1/*M/////////////////////////////////////////////////////////////////////////////////////// 2// 3// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. 4// 5// By downloading, copying, installing or using the software you agree to this license. 6// If you do not agree to this license, do not download, install, 7// copy or use the software. 8// 9// 10// License Agreement 11// For Open Source Computer Vision Library 12// 13// Copyright (C) 2000-2008, Intel Corporation, all rights reserved. 14// Copyright (C) 2009, Willow Garage Inc., all rights reserved. 15// Third party copyrights are property of their respective owners. 16// 17// Redistribution and use in source and binary forms, with or without modification, 18// are permitted provided that the following conditions are met: 19// 20// * Redistribution's of source code must retain the above copyright notice, 21// this list of conditions and the following disclaimer. 22// 23// * Redistribution's in binary form must reproduce the above copyright notice, 24// this list of conditions and the following disclaimer in the documentation 25// and/or other materials provided with the distribution. 26// 27// * The name of the copyright holders may not be used to endorse or promote products 28// derived from this software without specific prior written permission. 29// 30// This software is provided by the copyright holders and contributors "as is" and 31// any express or implied warranties, including, but not limited to, the implied 32// warranties of merchantability and fitness for a particular purpose are disclaimed. 33// In no event shall the Intel Corporation or contributors be liable for any direct, 34// indirect, incidental, special, exemplary, or consequential damages 35// (including, but not limited to, procurement of substitute goods or services; 36// loss of use, data, or profits; or business interruption) however caused 37// and on any theory of liability, whether in contract, strict liability, 38// or tort (including negligence or otherwise) arising in any way out of 39// the use of this software, even if advised of the possibility of such damage. 40// 41//M*/ 42 43#include "precomp.hpp" 44 45using namespace cv; 46using namespace cv::cuda; 47 48#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER) 49 50Ptr<cv::cuda::ORB> cv::cuda::ORB::create(int, float, int, int, int, int, int, int, int, bool) { throw_no_cuda(); return Ptr<cv::cuda::ORB>(); } 51 52#else /* !defined (HAVE_CUDA) */ 53 54namespace cv { namespace cuda { namespace device 55{ 56 namespace orb 57 { 58 int cull_gpu(int* loc, float* response, int size, int n_points); 59 60 void HarrisResponses_gpu(PtrStepSzb img, const short2* loc, float* response, const int npoints, int blockSize, float harris_k, cudaStream_t stream); 61 62 void loadUMax(const int* u_max, int count); 63 64 void IC_Angle_gpu(PtrStepSzb image, const short2* loc, float* angle, int npoints, int half_k, cudaStream_t stream); 65 66 void computeOrbDescriptor_gpu(PtrStepb img, const short2* loc, const float* angle, const int npoints, 67 const int* pattern_x, const int* pattern_y, PtrStepb desc, int dsize, int WTA_K, cudaStream_t stream); 68 69 void mergeLocation_gpu(const short2* loc, float* x, float* y, int npoints, float scale, cudaStream_t stream); 70 } 71}}} 72 73namespace 74{ 75 const float HARRIS_K = 0.04f; 76 const int DESCRIPTOR_SIZE = 32; 77 78 const int bit_pattern_31_[256 * 4] = 79 { 80 8,-3, 9,5/*mean (0), correlation (0)*/, 81 4,2, 7,-12/*mean (1.12461e-05), correlation (0.0437584)*/, 82 -11,9, -8,2/*mean (3.37382e-05), correlation (0.0617409)*/, 83 7,-12, 12,-13/*mean (5.62303e-05), correlation (0.0636977)*/, 84 2,-13, 2,12/*mean (0.000134953), correlation (0.085099)*/, 85 1,-7, 1,6/*mean (0.000528565), correlation (0.0857175)*/, 86 -2,-10, -2,-4/*mean (0.0188821), correlation (0.0985774)*/, 87 -13,-13, -11,-8/*mean (0.0363135), correlation (0.0899616)*/, 88 -13,-3, -12,-9/*mean (0.121806), correlation (0.099849)*/, 89 10,4, 11,9/*mean (0.122065), correlation (0.093285)*/, 90 -13,-8, -8,-9/*mean (0.162787), correlation (0.0942748)*/, 91 -11,7, -9,12/*mean (0.21561), correlation (0.0974438)*/, 92 7,7, 12,6/*mean (0.160583), correlation (0.130064)*/, 93 -4,-5, -3,0/*mean (0.228171), correlation (0.132998)*/, 94 -13,2, -12,-3/*mean (0.00997526), correlation (0.145926)*/, 95 -9,0, -7,5/*mean (0.198234), correlation (0.143636)*/, 96 12,-6, 12,-1/*mean (0.0676226), correlation (0.16689)*/, 97 -3,6, -2,12/*mean (0.166847), correlation (0.171682)*/, 98 -6,-13, -4,-8/*mean (0.101215), correlation (0.179716)*/, 99 11,-13, 12,-8/*mean (0.200641), correlation (0.192279)*/, 100 4,7, 5,1/*mean (0.205106), correlation (0.186848)*/, 101 5,-3, 10,-3/*mean (0.234908), correlation (0.192319)*/, 102 3,-7, 6,12/*mean (0.0709964), correlation (0.210872)*/, 103 -8,-7, -6,-2/*mean (0.0939834), correlation (0.212589)*/, 104 -2,11, -1,-10/*mean (0.127778), correlation (0.20866)*/, 105 -13,12, -8,10/*mean (0.14783), correlation (0.206356)*/, 106 -7,3, -5,-3/*mean (0.182141), correlation (0.198942)*/, 107 -4,2, -3,7/*mean (0.188237), correlation (0.21384)*/, 108 -10,-12, -6,11/*mean (0.14865), correlation (0.23571)*/, 109 5,-12, 6,-7/*mean (0.222312), correlation (0.23324)*/, 110 5,-6, 7,-1/*mean (0.229082), correlation (0.23389)*/, 111 1,0, 4,-5/*mean (0.241577), correlation (0.215286)*/, 112 9,11, 11,-13/*mean (0.00338507), correlation (0.251373)*/, 113 4,7, 4,12/*mean (0.131005), correlation (0.257622)*/, 114 2,-1, 4,4/*mean (0.152755), correlation (0.255205)*/, 115 -4,-12, -2,7/*mean (0.182771), correlation (0.244867)*/, 116 -8,-5, -7,-10/*mean (0.186898), correlation (0.23901)*/, 117 4,11, 9,12/*mean (0.226226), correlation (0.258255)*/, 118 0,-8, 1,-13/*mean (0.0897886), correlation (0.274827)*/, 119 -13,-2, -8,2/*mean (0.148774), correlation (0.28065)*/, 120 -3,-2, -2,3/*mean (0.153048), correlation (0.283063)*/, 121 -6,9, -4,-9/*mean (0.169523), correlation (0.278248)*/, 122 8,12, 10,7/*mean (0.225337), correlation (0.282851)*/, 123 0,9, 1,3/*mean (0.226687), correlation (0.278734)*/, 124 7,-5, 11,-10/*mean (0.00693882), correlation (0.305161)*/, 125 -13,-6, -11,0/*mean (0.0227283), correlation (0.300181)*/, 126 10,7, 12,1/*mean (0.125517), correlation (0.31089)*/, 127 -6,-3, -6,12/*mean (0.131748), correlation (0.312779)*/, 128 10,-9, 12,-4/*mean (0.144827), correlation (0.292797)*/, 129 -13,8, -8,-12/*mean (0.149202), correlation (0.308918)*/, 130 -13,0, -8,-4/*mean (0.160909), correlation (0.310013)*/, 131 3,3, 7,8/*mean (0.177755), correlation (0.309394)*/, 132 5,7, 10,-7/*mean (0.212337), correlation (0.310315)*/, 133 -1,7, 1,-12/*mean (0.214429), correlation (0.311933)*/, 134 3,-10, 5,6/*mean (0.235807), correlation (0.313104)*/, 135 2,-4, 3,-10/*mean (0.00494827), correlation (0.344948)*/, 136 -13,0, -13,5/*mean (0.0549145), correlation (0.344675)*/, 137 -13,-7, -12,12/*mean (0.103385), correlation (0.342715)*/, 138 -13,3, -11,8/*mean (0.134222), correlation (0.322922)*/, 139 -7,12, -4,7/*mean (0.153284), correlation (0.337061)*/, 140 6,-10, 12,8/*mean (0.154881), correlation (0.329257)*/, 141 -9,-1, -7,-6/*mean (0.200967), correlation (0.33312)*/, 142 -2,-5, 0,12/*mean (0.201518), correlation (0.340635)*/, 143 -12,5, -7,5/*mean (0.207805), correlation (0.335631)*/, 144 3,-10, 8,-13/*mean (0.224438), correlation (0.34504)*/, 145 -7,-7, -4,5/*mean (0.239361), correlation (0.338053)*/, 146 -3,-2, -1,-7/*mean (0.240744), correlation (0.344322)*/, 147 2,9, 5,-11/*mean (0.242949), correlation (0.34145)*/, 148 -11,-13, -5,-13/*mean (0.244028), correlation (0.336861)*/, 149 -1,6, 0,-1/*mean (0.247571), correlation (0.343684)*/, 150 5,-3, 5,2/*mean (0.000697256), correlation (0.357265)*/, 151 -4,-13, -4,12/*mean (0.00213675), correlation (0.373827)*/, 152 -9,-6, -9,6/*mean (0.0126856), correlation (0.373938)*/, 153 -12,-10, -8,-4/*mean (0.0152497), correlation (0.364237)*/, 154 10,2, 12,-3/*mean (0.0299933), correlation (0.345292)*/, 155 7,12, 12,12/*mean (0.0307242), correlation (0.366299)*/, 156 -7,-13, -6,5/*mean (0.0534975), correlation (0.368357)*/, 157 -4,9, -3,4/*mean (0.099865), correlation (0.372276)*/, 158 7,-1, 12,2/*mean (0.117083), correlation (0.364529)*/, 159 -7,6, -5,1/*mean (0.126125), correlation (0.369606)*/, 160 -13,11, -12,5/*mean (0.130364), correlation (0.358502)*/, 161 -3,7, -2,-6/*mean (0.131691), correlation (0.375531)*/, 162 7,-8, 12,-7/*mean (0.160166), correlation (0.379508)*/, 163 -13,-7, -11,-12/*mean (0.167848), correlation (0.353343)*/, 164 1,-3, 12,12/*mean (0.183378), correlation (0.371916)*/, 165 2,-6, 3,0/*mean (0.228711), correlation (0.371761)*/, 166 -4,3, -2,-13/*mean (0.247211), correlation (0.364063)*/, 167 -1,-13, 1,9/*mean (0.249325), correlation (0.378139)*/, 168 7,1, 8,-6/*mean (0.000652272), correlation (0.411682)*/, 169 1,-1, 3,12/*mean (0.00248538), correlation (0.392988)*/, 170 9,1, 12,6/*mean (0.0206815), correlation (0.386106)*/, 171 -1,-9, -1,3/*mean (0.0364485), correlation (0.410752)*/, 172 -13,-13, -10,5/*mean (0.0376068), correlation (0.398374)*/, 173 7,7, 10,12/*mean (0.0424202), correlation (0.405663)*/, 174 12,-5, 12,9/*mean (0.0942645), correlation (0.410422)*/, 175 6,3, 7,11/*mean (0.1074), correlation (0.413224)*/, 176 5,-13, 6,10/*mean (0.109256), correlation (0.408646)*/, 177 2,-12, 2,3/*mean (0.131691), correlation (0.416076)*/, 178 3,8, 4,-6/*mean (0.165081), correlation (0.417569)*/, 179 2,6, 12,-13/*mean (0.171874), correlation (0.408471)*/, 180 9,-12, 10,3/*mean (0.175146), correlation (0.41296)*/, 181 -8,4, -7,9/*mean (0.183682), correlation (0.402956)*/, 182 -11,12, -4,-6/*mean (0.184672), correlation (0.416125)*/, 183 1,12, 2,-8/*mean (0.191487), correlation (0.386696)*/, 184 6,-9, 7,-4/*mean (0.192668), correlation (0.394771)*/, 185 2,3, 3,-2/*mean (0.200157), correlation (0.408303)*/, 186 6,3, 11,0/*mean (0.204588), correlation (0.411762)*/, 187 3,-3, 8,-8/*mean (0.205904), correlation (0.416294)*/, 188 7,8, 9,3/*mean (0.213237), correlation (0.409306)*/, 189 -11,-5, -6,-4/*mean (0.243444), correlation (0.395069)*/, 190 -10,11, -5,10/*mean (0.247672), correlation (0.413392)*/, 191 -5,-8, -3,12/*mean (0.24774), correlation (0.411416)*/, 192 -10,5, -9,0/*mean (0.00213675), correlation (0.454003)*/, 193 8,-1, 12,-6/*mean (0.0293635), correlation (0.455368)*/, 194 4,-6, 6,-11/*mean (0.0404971), correlation (0.457393)*/, 195 -10,12, -8,7/*mean (0.0481107), correlation (0.448364)*/, 196 4,-2, 6,7/*mean (0.050641), correlation (0.455019)*/, 197 -2,0, -2,12/*mean (0.0525978), correlation (0.44338)*/, 198 -5,-8, -5,2/*mean (0.0629667), correlation (0.457096)*/, 199 7,-6, 10,12/*mean (0.0653846), correlation (0.445623)*/, 200 -9,-13, -8,-8/*mean (0.0858749), correlation (0.449789)*/, 201 -5,-13, -5,-2/*mean (0.122402), correlation (0.450201)*/, 202 8,-8, 9,-13/*mean (0.125416), correlation (0.453224)*/, 203 -9,-11, -9,0/*mean (0.130128), correlation (0.458724)*/, 204 1,-8, 1,-2/*mean (0.132467), correlation (0.440133)*/, 205 7,-4, 9,1/*mean (0.132692), correlation (0.454)*/, 206 -2,1, -1,-4/*mean (0.135695), correlation (0.455739)*/, 207 11,-6, 12,-11/*mean (0.142904), correlation (0.446114)*/, 208 -12,-9, -6,4/*mean (0.146165), correlation (0.451473)*/, 209 3,7, 7,12/*mean (0.147627), correlation (0.456643)*/, 210 5,5, 10,8/*mean (0.152901), correlation (0.455036)*/, 211 0,-4, 2,8/*mean (0.167083), correlation (0.459315)*/, 212 -9,12, -5,-13/*mean (0.173234), correlation (0.454706)*/, 213 0,7, 2,12/*mean (0.18312), correlation (0.433855)*/, 214 -1,2, 1,7/*mean (0.185504), correlation (0.443838)*/, 215 5,11, 7,-9/*mean (0.185706), correlation (0.451123)*/, 216 3,5, 6,-8/*mean (0.188968), correlation (0.455808)*/, 217 -13,-4, -8,9/*mean (0.191667), correlation (0.459128)*/, 218 -5,9, -3,-3/*mean (0.193196), correlation (0.458364)*/, 219 -4,-7, -3,-12/*mean (0.196536), correlation (0.455782)*/, 220 6,5, 8,0/*mean (0.1972), correlation (0.450481)*/, 221 -7,6, -6,12/*mean (0.199438), correlation (0.458156)*/, 222 -13,6, -5,-2/*mean (0.211224), correlation (0.449548)*/, 223 1,-10, 3,10/*mean (0.211718), correlation (0.440606)*/, 224 4,1, 8,-4/*mean (0.213034), correlation (0.443177)*/, 225 -2,-2, 2,-13/*mean (0.234334), correlation (0.455304)*/, 226 2,-12, 12,12/*mean (0.235684), correlation (0.443436)*/, 227 -2,-13, 0,-6/*mean (0.237674), correlation (0.452525)*/, 228 4,1, 9,3/*mean (0.23962), correlation (0.444824)*/, 229 -6,-10, -3,-5/*mean (0.248459), correlation (0.439621)*/, 230 -3,-13, -1,1/*mean (0.249505), correlation (0.456666)*/, 231 7,5, 12,-11/*mean (0.00119208), correlation (0.495466)*/, 232 4,-2, 5,-7/*mean (0.00372245), correlation (0.484214)*/, 233 -13,9, -9,-5/*mean (0.00741116), correlation (0.499854)*/, 234 7,1, 8,6/*mean (0.0208952), correlation (0.499773)*/, 235 7,-8, 7,6/*mean (0.0220085), correlation (0.501609)*/, 236 -7,-4, -7,1/*mean (0.0233806), correlation (0.496568)*/, 237 -8,11, -7,-8/*mean (0.0236505), correlation (0.489719)*/, 238 -13,6, -12,-8/*mean (0.0268781), correlation (0.503487)*/, 239 2,4, 3,9/*mean (0.0323324), correlation (0.501938)*/, 240 10,-5, 12,3/*mean (0.0399235), correlation (0.494029)*/, 241 -6,-5, -6,7/*mean (0.0420153), correlation (0.486579)*/, 242 8,-3, 9,-8/*mean (0.0548021), correlation (0.484237)*/, 243 2,-12, 2,8/*mean (0.0616622), correlation (0.496642)*/, 244 -11,-2, -10,3/*mean (0.0627755), correlation (0.498563)*/, 245 -12,-13, -7,-9/*mean (0.0829622), correlation (0.495491)*/, 246 -11,0, -10,-5/*mean (0.0843342), correlation (0.487146)*/, 247 5,-3, 11,8/*mean (0.0929937), correlation (0.502315)*/, 248 -2,-13, -1,12/*mean (0.113327), correlation (0.48941)*/, 249 -1,-8, 0,9/*mean (0.132119), correlation (0.467268)*/, 250 -13,-11, -12,-5/*mean (0.136269), correlation (0.498771)*/, 251 -10,-2, -10,11/*mean (0.142173), correlation (0.498714)*/, 252 -3,9, -2,-13/*mean (0.144141), correlation (0.491973)*/, 253 2,-3, 3,2/*mean (0.14892), correlation (0.500782)*/, 254 -9,-13, -4,0/*mean (0.150371), correlation (0.498211)*/, 255 -4,6, -3,-10/*mean (0.152159), correlation (0.495547)*/, 256 -4,12, -2,-7/*mean (0.156152), correlation (0.496925)*/, 257 -6,-11, -4,9/*mean (0.15749), correlation (0.499222)*/, 258 6,-3, 6,11/*mean (0.159211), correlation (0.503821)*/, 259 -13,11, -5,5/*mean (0.162427), correlation (0.501907)*/, 260 11,11, 12,6/*mean (0.16652), correlation (0.497632)*/, 261 7,-5, 12,-2/*mean (0.169141), correlation (0.484474)*/, 262 -1,12, 0,7/*mean (0.169456), correlation (0.495339)*/, 263 -4,-8, -3,-2/*mean (0.171457), correlation (0.487251)*/, 264 -7,1, -6,7/*mean (0.175), correlation (0.500024)*/, 265 -13,-12, -8,-13/*mean (0.175866), correlation (0.497523)*/, 266 -7,-2, -6,-8/*mean (0.178273), correlation (0.501854)*/, 267 -8,5, -6,-9/*mean (0.181107), correlation (0.494888)*/, 268 -5,-1, -4,5/*mean (0.190227), correlation (0.482557)*/, 269 -13,7, -8,10/*mean (0.196739), correlation (0.496503)*/, 270 1,5, 5,-13/*mean (0.19973), correlation (0.499759)*/, 271 1,0, 10,-13/*mean (0.204465), correlation (0.49873)*/, 272 9,12, 10,-1/*mean (0.209334), correlation (0.49063)*/, 273 5,-8, 10,-9/*mean (0.211134), correlation (0.503011)*/, 274 -1,11, 1,-13/*mean (0.212), correlation (0.499414)*/, 275 -9,-3, -6,2/*mean (0.212168), correlation (0.480739)*/, 276 -1,-10, 1,12/*mean (0.212731), correlation (0.502523)*/, 277 -13,1, -8,-10/*mean (0.21327), correlation (0.489786)*/, 278 8,-11, 10,-6/*mean (0.214159), correlation (0.488246)*/, 279 2,-13, 3,-6/*mean (0.216993), correlation (0.50287)*/, 280 7,-13, 12,-9/*mean (0.223639), correlation (0.470502)*/, 281 -10,-10, -5,-7/*mean (0.224089), correlation (0.500852)*/, 282 -10,-8, -8,-13/*mean (0.228666), correlation (0.502629)*/, 283 4,-6, 8,5/*mean (0.22906), correlation (0.498305)*/, 284 3,12, 8,-13/*mean (0.233378), correlation (0.503825)*/, 285 -4,2, -3,-3/*mean (0.234323), correlation (0.476692)*/, 286 5,-13, 10,-12/*mean (0.236392), correlation (0.475462)*/, 287 4,-13, 5,-1/*mean (0.236842), correlation (0.504132)*/, 288 -9,9, -4,3/*mean (0.236977), correlation (0.497739)*/, 289 0,3, 3,-9/*mean (0.24314), correlation (0.499398)*/, 290 -12,1, -6,1/*mean (0.243297), correlation (0.489447)*/, 291 3,2, 4,-8/*mean (0.00155196), correlation (0.553496)*/, 292 -10,-10, -10,9/*mean (0.00239541), correlation (0.54297)*/, 293 8,-13, 12,12/*mean (0.0034413), correlation (0.544361)*/, 294 -8,-12, -6,-5/*mean (0.003565), correlation (0.551225)*/, 295 2,2, 3,7/*mean (0.00835583), correlation (0.55285)*/, 296 10,6, 11,-8/*mean (0.00885065), correlation (0.540913)*/, 297 6,8, 8,-12/*mean (0.0101552), correlation (0.551085)*/, 298 -7,10, -6,5/*mean (0.0102227), correlation (0.533635)*/, 299 -3,-9, -3,9/*mean (0.0110211), correlation (0.543121)*/, 300 -1,-13, -1,5/*mean (0.0113473), correlation (0.550173)*/, 301 -3,-7, -3,4/*mean (0.0140913), correlation (0.554774)*/, 302 -8,-2, -8,3/*mean (0.017049), correlation (0.55461)*/, 303 4,2, 12,12/*mean (0.01778), correlation (0.546921)*/, 304 2,-5, 3,11/*mean (0.0224022), correlation (0.549667)*/, 305 6,-9, 11,-13/*mean (0.029161), correlation (0.546295)*/, 306 3,-1, 7,12/*mean (0.0303081), correlation (0.548599)*/, 307 11,-1, 12,4/*mean (0.0355151), correlation (0.523943)*/, 308 -3,0, -3,6/*mean (0.0417904), correlation (0.543395)*/, 309 4,-11, 4,12/*mean (0.0487292), correlation (0.542818)*/, 310 2,-4, 2,1/*mean (0.0575124), correlation (0.554888)*/, 311 -10,-6, -8,1/*mean (0.0594242), correlation (0.544026)*/, 312 -13,7, -11,1/*mean (0.0597391), correlation (0.550524)*/, 313 -13,12, -11,-13/*mean (0.0608974), correlation (0.55383)*/, 314 6,0, 11,-13/*mean (0.065126), correlation (0.552006)*/, 315 0,-1, 1,4/*mean (0.074224), correlation (0.546372)*/, 316 -13,3, -9,-2/*mean (0.0808592), correlation (0.554875)*/, 317 -9,8, -6,-3/*mean (0.0883378), correlation (0.551178)*/, 318 -13,-6, -8,-2/*mean (0.0901035), correlation (0.548446)*/, 319 5,-9, 8,10/*mean (0.0949843), correlation (0.554694)*/, 320 2,7, 3,-9/*mean (0.0994152), correlation (0.550979)*/, 321 -1,-6, -1,-1/*mean (0.10045), correlation (0.552714)*/, 322 9,5, 11,-2/*mean (0.100686), correlation (0.552594)*/, 323 11,-3, 12,-8/*mean (0.101091), correlation (0.532394)*/, 324 3,0, 3,5/*mean (0.101147), correlation (0.525576)*/, 325 -1,4, 0,10/*mean (0.105263), correlation (0.531498)*/, 326 3,-6, 4,5/*mean (0.110785), correlation (0.540491)*/, 327 -13,0, -10,5/*mean (0.112798), correlation (0.536582)*/, 328 5,8, 12,11/*mean (0.114181), correlation (0.555793)*/, 329 8,9, 9,-6/*mean (0.117431), correlation (0.553763)*/, 330 7,-4, 8,-12/*mean (0.118522), correlation (0.553452)*/, 331 -10,4, -10,9/*mean (0.12094), correlation (0.554785)*/, 332 7,3, 12,4/*mean (0.122582), correlation (0.555825)*/, 333 9,-7, 10,-2/*mean (0.124978), correlation (0.549846)*/, 334 7,0, 12,-2/*mean (0.127002), correlation (0.537452)*/, 335 -1,-6, 0,-11/*mean (0.127148), correlation (0.547401)*/ 336 }; 337 338 class ORB_Impl : public cv::cuda::ORB 339 { 340 public: 341 ORB_Impl(int nfeatures, 342 float scaleFactor, 343 int nlevels, 344 int edgeThreshold, 345 int firstLevel, 346 int WTA_K, 347 int scoreType, 348 int patchSize, 349 int fastThreshold, 350 bool blurForDescriptor); 351 352 virtual void detectAndCompute(InputArray _image, InputArray _mask, std::vector<KeyPoint>& keypoints, OutputArray _descriptors, bool useProvidedKeypoints); 353 virtual void detectAndComputeAsync(InputArray _image, InputArray _mask, OutputArray _keypoints, OutputArray _descriptors, bool useProvidedKeypoints, Stream& stream); 354 355 virtual void convert(InputArray _gpu_keypoints, std::vector<KeyPoint>& keypoints); 356 357 virtual int descriptorSize() const { return kBytes; } 358 virtual int descriptorType() const { return CV_8U; } 359 virtual int defaultNorm() const { return NORM_HAMMING; } 360 361 virtual void setMaxFeatures(int maxFeatures) { nFeatures_ = maxFeatures; } 362 virtual int getMaxFeatures() const { return nFeatures_; } 363 364 virtual void setScaleFactor(double scaleFactor) { scaleFactor_ = scaleFactor; } 365 virtual double getScaleFactor() const { return scaleFactor_; } 366 367 virtual void setNLevels(int nlevels) { nLevels_ = nlevels; } 368 virtual int getNLevels() const { return nLevels_; } 369 370 virtual void setEdgeThreshold(int edgeThreshold) { edgeThreshold_ = edgeThreshold; } 371 virtual int getEdgeThreshold() const { return edgeThreshold_; } 372 373 virtual void setFirstLevel(int firstLevel) { firstLevel_ = firstLevel; } 374 virtual int getFirstLevel() const { return firstLevel_; } 375 376 virtual void setWTA_K(int wta_k) { WTA_K_ = wta_k; } 377 virtual int getWTA_K() const { return WTA_K_; } 378 379 virtual void setScoreType(int scoreType) { scoreType_ = scoreType; } 380 virtual int getScoreType() const { return scoreType_; } 381 382 virtual void setPatchSize(int patchSize) { patchSize_ = patchSize; } 383 virtual int getPatchSize() const { return patchSize_; } 384 385 virtual void setFastThreshold(int fastThreshold) { fastThreshold_ = fastThreshold; } 386 virtual int getFastThreshold() const { return fastThreshold_; } 387 388 virtual void setBlurForDescriptor(bool blurForDescriptor) { blurForDescriptor_ = blurForDescriptor; } 389 virtual bool getBlurForDescriptor() const { return blurForDescriptor_; } 390 391 private: 392 int nFeatures_; 393 float scaleFactor_; 394 int nLevels_; 395 int edgeThreshold_; 396 int firstLevel_; 397 int WTA_K_; 398 int scoreType_; 399 int patchSize_; 400 int fastThreshold_; 401 bool blurForDescriptor_; 402 403 private: 404 void buildScalePyramids(InputArray _image, InputArray _mask); 405 void computeKeyPointsPyramid(); 406 void computeDescriptors(OutputArray _descriptors); 407 void mergeKeyPoints(OutputArray _keypoints); 408 409 private: 410 Ptr<cv::cuda::FastFeatureDetector> fastDetector_; 411 412 //! The number of desired features per scale 413 std::vector<size_t> n_features_per_level_; 414 415 //! Points to compute BRIEF descriptors from 416 GpuMat pattern_; 417 418 std::vector<GpuMat> imagePyr_; 419 std::vector<GpuMat> maskPyr_; 420 421 GpuMat buf_; 422 423 std::vector<GpuMat> keyPointsPyr_; 424 std::vector<int> keyPointsCount_; 425 426 Ptr<cuda::Filter> blurFilter_; 