ORBextractor.cc
Go to the documentation of this file.
1 
65 #include <opencv2/core/core.hpp>
66 #include <opencv2/highgui/highgui.hpp>
67 #include <opencv2/features2d/features2d.hpp>
68 #include <opencv2/imgproc/imgproc.hpp>
69 #include <vector>
70 #include <algorithm>
71 #include <iterator>
72 #include <set>
73 
74 #include "ORBextractor.h"
75 
76 
77 using namespace cv;
78 using namespace std;
79 
80 namespace rtabmap
81 {
82 
83 static float IC_Angle(const Mat& image, Point2f pt, const vector<int> & u_max, int halfPatchSize)
84 {
85  int m_01 = 0, m_10 = 0;
86 
87  const uchar* center = &image.at<uchar> (cvRound(pt.y), cvRound(pt.x));
88 
89  // Treat the center line differently, v=0
90  for (int u = -halfPatchSize; u <= halfPatchSize; ++u)
91  m_10 += u * center[u];
92 
93  // Go line by line in the circuI853lar patch
94  int step = (int)image.step1();
95  for (int v = 1; v <= halfPatchSize; ++v)
96  {
97  // Proceed over the two lines
98  int v_sum = 0;
99  int d = u_max[v];
100  for (int u = -d; u <= d; ++u)
101  {
102  int val_plus = center[u + v*step], val_minus = center[u - v*step];
103  v_sum += (val_plus - val_minus);
104  m_10 += u * (val_plus + val_minus);
105  }
106  m_01 += v * v_sum;
107  }
108 
109  return fastAtan2((float)m_01, (float)m_10);
110 }
111 
112 
113 const float factorPI = (float)(CV_PI/180.f);
114 static void computeOrbDescriptor(const KeyPoint& kpt,
115  const Mat& img, const Point* pattern,
116  uchar* desc)
117 {
118  float angle = (float)kpt.angle*factorPI;
119  float a = (float)cos(angle), b = (float)sin(angle);
120 
121  const uchar* center = &img.at<uchar>(cvRound(kpt.pt.y), cvRound(kpt.pt.x));
122  const int step = (int)img.step;
123 
124  #define GET_VALUE(idx) \
125  center[cvRound(pattern[idx].x*b + pattern[idx].y*a)*step + \
126  cvRound(pattern[idx].x*a - pattern[idx].y*b)]
127 
128 
129  for (int i = 0; i < 32; ++i, pattern += 16)
130  {
131  int t0, t1, val;
132  t0 = GET_VALUE(0); t1 = GET_VALUE(1);
133  val = t0 < t1;
134  t0 = GET_VALUE(2); t1 = GET_VALUE(3);
135  val |= (t0 < t1) << 1;
136  t0 = GET_VALUE(4); t1 = GET_VALUE(5);
137  val |= (t0 < t1) << 2;
138  t0 = GET_VALUE(6); t1 = GET_VALUE(7);
139  val |= (t0 < t1) << 3;
140  t0 = GET_VALUE(8); t1 = GET_VALUE(9);
141  val |= (t0 < t1) << 4;
142  t0 = GET_VALUE(10); t1 = GET_VALUE(11);
143  val |= (t0 < t1) << 5;
144  t0 = GET_VALUE(12); t1 = GET_VALUE(13);
145  val |= (t0 < t1) << 6;
146  t0 = GET_VALUE(14); t1 = GET_VALUE(15);
147  val |= (t0 < t1) << 7;
148 
149  desc[i] = (uchar)val;
150  }
151 
152  #undef GET_VALUE
153 }
154 
155 
156 static int bit_pattern_31_[256*4] =
157 {
158  8,-3, 9,5/*mean (0), correlation (0)*/,
159  4,2, 7,-12/*mean (1.12461e-05), correlation (0.0437584)*/,
160  -11,9, -8,2/*mean (3.37382e-05), correlation (0.0617409)*/,
161  7,-12, 12,-13/*mean (5.62303e-05), correlation (0.0636977)*/,
162  2,-13, 2,12/*mean (0.000134953), correlation (0.085099)*/,
163  1,-7, 1,6/*mean (0.000528565), correlation (0.0857175)*/,
