision the spherical and the flat projection The spherical projection identifies the projection plane as the spherical surface and the image point p is X oz However in the flat projection the point is projected on a plane with its image 1 p zboz Indeed since equality in projective geometry is an equality between directions both Planar Homography Let 1 l t points are on the same ray emanating from the origin and are thus not distinguished Raphael GCS will assume a calibrated camera but we do not distinguish between spherical or flat projections m 1 id coordinate frames 3 gt eR denote the Euclidean coordinates for the i visual feature on the landing surface relative to the camera at position and D respectively From the geometry between the EL my are related as follows 3 Also illustrated in figure 3 V 1 n eR denotes the known constant normal to the plane 7 expressed in 1 the coordinates of Ba and the constant 0 R denotes the distance of the landing surface
5. N m rad p air density slug ft pc agR Lock number Siftness number fixed shaft pitch angle rad distances of main rotor hub aft and above center of mass ft blade linear twist rad distances of tail rotor hub aft and above center of mass ft rotor speed gearing constant pitch flap coupling factor Bell Hiller Hiller Stabilizing Bar R Ro paddle starting and ending radii ft d Yf p Cf ao a swashplate and flybar linkage ratios Empennage lin Rtp distances of horizontal tailplane aft and above center of mass ft lt Rtn distances of vertical fin aft and above center of mass ft Arps Q0 lift curve slopes 1 rad Sip S fn areas ft XOrp tailplane zerp lift incidence angle rad BO jn fin zero lift sideslip angle rad Table 3 VI 5 Helicopter Parameters JE JE n Figure 3 V1 9 Helicopter parameters shown by Raphael GCS Wind model We have adopted a simple wind model It is a dynamic system that will generate random winds up to a maximum value in both vertical and horizontal directions Each wind component is modeled by the following transfer function in fixed body frame S o n S D 260 So amp ke Servo control input In this section we will provide a servo actuator model It is a second order dynamic model which is tunable via n wn and values The transfer funtion model of the servo used is as follows n 3 2 Hor u S 2 C0 Sto
6. T from the origin of the frame D t can be seem from figure X that for all i visual features the projection of ia 1 along the unit normal is given by b n Mid 4 Using equation 4 the relationship in eq 3 can be expressed in the following manner 1 s ne Re Lond m s dr a H where H ER represents a Euclidean Homography To express the above relationship in terms of the measurable image space coordinates of the visual features relative to the camera frame the 3 normalized Euclidean coordinates i Mia tleR for the visual features are defined as m M Mi Mid 6 3 where and Zid are the third coordinate elements in the vector 4 t and iia fl respectively The 2D homogeneous image coordinates of the visual features denoted by pit Pat aR expressed relative to B and Pa respectively are related to the normalized Euclidean coordinates by the pin hole model camera such that Pi Amy Pid Amid 7 where AeR isa known constant upper triangular and invertible intrinsic camera calibration matrix Hence the relationship in 5 can now be expressed in terms of image coordinates of the corresponding feature points in B and P as follows Lid 4 l ue m E R pop i A7 pia 8 et a Q C where lt aR de
7. of extracting relevant information from confusion data sets PCA has limitations that can be solved using other mathematical method such as Independent Component Analysis ICA The assumption that normaly PCA failure is when non gaussian or multi modal gaussian are presente on the data sets This is no a guarantee that the directions of maximum variance will contain good features for discrimination Example Given some input data as shown in the figure 4 11 7 The experiment is given by five input data The covariance matrix it was calculated Then we get the eigenvector and eigenvalues Finally the principal components are calculated Covariance matrix 1 S NUN TN 0 50 0 03 0 03 0 03 0 50 d s RM 0 03 0 04 0 51 0 05 0 03 in EU gt 0 03 0 04 0 05 0 52 0 03 NA rs 0 50 0 03 0 03 0 03 0 50 A A gt re a b Input signals Ordered eigen vectors Principal components 0 69 0 14 0 00 0 00 0 70 y 0 11 0 50 0 65 0 55 0 00 ENZ NA x NX 0 12 0 56 0 20 0 78 0 00 me d sp 0 12 20 62 0 72 0 27 0 09 T Ua TES A 0 69 0 14 0 00 0 00 0 70 u ER Pig gemas Ordered eigen values M LN MEVS ds LE Le an m 1 02 0 59 0 47 0 47 0 00 c d Figure 4 11 7 a Experimental input data b Covariance matrix c Eigenvectors and eigenvalues and d Principal components Independent Component Analysis ICA It is a statistical and computational technique for revealing hidden factors that underlie sets of ra
8. 3 VI 3 and 3 VI 4 In p ut Latitude Latitude rad Altitude Altitude ft Output 1 G Local gravity vector in earth frame gN gE gD ft s Local gravitation vector in earth frame GN GE GD ft s Table 3 VI 3 Gravity model parameters WGSB4 a 63781370 WGS 84 semimajor axis m WGS84 b 6356752 3142 WGS 84 semiminor axis m GM GM 3 9860015 10 m s f WGS84 a WG584 b WGS84 a Flattening of the Earth e v 2f P Eccentricity of the Earth Geocentric latitude rad C 0 48416685 x10 Pa V5 2 3 sinit 1 Ja V5 Cy w 2 x 24 6060 rad s R Distance from you to center of the Earth m N WGS84 a 1 c sin c Physical Parameters Table 3 VI 4 Ellipsoidal earth model The helicopter parameters are summarize in the table 3 VI 5 and figure 3 VI 9 Rigid Body and Fuselage Ig B RON 8 Y 13Q Main Rotor Specific Ys lois hys 0 w Tail Rotor Specific ltr her gt ka m mass slugs Idi moments of inertia slug ft Ls cross moment of inertia slug ft uri iQ aerodynamic reference areas ft Vm Va aerodynamic reference volumes ft HPiost HP lost in transmission Generic Rotor nb number of blades Q rotor speed rad s R rotor disk radius ft c blade chord ft ao Clo lift curve slope 1 rad and offset Cao Cdi Cd2 lift dependent profile drag coefficients Ig flap moment of inertia slug ft kg flapp spring stiffness
9. 513 498 329 481 146 463 962 446 778 429 594 412411 395 227 378 043 360 859 343 675 326 492 309 308 292 124 274 940 257 757 240 573 223 389 206 205 189 021 171 838 154 654 137 470 120 286 103 103 85 919 68 735 51 5514 34 368 17 184 4 ur KC SAB x SOAS S e CELA RECON LOOSE JA GATHERED CO NOSWUV2 Figure 3 IX 11 Litter system simulation for temperate forest model Carbon Cycle Simulation for Terrestrial Biomes Simulation of the Carbon Cycle in Terrestrial Biomes Spatially Distributed Simulation of Soil Nitrogen Dynamics Spatially Distributed Simulation of Soil Nitrogen Dynamics Terrain Analysis Channels There is a tool for digital terrain analysis For example taking the following DEM figure 3 1X 12 we calculated some of the maps listed below im e Fu Er t D t s gt 4 Figure 3 IX 12 DEM for channels calculations Channel Network This module derives a channel network based on gridded digital elevation data Use the initiation options to determine under which conditions channels shall start e D8Flow Analysis Deterministic 8 based flow network analysis a b Figure 3 IX 13 It shows a flow direction and b flow connectivity Overland Flow Distance to Channel Network This module calculates overland flow distances to a channel network based on gridded digital elevation data and channel network information The flow algorithm may be either Deterministic
10. 8 O Callaghan amp Mark 1984 or Multiple Flow Direction Freeman 1991 e Strahler Order A fast recursive algorithm for computing Strahler Stream Order Stream ordering is a useful property of every river network having a wide range of applications a i ee Pte LE FF truite TOA 134 HUE A eS AE HEH r UL o S II c4 a TI i eee M 9 P cw mW ST SF IN r nt ate ptet 1 Figure 3 IX 14 Terrain Analysis Strahler order calculation e Vertical Distance to Channel Network This module calculates the vertical distance to a channel network base level The algorithm consists of two major steps Interpolation of a channel network base level elevation grid Subtraction of this grid from the original elevations N e Watershed Basins e Watershed Basins extended Terrain Analysis Hydrology Cell Balance Cell Balance Calculation Uu Hati i Sle tie 12142 La m om lt ini ifiw 2c4 yO 3 Is i ML STE EE Y MEETS pry 187 c TRUE DA dT pa nord AT Pat e on ce a yu c LE IPY a m use mm 4 NU i t i od 4 Z l Ta gx 1 or La gt E ha B an ctm EN 2 Le cis Ur PL i yr wv NS edi Vh paa T 7 gt LT r T 4 u gt oye d T r du gt d T 2 TE EES 3 LI m I 4 n i t d G e Le a JUN VUE pet fl i
11. Photogrammetry capability is also implemented on Raphael GCS n addition it is integrate an advanced metadata edition tool compatible with most popular geographic information systems software Also it can support the most widely used file formats for import export metadata types Raphael GCS is powerful tool in pre processing and post processing stages from basic image tasks until the most specialise functionalities for example sum or differences between two grid sets until flight image refining stage capable of improving UAV images sufficiently Our system has the most advanced modules such as autopilot terrain analysis channels hydrology lighting visibility morphometry pre processing profiles simulation fire spreading analysis fire risk analysis hydrology analysis human impact on nature computer vision algorithm multispectral data toolkit spectral analysis PCA FastlCA and multidimensional display satellite access 3D reconstruction from multiple images coordinate transformations metadata editing and 3D realistic vision are some of the capabilities of Raphael GCS We offer a product with the capability to produce the sort of information that civilian end users such as real world environmental forestry planning building agencies require Logo Version AVNTK PR aS Ja Fu x 4 om Raphael Advanced Ground Control amp Post processing Station 2 Introduction Raphael GCS is the result of the re
12. a good example of what noise means When dealing with images instead of data grids filters are a basic tool that is constantly used to prepare those images before any further processing However you should be careful when applying filters to data grids since their effect can significantly alter their information For example applying a smoothing filter to a DEM is a somehow rude way of eliminating small sinks but is not recommended since it changes the height information contained in all of the cells while only some of them should be modified Although it removes noise a smoothing filter causes the filtered grid to have a lower lever of detail than the original one Always have this in mind when you use a filter The less you modify the grid assuming its information is reliable and comes from a good source the more accurate your results will be There are several simple filters available on Raphael GCS such as Average Sharp Edge Neon Sobel Laplace Difference of Gaussian Gaussian The degree of smoothing is determined by the standard deviation Exponential Nonlineal Multidirectional Lee This direction Lee Filter is a enhanced Lee Filter It looks into 16 directions for the direction with the minimum variance and applied a Lee Filter on this direction Uses this filter for remove speckle noise in SAR images or DTMs On DTMs this filters will preserves the slope and narrow valleys User defined Filter Interpolation o
13. arithmetical operations with grids and can be therefore used for a large variety of purposes Operations such as Addition subtraction multiplication division power trigonometric functions transcendental operations boolean operations Changing grid orientation Raphael GCS has its own module to do basic operations using grid such as rotating mirror stretch Flip or Invert an image e Grid Normalisation This is quite common when you deal with several grids and you want to use a model in which each parameter should be scaled according to a particular scale such as the trivial range 0 1 You can normalize a grid using the grid calculator If standard deviation method is selected then the standard deviation of the resulting normalized grid will equal 1 e Grid Volume Calculate the volume under the grid s surface This is mainly useful for Digital Elevation Models DEM e Random Field Create a grid with pseudo random numbers as grid cell values e Random Terrain Generation e Filtering grid A filter modifies all the cells in a grid applying different algorithms and formulas to generate a new version of that grid The grid you obtain after a filtering process represents the same variable and in the same units as the original grid The most common use for a filter is the elimination of noise Although the art of filter is widely used in detection circles lines or any other template Single cells with unexpectedly high values are
14. b t m Figure 3 1X 15 It shows cell balance map Downslope Area Interactive This interactive module allows you to specify source cells with a left mouse click for which the downslope area shall be identified For the Deterministic Infinity and Multiple Flow Direction algorithms which are able to simulate flow divergence the result will give the percentage of the source cells flow that drains through each cell Currently available algorithms are e Deterministic 8 Rho 8 Braunschweiger Reliefmodell Deterministic Infinity Multiple Flow Direction Kinematic Routing Algorithm DEMON Edge Contamination Flow Depth Interactive Flow Depth Calculation Flow Path Length This module calculates the average flow path length starting from the seeds that are given by the optional Seeds grid and optionally from cells without upslope contributing areas i e summits ridges Seeds will be all grid cells that are not no data values If seeds are not given only summits and ridges as given by the flow routing will be taken into account Available flow routing methods are based on the Deterministic 8 D8 Callaghan and Mark 1984 and the Multiple Flow Direction FD8 Freeman 1991 Quinn et al 1991 algorithms Figure 3 1X 16 Flow path length map Flow Sinuosity Interactive Flow sinuosity calculation Flow Tracing Flow tracing algorithms for calculations of flow accumulation and r
15. export raster format which uses information on how to interpret the actual data file Raster data can be created from point data using nearest neighbour triangulation and other interpolation techniques Modules for the construction and preparation of raster data allow the resampling closing of gaps and the value manipulation with user defined rules A number of filter algorithms exist for smoothing sharpening or edge detection Classifications can be performed using cluster analysis or a supervised procedure like Maximum Likelihood classification Data analyses cover image pattern and cost analysis Other standard operations are skeletonisations and bufferings However the fundamental task to do is the visualization of raster data Raphael GCS is capable of visualize raster data easily A typical Digital Elevation Model is shown in the figure 3 VIII 1 The image has great level of detail identifying high area and low areas The combination of shapes vector data and grid raster data has many advantages and Raphael GCS is perfectly capable of combining them and making it easy for the user to handle both data types without effort For example if you want to create a DEM from point data by interpolating the height information associated with those points Raphael GCS will have no problem For a detailed list of all available functions see Appendix 8 1 3 1X ADVANCED GIS ANALYSIS MODULES However final users such as aid agencies search and r
16. f Figure 4 11 12 It is shows the six principal components of the satellite image data a It is the first principal component From b to f it is shows the deceasing contribution to the total signal image Figure 4 11 13 Noisy image component We can see that the first component see figure 4 11 12 a has strong contribution to the total image set while last component see figure 4 11 13 is associated with the nosier image component Fast ICA method Once the centering and whitening step is done and using the minimization function performed we get the covariance matrix which is given by ENT y xe x 2 is Figure 4 11 14 Shows covariance matrix After the emer eence is reached the principal components is given by the following images see figure 4 11 15 763 136902 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 353 405670 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 1171 009521 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 1108 754150 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 6981 224609 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 622 828125 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 0 000000 2547 533203 d e f Figure 4 11 15 It is shown the principal components using the FastlCA algorithm a It is shows th
17. forces trust torque and power The main rotor is not only the dominant system but also the most complex mechanism It is the primary source of lift which counteracts the body weight and sustains the helicopter on air Additionally the main rotor generates other forces and moments that enable the control of the aircraft position orientation and velocity Raphael GCS uses the rotor dynamic model whose main building blocks are depicted in figure 3 VI 7 main rotor flapping A force amp dynamics moment a A gt u 9 downwash CES f Figure 3 VI 7 Main rotor block diagram Blade a Lift curve slope 1 rad R Blade radius ft w Rotor angular velocity rad s b Number of rotor blades c Rotor blade chord ft Ry Blade root cutout ft Blade collective pitch at 75 R rad 0 eui Blade twist for washout rad Cdo Profile drag cooeficent Oswald efficency factor Vi Perpendicular velocity to rotor disk in direction of T ft s p Air density slug ft T Output thrust Ibs Q Output torque lb ft P Output power Ib ft s avg Output average induced velocity ft s Table 3 VI 1 Structure for the blade element calculations Blade calculation In order to calculate thrust power torque and induced velocity of the main rotor and tail rotor we have used a combined blade element moment theory The table 3 VI 1 shows the structure for these calculations Bell Hiller stabilizing bar Cur
18. in order to get a more detailed 3D scene reconstruction model This 3D scene reconstruction work for multiple image views of a target of interest where the position of each viewing image is only roughly known is ideal when digital images including photographs are available for example because the phenomenon is unique or of difficult access such as an asteroid perhaps videos from a youtube website or a combination of aerial videos taken at different times from varying positions with a range of imaging systems and or ground images of a target of interest When a sophisticated expensive heavy sensor such as a Light Detection And Ranging technology LIDAR cannot be readily accessed due to cost availability or maximum payload constraints of a platform such as a tactical UAV even thus we can get an accurate 3D scene reconstruction using images This approach will be widely useful in coming years Initially our 3D scene reconstruction algorithm assumes multiple image views in the form of several digitally captured 2D images of a scene or target taken from different angles and under distinct weather conditions These images can be coded in bmp png jpg gif mpeg avi or any other digital format Our software automatically computes information about location and orientation of the cameras only from the images themselves using computer vision methods though providing rough GPS information certainly speeds up the process The algorithm develo
19. intensity flame length and scorch height of free burning surface fires for a given Digital Elevation Model DEM fuel model grid set of fuel moisture wind speed and its direction It is based on the BEHAVE fire modelling system supported by the U S Forest Service Fire and Aviation Management see at http fire org Input Data Name Type Label Description DEM Grid DEM Digital Elevation Model Fuel Model Grid FUEL Wind Speed Grid WINDSPD Wind speed m s Wind Direction Grid WINDDIR Wind direction degrees clockwise from north Dead Fuel Grid MIH MIOH MIOOH Moisture 1H 10H 100H Herbaceous Grid MHERB Fuel Mosture Wood Fuel Grid MWOOD Moisture Ignition Points Grid VALUE Table 3 IX 3 Fire Spreading Analysis input data Output data Name Type Label Description Time Grid TIME Time since initial ignition Flame Length Grid FLAME Flame length m Intensity Grid INTENSITY Intensity Kcal m Table 3 IX 4 Fire Spreading Analysis output data fune Hie piene kangh i ey wen Cocoa irxten Pores TE lue was Speed iima 13 wid Law ia deg Corinne Prem berth Figure 3 1X 5 Time spread mins Figure 3 1X 6 Left Flame Length m Rigth Intensity Kcal m Simulation Hydrology Overland Flow Kinematic Wave D8 A distributed hydrologic model known as the Terrestrial Hydrologic Model or THM was developed f