427 428 GpuMat d_keypoints_; 429 }; 430 431 static void initializeOrbPattern(const Point* pattern0, Mat& pattern, int ntuples, int tupleSize, int poolSize) 432 { 433 RNG rng(0x12345678); 434 435 pattern.create(2, ntuples * tupleSize, CV_32SC1); 436 pattern.setTo(Scalar::all(0)); 437 438 int* pattern_x_ptr = pattern.ptr<int>(0); 439 int* pattern_y_ptr = pattern.ptr<int>(1); 440 441 for (int i = 0; i < ntuples; i++) 442 { 443 for (int k = 0; k < tupleSize; k++) 444 { 445 for(;;) 446 { 447 int idx = rng.uniform(0, poolSize); 448 Point pt = pattern0[idx]; 449 450 int k1; 451 for (k1 = 0; k1 < k; k1++) 452 if (pattern_x_ptr[tupleSize * i + k1] == pt.x && pattern_y_ptr[tupleSize * i + k1] == pt.y) 453 break; 454 455 if (k1 == k) 456 { 457 pattern_x_ptr[tupleSize * i + k] = pt.x; 458 pattern_y_ptr[tupleSize * i + k] = pt.y; 459 break; 460 } 461 } 462 } 463 } 464 } 465 466 static void makeRandomPattern(int patchSize, Point* pattern, int npoints) 467 { 468 // we always start with a fixed seed, 469 // to make patterns the same on each run 470 RNG rng(0x34985739); 471 472 for (int i = 0; i < npoints; i++) 473 { 474 pattern[i].x = rng.uniform(-patchSize / 2, patchSize / 2 + 1); 475 pattern[i].y = rng.uniform(-patchSize / 2, patchSize / 2 + 1); 476 } 477 } 478 479 ORB_Impl::ORB_Impl(int nFeatures, 480 float scaleFactor, 481 int nLevels, 482 int edgeThreshold, 483 int firstLevel, 484 int WTA_K, 485 int scoreType, 486 int patchSize, 487 int fastThreshold, 488 bool blurForDescriptor) : 489 nFeatures_(nFeatures), 490 scaleFactor_(scaleFactor), 491 nLevels_(nLevels), 492 edgeThreshold_(edgeThreshold), 493 firstLevel_(firstLevel), 494 WTA_K_(WTA_K), 495 scoreType_(scoreType), 496 patchSize_(patchSize), 497 fastThreshold_(fastThreshold), 498 blurForDescriptor_(blurForDescriptor) 499 { 500 CV_Assert( patchSize_ >= 2 ); 501 CV_Assert( WTA_K_ == 2 || WTA_K_ == 3 || WTA_K_ == 4 ); 502 503 fastDetector_ = cuda::FastFeatureDetector::create(fastThreshold_); 504 505 // fill the extractors and descriptors for the corresponding scales 506 float factor = 1.0f / scaleFactor_; 507 float n_desired_features_per_scale = nFeatures_ * (1.0f - factor) / (1.0f - std::pow(factor, nLevels_)); 508 509 n_features_per_level_.resize(nLevels_); 510 size_t sum_n_features = 0; 511 for (int level = 0; level < nLevels_ - 1; ++level) 512 { 513 n_features_per_level_[level] = cvRound(n_desired_features_per_scale); 514 sum_n_features += n_features_per_level_[level]; 515 n_desired_features_per_scale *= factor; 516 } 517 n_features_per_level_[nLevels_ - 1] = nFeatures - sum_n_features; 518 519 // pre-compute the end of a row in a circular patch 520 int half_patch_size = patchSize_ / 2; 521 std::vector<int> u_max(half_patch_size + 2); 522 for (int v = 0; v <= half_patch_size * std::sqrt(2.f) / 2 + 1; ++v) 523 { 524 u_max[v] = cvRound(std::sqrt(static_cast<float>(half_patch_size * half_patch_size - v * v))); 525 } 526 527 // Make sure we are symmetric 528 for (int v = half_patch_size, v_0 = 0; v >= half_patch_size * std::sqrt(2.f) / 2; --v) 529 { 530 while (u_max[v_0] == u_max[v_0 + 1]) 531 ++v_0; 532 u_max[v] = v_0; 533 ++v_0; 534 } 535 CV_Assert( u_max.size() < 32 ); 536 cv::cuda::device::orb::loadUMax(&u_max[0], static_cast<int>(u_max.size())); 537 538 // Calc pattern 539 const int npoints = 512; 540 Point pattern_buf[npoints]; 541 const Point* pattern0 = (const Point*)bit_pattern_31_; 542 if (patchSize_ != 31) 543 { 544 pattern0 = pattern_buf; 545 makeRandomPattern(patchSize_, pattern_buf, npoints); 546 } 547 548 Mat h_pattern; 549 if (WTA_K_ == 2) 550 { 551 h_pattern.create(2, npoints, CV_32SC1); 552 553 int* pattern_x_ptr = h_pattern.ptr<int>(0); 554 int* pattern_y_ptr = h_pattern.ptr<int>(1); 555 556 for (int i = 0; i < npoints; ++i) 557 { 558 pattern_x_ptr[i] = pattern0[i].x; 559 pattern_y_ptr[i] = pattern0[i].y; 560 } 561 } 562 else 563 { 564 int ntuples = descriptorSize() * 4; 565 initializeOrbPattern(pattern0, h_pattern, ntuples, WTA_K_, npoints); 566 } 567 568 pattern_.upload(h_pattern); 569 570 blurFilter_ = cuda::createGaussianFilter(CV_8UC1, -1, Size(7, 7), 2, 2, BORDER_REFLECT_101); 571 } 572 573 void ORB_Impl::detectAndCompute(InputArray _image, InputArray _mask, std::vector<KeyPoint>& keypoints, OutputArray _descriptors, bool useProvidedKeypoints) 574 { 575 CV_Assert( useProvidedKeypoints == false ); 576 577 detectAndComputeAsync(_image, _mask, d_keypoints_, _descriptors, false, Stream::Null()); 