164  -2,-10, -2,-4/*mean (0.0188821), correlation (0.0985774)*/,
165  -13,-13, -11,-8/*mean (0.0363135), correlation (0.0899616)*/,
166  -13,-3, -12,-9/*mean (0.121806), correlation (0.099849)*/,
167  10,4, 11,9/*mean (0.122065), correlation (0.093285)*/,
168  -13,-8, -8,-9/*mean (0.162787), correlation (0.0942748)*/,
169  -11,7, -9,12/*mean (0.21561), correlation (0.0974438)*/,
170  7,7, 12,6/*mean (0.160583), correlation (0.130064)*/,
171  -4,-5, -3,0/*mean (0.228171), correlation (0.132998)*/,
172  -13,2, -12,-3/*mean (0.00997526), correlation (0.145926)*/,
173  -9,0, -7,5/*mean (0.198234), correlation (0.143636)*/,
174  12,-6, 12,-1/*mean (0.0676226), correlation (0.16689)*/,
175  -3,6, -2,12/*mean (0.166847), correlation (0.171682)*/,
176  -6,-13, -4,-8/*mean (0.101215), correlation (0.179716)*/,
177  11,-13, 12,-8/*mean (0.200641), correlation (0.192279)*/,
178  4,7, 5,1/*mean (0.205106), correlation (0.186848)*/,
179  5,-3, 10,-3/*mean (0.234908), correlation (0.192319)*/,
180  3,-7, 6,12/*mean (0.0709964), correlation (0.210872)*/,
181  -8,-7, -6,-2/*mean (0.0939834), correlation (0.212589)*/,
182  -2,11, -1,-10/*mean (0.127778), correlation (0.20866)*/,
183  -13,12, -8,10/*mean (0.14783), correlation (0.206356)*/,
184  -7,3, -5,-3/*mean (0.182141), correlation (0.198942)*/,
185  -4,2, -3,7/*mean (0.188237), correlation (0.21384)*/,
186  -10,-12, -6,11/*mean (0.14865), correlation (0.23571)*/,
187  5,-12, 6,-7/*mean (0.222312), correlation (0.23324)*/,
188  5,-6, 7,-1/*mean (0.229082), correlation (0.23389)*/,
189  1,0, 4,-5/*mean (0.241577), correlation (0.215286)*/,
190  9,11, 11,-13/*mean (0.00338507), correlation (0.251373)*/,
191  4,7, 4,12/*mean (0.131005), correlation (0.257622)*/,
192  2,-1, 4,4/*mean (0.152755), correlation (0.255205)*/,
193  -4,-12, -2,7/*mean (0.182771), correlation (0.244867)*/,
194  -8,-5, -7,-10/*mean (0.186898), correlation (0.23901)*/,
195  4,11, 9,12/*mean (0.226226), correlation (0.258255)*/,
196  0,-8, 1,-13/*mean (0.0897886), correlation (0.274827)*/,
197  -13,-2, -8,2/*mean (0.148774), correlation (0.28065)*/,
198  -3,-2, -2,3/*mean (0.153048), correlation (0.283063)*/,
199  -6,9, -4,-9/*mean (0.169523), correlation (0.278248)*/,
200  8,12, 10,7/*mean (0.225337), correlation (0.282851)*/,
201  0,9, 1,3/*mean (0.226687), correlation (0.278734)*/,
202  7,-5, 11,-10/*mean (0.00693882), correlation (0.305161)*/,
203  -13,-6, -11,0/*mean (0.0227283), correlation (0.300181)*/,
204  10,7, 12,1/*mean (0.125517), correlation (0.31089)*/,
205  -6,-3, -6,12/*mean (0.131748), correlation (0.312779)*/,
206  10,-9, 12,-4/*mean (0.144827), correlation (0.292797)*/,
207  -13,8, -8,-12/*mean (0.149202), correlation (0.308918)*/,
208  -13,0, -8,-4/*mean (0.160909), correlation (0.310013)*/,
209  3,3, 7,8/*mean (0.177755), correlation (0.309394)*/,
210  5,7, 10,-7/*mean (0.212337), correlation (0.310315)*/,
211  -1,7, 1,-12/*mean (0.214429), correlation (0.311933)*/,
212  3,-10, 5,6/*mean (0.235807), correlation (0.313104)*/,
213  2,-4, 3,-10/*mean (0.00494827), correlation (0.344948)*/,