20. marginal densities i e the definition of independence The following section is an example of applying PCA and ICA on satellite image data PCA method Modules Data f Maps E mee 4 Thumbnails T 3234x 2903y Ox Oy Figure 4 11 10 Lansat 7 as input data for PCA calculations The covariance matrix and eigenvalues calculated are given by 1 456865295 315781097 600 248657 528 777222 1602683105 622 828125 1011 075195 E EB 462 038818 353 405670 563 622437 420 210754 1087 736084 315 781097 706 635498 3 604848877 420210754 742812073 1108 754150 1911 196289 528 777222 1107 785522 IN 763 136902 462 038818 794 773621 604 848877 1563 328613 456 865295 1024 065796 5 794 773621 563 622437 1171 009521 742 812073 1960 635742 600 248657 1293 747803 EN MI 1024 065796 706 635498 1293 747803 1107 785522 3501 788086 1011 075195 2547 533203 1563 328613 1087 736084 1960 635742 1911196289 6981 224609 1602 683105 3501 788086 3046 861816 866 863525 493 075409 369 764099 144 417526 126 050346 34 527451 Figure 4 11 11 Shows covariance matrix and eigenvalues After applying PCA on this satellite data image set we have the following principal components a b c d e
21. materials and shall conform to its product specifications 12 13 14 15 16 17 Items marked with an asterisk on the face of the quotation are not produced by seller Seller is not responsible for the installation operation or warranty of these products although the manufactures of these products may provide their own warranties to you THIS WARRANTY IS EXCLUSIVE AND IN LIEU OF ALL OTHER EXPRESS AND IMPLIED WARRANTIES INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE AND ANY OTHER OBLIGATIONS OR LIABILITIES WHETER IN CONTRACT WARRANTY NEGLIGENCE OR OTHERWISE SELLER SHALL NOT BE LIABLE FOR ANY DIRECT INDIRECT SPECIAL PUNITIVE CONSEQUENTAL OR INCIDENTAL DAMAGES INCLUDING WITHOUT LIMITATION DAMAGES FOR LOSS OF BUSINESS PROFITS AND BUSINESS INTERRUPTION ARISING OUT OF THE USE OR INABILITY TO USE PRODUCTS OR SERVICES SOLD HEREUNDER Seller s liability shall in no event exceed the cost of the products sold under this contract SOFTWARE LICENSES Title to licensed products shall remain with seller and shall be subject to software license terms provided with software AUTHORITY TO EXPORT Buyer shall not export or re export technical data software or products supplied by seller in violation of applicable export regulations Buyer who exports from the U S Mexico or United Kingdom products purchased hereunder assumes all responsibility for obtaining any required exp
22. of Grids Coordinate Transformation for Grids Proj 4 Dialog List of Grids Proj 4 List of Shapes Coordinate Transformation for Shapes Proj 4 Dialog List of Shapes Proj 4 Shapes Coordinate Transformation for Shapes Proj 4 Dialog Shapes Projection Tools for the georeferencing of spatial data grids shapes Create Reference Points Interactive Digitize reference points for georeferencing grids images and shapes Click with the mouse on known locations in the map window and add the reference coordinates After choosing 4 or more points stop the iterative module execution by unclicking it in the modules menu Georeferencing Grids Either choose the attribute fields x y with the projected coordinates for the reference points origin or supply a additional points layer with correspondend points in the target projection Georeferencing Move Grids Interactive Georeferencing Shapes 10 11 9 General Terms and Conditions SELLER AVNTK S C ACCEPTANCE This quotation constitutes seller s offer to buyer and is expressly limited to the terms and conditions stated herein Any order received subsequent to this offer shall be an acceptance of the offer Seller objects to any additional and or different terms contained in buyer s response This offer expires sixty 60 days from its date unless products are subsequently shipped by seller and accepted by buyer PRICES AND TERMS OF PAYMENT The prices for the pro
23. of difficult access such as an asteroid perhaps videos or a combination of aerial videos taken at different times from varying positions with a range of imaging systems and or ground images of a target of interest When a sophisticated expensive heavy sensor such as a Light Detection and Ranging technology LIDAR cannot be readily accessed due to cost availability or maximum payload constraints of a platform such as a tactical UAV even then we can get an accurate 3D scene reconstruction This approach will be widely useful in coming years Our software automatically computes information about location and orientation of the cameras only from the images themselves using computer vision methods though providing rough GPS information certainly speeds up the process The algorithm developed works by breaking down the problem into several stages These are as follows camera model and calibration is presented Detected features points and the algorithm about matching point features between pairs of images An iterative Structure from Motion SfM procedure to recover the camera parameters is used Two methods about stereo matching of calibrated images are also used The Poisson reconstruction with oriented 3D reconstructed points set is employed The final texture 3D scene is also novel Finally we have developed an optional and complementary module that is useful to align multiple mission tracks under distinct weather condition and illumination conditions
24. projection center into the image plane It is a useful model that enables simple mathematical formulation for the relationship between object and image coordinates However it is not valid when high accuracy is required and therefore a more comprehensive camera model must be used The model contains parameters divided into extrinsic and intrinsic parameters Extrinsic parameters refer to camera pose and intrinsic parameters refer to image plane properties The origin of the camera coordinate system is in the projection center at the location xo yo zo with respect to the object coordinate system and the z axis of the camera frame is perpendicular to the image plane The rotation is represented using Euler angles w p and k that define a sequence of three elementary rotations around x y z axis respectively Rotations are performed clockwise first around the x axis then the y axis which is already once rotated and finally around the z axis that is twice rotated during the previous stages In order to express an arbitrary object point P at location X Y Z in image coordinates we first need to transform it to camera coordinates pg L This transformation consists of a translation and a rotation and it can be performed by using the following matrix equation T r11 712 713 X i TO y r21 T22 T23 Yi f Yo a r31 32 T33 Zi zu where rio SIN sin ip cos K cosu SIN K T1 COS Y COS K Tos SIN w SIN ip sim amp COS W cos K la
25. square method to solve for the translation vector and rotation matrix Furthermore we have included in the least square problem information about normal vectors over each matched pair point Finally we get an enlarged list of 3D orientated points with additional information about mixed cameras This new list of 3D orientated points is filtered out using the Poisson reconstruction process to get a new 3D surface reconstructed model with additional detail given by the combination of multiple and independent mission tracks Since we have more 3D oriented points with associated camera lists we can get a better texture 3D surface model Of course this extra images could include images taken from the ground thus being complementary to those taken from satellites LIDAR and unmanned vehicles in more than one mission and at different times in order to derive a high quality 3D reconstruction of any target of interest 1 Lok we Tq B amp i e N po Figure 4 1 9 1 The alignment between a oriented 3D reconstructed mesh red dots and the CAD model mesh blue dots Comparison between CAD model and 3D reconstructed scene We have tested this algorithm on a number of real life objects but here we show a study of a building of known dimensions in order to quantitatively estimate the algorithm s accuracy We have quantified our 3D reconstruction process through a typical 3D scene reconstruction with a well known CAD model Thi
26. t where K and 2 are coefficients for radial distortion and r i V i2 1 2 Typically one or two coefficients are enough to compensate for the distortion Curvature centers of surface lenses of lens surface are not always strictly collinear This introduces another common distortion type de centering distortion that has both a radial and tangential component The expression for the tangential distortion is often written in the following form Ages nhat 2pi V pa r 2 2 Avian ntial p r ov i 2pzuiVi where P and P are coefficients for tangential distortion A sufficiently accurate camera model for accurate calibration can be derived by combining the pinhole model with the correction for the radial and tangential distortion components js 4 radial 4 tangential b P tw d a li l te tt l ee Ui D u Av mnt od Av PERTE vo In this model the set of intrinsic parameters Jis Mg Vois augmented with the distortion coefficients ky k P and P These parameters are also known as physical camera parameters since they have a certain physical meaning Generally the objective of the explicit camera calibration procedure is to determine optimal values for these parameters based on image observations of a known 3D target In the case of self calibration the 3D coordinates of the target points are also included in the set of unknown parameters Since we have th
27. the case of work done for a final customer or shared with the UAV manufacturer if applicable UAV users To help develop a collaborative application base is beneficial to everyone as shown by cases such as the Neuron UCAV Airbus JSF ISS a single country company cannot succeed We are happy to support end users both with technology and also with consultancy services both to themselves and or to their prime contractors or other suppliers such as universities R amp D centres etc Revisions history Rev Date 2071072007 Preliminary Draft First draft 11 0101710 First Draft FR SSS 8 Appendices 8 1 DETAIL OF GIS FUNCTIONS The GIS module currently has basic operation with grids such as Elementary grid operations e Function Generate a grid based on a function expression The function interpreter uses an expression parser that offers the following operators Addition Subtraction Multiplication Division power sin a cos a tan a asin a acos a atan a atan2 a b abs a int a sqrt a mod a b gt a b return 1 if a greater b lt a b returns 1 if a lower b eq a b return 1 if a equal b The variable are x and y Examples sin x x y y x x y y e Geometric Figures Construct grids from geometric figures planes cones e Grid Calculator One of the most useful modules that you can find in Raphael GCS is the grid calculator Since it isa very flexible module it allows your to perform all kind of
28. the exponential conductivity Green Ampt model of Beven HSJ 1984 but if infiltration excess does occur it does so over whole area of a subcatchment Spatial variability in conductivities can however be handled by specifying a parameter values for different subcatchments even if they have the same In Ter and routing parameters i e to represent different parts of the area Note that time step calculations are explicit ie SBAR at start of time step is used to determine contributing area Thus with long daily time steps contributing area depends on initial value together with any volume filling effect of daily inputs Also base flow at start of time step is used to update SBAR at end of time step Input Name Type Label Description a Grid ATANB tan B Climate Data P EP Table CLIMATE Table 3 1X 5 TOPMODEL Input data Optimal simulation parameters Name Label Value Unit Time Step DTIME 1 h Number of Classes NCLASSES 30 Initial subsurface P_QSO m h flow per unit area a agio gt Areal average of In TO In Te P LNTE 5 m in Pt Model parameter P MODEL 0 032 m Initial root zone storage deficit P SRO 0 002 m Maximum root zone storage deficit P SRZMAX 0 05 m Unsaturated zone time delay P SUZ TD 50 0 h per unit storage deficit Main channel routing velocity P VCH 3600 0 m h Internal subcatchme
29. the software as previously described can be combined with hardware into a standard package containing the following items e Basic Raphael GCS software suite e Manual radio control e Telemetry transmitter receiver e Directional antenna mounted on a gimbal able to communicate with a vehicle at up to 30 Km in direct line of sight e Ruggerized laptop User manual 6 Pricing details The following are the launch costs associated with the Raphael GCS software which compare very favourably with the costs of in house development and or using hardware such as a LIDAR for doing 3D reconstruction for instance The cost of each module is as follows e Basic System 6 500 for a site license irrespective of the number of users and 500 per yearly update e FMV to 3D module 3 500 for a site license irrespective of the number of users including the CAD to 3D object comparison routine e Multi spectral analysis 2 500 for a site license irrespective of the number of users including multivariate analysis and image fusion routines e Satellite access 2 500 for a site license irrespective of the number of users currently supported satellite sensors include AIRS MODIS and LANDSAT 7 Collaboration Structure We believe that in order to establish Raphael GCS as the leading GCS in the market it is important to accommodate the need of many companies of not only acquiring technology but also being able to add value by adding thei
30. 30 sun gilet destsece LIB AMSU 16 tet integer 45x30 topog LIB AMSU 32 bit fleating voin 45x30 t0pog err 118 AMEL 32 bit fostng pont 45x30 anra LIS ANSU 32 tit foating point 145 30 landFrac er LIB AMSU 32 Dit floating port 145x 30415 antenne temp LIS AMSU 3 bit foating poire 145 30415 brightness temp LIB AMSU CX2 bir faatino coir 45x 30x15 brightness temp ee LIN AMS 32 bit ieaceg point 2115 bio sonas min 118 ANSU 57 211 Nosting poinm 2 15 to sonis max LIB AM SU CX7 btt hosting porrm 215 bi gnas mean LIB AMSU 132 b fioeting porti 2215 b gns dev LIB AMSU G2 b hating core 215 Do Sones rum LIS AMSU 32 bit Integer 2115 b ogres num bad LIS ANSU 32 5t integer 12x15 bb Zonas max track LIS AMSA 32 5 integer 12215 b uonals max track LIS AMSU 32 bit integer 122251 bb Sonais min track LIB AMSU 22 bit integer 2115 b Sonan min xtrack LIS AMEL 32 bit integer Gne min 18 ANSU 32 bir floabeg pod J 1x15 Space Sonas max LS AMSU 32 bit fiobat ng poim 2215 spece snae meen LIB AMSU L32 bst eating pointi 2115 space sons dev L2B AMSU 22 68 fioating poirt u tel owe va IO amp t 4TT13 LU orem Figure 4 IIll 1 Header of the hdf file format of AIRS instrument aboard Aqua satellite Figure 4 111 2 World map representation of Lansat 7 data It uses our 3D visualization toolkit 5 Standard Raphael Ground Control Station The standard specification of
31. 5 Peucker and Douglas Wind Effect Windward Leeward Index This module calculates the wind effect of a given DEM The input parameter are wind direction and maximum distance Terrain Analysis Preprocessing Fill Sinks Planchon Darboux 2001 Depression filling algorithm given Olivier Planchon amp Frederic Darboux 2001 This algorithm consists of increasing the elevation of pixels in closed depressions until the sink disappears and a mininum slope angle of minslope default 0 01 degree is established Fill Sinks Wang amp Liu This module uses an algorithm proposed by Wang amp Liu to identify and fill surface depressions in digital elevation models The method was enhanced to allow the creation of hydrologic sound elevation models i e not only to fill the depression s but also to preserve a downward slope along the flow path If desired this is accomplished by preserving a minimum slope gradient and thus elevation difference between cells This is the fully featured version of the module creating a depression less DEM a flow path grid and a grid with watershed basins If you encounter problems processing large data sets e g LIDAR data with this module try the basic version Fill Sinks XXL Fill Sinks XXL Wang amp Liu This module uses an algorithm proposed by Wang amp Liu to identify and fill surface depressions in digital elevation models The method was enhanced to allow the creation of hydrologic sound elevation
32. 7 used on a standard CCD camera gives the following results Parameter Value Principal Point 321 9588 237 3126 pixel Scale Factor 0 9989 Effective focal length 3 2864 mm Radial distortion k 2 2 422831 x10 mm k 1 234360 x 10 mm Tangential distortion pi 1 343740 x 10 a mm l p 7 714951 x 10 mm Table 3 VI 1 It shows calibration parameter for a typical digital camera Me Feet bons af te tre twee l art of Figure 3 Vl 1 A window of the camera calibration module Figure 3 VI 2 The radial and tangential effects are shown The effect has been exaggerated 10 times Orthorectification Cartography having properly corrected source UAV images enables a successful and highly accurate ortho rectification and tiling process Direct Linear Transformation Projective 2D Polygonal 3D Polygonal Linear General Affine Non Linear Rigid Body Non linear Orthogonal and Linear Isogonal Figure 3 Vl 3 shows a typical orthorectificaiton for a pair of images View A View B Figure 3 Vl 3 View A shows a image taken with a certain camera pose while View B shows another view taken with different camera pose The ortho rectified image is shown too One of the challenges of full motion video exploitation lies in how to present the images to the user in such a way as to maximize comprehension of the information One of the effects that minimizes comprehension is having only a localized v
33. AVNTK RAPHAEL GROUND CONTROL AND POST PROCESSING STATION GENERAL DESCRIPTION PRODUCED BY THE AEROSPACE TECHNOLOGY GROUP ARDITA AERONAUTICA DIVISION Technical POC Fidel Guti rrez Resendiz Ph D Aerospace Remote sensing Group Leader Guadalajara Jalisco Mexico E mail fidel ardita aeronautica com Contents Summary Introduction Basic System Description 3 1 Introduction 3 1 Dynamic Mission Planning and Tasking 3 1 Synthetic Vision 3 IV Interface with onboard avionics 3 V Simulator amp Autopilot Automatic Takeoff and Landing Takeoff Performance Stellar navigation Autopilot helicopters blimps fixed wings Control Panel Helicopter Dynamic Model Main rotor Blade calculation Bell Hiller stabilizing bar Gravitation Model Physical Parameters Wind model Servo control input Atmosphere Earth Model 3 VI Photogrammetry Module Camera Calibration Orthorectification 3 VII Image restoration Increase Depth of Field 3 VIII Raphael Geographical Information System 3 IX Advanced GIS Analysis Modules Simulation Cellular Automata Simulation Fire Risk Analysis Simulation Fire Spreading Analysis Simulation Hydrology CON OO OO OF SH 12 15 20 21 22 22 22 29 23 24 25 21 24 28 28 29 30 32 34 35 37 37 40 41 42 Simulation Identification of unit hydrographs and component flows from rainfall evaporation and streamflow data IHACRES Simulation Mode
34. ERY DATE The delivery date stated on the face of this contract is an estimate only Seller will make reasonable efforts to meet the estimated delivery date however seller shall not be liable for failure to meet it unless otherwise stipulated FORCE MAJEURE Seller shall not be liable for any failure to perform due to causes beyond its reasonable control including but not limited to fire accident war pandemics labor dispute shortages embargo delayed delivery by suppliers delay in transportation inability to secure labor or materials acts of government whether foreseen or unforeseen Should any of these events occur seller may at its option cancel buyer s purchase order with respect to undelivered products or extend the delivery date for a period equal to the time lost because of the delay DESCRIPTIVE LITERATURE Specifications are subject to change without notice unless otherwise stipulated FEDERAL ACQUISITION REGULATIONS If seller supplies certain products to the U S federal government under contract to the General Services Administration applicable Federal Acquisition Regulations FAR apply to such contracts WARRANTY AND EXCLUSIONS The warranty for products produced by seller and supplied under this contract is limited to the system warranty incorporated by reference a copy of which is provided in the product documentation and or upon request Seller warrants that each product it sells buyer is free from defects in labor and