578 convert(d_keypoints_, keypoints); 579 } 580 581 void ORB_Impl::detectAndComputeAsync(InputArray _image, InputArray _mask, OutputArray _keypoints, OutputArray _descriptors, bool useProvidedKeypoints, Stream& stream) 582 { 583 CV_Assert( useProvidedKeypoints == false ); 584 585 buildScalePyramids(_image, _mask); 586 computeKeyPointsPyramid(); 587 if (_descriptors.needed()) 588 { 589 computeDescriptors(_descriptors); 590 } 591 mergeKeyPoints(_keypoints); 592 } 593 594 static float getScale(float scaleFactor, int firstLevel, int level) 595 { 596 return pow(scaleFactor, level - firstLevel); 597 } 598 599 void ORB_Impl::buildScalePyramids(InputArray _image, InputArray _mask) 600 { 601 const GpuMat image = _image.getGpuMat(); 602 const GpuMat mask = _mask.getGpuMat(); 603 604 CV_Assert( image.type() == CV_8UC1 ); 605 CV_Assert( mask.empty() || (mask.type() == CV_8UC1 && mask.size() == image.size()) ); 606 607 imagePyr_.resize(nLevels_); 608 maskPyr_.resize(nLevels_); 609 610 for (int level = 0; level < nLevels_; ++level) 611 { 612 float scale = 1.0f / getScale(scaleFactor_, firstLevel_, level); 613 614 Size sz(cvRound(image.cols * scale), cvRound(image.rows * scale)); 615 616 ensureSizeIsEnough(sz, image.type(), imagePyr_[level]); 617 ensureSizeIsEnough(sz, CV_8UC1, maskPyr_[level]); 618 maskPyr_[level].setTo(Scalar::all(255)); 619 620 // Compute the resized image 621 if (level != firstLevel_) 622 { 623 if (level < firstLevel_) 624 { 625 cuda::resize(image, imagePyr_[level], sz, 0, 0, INTER_LINEAR); 626 627 if (!mask.empty()) 628 cuda::resize(mask, maskPyr_[level], sz, 0, 0, INTER_LINEAR); 629 } 630 else 631 { 632 cuda::resize(imagePyr_[level - 1], imagePyr_[level], sz, 0, 0, INTER_LINEAR); 633 634 if (!mask.empty()) 635 { 636 cuda::resize(maskPyr_[level - 1], maskPyr_[level], sz, 0, 0, INTER_LINEAR); 637 cuda::threshold(maskPyr_[level], maskPyr_[level], 254, 0, THRESH_TOZERO); 638 } 639 } 640 } 641 else 642 { 643 image.copyTo(imagePyr_[level]); 644 645 if (!mask.empty()) 646 mask.copyTo(maskPyr_[level]); 647 } 648 649 // Filter keypoints by image border 650 ensureSizeIsEnough(sz, CV_8UC1, buf_); 651 buf_.setTo(Scalar::all(0)); 652 Rect inner(edgeThreshold_, edgeThreshold_, sz.width - 2 * edgeThreshold_, sz.height - 2 * edgeThreshold_); 653 buf_(inner).setTo(Scalar::all(255)); 654 655 cuda::bitwise_and(maskPyr_[level], buf_, maskPyr_[level]); 656 } 657 } 658 659 // takes keypoints and culls them by the response 660 static void cull(GpuMat& keypoints, int& count, int n_points) 661 { 662 using namespace cv::cuda::device::orb; 663 664 //this is only necessary if the keypoints size is greater than the number of desired points. 665 if (count > n_points) 666 { 667 if (n_points == 0) 668 { 669 keypoints.release(); 670 return; 671 } 672 673 count = cull_gpu(keypoints.ptr<int>(cuda::FastFeatureDetector::LOCATION_ROW), keypoints.ptr<float>(cuda::FastFeatureDetector::RESPONSE_ROW), count, n_points); 674 } 675 } 676 677 void ORB_Impl::computeKeyPointsPyramid() 678 { 679 using namespace cv::cuda::device::orb; 680 681 int half_patch_size = patchSize_ / 2; 682 683 keyPointsPyr_.resize(nLevels_); 684 keyPointsCount_.resize(nLevels_); 685 686 fastDetector_->setThreshold(fastThreshold_); 687 688 for (int level = 0; level < nLevels_; ++level) 689 { 690 fastDetector_->setMaxNumPoints(0.05 * imagePyr_[level].size().area()); 691 692 GpuMat fastKpRange; 693 fastDetector_->detectAsync(imagePyr_[level], fastKpRange, maskPyr_[level], Stream::Null()); 694 695 keyPointsCount_[level] = fastKpRange.cols; 696 697 if (keyPointsCount_[level] == 0) 698 continue; 699 700 ensureSizeIsEnough(3, keyPointsCount_[level], fastKpRange.type(), keyPointsPyr_[level]); 701 fastKpRange.copyTo(keyPointsPyr_[level].rowRange(0, 2)); 702 703 const int n_features = static_cast<int>(n_features_per_level_[level]); 704 705 if (scoreType_ == ORB::HARRIS_SCORE) 706 { 707 // Keep more points than necessary as FAST does not give amazing corners 708 cull(keyPointsPyr_[level], keyPointsCount_[level], 2 * n_features); 709 710 // Compute the Harris cornerness (better scoring than FAST) 711 HarrisResponses_gpu(imagePyr_[level], keyPointsPyr_[level].ptr<short2>(0), keyPointsPyr_[level].ptr<float>(1), keyPointsCount_[level], 7, HARRIS_K, 0); 712 } 713 714 //cull to the final desired level, using the new Harris scores or the original FAST scores. 