214  -13,0, -13,5/*mean (0.0549145), correlation (0.344675)*/,
215  -13,-7, -12,12/*mean (0.103385), correlation (0.342715)*/,
216  -13,3, -11,8/*mean (0.134222), correlation (0.322922)*/,
217  -7,12, -4,7/*mean (0.153284), correlation (0.337061)*/,
218  6,-10, 12,8/*mean (0.154881), correlation (0.329257)*/,
219  -9,-1, -7,-6/*mean (0.200967), correlation (0.33312)*/,
220  -2,-5, 0,12/*mean (0.201518), correlation (0.340635)*/,
221  -12,5, -7,5/*mean (0.207805), correlation (0.335631)*/,
222  3,-10, 8,-13/*mean (0.224438), correlation (0.34504)*/,
223  -7,-7, -4,5/*mean (0.239361), correlation (0.338053)*/,
224  -3,-2, -1,-7/*mean (0.240744), correlation (0.344322)*/,
225  2,9, 5,-11/*mean (0.242949), correlation (0.34145)*/,
226  -11,-13, -5,-13/*mean (0.244028), correlation (0.336861)*/,
227  -1,6, 0,-1/*mean (0.247571), correlation (0.343684)*/,
228  5,-3, 5,2/*mean (0.000697256), correlation (0.357265)*/,
229  -4,-13, -4,12/*mean (0.00213675), correlation (0.373827)*/,
230  -9,-6, -9,6/*mean (0.0126856), correlation (0.373938)*/,
231  -12,-10, -8,-4/*mean (0.0152497), correlation (0.364237)*/,
232  10,2, 12,-3/*mean (0.0299933), correlation (0.345292)*/,
233  7,12, 12,12/*mean (0.0307242), correlation (0.366299)*/,
234  -7,-13, -6,5/*mean (0.0534975), correlation (0.368357)*/,
235  -4,9, -3,4/*mean (0.099865), correlation (0.372276)*/,
236  7,-1, 12,2/*mean (0.117083), correlation (0.364529)*/,
237  -7,6, -5,1/*mean (0.126125), correlation (0.369606)*/,
238  -13,11, -12,5/*mean (0.130364), correlation (0.358502)*/,
239  -3,7, -2,-6/*mean (0.131691), correlation (0.375531)*/,
240  7,-8, 12,-7/*mean (0.160166), correlation (0.379508)*/,
241  -13,-7, -11,-12/*mean (0.167848), correlation (0.353343)*/,
242  1,-3, 12,12/*mean (0.183378), correlation (0.371916)*/,
243  2,-6, 3,0/*mean (0.228711), correlation (0.371761)*/,
244  -4,3, -2,-13/*mean (0.247211), correlation (0.364063)*/,
245  -1,-13, 1,9/*mean (0.249325), correlation (0.378139)*/,
246  7,1, 8,-6/*mean (0.000652272), correlation (0.411682)*/,
247  1,-1, 3,12/*mean (0.00248538), correlation (0.392988)*/,
248  9,1, 12,6/*mean (0.0206815), correlation (0.386106)*/,
249  -1,-9, -1,3/*mean (0.0364485), correlation (0.410752)*/,
250  -13,-13, -10,5/*mean (0.0376068), correlation (0.398374)*/,
251  7,7, 10,12/*mean (0.0424202), correlation (0.405663)*/,
252  12,-5, 12,9/*mean (0.0942645), correlation (0.410422)*/,
253  6,3, 7,11/*mean (0.1074), correlation (0.413224)*/,
254  5,-13, 6,10/*mean (0.109256), correlation (0.408646)*/,
255  2,-12, 2,3/*mean (0.131691), correlation (0.416076)*/,
256  3,8, 4,-6/*mean (0.165081), correlation (0.417569)*/,
257  2,6, 12,-13/*mean (0.171874), correlation (0.408471)*/,
258  9,-12, 10,3/*mean (0.175146), correlation (0.41296)*/,
259  -8,4, -7,9/*mean (0.183682), correlation (0.402956)*/,
260  -11,12, -4,-6/*mean (0.184672), correlation (0.416125)*/,
261  1,12, 2,-8/*mean (0.191487), correlation (0.386696)*/,
262  6,-9, 7,-4/*mean (0.192668), correlation (0.394771)*/,