35. GCS Wetness Index is as the name says similar to the Topographic Wetness Index TWI but it is based on a modified catchment area calculation Modified Catchment Area which does not think of the flow as very thin film As result it predicts for cells situated in valley floors with a small vertical distance to a channel a more realistic higher potential soil moisture compared to the standard TWI calculation Slope Length Topographic Indices Calculation of slope and catchment area based topographic indices Topograhic Wetness Index TWI USLE LS factor Stream Power Index Upslope Area This module allows you to specify target cells for which the upslope contributing area shall be identified The result will give for each cell the percentage of its flow that reaches the target cell s Available Upslope Area methods are based on Deterministic 8 Deterministic Infinitum Multiple Flow Direction e Upslope Area Interactive This module allows you to specify target cells for which the upslope contributing area shall be identified The result will give for each cell the percentage of its flow that reaches the target cell s Available Upslope Area methods are based on Deterministic 8 Deterministic Infinitum Multiple Flow Direction e Flow Width Flow width and specific catchment area SCA calculation Available flow width methods are based on Deterministic 8 Multiple Flow Direction and Aspect Terrain Analysis Lighting Visibilit
36. H ONBOARD AVIONICS Our ground control station has been used with a wide range of unmanned vehicles including fixed wing UAVs blimps and helicopters A range of communication protocols are available and have been tested including WiFi using Matchport hardware GSM GPRS using Fargo Maestro 100 hardware and RF High Speed Download Packet Access HSDPA is also an option but not yet tested because it is too expensive to implement due to the costs of the hardware and the monthly subscription costs for the use of 3G services As a commercial large scale implementation rather than small scale vehicles this would be the ideal choice of communications standard due to its high data rates and large range The large coverage is due to the wide availability of 3G it is available in most urban areas making it ideal for surveillance use RF is cheap to implement and has a range of up to 10 km The range is large enough to control a quadcopter for instance as the battery life is only 10 15 minutes typically which limits the distance that it can cover The main problem with RF is that it is not as readily available as other standards so only covers the distance from an RF transmitter Another issue is that the data rate is only up to 10 kbps which is not high enough for good quality video transmission The MikroKopter is equipped with a basic RF receiver that can be used for control but was not be upgraded for data communications GSM GPRS is cheap to i
37. Memory BNA CSV GML GPX KML GeoJSON GMT XPlane AVCBin AVCEO00 Geoconcept Table 3 11 5 Export vector data OGR Import vector data This modules imports vector data from various file database formats using the GDAL library see Table X Import Export Grids Tools for the import and export of gridded data currently available on Raphael GCS Export ESRI Arc Info Grid Export grid to ESRI s Arc Info grid format Binary or ASCII format Export Grid to XYZ Export grid to a table text format that constains for each grid cell the x y coordinates and additionally data from selected grids Export Surfer Grid Export grid to Golden Software s Surfer grid format Export True Color Bitmap Export red green blue coded image grids to MS Windows true color bitmaps This module writes the data directly to the file and is hence particularly suitable for very large data sets Import Binary Raw Data Imports grid from binary raw data Import ESRI Arc Info Grid Import grid from ESRI s Arc Info grid format Import Erdas LAN GIS Import Erdas LAN GIS files Import Grid from Table Imports grid from table Import Grid from XYZ Import grid from a table text format that constains for each grid cell the x y z coordinates and additional data from selected grids Import MOLA Grid MEGDR Import Maras Orbit Laser Altimer MOLA grids of the Mars Global Survey MGS Mission Topographic maps Mission Experiment Gridded Data Recor
38. Ordered weighted averaging OWA Ordered Weighted Averaging OWA Pattern analysis Polar to Rect Polar to Rect conversion for paired vector data Rect to Polar Rect to Polar conversion for paired vector data Soil texture classification Define soil texture with USDA scheme from sand and clay contents Classification Clay soils Silt soils SiltyClay soils SiltLoam soils SiltyClayLoam soils Loam soils SandyClay soils Sand soils SandyClayLoam soils LoamySand soils ClayLoam soils SandyLoam Table 3 11 1 Soil texture classification e Vegetation Index distance based Distance based vegatation indexes 1 Richardson Wiegand PVI 2 Perry and Lautenschlager 3 Walter and Shabaani 4 Qi e Vegetation Index slope based Slope based vegetation indexes 1 Normalized difference vegetation 2 Ratio vegetation index 3 Transformed vegetation index 4 Corrected transformed vegetation 4 Thiam s transformed vegetation 5 Normalized ratio vegetation index Import and Export between raster vector data formats Import and Export of Data Fundamental for the work with spatial data are interfaces to the countless data file formats Particularly the data exchange between different programs usually requires a suite of import and export filters Raphael GCS offers several filters to common data formats including various image Most flexible is a raster data import tool that uses the Geospatial Data Abst
39. ac LIS ANSU 32 bit floating point 145420 fandFrac er LIB AMSU 32 bit footing port 145x 30415 antenea temp LIS AMSU 32 bit Noating point 45130115 brightness temp LIB AMSL 32 bit fieating pelnt 145 30415 bxightness temp ar LIA AMSJ 32 bit fiesceg por 12215 bib Sele min 16 AMS 37 bit flospng poie 2115 tib sursis max LIG AM SU X2 bit flosting peinti 2x15 bio signas mean L1B AMSU 32 b floating port 2115 bi s gna s dev LiB AMSU 2 b flonting poirti 12x45 D5 wgnais rum LIS ANSU 32 bit imeger 12x25 io Sgnais num bed LIS ANSU 32 5t integer 12x15 bb zonas max track LIS AMSA 32 5 integer 12x15 b Signals max Wrack LIS AMSU 32 bit Integer 12215 bb signas min track LIB AMSU 32 bit integer 12115 t gnas min xtrack LIS AMSU 32 SiE integer 12215 pece sonas min 18 AMEL 32 51 flaatieg pom 2115 Space sanas max LIB AMSU 372 bit fipatng point 2x15 space signes meen LIB AMSU bi fiesting pointi 2215 spsce sons dev L2B AMSU 32 bit flonting gotr t t m P oem rr E m i 2 wet ieme e oom ee edi we I1 j ILILIEJ LILIDDE LJLJL JL J JOOOJ I IL JL JI JUJJA 4 4 GA amp mu re e T i mn it This module mainly works with multispectral satellite metadata Satellite information is normally written using hdf format version 4 or 5 The hdf header contains information about
40. al flow for a sparse feature set using the iterative Lucas Kanade method with pyramids Compute dense optical flow using Gunnar Farneback s algorithm a function implementing the CAMSHIFT object tracking algorithm etc All these algorithms are based on openCV as well as Vigra software for example object tracking for which we have implemented a tracking algorithm to follow an objective Feature detection and match correspondence is a crucial step to make a 3D reconstruction model By itself tracking is an essential step in the mapping of objectives The virtual reality environment has a number of 3D libraries to help generate realistic environments including a range of surface textures bricks cement doors tiles windows wood etc vegetation primitives including a wide range of trees city items such as lights and signals cars posts including high voltage posts and the like In global terms the environment correctly places the Sun in relation to the Earth at any one time and any geographical DEM position can be freely downloaded from http www gdem aster ersdac or jp with a resolution of 30 meters and used as a reference for UAV derived imaging comparison with satellite data or many other forms of GIS information processing r pe me re we Dern ee m om Figure 3 IIl 2 The Colima Volcano in Jalisco Mexico using synthetic data i e we have used a DEM with 30 meters resolution and Raphael GCS texture 3 1V INTERFACE WIT
41. alization that most UAV manufacturers have concentrated on hardware and functionality and not enough on post processing of the data obtained On the other hand some potential users have advanced geographical information systems such as both ArcView or ArcGIS on windows and GRASS in a UNIX operating system but these assume high input image quality that is not easily obtainable from vibrating low resolution uncorrected UAV cameras and cannot analyze UAV imaging even with quite advanced GIS Moreover civilian customers require advanced analysis tools to derive useful results such as fire predictions hydrology maps cartography 3D reconstruction from full moving video FMV build maps from interpolate random samples points super resolution images segmentations transformation between multiple earth coordinate systems visualization of 3D satellite metadata multifrequency analysis etc Therefore there is a need for a software tool capable of performing these tasks to be combined with existing software so as to offer an attractive product to potential civilian end users We have developed a leading edge ground control system for UAVs that includes a mission planning and re tasking module synthetic vision in a 3D virtual reality environment an autopilot autonomous take off and landing using a single camera a robust wind compensating algorithm camera correction and calibration a photogrammetry module a complete geographical information system i
42. an must accommodate a range of possible emergencies and other unplanned in flight events like pop up threats or a critical aircraft system failure Current in flight mission re planning systems do not have sufficient capability for operators to effectively handle the full range of surprises commonly encountered in flight operations Automation is commonly employed to reduce this high workload on human operators Our dynamic mission planning module overcomes a variety of common operational situations in HAE UAV reconnaissance that necessitate more direct human involvement in the aircraft control process than is currently acknowledged or allowed A state of the art mission planning software package OPUS can be used to demonstrate the current capability of conventional mission planning systems This current capability can be extrapolated to depict the near future capability of highly automated HAE reconnaissance UAV in flight mission replanning Many scenarios exist in which current capabilities of in flight replanning fall short Our dynamic mission planning module has been developed and implemented in Raphael GCS and when the same problematic scenarios are revisited with it improved replanning results can be demonstrated particularly being able to reroute in the light of new information and threats slack time available interpretation rating scale of points of interest and a given route survivability estimate Capabilities include e Survivabili
43. apes Tools for the manipulation of point vector data Shapes Tools for Polygons Shapes Tools for the manipulation of vector data Table This module is designed for table calculations Table Tools TIN Tools for triangular Irregular Network TIN processing Point Cloud Tools for points clouds Point Cloud Point Cloud Viewer Projection Tools for coordinate transformation based on PROJ 4 Projection Tools for the georeferencing of spatial data grids shapes General Terms and Conditions 57 57 58 61 62 64 66 66 69 70 71 74 79 81 82 83 84 84 84 85 86 87 87 88 89 90 92 93 93 93 94 94 94 95 95 96 96 96 96 97 98 1 Summary Raphael Ground Control Station Raphael GCS is an advanced system developed under LINUX operating system it is based on C programming Raphael GCS integrate several packages in only one system For example normal functions associated with a Ground Control Station are already implemented such as mission planning autopilot remote control Raphael GCS has integrate a scale helicopter simulator characterize by 80 physical parameters It solves aerodynamic equations of a rigid body embedded on the earth gravitational potential Synthetic vision module is a sophisticated virtual reality tool that is also implemented The 3D synthetic vision is widely used on metadata visualization for example multi frequency satellite raster maps
44. are 7 Binary Grid Grid grd Grid grd grd COSAR COSAR Annoted Binary TSX TerraSAR X Product COASP DRDC COASP SAR Processor Matrix TerraSAR X Raster PNM Portable PixMap Format D0QI USGS DOQ old style D0Q2 USGS DOQ new style netpbm ENVI ENVI hdr EHDR ESRI hdr GenBin Generic Binary hdr PAux PCI aux MFF Vexcel MFF Raster MFF2 Vexcel MFF2 HKV Raster FujiBAS Fuji BAS Scanner Image GSC GSC Geogrid FAST EOSAT FAST Format BT VTP bt Binary Terrain LAN Erdas lan gis CPG Convair PoIGASP 1 3 IDA Image Data and Analysis NDF NLAPS Data Format EIR Erdas Imagine Raw DIPEx DIPEx LCP FARSITE v 4 Lascape File RIK Swedish Grid RIK rik Icp USGSDEM USGS Operational ASCII GXF GeoSoft Grid Exchange HTTP HTTP Fetching Wrapper DEM and CDED Format HDF5 Hierarchical Data Format HDFImage HDF5 Dataset ADRG ARC Digitized Raster Graphics Release 5 BLX Magallan topo blx Table 3 11 4 Import raster support formats continued Export Raster to GeoTiff via GDAL This module exports one o more grids to a Geocoded Tagged Image File Format using the Geospatial Data Abstraction Library GDAL OGR Export vector data This module exports vector data to various file database formats using the GDAL library Currently vector data supported formats are Name Name Name Name ESRI Shapefile MapInfo File UK NTF SDTS TIGER S57 DGN VRT REC
45. at explicitly includes the effects of the main rotor stabilizing bar tail rotor fuselage horizontal tailplane vertical fin and gravity Also it includes a module to make the algorithm robust and linearised The control is effected through a Kalman filter as standard The model used for the atmospheric parameters such as temperature pressure density speed of sound gravity etc is that of the NASA 1976 standard atmosphere with a variation of less than 1 compared to tabulated values The autopilot model includes models of the systems response of the two types of servo used in our UAV hardware The autopilot is able to navigate in three modes GPS inertial and stellar For the stellar navigation a basic set of bright stars is used based on the 5th revised edition of the Yale Bright Star Catalog 1991 The star tracking algorithm is a high accuracy proprietary method able to resolve position and time of a craft anywhere in the world down to 100 meters Control Panel The control panel for the simulator is identical to that on the UAV control module as the graphical interface between the flight simulator and the user It displays useful information about simulation state like Ground Speed Altimeter Trimming Battery Coordinates Air Temperature Compass Helicopter Dynamic Model Figure 3 VI 6 shows the general structure of the helicopter model where is the gravitational force J and the remaining external force and moment vect
46. at must be simple using hints and pointing methods for the user Clearly good spatial abilities are also important in reporting because of the need to translate activity from the viewpoint of the sensor to that of personnel on the ground whereas we had developed pretty much the same sort of capabilities but for manufacturing processes In recent years we have extended this capability to customised sensor and imaging exploitation for monitoring surveillance purposes At the moment we are also building a high power pylon multi spectral monitoring system deployed aboard a blimp for the Mexican electricity Board and UAV micro avionics compatible with Raphael GCS to be offered soon The following is a brief description of the toolkit and its current capabilities 3 Basic System Description 3 1 INTRODUCTION The Raphael GCS GUI Graphical User Interface represents the linking element between the user and the Raphael GCS modules The GUI has simple structure that allows working with many different data sources and results while keeping all of them correctly organized The main Raphael GCS window has a look like the one shown below these lines see figure 3 1 1 Raphael GCS File Mission Planning Flight Control Launch Page Documentation Data Processing Data Analysis Final Reports Show Help w OBO 7 Raphes GCS Wartaoace Mocule Li tar ares R Modules Data Q Maps D Settings Description Show Logo at onty Guning start gt Contributions
47. ata from GPS eXchange format GPX Import GStat Shapes Gstat shapes format import Import Shapes from XYZ Point shapes import from text formated XYZ table Import Stereo Lithography File STL Import Surfer Blanking Files Import polygons polylines from Golden Software s Surfer Blanking File format Import WASP terrain map File Point Cloud from File Tools for the import and export of tables Export Text Table Import Text Table Import Text Table with Number only Shapes Tools related to gridded and vector data Add Grid Values to Points Retrives information from the selected grids at the positions of the points of the selected points layer and adds it to the resulting layer Clip Grid with Polygon Clips the input grid with a polygon shapefile Select polygons from the shapeflile prior to module execution in case you like to use only a subset from the shapefile for clipping Contour Lines from Grid Create contour lines isolines from grid values Get Grid Data for Shapes Retrives information from the selected grids at the positions of the shapes of the selected shapes layers and adds it to the resulting shapes layer Gradient from Grid Create lines indicanting the gradient Grid Statistics for Polygons For each polygon statistics of the contained grid values will be generated Grid Values to Points This modules saves grid values to a point shape Optionally only points can be saved which are contained by polygons of t
48. ch in any way contravenes the law of any country in which this contract 1s effective shall to the extent of such contravention of law be deemed separable and shall not affect any other provision or the validity of this contract Prepared by MFG Wednesday February 10 2010 Specifications are subject to changes without prior notice Third party brands and names are the property of their respective owners O Copyright AVNTK S C 2010
49. d a table of 3D oriented points with a list of associated cameras If we merged the two previous algorithms we have a better representation of the 3D scene reconstruction Now we use a Poisson reconstruction algorithm in order to get the final 3D mesh associated with the final reconstructed model 4 1 7 Poisson Surface Reconstruction We follow the work done by Kazhdan et al 2006 We assume the surface reconstruction from oriented points can be cast as a spatial Poisson problem This Poisson formulation considers all the points at once without resorting to heuristic spatial partitioning or blending and is therefore highly resilient to data noise Unlike radial basis function schemes the Poisson approach allows a hierarchy of locally supported basis functions and therefore the solution reduces to a well conditioned sparse linear system Poisson surface reconstructions generate a mesh from a set of surface samples After Poisson reconstruction is performed for the previous stereo matching step we get a refined surface Figure 7 shows the surface obtained from the 3D oriented points given by the merged stereo matches techniques above Figure 4 1 7 1 Poisson surface reconstruction of the previous stereo matched mesh 4 1 8 Textured 3D reconstructed scene The texture is taken from the camera views and combined to form a multi texture surface In order to take into account the appropriate texture of the 3D reconstructed model we follow