715 cull(keyPointsPyr_[level], keyPointsCount_[level], n_features); 716 717 // Compute orientation 718 IC_Angle_gpu(imagePyr_[level], keyPointsPyr_[level].ptr<short2>(0), keyPointsPyr_[level].ptr<float>(2), keyPointsCount_[level], half_patch_size, 0); 719 } 720 } 721 722 void ORB_Impl::computeDescriptors(OutputArray _descriptors) 723 { 724 using namespace cv::cuda::device::orb; 725 726 int nAllkeypoints = 0; 727 728 for (int level = 0; level < nLevels_; ++level) 729 nAllkeypoints += keyPointsCount_[level]; 730 731 if (nAllkeypoints == 0) 732 { 733 _descriptors.release(); 734 return; 735 } 736 737 ensureSizeIsEnough(nAllkeypoints, descriptorSize(), CV_8UC1, _descriptors); 738 GpuMat descriptors = _descriptors.getGpuMat(); 739 740 int offset = 0; 741 742 for (int level = 0; level < nLevels_; ++level) 743 { 744 if (keyPointsCount_[level] == 0) 745 continue; 746 747 GpuMat descRange = descriptors.rowRange(offset, offset + keyPointsCount_[level]); 748 749 if (blurForDescriptor_) 750 { 751 // preprocess the resized image 752 ensureSizeIsEnough(imagePyr_[level].size(), imagePyr_[level].type(), buf_); 753 blurFilter_->apply(imagePyr_[level], buf_); 754 } 755 756 computeOrbDescriptor_gpu(blurForDescriptor_ ? buf_ : imagePyr_[level], keyPointsPyr_[level].ptr<short2>(0), keyPointsPyr_[level].ptr<float>(2), 757 keyPointsCount_[level], pattern_.ptr<int>(0), pattern_.ptr<int>(1), descRange, descriptorSize(), WTA_K_, 0); 758 759 offset += keyPointsCount_[level]; 760 } 761 } 762 763 void ORB_Impl::mergeKeyPoints(OutputArray _keypoints) 764 { 765 using namespace cv::cuda::device::orb; 766 767 int nAllkeypoints = 0; 768 769 for (int level = 0; level < nLevels_; ++level) 770 nAllkeypoints += keyPointsCount_[level]; 771 772 if (nAllkeypoints == 0) 773 { 774 _keypoints.release(); 775 return; 776 } 777 778 ensureSizeIsEnough(ROWS_COUNT, nAllkeypoints, CV_32FC1, _keypoints); 779 GpuMat& keypoints = _keypoints.getGpuMatRef(); 780 781 int offset = 0; 782 783 for (int level = 0; level < nLevels_; ++level) 784 { 785 if (keyPointsCount_[level] == 0) 786 continue; 787 788 float sf = getScale(scaleFactor_, firstLevel_, level); 789 790 GpuMat keyPointsRange = keypoints.colRange(offset, offset + keyPointsCount_[level]); 791 792 float locScale = level != firstLevel_ ? sf : 1.0f; 793 794 mergeLocation_gpu(keyPointsPyr_[level].ptr<short2>(0), keyPointsRange.ptr<float>(0), keyPointsRange.ptr<float>(1), keyPointsCount_[level], locScale, 0); 795 796 GpuMat range = keyPointsRange.rowRange(2, 4); 797 keyPointsPyr_[level](Range(1, 3), Range(0, keyPointsCount_[level])).copyTo(range); 798 799 keyPointsRange.row(4).setTo(Scalar::all(level)); 800 keyPointsRange.row(5).setTo(Scalar::all(patchSize_ * sf)); 801 802 offset += keyPointsCount_[level]; 803 } 804 } 805 806 void ORB_Impl::convert(InputArray _gpu_keypoints, std::vector<KeyPoint>& keypoints) 807 { 808 if (_gpu_keypoints.empty()) 809 { 810 keypoints.clear(); 811 return; 812 } 813 814 Mat h_keypoints; 815 if (_gpu_keypoints.kind() == _InputArray::CUDA_GPU_MAT) 816 { 817 _gpu_keypoints.getGpuMat().download(h_keypoints); 818 } 819 else 820 { 821 h_keypoints = _gpu_keypoints.getMat(); 822 } 823 824 CV_Assert( h_keypoints.rows == ROWS_COUNT ); 825 CV_Assert( h_keypoints.type() == CV_32FC1 ); 826 827 const int npoints = h_keypoints.cols; 828 829 keypoints.resize(npoints); 830 831 const float* x_ptr = h_keypoints.ptr<float>(X_ROW); 832 const float* y_ptr = h_keypoints.ptr<float>(Y_ROW); 833 const float* response_ptr = h_keypoints.ptr<float>(RESPONSE_ROW); 834 const float* angle_ptr = h_keypoints.ptr<float>(ANGLE_ROW); 835 const float* octave_ptr = h_keypoints.ptr<float>(OCTAVE_ROW); 836 const float* size_ptr = h_keypoints.ptr<float>(SIZE_ROW); 837 838 for (int i = 0; i < npoints; ++i) 839 { 840 KeyPoint kp; 841 842 kp.pt.x = x_ptr[i]; 843 kp.pt.y = y_ptr[i]; 844 kp.response = response_ptr[i]; 845 kp.angle = angle_ptr[i]; 846 kp.octave = static_cast<int>(octave_ptr[i]); 847 kp.size = size_ptr[i]; 848 849 keypoints[i] = kp; 850 } 851 } 852} 853 854Ptr<cv::cuda::ORB> cv::cuda::ORB::create(int nfeatures, 855 float scaleFactor, 856 int nlevels, 857 int edgeThreshold, 858 int firstLevel, 859 int WTA_K, 860 int scoreType, 861 int patchSize, 862 int fastThreshold, 863 bool blurForDescriptor) 864{ 865 return makePtr<ORB_Impl>(nfeatures, scaleFactor, nlevels, edgeThreshold, firstLevel, WTA_K, scoreType, patchSize, fastThreshold, blurForDescriptor); 866} 867 868#endif /* !defined (HAVE_CUDA) */ 869