263  2,3, 3,-2/*mean (0.200157), correlation (0.408303)*/,
264  6,3, 11,0/*mean (0.204588), correlation (0.411762)*/,
265  3,-3, 8,-8/*mean (0.205904), correlation (0.416294)*/,
266  7,8, 9,3/*mean (0.213237), correlation (0.409306)*/,
267  -11,-5, -6,-4/*mean (0.243444), correlation (0.395069)*/,
268  -10,11, -5,10/*mean (0.247672), correlation (0.413392)*/,
269  -5,-8, -3,12/*mean (0.24774), correlation (0.411416)*/,
270  -10,5, -9,0/*mean (0.00213675), correlation (0.454003)*/,
271  8,-1, 12,-6/*mean (0.0293635), correlation (0.455368)*/,
272  4,-6, 6,-11/*mean (0.0404971), correlation (0.457393)*/,
273  -10,12, -8,7/*mean (0.0481107), correlation (0.448364)*/,
274  4,-2, 6,7/*mean (0.050641), correlation (0.455019)*/,
275  -2,0, -2,12/*mean (0.0525978), correlation (0.44338)*/,
276  -5,-8, -5,2/*mean (0.0629667), correlation (0.457096)*/,
277  7,-6, 10,12/*mean (0.0653846), correlation (0.445623)*/,
278  -9,-13, -8,-8/*mean (0.0858749), correlation (0.449789)*/,
279  -5,-13, -5,-2/*mean (0.122402), correlation (0.450201)*/,
280  8,-8, 9,-13/*mean (0.125416), correlation (0.453224)*/,
281  -9,-11, -9,0/*mean (0.130128), correlation (0.458724)*/,
282  1,-8, 1,-2/*mean (0.132467), correlation (0.440133)*/,
283  7,-4, 9,1/*mean (0.132692), correlation (0.454)*/,
284  -2,1, -1,-4/*mean (0.135695), correlation (0.455739)*/,
285  11,-6, 12,-11/*mean (0.142904), correlation (0.446114)*/,
286  -12,-9, -6,4/*mean (0.146165), correlation (0.451473)*/,
287  3,7, 7,12/*mean (0.147627), correlation (0.456643)*/,
288  5,5, 10,8/*mean (0.152901), correlation (0.455036)*/,
289  0,-4, 2,8/*mean (0.167083), correlation (0.459315)*/,
290  -9,12, -5,-13/*mean (0.173234), correlation (0.454706)*/,
291  0,7, 2,12/*mean (0.18312), correlation (0.433855)*/,
292  -1,2, 1,7/*mean (0.185504), correlation (0.443838)*/,
293  5,11, 7,-9/*mean (0.185706), correlation (0.451123)*/,
294  3,5, 6,-8/*mean (0.188968), correlation (0.455808)*/,
295  -13,-4, -8,9/*mean (0.191667), correlation (0.459128)*/,
296  -5,9, -3,-3/*mean (0.193196), correlation (0.458364)*/,
297  -4,-7, -3,-12/*mean (0.196536), correlation (0.455782)*/,
298  6,5, 8,0/*mean (0.1972), correlation (0.450481)*/,
299  -7,6, -6,12/*mean (0.199438), correlation (0.458156)*/,
300  -13,6, -5,-2/*mean (0.211224), correlation (0.449548)*/,
301  1,-10, 3,10/*mean (0.211718), correlation (0.440606)*/,
302  4,1, 8,-4/*mean (0.213034), correlation (0.443177)*/,
303  -2,-2, 2,-13/*mean (0.234334), correlation (0.455304)*/,
304  2,-12, 12,12/*mean (0.235684), correlation (0.443436)*/,
305  -2,-13, 0,-6/*mean (0.237674), correlation (0.452525)*/,
306  4,1, 9,3/*mean (0.23962), correlation (0.444824)*/,
307  -6,-10, -3,-5/*mean (0.248459), correlation (0.439621)*/,
308  -3,-13, -1,1/*mean (0.249505), correlation (0.456666)*/,
309  7,5, 12,-11/*mean (0.00119208), correlation (0.495466)*/,
310  4,-2, 5,-7/*mean (0.00372245), correlation (0.484214)*/,
311  -13,9, -9,-5/*mean (0.00741116), correlation (0.499854)*/,