50. ds MEGDRs Import SRTM30 DEM Extracts elevation grids from SRTM30 data Import Surfer Grid Import grid from Golden Software s Surface grid format Import USGS SRTM Grid Import grid from USGS SRTM Shuttle Radar Topographic Mission data Image Import and Export Export Image Saves a grid as image using display properties as used by the graphical user interface bmp jpg png Import Image Loads an image in formats bmp jpg png tif tiff pnm xpm Import and Export of vector data Export Atlas Boundary File Export GPX Exports vector data points to GPS eXchange format GPX Export GStat Shapes GSTat shapes format export Export Shapes to Generate Export generate shapes format Export Shapes to XYZ XYZ export filter for shapes Export Surfer Blanking File Export shapes to Golden Software s Surfer Blanking File format Export TIN to Stereo Lithography File STL An StL StereoLithography file is a triangular representation of a 3 dimensional surface geometry The surface is tessellated or broken down logically into a series of small triangles facets Each facet is described by a perpendicular direction and three points representing the vertices corners of the triangle These data are used by a slicing algorithm to determine the cross sections of the 3 dimensional shape to be built by the fabber A fabber short for digital fabricator Export WASP terrain map file Import Atlas Boundary File Import GPX Import GPS d
51. ducts and or services are stated on the face of this contract Payment shall be made unless otherwise stipulated in cash in full no later than thirty 30 days from the date of shipment Payment not made when due shall bear interest at the rate of one and one half percent 1 596 per month from date of shipment until paid in full If buyer defaults in any payment when due or in the event any bankruptcy or insolvency proceedings involving buyer are initiated by or against buyer then the whole contract price shall immediately become due and payable upon demand or seller at its option without prejudice to its other lawful remedies may defer delivery or cancel this contract or any unfulfilled orders CACELLATION AND OR CHANGES If buyer returns a product for credit buyer shall be subject to a twenty percent 20 restocking charge If buyer changes the delivery date to exceed six 6 months from the original date seller may charge the price of the products TAXES AND OTHER CHARGES Buyer shall pay any taxes imposed by any governmental authority in addition to the prices quoted herein unless buyer gives seller a signed exemption certificate or direct pay permit SHIPMENT AND INSURANCE Buyer shall be responsible for all loading freight shipping forwarding and handling charges licenses import duties and taxes applicable to the delivery of the products Seller will hold title to the products and bear risk of loss until buyer receives products DELIV
52. e Change Grid Values Flood Fill Interactive Interactively use the flood fill method to replace a grids cell values If the target is not set the changes will be stored to the original grid Close Gaps Close gaps of a grid data set i e Eliminate no data values If the target is not set the changes will be stored to the original grid Close One Cell Gaps Close one cell gaps using the mean value of the surrounding cell values If the target is not set the changes will be stored to the original grid Combine Grid Combine two given grids Convert Data Storage Type Changes the storage data type of a grid e g From 4 byte floating point to 2 byte signed integer If the target is not set the original grids storage type will be changed Create Constant Grid Create Grid System This module creates a new user Grid System for use with other modules Crop to Data Crop grids to valid data cells Cutting Interactive Create a new grid from interactively selected cut of an input grid Grid Buffer Grid buffer creation Grid Orientation Copy mirror horizontally or vectically and invert grid values If the target is not set the changes will be stored to the original grid Grid Proximity Buffer This module calculates the euclidean distance within a buffer distance from all NoData cells to the nearest valid neighbor in a source grid Additionally the source cells define the zones that will be used in the euclidean allocation calculati
53. e first component From b to f it is shows the descending contribution image to the original image data set Finally the weakest component is give by figure 4 11 16 Figure 4 11 16 It shows the weakest component 4 lll SATELLITE DATA ACCESS We currently have access to data from only two satellites through their instrument aboard MODIS aboard Terra and AIRS aboard Aqua in a format compatible with our system but we hope to provide access to further sources over the coming months This satellite information is read written using the standard files hdf version 4 and 5 The multispectral module is fed through this satellite data access too In this way UAV imaging can be overlaid on satellite data thus providing fine detail combined with the global context of areas under study Satellite information is normally written using hdf format version 4 or 5 Raphael GCS is compatible to read write this kind of metadata for example the following dialog window is from the hdf4 header of AIRS L Ronan n F E E Ok 45x30 sarang LIB AMSU 132 b f aating cov 145x395 tptgeoqga LIB AMSU 32 bt unsigned wtege Cane 145430 zengcogs LIB AMSU 16 bt unsigned tee x 1453 30 Gemoeons LIA AMEL 16 bt unsigned integer 145230 at LIB AMA 32 bx haing pot Load 44290 sata LIN AMS 1232 08 fatine aoid Save 45x 30 soren LID ANSU 132 bit l oabng porrm 45a 30 solazi LIB AMSU 32 bit fiosbeg pormb 45x
54. e pinhole camera model for the captured images and we have calibrated each camera we created an image list with information about image sizes and camera focal length Since images had been captured under different conditions this last data could not always exist however our system may calculate an initial estimation for it or posses a database with a family of imaging systems characterized as used in any particular theater of operations 4 1 3 Detection Features In order to find feature points in each image we use feature matching which is a fundamental step in 3D scene reconstruction We have used the well known SIFT method which is Scale Invariant Feature Transform to detect and describe local features keypoints in images Lowe 2004 A typical image contains several thousand SIFT keypoints However sometimes the number of keypoints falls below the minimum number of keypoints to get a good estimation of initial camera pose Then we have adapted an interface editor to increase or modify the number of keypoints the user can do this by hand Each keypoint is invariant to image scale rotation affine distortion change in 3D viewpoint in addition of noise and changes in illumination The SIFT method ensures that a single feature can be correctly matched with high probability against a large database of feature from many images The following are the major stages of computation used to generate the set of keypoints 1 Scale space extreme de
55. elated parameters These algorithms trace the flow of each cell in a DEM separately until it finally leaves the DEM or ends in a sink Available flow routing methods are based on Rho8 Kinematic Routingi Algorithm and DEMON Isochrones Constant Speed Interactive Isochrones calculation with constant speed Isochrones Variable Speed Interactive C lculo del tiempo de salida con velocidad variable Lake Flood This module can be used to simulate the extent and volume of a lake for a specified water depth in a seed cell Lake Flood Interactive This module works interactively and can be used to simulate the extent and volume of a lake for a given water depth Execute the module and use the action tool on a cell of the digital elevation model to fill the lake starting from this location Execute the module again to terminate module operation Parallel Processing Parallel processing of cells for calculation of flow accumulation and related parameters This set of algorithms processes a DEM downwards from the highest to the lowest cell Recursive Upward Processing Recursive upward processing of cells for calculation of flow accumulation and related parameters This set of algorithms processes recursively all upwards connected cells until each cell of the DEM has been processed Available Recursive Upward Processing methods are based on Deterministic 8 Rho 8 Deterministic Infinity Multiple Flow Direction Raphael GCS Wetness Index The Raphael
56. en identified orientation and attitude can be inferred based on which stars are in view and how these stars are arranged in the star image Welcome to Figure 3 V 4 It shows the interface for stellar navigation avai lable on Raphael GCS fps 789 Cursor 0 00 W 0 005 Raphael GCS has several algorithms for star Star location data is available from the 5th revised edition of the Yale Bright Star Catalog 1991 in celestial coordinates which are a kind of polar coordinates Setting a maximum magnitude minimum brightness of 5 5 gives 2887 stars while 5 8 gives 4103 stars As these magnitudes are barely visible under the best of conditions one can see that a relatively small number of stars is sufficient to draw an accurate sky i Ble Job See Wen tripie eran Beth rep ha 7 E Orman Not a gouri Mat on gnane Figure 3 V 5 A sample night sky available on Raphael GCS Autopilot helicopters blimps fixed wings The autopilot was tested on fixed wing aircraft and blimps but undoubtedly the most demanding application is that of a helicopter Raphael GCS has implemented a sophisticated 80 parameter helicopter simulator We have modelled this non lineal dynamic system using advanced routines that calculate in detail every component of our own flight simulator In the following sections we describe the dynamics of the system using conventional six degree of freedom rigid body model driven by forces and moments th
57. es statistical properties arithmetic mean minimum maximum variance standard deviation for each cell position for the values of the selected grids Zonal grid statistics The module can be used to create a contigency table of unique condition units UCUs These units are delineated from a zonal grid eg Sub catchments and optional catergorial grids eg Lancover soil It is posible to calculate simple statistics min max mean standard deviation and sum for each UCU from optional grids with continuos data eg Slope The number of input grids is only limited by available memory The module has four different modes of applications 1 only a zonal grid is used as input This results is a simple contingency table with the number of grid cells in each zone 2 a zonal grid and additional categorial grids are used as input This results in a contingency table with the number of cells in each UCU 3 a zonal grid and additional grids with continuous data are used as input This results in a contingency table with the number of cells in each zone and some simple statistics for each zone The statistics are calculated for each continuous grid 4 a zonal grid additional categorial grids and additional grids with continuous data are used as input This results in a contingency table with the number of cells in each UCU and the corresponding statistics for each continuous grid Discretization This is a tool for the discretization classificati
58. escue agriculture planning disaster relief units etc often require more sophisticated reports In the following section we present advanced modules already implemented on Raphael GCS Simulation Cellular Automata Conway s Life It would be incomplete to explain without an example The history of cellular automata dates back to the forties with Stanislas Ulam This mathematician was interested in the evolution of graphic constructions generated by simple rules The base of his construction was a two dimensional space divided into cells a sort of grid Each of these cells could have two states ON or OFF Starting from a given pattern the following generation was determined according to neighbourhood rules For example if a cell was in contact with two ON cells it would switch on too otherwise it would switch off Ulam who used one of the first computers quickly noticed that this mechanism permitted to generate complex and graceful figures and that these figures could in some cases self reproduce Extremely simple rules permitted to build very complex patterns There are three fundamental properties of cellular automata e Parallelism A system is said to be parallel when its constituents evolve simultaneously and independently In that case cells update are performed independently of each other e Locality The new state of a cell only depends on its actual state and on the neighbourhood e Homogeneity The laws are universa
59. f Takeoff Takeoff M TT r Ms Distance X ft Height ft 15 4 39 35 Time s 19 15 n 15 39 Time 5 a b Takeoff Takeoff 100 t 1 i Velocity X ft s Velocity Z Ris 0 3 10 13 s 1 15 2 30 Time 5 Time 5 c d Figure 3 V 3 It shows the normal takeoff of a DC9 aircraft Normal Take off Summary Name Value Unit Rotation Velocity 219 912 ft s Lift off velocity 242 080 ft s Velocity over obstacle 256 451 ft s Rotation distance 2862 147 ft Lift off distance 3555 39 ft Distance to obstacle 4260 314 ft Rotation time 24 036 S Lift off time 27 936 S Time to obstacle 30 759 S OEI Take Off Summary Name Value Unit Critical Velocity 203 886 ft s Decision Velocity 212 381 ft s Velocity over obstacle 233 520 ft s Critical Distance 2419 709 ft Decision Distance 3044 195 ft Balanced Field Length 5308 148 ft Critical Time 22 849 S Decision Time 25 849 S OEI Takeoff Time 36 300 S Stellar navigation The on board camera can be used as a star sensor for star image acquisition system The algorithm employed is to determine star identification identification a correlation of known star position from an empirical star catalog with the unindentified stars in the star image b Pyramid algorithm c triangulation algorithm Once the stars in an image have be
60. f grids e B Spline Approximation Calculates B spline functions for choosen level of detail This module serves as basis for the Multilevel B Spline Interpolation and is not suited as it is for spatial data interpolation from scattered data e Multilevel B Spline Interpolation Multilevel B spline algorithm for spatial interpolation of scattered data The algorithm makes use of a coarse to fine hierarchy of control lattices to generate a sequence of bicubic B spline functions whose sum approaches the desired interpolation function Large performance gains are realized by B spline refinament to reduce the sum of these functions into one equivalent B spline function The maximum level determines the maximun size of the final B spline matrix and increases exponential with each level Where level 10 requieres about 1MB level 12 needs about 16MB and level 14 about 256MB of additonal memory e Multilevel B Spline Interpolation from grid e Thin Plate Spline Global e Thin Plate Spline Local e Thin Plate Spline TIN Geostatistics Kriging is a complex geostatistical technique used to create regular grids from irregularly spaced data points There are several Variogram Model to fit the data such as Spherical Exponential Gaussian Linear Regression Exponential Regression Power Function Regression Ordinary Kriging Ordinary Krining for grid interpolation from irregular samples points Ordinary Kriging Global This implementa
61. for a set of geometrically consistent matches between each image pair in an incremental scheme to reconstruct an initial sparse 3D point cloud and camera pose Finally we remove matches that are outliers to the epipolar geometry constraint n addition we recover an initial set of camera parameters and 3D scene reconstructed point cloud In order to get refinement for new camera poses and 3D scene point cloud we take the back projection error as the cost function to minimize it This is a much more computationally intensive approach than previously reported but it is robust and able to cope with real life images 4 1 5 Bundle Adjustment Once the last frame is added a regularization method for final camera poses is calculated by a generic Bundle Adjustment method based on the sparse Levenberg Marquardt algorithm given by Lourakis and Argyros 2004 The Bundle Adjustment is calculated till back projection errors convergence The initial 2D to 3D correspondence is refined by the Bundle Adjustment scheme and it is used to compute the final camera pose The 3D point cloud scene reconstruction is carried out using all 2D camera views and 3D camera specific information A typical output is as shown in figure 4 1 5 1 c d Figure 4 1 5 1 Shows output from a typical Bundle Adjustment run Figures a b c and d show 3D point cloud reconstruction in blue dots and camera pose in red line We have included 10 camera poses only for illustratio
62. hat yields a stable stereo match The optimization stage for surface normals is within a photometry consistency calculation what significantly improves the matching results We include reflectance coefficients between pair images in order to model changes in illumination After we have run this stage we have oriented points around SIFT matched points within a list of associated cameras each potentially with different characteristics ii Patch based Multi view Stereo PMVS Furukawa and Ponce 2007 2009 The patch based multi view stereo matching is complementary to the above stereo matching technique i e we obtain additional oriented points where the other method fails PMVS is implemented as a match expand and filter procedure starting from a sparse set of SIFT matched points and repeatedly expanding these to nearby pixel correspondences before using visibility constraints to filter away false matches 1 Matching Features found by Harris and Difference of Gaussians operators are matched across multiple images yielding a sparse set of oriented patches associated with salient image regions Given these initial matches the following two steps are repeated 3 times 2 Expansion The technique is used to spread the initial matches to nearby pixels and obtain a dense set of oriented patches 3 Filtering Visibility constraints are used to eliminate incorrect matches lying either in front or behind the observed surface The two methods yiel
63. he specified shapes layer In addition it is possible to exclude all cells that are coded NoData in the first grid to the grid list Grid Values to Points randomly Extract randomly points from gridded data Local Minima and Maxima Extracts local grid value minima and maxima of to vector points Vectorising Grid Classes Vectorising grid classes Shapes Tools for lines Convert Points to Line Convert Polygons to Lines Converts polygons shapes to lines shapes Line Properties Line properties such as length number of vertices Shapes Tools for the manipulation of point vector data Add Coordinates to Points Clip Points with Polygons Convert a Table to Points Create Point Theme From Table Count Points in Polygons Create Point Grid Creates a regular grid of points Distance Matrix Computes distance between pairs of points in a layer Fit N Points to Shape Points from Lines Converts a line theme to a points theme Optionally inserts additional points in user defined distances Remove Duplicate Points Separates Points by Direction Separate points by direction Direction is determined as average direction of three consecutive points A B and C If the angle between the directions of A B and B C is higher than given tolerance angle the point is dropped This modules has been designe to separate GPS tracks recorded by tractors while preparing a field Shapes Tools for Polygons Convert Lines to Polygons Converts
64. hout Figure 3 VII 3 a Focused image on first plane b focused image on second plane and c increased depth of field image Blurring Another common problem with UAV derived imaging is that due to vibration making images blurred very often So an advanced automatic blurring correction tool has been implemented ee oe ope Figure 3 VII 4 Automatic de blurring 3 VIII RAPHAEL GEOGRAPHICAL INFORMATION SYSTEM A problem often encountered by users is that the quality of unmanned vehicle images is simply not of sufficient quality to analyze the data in a GIS Moreover the interface between UAV derived images and a GIS is extremely difficult due to the image quality and referencing requirements of a GIS The image quality restoration available in Raphael GCS makes it possible to combine UAV aerial imagery with existing data such as geological topological land use rain fall statistics etc The GIS module currently has the following basic functions Analysis of both vector and raster images User friendly Import export of popular data formats On screen and tablet digitizing Comprehensive set of image processing tools Ortho photo Cartography Image geo referencing and registration Transformation rectification and mosaicing Advanced modeling and spatial data analysis Rich projection and coordinate system library Geo statistical analysis with Kriging for improved interpolation Production and visualization
65. hydrographs that characterise the quick and slow components of streamflow This feature allows hydrographs to be separated into their dominant quick and slow flow components and provides a Slow Flow Index SFI analogous to the well known Baseflow Index BFI 6 parameters or DRCs quick flow response decay time C volume of a conceptual catchment wetness storage mm r slow flow response decay time r catchment drying time _ _ _ ee SFI i analogous to BFI v8 proportional volumetric contribution of slow flow to streamflow f temperature modulation factor C1 Figure 3 1X 10 Description of 6 parameters input output of IHACRES module Simulation Modelling the Human Impact on Nature A simple litter system It is a model using the euler method Carbon storage C is calculated in dependency of litter fall rate C_input and rate constant for litter loss C_loss as C Bc 1C EID Oppa C jogs C Pt Typical values Model Parameter Value Unit Tropical Rainforest Litter fall rate 500 g m a Litter loss rate 2 0 l a Temperate forest Litter fall rate 240 g m a Litter loss rate 0 4 l a Boreal forest Litter fall rate 50 g m a Litter loss rate 0 05 I a Table 3 1X 8 Typical parameter for several litter system model TIME 584 248 567 064 549 881 532 697 515
66. ich the aircraft can operate The stall velocity is the aerodynamically limited velocity at which the aircraft can produce enough lift to balance the aircraft weight This velocity occurs at the maximum aircraft lift coefficient Cie and is defined as The minimum control velocity V uc is the lowest airspeed at which it has proved to be possible to recover control of the airplane after engine failure This is an aerodynamic limit which is difficult to predict during the preliminary design stage but may be obtained from wind tunnel data during later phases of design The minimum unstick velocity V mu is the airspeed at and above which it can be demonstrated by means of flight tests that the aircraft can safely leave the ground and continue takeoff This velocity is usually very close to the stall velocity of the aircraft With these reference velocities defined the FAR places the following broad requirements on the velocities of takeoff V gt 1 05 Vyc Vio 2 1 1 Vy Because the minimum unstick velocity is usually very close to the stall velocity the liftoff velocity is often referenced as greater than 1 1 times the stall velocity rather than 1 1 times the minimum unstick velocity Engine failure brings another level of complexity to the definitions and requirements of takeoff Usually takeoff of this nature is categorized by the failure of one engine or one engine inoperative OEI OEI takeoff includes an AEO ground run to engi
67. iew of the video information presented in each frame To reduce this effect a portion of video can be viewed as a mosaic of images This birds eye view increases spatial comprehension of the data allowing more accurate analysis of the information This module is currently being added and will include super resolution increased depth of field blurring correction as required b Figure 3 VI 4 Images a and b show the first and 180th frames of the Predator F sequence The vehicle near the center moves as the camera pans across the scene in the same general direction Poor contrast is evident in the top right of a and in most of b The use of basis functions for computing optical flow pools together information across large areas of the sequence thereby mitigating the effect of poor contrast Likewise the iterative process for obtaining model parameters successfully eliminates outliers caused by the moving vehicle The mosaic constructed from this sequence is shown in c 3 VII IMAGE RESTORATION Our sophisticated photogrammetry module allows us to make a number of improvements to restore images unavailable in other products such as automatic blurring removal and another common problem with UAV derived imaging that we often have 5 10 low quality images of a target and require a single high resolution image So an advanced sub pixel super resolution resolution algorithm using wavelets was implemented Super resolution Another c
68. ion since the new images can be seen from different points of view and different illumination conditions even using distinct imaging devices This fact allows us to write down an extra module that aligns independent 3D reconstructions given by distinct mission tracks and it gets better detailed 3D reconstruction with appropriate texture with reduced illumination effects In order to overlap two 3D oriented point sets we have used a modified version of the Iterative Closest Point ICP algorithm developed by Besl and McKay 1992 Chen and Medioni 1992 and Zhang 1994 A classical ICP is based on the search for pairs of nearest points in two data sets and estimates the rigid body transformation that aligns them The rigid body transformation is then applied to the points of one set and the procedure is iterated until convergence is achieved This point to point scheme has low convergence Thus we have used a modified version of the ICP the point to tangent plane approach originally proposed by Potmesil 1983 The search for pairs of nearest points in two data sets is done using a kd trees structure Thus the Approximate Nearest Neighbors package of Arya et al 1998 is used extensively is this process On the other hand the rigid body transformation is given by a translation vector T x y z and a matrix of rotation R o p Kk In order to find the correct rigid body transformation we have used a Levenberg Marquardt algorithm as the least
69. l thats to say common to the whole space of cellular automata The rules are quite simple One inactive cell surrounded by three active cells becomes active its born One active cell surrounded by 2 or 3 active cells remains active In any other case the cell dies or remains inactive UJ NN G amp a P Le w ET ES de A J E X P Wow L Figure 3 IX 1 Simulation of Conway s Life cellular automata Wa Tor An ecological simulation of predator prey populations It simulates the hypothetical toroidal Planet Wa Tor Water Torus whose surface is completely covered with water occupied by two species fish and sharks The sharks are the predators They eat the fish The fish exist on a never ending supply of plankton Both sharks and fish live according to a strict set of rules This simulation of a simple ecology is highly dynamic as both species are walking a thin line between continuing life and extinction Rules In general the simulation is based on discrete time steps The simulation runs on a rectangular grid To represent the toroidal world opposing sides of the grid are connected If an individual moves out on one side of the simulation domain it reenters immediately on the opposing side Fish and shark move every time step if possible and interact according to the following set of rules Rules for fish In every time step a fish moves randomly to one of the four neighboring fields provided it is empty E
70. l and pitch rates However unlike the main rotor the flybar is not responsible for providing lift or maneuvering ability Thus it can be designed to have a slower response and provide the desired stabilization effect The notation used to describe the Bell Hiller system is presented in figure 3 VI 8 where the mechanical arrangement is reproduced Gravitation Model The six degree of freedom rigid body dynamic equations could be solved under two cases 1 Flat earth approximation and 2 WGS 84 earth model In the flat earth approximation no information is needed on the home latitude and longitude In this case it is only needed the starting position On the other hand WGS 84 model needs latitude longitude and altitude which ECEF is generated from this see table 3 VI 2 Genenc starting position BI Et E EY Starting body position ft uvwlu v w Starting body translational velocity ft s pqr P Q R Starting body angular rates rad s THETA o 0 w Starting body attitude rad For the WGS 84 model Latitude Starting latitude rad Longitude Starting longitude rad Altitude Starting altitude ft Table 3 VI 2 The starting position for both modes 1 Flat earth approximation and 2 WGS84 earth model In both cases the six degree of freedom rigid body equations need to be initialized with the desired starting values The gravity and gravitational model it is based on the WGS 84 ellipsoid model of the Earth see table
71. lected elevation values Flow Accumulation Trace Calculates the catchment area based on the selected elevation values Gradient Calculates the gradient based on the values of each triangle s points Grid to TIN Creates a TIN from grid points NoData values will be ignored Grid to TIN Surface Specific Points Creates a TIN by identifying surface specific points of a grid Shapes to TIN Convert a shapes layer to a TIN TIN to Shapes Convert a TIN data set to shapes layers Point Cloud Tools for points clouds Point Cloud Cutter Point Cloud Cutter iteractive Point Cloud from Grid Points Point Cloud from Shapes Point Cloud to Grid Point Cloud to Shapes Point Cloud Point Cloud Viewer Point Cloud Viewer Projection Tools for coordinate transformation based on PROJ 4 Projections Once data have been imported the next necessary step is in most cases to georeference or to project it so that all spatial data sets of a project belong to one single coordinate system Coordinate transformation of Raphael GCS is based on two alternative cartographic projection libraries the GeoTrans library developed by the National Imagery and Mapping Agency and the Proj 4 library initiated by the U S Geological Survey Both libraries work for raster and vector data and provide various projections for free definable cartographic parameters Proj 4 Grid Coordinate Transformation for Grids Proj 4 Dialog Grid Proj 4 List
72. lines shapes to polygon shapes Line arcs are closed to polygons simple by connecting the last point with the first Convert Polygon Line Vertices to Points Geometrical Properties of Polygons Polygon Centroids Creates a points layer containing the centroids of the input polygon layer Polygon Intersection Intersection of polygon layers Uses General Polygon Clipper version 2 3 Polygon Shape Indices Various indices describing the shapes of polygons Based on area perimeter maximum distance between the vertices of a polygon Polygon Union The union of polygons which have the same attribute value Polygon Statistics from Points Calculates statistics using a polygon and a points layer Shapes Tools for the manipulation of vector data Create Chart Layer bars sectors Create Empty Shapes Layer Creates a new empty shapes layer of given type which might be either point multipoint line or polygon Available field types for the attributes table are string 1 2 and 4 byte inter 4 and 8 floating point Create PDF Report for Shapes Layer Create Web Content Interactive Create graticule Cut Shapes Layer Cut Shapes Layer Interactive Get Shapes Extents Join a Table Joins a table with shapes layer s attributes Merge Shapes Layers New Layer from Selected Shapes QuadTree Structure to Shapes Query Builder for Shapes Search in Attributes Tables Searches for an expression in the attributes table and selects reco
73. ll require an aborted takeoff while engine failure after the critical velocity has been reached will require a continued takeoff With the above definitions in place FAR 25 imposes additional requirements for OEI takeoff V gt V r LI Vio gt 1 05 Vy U Note that although other standards for aircraft takeoff exist most use the same four velocities in their takeoff analysis VoVo ine YD Velocity 27 713 SO a re Vlol Vi Meus 1 One Engine Out All Engine Takeoff OEI Takeoff AEO Accelerate Stop ae Xcrit X Xlo Xobs Ground Run Climb Distance Figure 3 V 2 A normal takeoff involves a definition of each velocity and distances Raphael GCS has implemented the simplified method proposed by Powers and a modified version of a method proposed by Krenkel and Salzman The simplified Powers method requires 13 input parameters and solves the governing equation analytically for takeoff times and distances The method assumes constant thrust throughout the takeoff run and climb phase aerodynamics are the same as in the eround roll The two major problems with the Powers methodology are the lack of a rotation phase and the use of ground roll equations of motion to predict the climb phase of takeoff The lack of a rotation phase causes the method
74. lling the Human Impact on Nature Terrain Analysis Channels Terrain Analysis Hydrology Terrain Analysis Lighting Visibility Terrain Analysis Morphometry Terrain Analysis Preprocessing Terrain Analysis Profiles 3D Visualisation 44 46 47 49 51 53 54 55 56 4 eoi EN AM 9 Optional Modules 4 3D Reconstruction from Full Moving Video FMV 4 1 2 Camera Model and calibration 4 1 3 Detection Features 4 1 4 SIFT matching between pair images 4 1 5 Bundle Adjustment 4 1 6 3D Scene Surface Reconstruction 4 1 7 Poisson Surface Reconstruction 4 1 8 Textured 3D reconstructed scene 4 1 9 Alignment between independent mission tracks Comparison between CAD model and 3D reconstructed scene CAD Data vs Multiple Images 4 Multispectral Analysis Multivariate Methods 4 1 Satellite Data Access Standard Raphael Ground Control Station Pricing details Collaboration Structure Appendices 8 Detail of GIS Functions Elementary grid operations Interpolation of grids Geostatistics Discretization Gridding Advanced grid operations I Advanced grid operations I Import and Export between raster vector data formats Geospatial Data Abstraction Library Import Export Grids Image Import and Export Import and Export of vector data Tools for the import and export of tables Shapes Tools related to gridded and vector data Shapes Tools for lines Sh
75. matching feature If the ratio of closest distance to 2 closest distance greater than 0 8 then reject as a false match The figure 4 1 4 1 shows a typical example of SIFT matching between two arbitrary chosen images A cyan line marks the correspondence between two keypoints However we have included epipolar geometry constraints in order to do a best matching between keypoints as describe below iua n mauu i Prepir wf ie ee ay mg Figure 4 1 4 1 SIFT matching between image 1 and image 2 The cyan line marks the same point seen on different frames Once initial matching is done using a local description vector for each keypoint then robustly we estimate the fundamental matrix F for each pair image using the RANSAC method Fischler and Bolles 1987 As is well known in computer vision the fundamental matrix F is a 3 x 3 matrix which relates 2D corresponding points in stereo images The fundamental matrix relates 2D coordinate images X Fx 0 between two stereo images Furthermore if we know image camera parameters we have T the so called essential matrix E K FK that relates to rotation and translation between cameras E Rt x During each RANSAC iteration we compute a candidate fundamental matrix F using the eight point algorithm Hartley and Zisserman 2004 followed by a non linear refinement The epipolar constraint is applied intensively around the SIFT matching to ensure better feature matches We look
76. models i e not only to fill the depression s but also to preserve a downward slope along the flow path If desired this is accomplished by preserving a minimum slope gradient and thus elevation difference between cells This version of the module is designed to work on large data sets e g LIDAR data with smaller datasets you might like to check out the fully featured standard version of the module Sink Drainage Route Detection This module implement the sink drainage route detection algorithm For non sinks obtain a value of 0 sinks are assigned an integer between 0 and 8 indicating the direction to which flow from this sink should be routed 0 north 1 northeast 2 east 7 northwest ASTGTM N1TW194 2 imi a b Figure 3 IX 21 Sink Drainage Route Detection a DEM image b Sink Route e Sink Removal Remove sinks from a digital elevation model by deepening drainage routes or filling sinks Terrain Analysis Profiles e Cross Profiles Create cross profiles from a grid based DEM for given lines e Cross Sections Creates cross section from a grid based DEM for given lines e Flow Path Profile Interactive Create interactively flow path profiles from a grid based DEM Use a left mouse button click to create a flow profile starting from the clicked point e Profile Interactive Create interactively profiles from a grid based DEM Use left mouse button clicks into a map window to add profile points A right
77. mouse button click will finish the profile e Profile from Points Creates a profile from point coordinates stored in a table e Profiles from Lines Create profiles from a grid based DEM for each line of a lines layer e Swath Profile Interactive Create interactively swath profiles from a grid based DEM Use left mouse button clicks into a map window to add profile points A right mouse button click will finish the profile Generated outputs for the swath profile are arithmetic mean minimum maximum values and the standard deviation 3D Visualisation All the previous images and indeed any DEM data can also be displayed in 3D as shown below for a few previously displayed cases CETTE CET ENT eli Figure 3 1X 22 It shows catchment amp height esp GES SO Figure 3 1X 23 Total contribution of solar radiation spread over the terrain 4 Optional Modules 4 3D RECONSTRUCTION FROM FULL MOVING VIDEO FMV This is an automatic algorithm capable of extracting a 3D target surface model from FMV It automatically calculates relative camera position at each frame and tracks multiple features in order to generate a points cloud from which a 3D surface is formed Multiple video feeds can be used to carry out the reconstruction of 3D target data Occlusion Obscuration is also contemplated and be accommodated in cases of partial obscuration as these lead to a lack of correspondence of features being tracked Although we ha
78. mplement due to the maturity of the technology The range of GPRS is large as it works over the cellular network so will work anywhere there is a GSM signal most urban areas have very good signal quality The data rate of the GPRS class 10 modems is 16 24kbps upload and 32 48kbps download In theory this will be large enough to send a decent quality video stream GPRS offers the ideal balance of range cost and availability WiFi is a good choice as proof of concept as it is relatively cheap to implement and has a wide enough range for testing purposes The data rate for the WiFi transceiver is 11Mbps which is large enough for good quality video transmission For proof of concept this should be used before GPRS is implemented as it has a wide enough range for testing WiFi is more suited to close range applications and is therefore easier to demonstrate at a Trade Fair for example Figure 3 IV 1 Some of the platforms that Raphael GCS has been configured to control The control panel includes all the typical functions found in standard GCS platforms It is split into two screens a flight control screen and a visual onboard camera driven screen Rseese ott 80 Vire a con E Z We TU S202 Figure 3 1V 2 Raphael GCS UAV control panel and onboard camera view with all optional displays on 3 V SIMULATOR amp AUTOPILOT Automatic Takeoff and Landing Raphael GCS has implemented a strategy for an autonomous both take
79. n purposes See the full video on http www thegoodwillcompany co uk 4 1 6 3D Scene Surface Reconstruction So far we have constructed a sparse 3D point cloud of the viewed scene and to simultaneously recover camera poses as shown in figure 6 n order to get a dense matching between multiple view images we have combined two methods for this purpose In order to get dense 3D oriented points we combine two multi view stereo matching methods Iteratively grown surface from matched points Goesele et al 2007 and Patch based Multi view Stereo Furukawa and Ponde 2007 2009 After the dense matching is done we solve for the iso surface solution in a Poisson surface reconstruction algorithm as given by Kazhdan et al 2006 Multi view stereo matching i Iteratively grows surface from matched points Goesele et al 2007 This stereo matching technique takes as input sparse 3D points reconstructed from bundle adjustment method and iteratively grows surfaces from these points The main algorithm looks at both depth and normal vector starting from an initial SIFT matched points or copied from previously computed neighbors In order to find optimal surfaces around SIFT matches we apply a two level view selection camera i at the image level global view selection identifies for each reference view a set of good neighborhood images to use for stereo matching ii at the pixel level local view selection determines a subset of these images t