312  7,1, 8,6/*mean (0.0208952), correlation (0.499773)*/,
313  7,-8, 7,6/*mean (0.0220085), correlation (0.501609)*/,
314  -7,-4, -7,1/*mean (0.0233806), correlation (0.496568)*/,
315  -8,11, -7,-8/*mean (0.0236505), correlation (0.489719)*/,
316  -13,6, -12,-8/*mean (0.0268781), correlation (0.503487)*/,
317  2,4, 3,9/*mean (0.0323324), correlation (0.501938)*/,
318  10,-5, 12,3/*mean (0.0399235), correlation (0.494029)*/,
319  -6,-5, -6,7/*mean (0.0420153), correlation (0.486579)*/,
320  8,-3, 9,-8/*mean (0.0548021), correlation (0.484237)*/,
321  2,-12, 2,8/*mean (0.0616622), correlation (0.496642)*/,
322  -11,-2, -10,3/*mean (0.0627755), correlation (0.498563)*/,
323  -12,-13, -7,-9/*mean (0.0829622), correlation (0.495491)*/,
324  -11,0, -10,-5/*mean (0.0843342), correlation (0.487146)*/,
325  5,-3, 11,8/*mean (0.0929937), correlation (0.502315)*/,
326  -2,-13, -1,12/*mean (0.113327), correlation (0.48941)*/,
327  -1,-8, 0,9/*mean (0.132119), correlation (0.467268)*/,
328  -13,-11, -12,-5/*mean (0.136269), correlation (0.498771)*/,
329  -10,-2, -10,11/*mean (0.142173), correlation (0.498714)*/,
330  -3,9, -2,-13/*mean (0.144141), correlation (0.491973)*/,
331  2,-3, 3,2/*mean (0.14892), correlation (0.500782)*/,
332  -9,-13, -4,0/*mean (0.150371), correlation (0.498211)*/,
333  -4,6, -3,-10/*mean (0.152159), correlation (0.495547)*/,
334  -4,12, -2,-7/*mean (0.156152), correlation (0.496925)*/,
335  -6,-11, -4,9/*mean (0.15749), correlation (0.499222)*/,
336  6,-3, 6,11/*mean (0.159211), correlation (0.503821)*/,
337  -13,11, -5,5/*mean (0.162427), correlation (0.501907)*/,
338  11,11, 12,6/*mean (0.16652), correlation (0.497632)*/,
339  7,-5, 12,-2/*mean (0.169141), correlation (0.484474)*/,
340  -1,12, 0,7/*mean (0.169456), correlation (0.495339)*/,
341  -4,-8, -3,-2/*mean (0.171457), correlation (0.487251)*/,
342  -7,1, -6,7/*mean (0.175), correlation (0.500024)*/,
343  -13,-12, -8,-13/*mean (0.175866), correlation (0.497523)*/,
344  -7,-2, -6,-8/*mean (0.178273), correlation (0.501854)*/,
345  -8,5, -6,-9/*mean (0.181107), correlation (0.494888)*/,
346  -5,-1, -4,5/*mean (0.190227), correlation (0.482557)*/,
347  -13,7, -8,10/*mean (0.196739), correlation (0.496503)*/,
348  1,5, 5,-13/*mean (0.19973), correlation (0.499759)*/,
349  1,0, 10,-13/*mean (0.204465), correlation (0.49873)*/,
350  9,12, 10,-1/*mean (0.209334), correlation (0.49063)*/,
351  5,-8, 10,-9/*mean (0.211134), correlation (0.503011)*/,
352  -1,11, 1,-13/*mean (0.212), correlation (0.499414)*/,
353  -9,-3, -6,2/*mean (0.212168), correlation (0.480739)*/,
354  -1,-10, 1,12/*mean (0.212731), correlation (0.502523)*/,
355  -13,1, -8,-10/*mean (0.21327), correlation (0.489786)*/,
356  8,-11, 10,-6/*mean (0.214159), correlation (0.488246)*/,
357  2,-13, 3,-6/*mean (0.216993), correlation (0.50287)*/,
358  7,-13, 12,-9/*mean (0.223639), correlation (0.470502)*/,
359  -10,-10, -5,-7/*mean (0.224089), correlation (0.500852)*/,
360  -10,-8, -8,-13/*mean (0.228666), correlation (0.502629)*/,
361  4,-6, 8,5/*mean (0.22906), correlation (0.498305)*/,
362  3,12, 8,-13/*mean (0.233378), correlation (0.503825)*/,
363  -4,2, -3,-3/*mean (0.234323), correlation (0.476692)*/,
364  5,-13, 10,-12/*mean (0.236392), correlation (0.475462)*/,
365  4,-13, 5,-1/*mean (0.236842), correlation (0.504132)*/,
366  -9,9, -4,3/*mean (0.236977), correlation (0.497739)*/,
367  0,3, 3,-9/*mean (0.24314), correlation (0.499398)*/,
368  -12,1, -6,1/*mean (0.243297), correlation (0.489447)*/,
369  3,2, 4,-8/*mean (0.00155196), correlation (0.553496)*/,
370  -10,-10, -10,9/*mean (0.00239541), correlation (0.54297)*/,
371  8,-13, 12,12/*mean (0.0034413), correlation (0.544361)*/,
372  -8,-12, -6,-5/*mean (0.003565), correlation (0.551225)*/,
373  2,2, 3,7/*mean (0.00835583), correlation (0.55285)*/,
374  10,6, 11,-8/*mean (0.00885065), correlation (0.540913)*/,
375  6,8, 8,-12/*mean (0.0101552), correlation (0.551085)*/,
376  -7,10, -6,5/*mean (0.0102227), correlation (0.533635)*/,
377  -3,-9, -3,9/*mean (0.0110211), correlation (0.543121)*/,
378  -1,-13, -1,5/*mean (0.0113473), correlation (0.550173)*/,
379  -3,-7, -3,4/*mean (0.0140913), correlation (0.554774)*/,
380  -8,-2, -8,3/*mean (0.017049), correlation (0.55461)*/,
381  4,2, 12,12/*mean (0.01778), correlation (0.546921)*/,
382  2,-5, 3,11/*mean (0.0224022), correlation (0.549667)*/,
383  6,-9, 11,-13/*mean (0.029161), correlation (0.546295)*/,
384  3,-1, 7,12/*mean (0.0303081), correlation (0.548599)*/,
385  11,-1, 12,4/*mean (0.0355151), correlation (0.523943)*/,
386  -3,0, -3,6/*mean (0.0417904), correlation (0.543395)*/,
387  4,-11, 4,12/*mean (0.0487292), correlation (0.542818)*/,
388  2,-4, 2,1/*mean (0.0575124), correlation (0.554888)*/,
389  -10,-6, -8,1/*mean (0.0594242), correlation (0.544026)*/,
390  -13,7, -11,1/*mean (0.0597391), correlation (0.550524)*/,
391  -13,12, -11,-13/*mean (0.0608974), correlation (0.55383)*/,
392  6,0, 11,-13/*mean (0.065126), correlation (0.552006)*/,
393  0,-1, 1,4/*mean (0.074224), correlation (0.546372)*/,
394  -13,3, -9,-2/*mean (0.0808592), correlation (0.554875)*/,
395  -9,8, -6,-3/*mean (0.0883378), correlation (0.551178)*/,
396  -13,-6, -8,-2/*mean (0.0901035), correlation (0.548446)*/,
397  5,-9, 8,10/*mean (0.0949843), correlation (0.554694)*/,
398  2,7, 3,-9/*mean (0.0994152), correlation (0.550979)*/,
399  -1,-6, -1,-1/*mean (0.10045), correlation (0.552714)*/,
400  9,5, 11,-2/*mean (0.100686), correlation (0.552594)*/,
401  11,-3, 12,-8/*mean (0.101091), correlation (0.532394)*/,
402  3,0, 3,5/*mean (0.101147), correlation (0.525576)*/,
403  -1,4, 0,10/*mean (0.105263), correlation (0.531498)*/,
404  3,-6, 4,5/*mean (0.110785), correlation (0.540491)*/,
405  -13,0, -10,5/*mean (0.112798), correlation (0.536582)*/,
406  5,8, 12,11/*mean (0.114181), correlation (0.555793)*/,
407  8,9, 9,-6/*mean (0.117431), correlation (0.553763)*/,
408  7,-4, 8,-12/*mean (0.118522), correlation (0.553452)*/,
409  -10,4, -10,9/*mean (0.12094), correlation (0.554785)*/,
410  7,3, 12,4/*mean (0.122582), correlation (0.555825)*/,