80. n of TOPMODEL Based on the TOPMODEL demonstration program v95 02 by Keith Beven Centre for Research on Environmental Systems and Statistics Institute of Environmental and Biological Sciences Lancaster University Lancaster LA1 4YQ UK and the C translation of the Fortran source codes implemented in GRASS This program allows single or multiple subcatchment calculations but with single average rainfall and potential evapotranspiration inputs to the whole catchment Subcatchment discharges are routed to the catchment outlet using a linear routing algorithm with constant main channel velocity and internal a subcatchment routing velocity The program requires In ET distributions for each subcatchment These may be calculated using the GRIDATB program which requires raster elevation data as input It is recommended that those data should be 50 m resolution or better NOTE TOPMODEL is not intended to be a traditional model package but is more a collection of concepts that can be used where appropriate It is up to the user to verify that the assumptions are appropriate see discussion in Beven et al 1994 This version of the model will be best suited to catchments with shallow soils and moderate topography which do not suffer from excessively long dry periods ldeally predicted contributing areas should be checked against what actually happens in the catchment It includes infiltration excess calculations and parameters based on
81. ncluding modules such as orthorectification georeferencing image fusion image mosaicing 1S019 115 compliant metadata editing satellite access 3D visualization of multiple layers multifrequency analysis terrain analysis channels hydrology lighting visibility morphometry preprocessing profiles simulation fire spreading analysis fire risk analysis hydrology analysis Human impact on nature computer vision algorithm multispectral data toolkit spectral analysis PCA FastICA and multidimensional display and more This software was first shown at Farnborough 2008 and has been recently acquired by Mexican companies and a British company The Goodwill Company Ltd for defense applications and is being offered to UAV manufacturers worldwide This station has been successfully employed to control fixed wing blimps and rotary wing unmanned vehicles in applications such as power line monitoring surveillance and marine SAR Current work includes enhancements such as the ability to analyze multi spectral images 3D reconstruction derived from real time video and persistent wide are surveillance Our system has been developed from modules originally produced for industrial applications where throughput is high the operator often undergoes fatigue repetitive tasks and typically has a relatively low degree of specialization so that interfaces need to be user friendly This is somewhat similar to soldier interfaces of advanced technology th
82. ndom variables measurements or signals ICA defines a generative model for the observed multivariate data which is typically given as a large database of samples In the model the data variables are assumed to be linear mixtures of some unknown latent variables and the mixing system is also unknown The latent variables are assumed nongaussian and mutually independent and they are called the independent components of the observed data These independent components also called sources or factors can be found by ICA ICA is superficially related to principal component analysis and factor analysis ICA is a much more powerful technique however capable of finding the underlying factors or sources when these classic methods fail completely The data analyzed by ICA could originate from many different kinds of application fields including digital images document databases economic indicators and psychometric measurements In many cases the measurements are given as a set of parallel signals or time series the term blind source separation is used to characterize this problem Typical examples are mixtures of simultaneous speech signals that have been picked up by several microphones brain waves recorded by multiple sensors interfering radio signals arriving at a mobile phone parallel time series obtained from some industrial process or multispectral satellite data image Example Bling Source Serparation BSS Problem Given two linear mi
83. ne failure an OEI ground run to liftoff and a climb to 35 ft also with OEI as illustrated in figure X A takeoff with an engine out will take a longer distance than AEO takeoff due to the lower acceleration produced by the remaining engines The obvious questions to ask are if the OEI takeoff field length required is longer than the field length available and if the distance to brake to a stop after engine failure is longer than the available field length These questions are often answered by solving for a critical or balanced field length CFL or BFL the distance at which OEI takeoff distance equals the distance needed to brake to a full stop after engine failure Defining the CFL leads back to the time or more specifically the velocity at which engine failure occurs As it turns out by imposing the CFL definition there is an engine failure velocity which uniquely defines the critical field length This velocity is often called the critical velocity A Mt must be noted that during an aborted takeoff some amount of time will be required after the engine fails for the pilot to actually begin braking both because of the pilot s reaction time and the mechanics of the aircraft During this passage of time the aircraft continues to accelerate on the remaining engines and will finally reach the decision velocity Vi Careful inspection of the above definitions will show that engine failure at a velocity lower than the critical velocity wi
84. notes the depth ratio The matrix Gt aR in 8 is a full rank homogeneous colineation matrix defined up to a scale factor and contains the motion parameters Pl and R between the frames B and Pa Given pairs of image correspondences Pi EL Pi t Ufor four feature points O at least three of which are non collinear the set of linear equation in 8 can be solved to compute a unique G t up to a scale factor When more than four feature point correspondences are available G f can also be recovered again up to a scale factor using techniques such as least squares minimization G It matrix can then be used to uniquely determine H it taking into account its known structure to eliminate the scale factor and the fact that the intrinsic camera calibration matrix is assumed to be known By utilizing epipolar geometry among many others methods H fl can be decomposed to I recover the rotational component and the scaled translational component 3 pi In summary Raphael GCS has two modules about automatic Takeoff and Landing of UAV Takeoff This module takes the craft from the current position to the First Waypoint Before Autopilot FWBA position from which the autopilot takes over This point needs to be specified in the Mission Planning module Landing This module receives craft control from the autopilot and takes the craft f
85. nt routing velocity P VR 3600 m h Surface hydraulic conductivity P KO 1 m h Wetting front suction P_PSI 0 02 m Water content change P DTHETA 0 1 across the wetting front Green Ampt Infiltration BINF TRUE boolean Table 3 1X 6 Initial parameter values Output Name Type Label Description Simulation Output Table TABLE Soil Moisture Deficit Grid MOIST Table 3 IX 7 TOPMODEL Output data Water Retention Capacity It is another advanced hydrology module implemented on Raphael GCS Simulation Identification of unit hydrographs and component flows from rainfall evaporation and streamflow data IHACRES This is a catchment scale rainfall stream flow modeling methodology developed collaboratively by the Institute of Hydrology IH United Kingdom and the Centre for Resource and Environmental Studies Australian National University CRES at ANU Canberra Its purpose is to assist the hydrologist or water resources engineer to characterize the dynamic relationship between basin rainfall and stream flow Possible applications include e Identification of unit hydrographs e Continuous time series stream flow modeling e Environmental change hydrological regime studies Runoff event modelling Hydrograph separation for example to assist with water quality investigations Derivation of a Slow Flow Index SFI Derivation of Dynamic Response Characteristics DRCs Investigation of relationships be
86. number of channels frequency or wavelength instrument s name longitude latitude of each pixel integration time gain factors etc Raphael GCS is compatible to read write this kind of metadata written on hdf Figure 4 11 1 shows few metadata about satellite multispectral maps rsen a The whole GIS operations defined in section 3 ll can be apply on multispectral metadata Particularly algebraic operations user defined on map space as shown in figure 4 1l 3 bremen Tung amp LE e i Figure 4 11 3 It is shows channel combination tool which is enable to compute algebraic operations on the multispectral maps For example algebraic operation like Band l Band 2 or Band 5 Band can be performed The result is shown in figure 4 11 4 bet nent TX HX TIL un pt Oih un hee Uitte love 1 Depew Made LIB AZISILCU 25 21408 jet Leto love 2 Depew Moussa LOOT 2 529309 Figure 4 11 4 Two linear combination maps from AIRS data bands Left Band 1 Band 2 and Right Band 3 Band 1 Band Band 2 Band 3 Band However we can design more complex calculations such as Rd Additionally we can plot scatter data using the scatter section It is useful when we have see if there is correlate uncorrelate information between multiple channels Furthermore we can use principal component analysis to extract information at high dimensional level see section below The figure 4 11 5 shows an example
87. of scatter data and RGB combine task 1 12 2 41 Figure 4 11 5 It is shows scatter plot of two satellite maps left It is shows a RGB combination of selected bands right Remember that we can perform the whole GIS operations set as defined in the Appendix However we mention a particular operation over multispectral metadata such as Filter to detect vegetation or Normalized Difference Vegetation Index NDVI Raphael GCS can calculate this kind of coefficients and all its variants The figure 4 1l 6 shows the NDVI index around Guadalajara City Green colour mark the distribution of vegetation while cyan colour corresponds to areas of rock sand or snow Figure 4 11 6 NDVI index around Guadalajara City using Lansat 7 data Multivariate Methods Principal components analysis PCA It is a mathematical procedure that transforms a number of possibly correlated variables int a smaller number of uncorrelated variables called principal components The first principal component accounts for how much contribution in the data as possible and each succeding compoent accounts for as much of the remaining weigth as possible In this way PCA is often used to reduce the number of variables or dimensions in a data set in order to simplify analysis or aid in visualization PCA is for compressing a set of high dimensional vectors into a set of lower dimensional vectors and then reconstructing the original set It is non parametric method
88. of stereo image pairs Spatial multiple criteria evaluation In principle data in the GIS module is organized according to its nature five different types of data objects can be addressed tables shapes vector data TIN Triangular Irregular Network Point Cloud and grid raster data All data object classes derive from the same base class which defines basic properties name description data history file path and methods load save copy clear Each of the derived classes has specific methods for data access and manipulation Tables Table objects are collections of data records which hold the data itself in data fields Data fields can be of different type e g character string or numerical value Tables are a very powerful tool for displaying data Tables are important because they constitute a linking element between Raphael GCS and other application such as spreadsheets Not all Raphael GCS modules or functions generate new grids as result Some of them return tabular data instead of new grids and you should know how to handle this kind of result Also tables are required sometimes as input File access is supported for text and DBase formats Shapes While grid raster data is contained in the grid itself vector data objects needs to be stored in a database This causes a vector layer to be separated in two different entities the shapes points lines polygons where the spatial information is kept and the database whe
89. off and landing of an Unmanned Aerial Vehicle for both rotor wings and fixed wings configuration Given that autonomous vehicles such as underwater vehicles aircrafts and helicopters are highly non linear dynamic systems The challenges involved in the design of a control strategy for such dynamic system there exits the problem of accurate position measurement in such machines Flying machines are usually equipped with on board inertial sensors which only measure the rate of motion The position is thus obtained from time integration of rate data resulting in potential drift over time due to sensor noise To overcome this problem Raphael GCS uses vision sensors and computer vision algorithms within the feedback loop control system The strategy for the autonomous takeoff landing maneuver is using data taken from the position information system obtained from a single monocular on board camera and inertial measurement unit IMU Thus the proposed vision based control algorithm is build upon homography based techniques and Lyapunov design methods in order to estimate position velocity and attitude of the flying machine during its navigation Without loss of generality the following method can be seen as takeoff phase or landing phase We take landing point of view to describe the methodology of autonomous control Homography determination of position and pose during landing approx The UAV is assumed to be equipped with Inertial Measurements Uni
90. ommon problem with UAV derived imaging is that we often have 5 10 low quality images of a target and require a single high resolution image So an advanced sub pixel resolution algorithm using wavelets was implemented C d Figure 3 VIl 1 Example of our super resolution algorithm a Original image b Super resolution image c Single image zoom and d Super resolution detail a b c d e f Figure 3 VII 2 A demonstration for performing motion super resolution is presented here The Predator B sequence data is gathered from an aerial platform the predator unmanned air vehicle and compressed with loss One frame of this sequence is shown in a Forty images of this sequence are co registered using an affine global motion model upsampled by a factor of 4 combined and sharpened to generate the super resolved image b and d show the car and truck present in the scene at the original resolution while e shows the truck image upsampled by a factor of 4 using a bilinear interpolator The super resolved images of the car and truck are shown in c and f respectively The significant improvement in visual quality is evident Increase Depth of Field Digitaly increased depth of field Where ambient conditions require a low f sometimes only a portion of the image is in focus so we developed a digital means to combine a series of images with a varying focal plane so as to obtain a single in focus image throug
91. on of metric gridded data Cluster analysis for grids An approach for solving the taxonomy problem Grid segmentation Segmentation with the local maximum method Grid segmentation Option B Segmentation with the Local minimu maximum method Grid skeletonization Skeletonisation methods Stadard Hilditch Channel Skeleton Supervised classification Minumum distance Maximum Likelihood Gridding Tools for the gridding of points and other verctor data Inverse Distance Inverse distance to a power method for grid interpolation from irregular distributed points Modified Quadratic Shepard Modified quadratic shepard method for grid interpolation from irregular distributed points Nearest Neighbor Natural neighbor method for grid interpolation from irregular distributed points Shapes to Grid Gridding of a shapes layer If some shapes are selected only these will be gridded Triangulation Gridding od a shapes layer using Delaunay Triangulation Advanced grid operations These modules are for edit delete mix fill overlap create manipulation of gridded data Aggregate For a given i j cell assign max min sum within a given window of radius r Change Grid Values Chabges values of a grid according to the rules of a user defined lookup table Values of value ranges that are not listed in the lookup table remain unchanged If the target is not set the changes will be stored to the orginal grid Change Cell Values Interactiv
92. ons Cell values in the source grid are treated as Ids integer and used in the allocation grid to identify the grid value of the closest source cell If a cell is at an equal distance to two or more sources the cell is assigned to the source that is first encountered in the modules scanning proces The buffer grid is reclassification of the distance grid using a user specified equidistance to create a set of discrete disntace buffers from sources features The buffer zones are coded with the maximum distance grid is floating point The output values for the allocation and buffer grid are of type integer The duration of module execution is dependent on the number of sources cells and the buffer distance Grid Value Request Interactive Gris value Request It is uses a interpolation method Grids from classified grid and table Creates several grids using a classified grid and a table with data values for each class Invert Data No Data Merging Patching Fill gaps of a grid with data from another grid Reclassify Grid Values The module can be used to reclassify the values of a grid It provides three different options a reclassification of single values b reclassification of a range of values and c reclassification of value ranges specified in a lookup table In addition to these methods two special cases NoData values and values not included in the reclassification setup are supported In mode a and b the NoData option is e
93. or use with rasterized databases to simulate surface runoff Computations are performed on a pixel by pixel basis and all physical drainage basin properties including area slope stream length and stream order are obtained or estimated from a digital elevation model DEM Other data sets such as curve numbers or infiltration rates are required for estimating the hydrologic abstractions Precipitation is supplied in the form of gage input uniform distributions or raster data At the present time hydrologic abstractions can be estimated by any of three methods a constant infiltration rate the Soil Conservation Service curve number method or solution of the more physically based Green Ampt equation Overland flow is computed by a kinematic wave approximation and channel routing is performed using the Muskingum Cunge method niom Los p a aa ui ms s O Be an 2 21 font y b Te r i tr LI e ve LA f Figure 3 IX 7 It is shows the Runoff of a DEM Soil Moisture Content The WEELS Wind Erosion on European Light Soils soil moisture model dynamically calculates the soil moisture based on the rules proposed by the DVWK 1996 with input data about e Soil properties Field capacity and permanent wilting point grid type e Land use crop types grid type e Climate Daily values of precipitation temperature and air humidity table type TOPMODEL This Simple Subcatchment Versio
94. ormat Transmission Format RPFTOC Raster Product SDTS SDTS Raster BSB Maptech BSB Nautilus Charts Format toc HFA Erdas Imagine DTED DTED Elevation Raster XPM X11 PixMap Format Images img SAR_CEOS CEOS SAR Image PNG Portable Network Graphics BMP MS Windows Device Independent Bitmap CEOS CEOS Image JPEG JPEG JFIF DIMAP SPOT DIMAP JAXAPALSAR JAXA PALSAR MEM In Memory Raster AirSAR AirSAR Polarimetric Image Product Reader level 1 1 1 5 GFF Ground based SAR JDEM Japanese DEM mem RS2 RadarSat 2 XML Product applications Testbed File Format gff PCIDSK PCIDSK Database PCRaster PCRaster Raster File ILWIS ILWIS Raster Map File SGI SGI SRTMHGT File Leveller Leveller heightfield Terragen Terragen heightfield Image File Format 1 0 Format Table 3 11 4 Import raster support formats ID Name ID Name ID Name HDF4 Hierarchical Data Format HDF4Image HDF4 Dataset ISIS3 USGS Astrogeology ISIS cube Release 4 version 3 ISIS2 USGS Astrogeology ISIS PDS NASA Planetary Data ERS ERMapper ers cube version 2 System LIB NOAA Polar Orbiter Level FIT FIT image GRIB GRIdded Binary grb 1b Data Set RMF Raster Matrix Format WCS OGC Web Coverage WMS OGC Web Map Service Service MSGN EUMESAT Archive nat RST Idrisi Raster A 1 INGR Intergraph Raster GSAG Golden Software ASCII GSBG Golden Software Binary GS7BG Golden Softw
95. ors respectively and uw Co t NOM the command vector that consists of the main rotor collective input o main rotor and flybar cyclic inputs dk and is and tail rotor collective input Oot Gravity Rigid body 5 u Helicopter JT n i Dynamics Kinematics components n y e p i i Figure 3 Vl 6 Helicopter dynamic model block diagram The total force and moment vectors account for the contributions of all helicopter components and can be decomposed as TF mr Tn T LU RER th Tn Ty TH fn where subscript mr stands for main rotor tr for tail rotor fs for fuselage tp for horizontal tailplane and fn for vertical fin As the primary source of lift propulsion and control the main rotor dominates helicopter dynamic behaviour The Bell Hiller stabilizing bar improves the stability characteristics of the helicopter The tail rotor located at the tail boom provides the moment needed to counteract the torque generated by the aerodynamic drag forces at the rotor hub The remaining components have less significant contributions and simpler models as well In short the fuselage produces drag forces and moments and the empennage components horizontal tailplane and vertical fin act as wings in forward flight increasing flight efficiency Main rotor A rotary wing aircraft flies via an engineering process called blade element theory which involves breaking the main rotor and tail rotor in many small elements and then finding
96. ort authorization and payment of applicable fees TECHNICHAL INFORMATION Any sketches models samples designs algorithms production processes or techniques submitted or disclosed by seller shall remain the property of seller and shall be treated as confidential information No use or disclosure of such information shall be made without the express written consent of seller GOVERNING LAW AND FORUM These terms and conditions shall be governed by the laws of the United Kingdom of Great Britain and Northern Ireland where goods or services originated in full or in part therein The rights and obligations of the parties shall not be governed by the 1980 U N Convention on Contracts for the International Sales of Goods Buyer consents to venue and jurisdiction over any action related to this contract at the relevant United Kingdom Court for goods or services originated in full or in part in the United Kingdom of Great Britain and Northern Ireland For goods or services originated in full from other countries where the seller or associated companies market its products or services these terms and conditions shall be governed by the laws of the country where this contract is agreed to by both the buyer and seller MODIFICATION No addition to or modification of any provision of this contract shall be binding upon seller unless made in writing and signed by a duly authorized representative of seller SEVERABILITY Any provision s of this contract whi
97. ped works by breaking down the problem into several stages These will be covered as follows in section 4 1 2 camera model and calibration is presented Detected features points are described on section 4 1 3 and section 4 1 4 presents the algorithm about matching point features between pairs of images Section 4 1 5 refers to an iterative Structure from Motion SfM procedure to recover the camera parameters Section 4 1 6 presents two methods about stereo matching of calibrated images The Poisson reconstruction with oriented 3D reconstructed points set is presented in section 4 1 7 The final texture 3D scene is described in Section 4 1 8 Finally section 4 1 9 shows an optional and complementary module that is useful to alien multiple mission tracks under distinct weather condition and illumination conditions in order to get a more detailed 3D scene reconstruction model Figure 4 1 1 shows images taken by the synthetic vision module of our ground control station Raphael GCS Guti rrez Resendiz and Funes Gallanzi 2009 Figure 4 1 1 Four images taken from the synthetic vision toolkit of Raphael GCS are shown 4 1 2 Camera Model and calibration As a first step we consider a camera as an electronic device that records an image digitally The pinhole camera model is widely used as the real camera projection This model is based on the principle of collinearity where each point in the object space is projected by a straight line through the
98. r own modules and further developments of sensors and other peripherals to the basic platform Using a common GCS platform will help developments by implementing individual contributions EO sensors control algorithms communications propulsion energy management etc on a common interoperable system that can be used by or marketed to third parties AVNTK can contribute by making available an advanced software product which has taken us 4 years to develop that will be cost effective reduce development costs shorten time to market and reduce duplication of effort on a crucial area In this context we offer the following facilities to the various market participants UAV manufacturers We are able to provide engineering services on an outsourcing basis to complement the customer software suites currently being used with the system herein described The customer needs to get a licence for using our software and will have full ownership of any software done specifically for him Where the software is sold on to a final customer we will pay a 25 Commission to the UAV manufacturer Engineering companies and UAV resellers We would be prepared to share some of the software work with the collaborating company Each entity would own their own developments The customer needs to get a license for using our software and will have full ownership of any software done specifically for him The 25 commission would be payable to the collaborating company in
99. raction Library GDAL which alone supports more than 40 different file formats e GPSBabel An interface to the GPSBabel software GPSBabel converts waypoints tracks and routes between popular GPS receivers and mapping programs It also has powerful manipulation tools for such data Geocaching com loc GPSman GPX XML Magellan protocol Magellan Mapsend Garmin PCX5 Garmin Mapsource gpsutil U S Census Bureau Tiger Comma separated values Delorme Topo USA4 XMap Navitrak DNA marker Mapping Service Conduit format MS PocketStreets 2002 Cetus for Pals OS GPSPilot Tracker for Magellan NAV companion Pushpin Palm OS for PalmOS Garmin serial protocol MapTech Exchange Format Holoux gm 100 wpo OziExplorer Waypoint Format National Geographic Topo Lpg TopoMapPro Places File Table 3 11 2 Available formats for PGS Babel interface e GPX to shapefile Converts a GPX file into a shapes shp e Import DXF Files This module imports DXF files using the free dxflib library e Import ESRI EQO File Import data sets from ESRI s E00 interchange format This import filter is based on the E00 format analysis of the GRASS GIS module m in e00 e Import GRIB2 record Import a GRIB2 record Geospatial Data Abstraction Library Raphael GCS has implemented an interface to import and export the data formats available through the GDAL library e Export Raster The GDAL Raster Export module export
100. rds where the expression in found Select by Theme Currently both input layers have to be of type polygon Separate Shapes Shapes Buffer A vector based buffer construction partly based on the method supported on An effective buffer generation method in GIS Split Shapes Layer Split Shapes Layer Randomly Randonly splits one shapes layer into to new shapes layers Useful to create a control group for model testing Split Table Shapes by Attribute Summary Transform Shapes Use this module to move rotate and or scale shapes Table This module is designed for table calculations e Function Fit Calculates the least square fit e Running Average Table Calculator The table calculator creates a new column based on the existing columns and a mathematical formula The columns are addressed by single characters a z which correspond in alphabetical order to the columns a first b second columns etc Table Calculator for Shapes Calculation of the new values in attributes tables Trend for Shapes Data Trend for Table Data Table Tools Create Empty Table Creates a new empty table Enumerate a Table Attribute Enumerate attribute of a table i e Assign to identical values of choosen attribute field unique identifiers Rotate Table Rotate a table i e Swap rows and columns TIN Tools for triangular Irregular Network TIN processing Flow Accumulation Parallel Calculates the catchment area based on the se
101. re being fused together This module needs to be upgraded for use by the MOD by enabling the calibration procedure to use known targets such as a helicopter or aircraft instead of the usual dot patterns as shown below n this way as a UAV takes off its imaging system can take a few frames of a known target and later these will be used to calibrate the imaging system and correct for distortion In order to do the more advanced tasks such as cartography ortho rectification target acquisition DEM modeling mosaicing etc high quality images must be produced from UAV derived imaging which requires modules such as a 6 degree pixel calibration camera correction tool capable of correcting for camera and lens distortion Other functions available in this module include normalization photo triangulation stereo plotter anaglyph rectification interior and exterior orientation correction In summary there are functions for e Camera Calibration e Rectification e Interior orientation e Exterior orientation e Photo triangulation e Stereo plotter anaglyph Camera Calibration Numerical example for camera calibration module In order to calibrate any camera we have used a chessboard pattern of known dimensions Each black white square on the chessboard has 3 cm x 3 cm of size Thus we have used 10 images with 640 x 512 pixel size These images were taken from distinct angular directions The calibration algorithm given by Heikkila 199
102. re the information about those shapes is stored There are many modules for manipulation and analysis of vector data like merging of layers querying of shapes attribute table manipulation type conversion and automated document creation Standard operations on vector data are polygon layer intersections and vector data creation from raster data e g of contour lines The built in file format is the ESRI Shape File format ESRI 1998 TIN It is a special vector data structure for point data for which the neighborhood relations of the points are defined by Triangular Irregular Network TIN using Delaunay Triangulation Similar to shapes TIN has associated table object for the storage of additional attribute data TIN can be loaded and saved as points in the ESRI Shape File format ie Mission Manning Plige Comal Lancmor Documentation Datal eceseng Data A wyss final Reports Show ep amp Oo rt HHan rga Be LOL Mee Figure 3 VIII 1 It is shows Digital Elevation Model of the Colima Volcano in Jalisco Mexico Grids Raster or grid data objects are matrix stored in numerical values Possible data types are 1 bit 1 2 4 byte integers 4 and 8 byte floating point values Raphael GCS contains standard tools for grid manipulation for example Grid calculator where a user defined formula is used to combine an arbitrary number of raster layers The raster data access methods supported by Raphael GCS have plenty of import
103. rently almost all model scale helicopters are equipped with a Bell Hiller stabilizing bar a mechanical blade pitch control system that improves helicopter stability From a control point of view the stabilizing bar can be interpreted as a dynamic feedback system for the roll and pitch rates The system consists of a so called flybar a teetering rotor placed at a 90 rotation interval from the main rotor blades and tipped on both ends by aerodynamic paddles and a mixing device that combines the flybar flapping motion with the cyclic inputs to determine the cyclic pitch angle applied to the main rotor blades C nn a 2 y fa fda t m2 7 paddle wa Bn u a f F M Bell Hiller mixing level roto blade chordwise view swashplate Odw RP Figure 3 VI 8 Bell Hiller system with angular displacements The system derives from a combination of the Bell stabilizing bar fitted with a mechanical damper and weights at each tip and the Hiller stabilizing bar which instead of weights uses small airfoils with incidence commanded by the cyclic inputs In the Hiller system the blade pitch angle is determined by the flybar flapping only The Bell Hiller system introduces the mixing device that allows some of the swashplate input to be directly applied to the blades The flybar and main rotor flapping motions are governed by the same effects namely the gyroscopic moments due the helicopter rol
104. rom the Last Waypoint Before Landing LWBL position to an end of runway position from which the operator takes over This point needs to be specified in the Mission Planning module Takeoff Performance Many standards are used to define the stages of an aircraft takeoff run depending on the country and type of aircraft We have taken the Federal Avitation Regulation FAR for illustration propose to use the definition of takeoff Under FAR 25 an aircraft taking off performs a ground roll to rotation velocity rotates to liftoff attitude lifts off and climbs to a height of 35 ft This definition can be applied to two types of takeoff takeoff with all engines operating AEO and takeoff with engine failure usually prescribed as one engine inoperative OEI Each of these types of takeoff will be discussed in turn Takeoff with all engines operating is the type dealt with in most day to day situations The aircraft accelerates from a stop or taxi speed to the velocity of rotation v rotates to the liftoff attitude with corresponding velocity Vig and climbs over an obstacle of 35 feet as shown in figure X The velocity at the end of the 35 ft climb is usually called the takeoff safety speed and given the designation V gt FAR 25 prescribes limits to these velocities based on the stall velocity is the minimum control velocity V uc and the minimum unstick velocity V mu These three velocities are physical minimum velocities under wh
105. round decrease energy by one point E die if energy drops below zero Figure 3 IX 3 Shark movement for Wa tor cellular automata The figure 3 IX 4 below shows the simulation of a Wa tor Automata given by Raphael GCS Figure 3 IX 4 The wa tor cellular automata Simulation Fire Risk Analysis Fire Risk Analysis This module predicts danger compound probability and priority index for a given Digital Elevation Model DEM fuel model grid set of fuel moisture wind speed and its direction It is based on the BEHAVE fire modelling system supported by the U S Forest Service Fire and Aviation Management see at http fire org Input Data Name Type Label Description DEM Grid DEM Digital Elevation Model Fuel Model Grid FUEL Wind Speed Grid WINDSPD Wind speed m s Wind Direction Grid WINDDIR Wind direction degrees clockwise from north Dead Fuel Grid MIH M10H M100H Moisture 1H 10H 100H Herbaceous Grid MHERB Fuel Mosture Wood Fuel Grid MWOOD Moisture Value Grid optional VALUE Base Probability Grid optional BASEPROB Table 3 IX 1 Fire Risk Analysis input data Output data Name Type Label Danger Grid DANGER Compound Probability Grid COMPPROB Priority Index Grid PRIORITY Table 3 1X 2 Fire Risk Analysis output data Simulation Fire Spreading Analysis Fire spread analysis This module predicts the spread rate
106. s is done within the synthetic vision module of Raphael GCS in order to compare the 3D surface reconstruction and the original 3D mesh model point set We have taken the template and fixed and fitted the 3D reconstructed mesh using a modified version of the ICP algorithm as above We search in an iterative process for a rigid body transformation that includes normal vector information between this two point sets and we get the final overlapping mesh between template and 3D reconstructed model In figure 4 1 9 1 we show the overlap between the 3D reconstructed model red dots and the its well known CAD model blue dots We read in the CAD model from an OBJ file As we can see the agreement is excellent Thus we calculated a histogram of error associated over each axis Figure 11 shows a histogram of the ratio between the shortest distance between both model and 3D reconstruction divide by a characteristic scale in this case the building height The average standard deviation about the three axes is around 1 0 CAD Data vs Multiple Images In many applications such as aircraft ID once a 3D object has been reconstructed it needs to be positively identified typically against a library of possible candidates This module is able to compare objects embeded within multiple images with a library of CAD objects to enable the determination of a specific item and item orientation The Raphael GCS module is based on the SoftPOSIT algorithm which is a fa
107. s one or more GIS to various file formats using the Geospatial Data Abstraction Library GDAL Currently raster supported formats are ID Name ID Name VRT Virtual Raster ERS ERMapper ers GTiff GeoTIFF RMF Raster Matrix Format NIFT National Imagery Transmission Format RST drisi Raster A 1 HFA Erdas Imagine Images img INGR Intergraph Raster ELAS ELAS GSBG Golden Software Binary Grid grd MEM In Memory Raster PNM Portable Pixmap Format BMP MS Windows Device Independent Bitmap ENVI ENVI hdr Table 3 11 3 Export raster support formats PCDISK PCDISK Database File EHdr ESRI hdr ILWIS ILWIS Raster Map PAux PCI aux SGI SGI Image File Format 1 0 MFF Vexcel MFF Raster Leveller Leveller Heightfield MFF2 Vexcel MFF2 HKV Raster Terragen Terragen Heightfield BT VTP bt Binary Terrain 1 3 Format HDF4Image HDF4 Dataset IDA Image Data and Analysis ADRG ARC Digitized Raster Graphics Table 3 11 3 Export raster support formats Continue e Import Raster This module imports grid data from various file formats using the GDAL library Currently raster supported formats are ID Name ID Name ID Name VRT Virtual Raster ELAS ELAS GIF Graphics Interchange Format gif GTiff GeoTIFF AIG Arc Info Binary Grid BIGGIF Graphics Interchange Format gif NIFT National Imagery AAIGrid Arc Info ASCI Grid ESAT Envisat Image F
108. st iterative method for estimate the pose of a 3D object using multiple images even in the case that correspondences between model features and image features are unknown c Figure 4 1 9 2 a Shows a single image b Shows initial pose of the CAD model over the image and c shows the estimate pose of the CAD model versus a single image 4 1l format Figure 4 11 2 It is shows Brightness Temperature maps Left Band 1 at 25 8 um Right Band 3 50 3 um In principle this module can displays satellite metadata likes maps The physical quantity that may contains this kind of metadata can be Brightness Temperature Antenna Temperature Chemical Concentration of some substance etc The figure 4 11 2 shows two channels of AIRS metadata mes LL w kia be Figure 4 11 1 Metadata taken from the AIRS instrument version L1B E IBI MULTISPECTRAL ANALYSIS 145x30 sarang LIB AMSU 32 b Tooting coti 145x395 tptgeoqga LIB AMSU 32 b amp t unsigned mtegeri 145230 2engeoga LIB AM SU 16 bt unsigned eteger 1452 30 demgeoge LIA AMSU 16 bit unsigned integer 145x330 salve LIB AMSU 932 Dx fhaing post 145230 Sata LOM AMS L32 bst flasting point 45x 30 solzwen 118 AM SU 32 bir floabng point 45230 solazi LIB AMSU 32 bit Aostan port 45630 sum_girt_dstence L1B AMSU 16 tet integer 45230 topog LIB AMSU 37 bit fieating pointi 145x30 topeg er 118 AMS 32 mt fosting pant 145430 lanciFr
109. t IMU from which velocity information can be deduced A homography based approach has been utilized for the determination of position and pose of the UAV with respect to the landing pad The homosgraphy based approach is well suited general application since all visual features are embedded on a flat planar landing pad On the other hand a constant design vector is integrated within the filtered regulation error signal resulting in an input matrix that facilities an advantageous coupling of translation dynamics of the UAV to the rotational torque inputs Additionally the null space of this input matrix is used to achieve a secondary control objective of damping the orientation error signal of the UAV to within a neighborhood about zero which can be made arbitrarily small through the proper selection of design parameters In the next section we will present a brief discussion of the camera projection model and then introduce the homography relations further camera model and camera calibration see section 3 VI and section 4 1 Projection models Visual information is a projection from the 3D world to the 2D camera image surface The pose of the camera determines a rigid body transformation from the current camera fixed frame P to the reference frame and subsequently from the desired image frame D4 to I One has Ar su E 1 XT Ra A T Ed as a relation between the coordinates of the same point in the current body fixed frame X
110. tection The first stage of computation searches over all scales and image locations It is implemented efficiently by using a difference of Gaussian function to identify interest points that are invariant to scale and orientation 2 Keypoint location At each candidate location a detailed model is fitted to determine location and scale Keypoints are selected based on measures of their stability 3 Orientation assignment One or more orientations are assigned to each keypoint location based on local image gradient directions All future operations are performed on image data that have been transformed relatively to the assigned orientation scale and location for each feature thereby providing invariance to these transformations 4 Keypoint descriptor The local image gradients are measured at the selected scale in the region around each keypoint These are transformed into a representation that allows for significant levels of local shape distortion and change in illumination In addition to the keypoints location themselves SIFT provides a local descriptor for each keypoint in the following sense e Consider a 16x16 patch centered on the keypoint e Decompose this patch into 4x4 pixel tiles e For each such tile we compute a histogram of its pixels gradient orientations with 8 bins each covering 45 degrees Actually we specify these gradient orientations relative to the keypoints dominant orientation We weigh the contribu
111. tion Topographic correction for differential illumination effects Available methods such as 1 Cosine Correction Teillet et al 1982 2 Cosine Correction Civco 1989 3 Minnaert Correction 4 Minnaert Correction with Slope Riano et al 2003 5 Minnaert Correction with Slope Law and Nichol 2004 6 C Correction Terrain Analysis Morphometry Convergence Index This module calculates the convergence index of a given DEM file Available methods such as Aspect or Gradient Convergence Index rearch radius This module calculates the convergence index of a given DEM file Available methods such as 1 Standard 2 Distance Weighted Linear and 3 Distance Weighted Inverse Curvature Classification Surface curvature based terrain classification V V GE V X V V GR GE GR X GR V X GE X X X Table 3 1X 9 Curvature classification CO IA tA 4d OO D e C Diurnal Anisotropic Heating This module calculates the diurnal anisotropic heating of a given DEM file Downslope Distance Gradient Calculation of a new topographic index to quantify downslope controls on local drainage Effective Air Flow Heights This module calculates the effective air flow heights of a given DEM file It is suppose a wind direction Lee fector Luv factor and maximun distance Hypsometry Calculates the hypsometric curve for a given DEM Land Surface Temperature This module calculates the land surface tempera
112. tion does not use a maximum search radius The weighting matrix is generated once globally for all points Ordinary Kriging VF Kriging with Variogram Fit Ordinary Kriging VF Global Kringing with Variogram Fit The weighting matrix is generated once globally for all points Universal Kriging Universal Kriging Global Universal Kriging VF Universal Kriging VF Global Semi Varigram Points Point Analysis Ripley s K variation ellipse etc Multiple regression analysis Linear regression analysis of point attributes with multiples grids Details of the regression correlation analysis will be saved to a table The regression functions is used to create a new grid with extrapolated values The multiple regression analysis uses forward selection procedure Radius of variance Find the radius within which the cell values exceed the given variance criterium This modules is closely related to the representativeness calculation variance within given search radius For easier usage the variance criterium is entered as standard deviation value Regression analysis Regression analysis of point attributes with grid values The regression function is used to create a new grid with extrapolated values Representativeness Calculation of the variance within a given search radius Residual analysis Relations of each grid cell to its neighborhood This calculations can be used on terrain analysis Statistics for grids Calculat
113. tion to the bin by the gradient magnitude We use a circular Gaussian falloff from the keypoint center about 8 pixels half the descriptor window e This gives 16 tiles x 8 histogram bins tile a 128 dimensional vector representing our a SIFT feature descriptor e Normalize this to unit length e For better illumination invariance a threshold value in this normalized vector to be no larger than 0 2 and then renormalize e This vector normalized is invariant to location scale rotation share and changes in illumination The figure 4 1 3 1 shows a typical output of a SIFT detection process Each image contains thousands of SIFT features Keypoints are shown as vectors in cyan with magnitude and direction only Figure 4 1 3 1 Figures show keypoints Left The 798 x 510 pixel image has 4467 keypoints Rigth The 798 x 510 pixel image has 3824 keypoints The keypoints are displayed as vectors indicating scale orientation and locations 4 1 4 SIFT matching between pair images Since each keypoint is described by a vector 128 dimensional We have used the SIFT method to get an automatic feature matching over all image pairs using the Approximate Nearest Neighbors package of Arya et al 1998 e We take Euclidian distance in the space of 128 dimensional vectors e We use kd trees for efficient near neighbor search e We compare the distance to best matching SIFT feature in the database and the distance to the 2 best
114. to under predict ground roll at times and the climb phase is often over predicted The modified Krenkel and Salzman method requieres 25 input parameters and allows thrust vectoring and assumes thrust varying with velocity A modification was made to assume thrust varied quadratically with velocity Originally the method solved the equations of motion both nonlinear ordinary differential equations parametrically Due to consistent under prediction of the eround roll the method was also modified to include a rotation phase a continuation of the ground roll for a user defined amount of time As with the simplified Powers method the modified Krenkel and Salzman method iterated from an initial guess critical engine failure velocity to predict the BFL Unlike the Powers method the Krenkel and Salzman method increased the engine out rotation velocity to allow the aircraft to take off with reduced thrust Takeoff Example Name Value Unit Atmospheric density 0 sI ft 2 Aircraft Weight 95000 Ibs Wing area 1000 ft 2 Maximum lift coefficient 2 Ground lift coefficient 0 3 Air lift coefficient 1 65 Ground drag coefficient 0 08 Air drag coefficient 0 121 Rolling friction coeff 0 025 Braking friction coeff 0 3 Angle of thrust 0 00 rad Stall margin 1 1 Descision time 3 00 sec Obstacle height 35 00 ft OEI power remaining 0 5 Thrust 31450 0 17 263404 V 0 025019 V 2 Normal Takeof
115. ture of a given DEM but it supposes short ware radiation grid kW m and leaf area index grid Local Morphometry Calculates local morphometric terrain attributes i e slope aspect and curvatures Available methods such as Available Choices 1 Maximum Slope Travis et al 1975 2 Maximum Triangle Slope Tarboton 1997 3 Least Squares Fitted Plane Horn 1981 Costa Cabral amp Burgess 1996 4 Fit 2 Degree Polynom Bauer Rohdenburg Bork 1985 5 Fit 2 Degree Polynom Heerdegen amp Beran 1982 6 Fit 2 Degree Polynom Zevenbergen amp Thorne 1987 7 Fit 3 Degree Polynom Haralick 1983 Mass Balance Index This module calculates the mas balance index of a given DEM and optionally as input data the vectical disntace to channel network is taken into account Morphometric Protection Index This module calculates the protection index of a given DEM Multiresolution Index of Valley Bottom Flatness MRVBF Calculation of the multiresolution index of valley bottom flatness MRVBF and the complementary multiresolution index of the ridge top flatness MRRTF of a given DEM Real Area Calculation Calculates real not proyected cell area Relative Heights and Slope Positions Surface Specific Points Detection of surface specific points by local parallel processing of discrete terrain elevation data Available methods such as 1 Mark Highest Neighbour 2 Opposite Neighbours 3 Flow Direction 4 Flow Direction up and down and
116. tween DRCs and Physical Catchment Descriptors PCDs Teaching unit hydrograph theory and its applications Hydrometric data quality assurance control QA QC Infilling missing streamflow data The only field data required are time series of rainfall and streamflow and a third variable by which evapotranspiration effects can be approximated The third variable can be air temperature but pan evaporation or potential evaporation PE derived from hydrometeorological measurements can also be used as alternatives if they are available Rainfall is converted to effective rainfall x by a non lineal loss module followed by a linear unit hydrograph UH module to convert effective rainfall to streamflow x See figure 3 1X 8 non linear linear pa loss module SVAT unit hydrograph Figure 3 1X 8 IHACRES block process effective rain model fit 50 observed modeted et am rer e i pe mr f e rn Age af l s ri an t d e 20 dE total quick REW modeled siva t5 zol je ere pe uad 3 8 7 9 111935 17 19 Nuts ue e ge gs air temperature unit hydrographs hydrograph separation Figure 3 IX 9 Data time steps from six minutes to one month have been employed successfully by IHACRES for catchments varying in size from 490 m in China to nearly 10 000 km in the UK The methodology yields a unit hydrograph UH for total streamflow This can often be resolved into separate unit
117. two steps i Each reconstructed 3D point has a list of associated cameras given by the SIFT matching process and the bundle adjustment process and ii each 3D reconstructed point has a list of visibility after the stereo matching process is done Thus we have at least two criteria for the visibility for a given 3D reconstructed point Since the stereo matching process has an implicit high degree of photometric consistency we combine visibility information to take the camera that has the maximum color cross correlation within the cameras list in a small oriented patch around each 3D reconstructed point We check RGB color continuity around each orientated patch in order to assign texture on it The figures 4 1 8 1 shows the textured 3D reconstructed model directly compared to the original CAD model In addition the figure 4 1 8 2 shows the 3D scene reconstruction and it displays some of the camera poses The figure 4 1 8 3 shows the whole 3D scene reconstruction a b Figure 4 1 8 1 Original CAD building a compared to its 3D reconstruction from FMV b a b Figure 4 1 8 2 Four views of the FMV camera pose sequences a and whole scene textured view b Figure 4 1 8 3 Final 3D reconstruction 4 1 9 Alignment between independent mission tracks We also take into account that when we have independent mission tracks under distinct operational conditions we have additional information to get a more detailed 3D reconstruct
118. ty estimate SU vivam e Threats He e Sensor dependence bas A e Imaging quality 37 D e Route details e Re planning limitations Al targets mate Durabon Weck Terme i Maer Mawr Evaluate Route Figure 3 11 1 Route Planning GUI 3 1 SYNTHETIC VISION Raphael GCS has an embedded 3D virtual reality environment This module can capture the general context of each position on earth surface The user can choose any location on earth and it will shows its topology under user friendly environment The 3D environment includes stellar information solar positioning so shading and illumination conditions can be predicted and accounted for as well as an extensive library of 3D objects In summary the 3D virtual reality environment has the following item Advanced world topological DEM Buildings Roads Rivers lakes Vegetation trees grass Urban furniture cars transit signals Textures Snow Sky Water Fog The synthetic vision available on Raphael GCS has typical screenshot given by the figure 3 11 1 Figure 3 111 1 Typical screenshots using the synthetic vision available on Raphael GCS with and without navigational aids Raphael GCS has implemented numerous of computer vision algorithms such as the Fourier Transform Feature Detection Motion Analysis Object Tracking Canny detector Harris detector Hough transform for line detector Hough transform for circle detector SIFT algorithm Calculate the optic
119. valuated before the method settings in mode c the option is evaluated only if the NoData values is not included in the lookup table The other values option is always evaluated after checking the method settings Resampling Resampling of grids Sort Grid Threshold Buffer Threshold buffer creation Advanced grid operations II Accumulated Cost Anisotropic Accumulated Cost Isotropic Aggregation Index It is useful to quantify spatial patterns of landscapes Analytical Hierarchy Process Change Vector Analysis Covered Distance Cross Classification and Tabulation Fragmentation Alternative 1 interior if Density 1 0 2 undetermined if Density gt 0 6 and Density Connectivity 3 perforated if Density gt 0 6 and Density Connectivity gt 0 4 edge if Density gt 0 6 and Density Connectivity 5 transitional if 0 4 6 patch if Density Fragmentation Standard Grid fragmentation analysis given by Riiters method Fragmentation classes from density and connectivity Fuzzify Prepares a grid for fuzzy logic analysis Fuzzy intersection grid Calculates the intersection for each grid cell of the selected grids Fuzzy union grid Calculates the union for each grid cell of the selected grids Layer of extreme value It creates a new grid containing the ID of the grid with the maximun minimum value Least cost path interactive Creates a least cost past profile using an accumulated cost surface
120. ve produced a number of robotics systems for such companies as Hitachi Global Storage Technologies in our view stereo vision should only be employed where distance to object is short and relative speeds high as in in door robotics In cases where there is large relative movement and distance between the imaging system and the object of interest we have found that an automatic 3D reconstruction algorithm is better than a stereoscopic approach as the distance between any two views can be varied whereas in a stereo system this is fixed which causes large errors as these are inversely related to the angle subtended between the two cameras and the object We have developed an advanced system for 3D scene reconstruction from multiple image views such as those obtain from full moving videos FMV onboard unmanned vehicles UAVs If a scene of interest is seen from several viewing cameras with different calibration parameters and physical characteristics positioned around the target even distinct illumination degrees occlusion scale image resolution share effect or motion we have the only commercially available computational package capable of getting a high definition 3D scene reconstruction This 3D scene reconstruction work for multiple image views of a target of interest where the position of each viewing image is only roughly known is ideal when digital images including photographs are available for example because the phenomenon is unique or
121. very fish has a predefined breed time On exceeding this time it gives birth to a new fish in one of the neighboring cells provided this randomly selected cell is free If not nothing happens Breed time counter of both the original and the descendant fish will be reset to zero Technically fish never die They live until they reach the breed time then they clone and both parent as well as offspring restart their life cycle The following picture shows options for prey movement Arrows indicate possible movements Fish are not allowed to move to cells occupied by sharks If there are no free neighboring cells no movement occurs Fish movement gt create offspring in free neighboring cells VA gt only move to free cells t Figure 3 IX 2 Fish movement for Wa tor cellular automata Rules for sharks Sharks move randomly to fields that are either free or occupied by fish Every round they lose one point of energy If they enter a field occupied by a fish they eat the fish and gain a defined amount of energy If the energy level drops below zero the shark dies If the energy exceeds a predefined value sharks create an offspring in a free neighboring field The energy is split evenly between the parent and the child Shark movement E only move to free or fish cells E when moving to fish cells eat fish and increase energy BI E if enough energy is accumulated create offspring GF in free neighboring cells 4 E every
122. xtures of two source signals which we know to be independent of each other i e observing the value of one signal does not give any information about the value of the other The BSS problem is then to determine the source signals given only the mixtures This figure 4 11 8 below shows the signal mixtures on the left and the corresponding joint density plot on the right That is at a given time instant the value of the top signal is the first input component and the value of the bottom signal is the corresponding second input component The plot on the right is then simply plotting both signal along first and second input component The marginal densities are also shown at the edge of the plot SIGNALS JOINT DENSITY Input signals and density Figure 4 11 8 Left Input signals components measured signals Right Joint density of both input components There are many algorithms for performing ICA but the most efficient to date is the FastlCA fixed point algorithm which is implemented on Raphael GCS The plot below shows the result after convergence of the FastlCA algorithm SIGNALS JOINT DENSITY zB AU d M d aisre ivt 7M SI Separated signals after 5 steps of FastlCA Figure 4 11 9 Result after FastlCA is applied The source signals in this example were a sinusoid and impulsive noise as can be seen in the left part of the plot below The right plot shows the joint density which can be seen to be the product of the
123. y Lighting and visibility calculations for digital terrain models e Analysis Hillshading Analytical hillshading calculation It has as input data to file elevation DEM and it yields a grid with angle value between the surface and the incoming light beams measured in radians Available methods such as Standard Standard max 90 degree Combined shading ray tracing e Incoming Solar Radiation Calculation of the potential incoming solar radiation The model calculates the total solar radiation KWh mA and the duration of insolation hours for a given range of days Duration of insolation ini Solar Radiation OLIS a b Figure 3 1X 18 Incoming solar radiation a The solar radiation distribution spread on the DEM b Duration of insolation spread on the DEM e Insolation Calculation of incoming solar radiation insolation Based on the SADO System for the Analysis of Discrete Surfaces routines developed by Boehner amp Trachinow It calculates the direct insolation diffuser insolation and the total insolation Due insolation c Figure 3 1X 20 Insolation given the de SADO method a Direct insolation spread over the terrain b Diffuse component spread over the terrain c Total contribution of solar radiation spread over the terrain Sky View Factor This module calculates the Visible Sky and Sky View factor for a given DEM file Visibility single point Interactive Topographic Correc

the aerospace technology group, ardita aeronautica division

Contents

Download Pdf Manuals

Related Search

Related Contents