411  9,-7, 10,-2/*mean (0.124978), correlation (0.549846)*/,
412  7,0, 12,-2/*mean (0.127002), correlation (0.537452)*/,
413  -1,-6, 0,-11/*mean (0.127148), correlation (0.547401)*/
414 };
415 
416 ORBextractor::ORBextractor(int _nfeatures, float _scaleFactor, int _nlevels,
417  int _iniThFAST, int _minThFAST, int _patchSize, int _edgeThreshold):
418  nfeatures(_nfeatures), scaleFactor(_scaleFactor), nlevels(_nlevels),
419  iniThFAST(_iniThFAST), minThFAST(_minThFAST), patchSize(_patchSize), edgeThreshold(_edgeThreshold)
420 {
422  mvScaleFactor.resize(nlevels);
423  mvLevelSigma2.resize(nlevels);
424  mvScaleFactor[0]=1.0f;
425  mvLevelSigma2[0]=1.0f;
426  for(int i=1; i<nlevels; i++)
427  {
430  }
431 
432  mvInvScaleFactor.resize(nlevels);
433  mvInvLevelSigma2.resize(nlevels);
434  for(int i=0; i<nlevels; i++)
435  {
436  mvInvScaleFactor[i]=1.0f/mvScaleFactor[i];
437  mvInvLevelSigma2[i]=1.0f/mvLevelSigma2[i];
438  }
439 
440  mvImagePyramid.resize(nlevels);
441 
442  mnFeaturesPerLevel.resize(nlevels);
443  float factor = 1.0f / scaleFactor;
444  float nDesiredFeaturesPerScale = nfeatures*(1 - factor)/(1 - (float)pow((double)factor, (double)nlevels));
445 
446  int sumFeatures = 0;
447  for( int level = 0; level < nlevels-1; level++ )
448  {
449  mnFeaturesPerLevel[level] = cvRound(nDesiredFeaturesPerScale);
450  sumFeatures += mnFeaturesPerLevel[level];
451  nDesiredFeaturesPerScale *= factor;
452  }
453  mnFeaturesPerLevel[nlevels-1] = std::max(nfeatures - sumFeatures, 0);
454 
455  const int npoints = 512;
456  const Point* pattern0 = (const Point*)bit_pattern_31_;
457  std::copy(pattern0, pattern0 + npoints, std::back_inserter(pattern));
458 
459  //This is for orientation
460  // pre-compute the end of a row in a circular patch
461  umax.resize(halfPatchSize + 1);
462 
463  int v, v0, vmax = cvFloor(float(halfPatchSize) * sqrt(2.f) / 2 + 1);
464  int vmin = cvCeil(float(halfPatchSize) * sqrt(2.f) / 2);
465  const double hp2 = halfPatchSize*halfPatchSize;
466  for (v = 0; v <= vmax; ++v)
467  umax[v] = cvRound(sqrt(hp2 - v * v));
468 
469  // Make sure we are symmetric
470  for (v = halfPatchSize, v0 = 0; v >= vmin; --v)
471  {
472  while (umax[v0] == umax[v0 + 1])
473  ++v0;
474  umax[v] = v0;
475  ++v0;
476  }
477 }
478 
479 static void computeOrientation(const Mat& image, vector<KeyPoint>& keypoints, const vector<int>& umax, int halfPatchSize)
480 {
481  for (vector<KeyPoint>::iterator keypoint = keypoints.begin(),
482  keypointEnd = keypoints.end(); keypoint != keypointEnd; ++keypoint)
483  {
484  keypoint->angle = IC_Angle(image, keypoint->pt, umax, halfPatchSize);
485  }
486 }
487 
489 {
490  const int halfX = ceil(static_cast<float>(UR.x-UL.x)/2);
491  const int halfY = ceil(static_cast<float>(BR.y-UL.y)/2);
492 
493  //Define boundaries of childs
494  n1.UL = UL;
495  n1.UR = cv::Point2i(UL.x+halfX,UL.y);
496  n1.BL = cv::Point2i(UL.x,UL.y+halfY);
497  n1.BR = cv::Point2i(UL.x+halfX,UL.y+halfY);
498  n1.vKeys.reserve(vKeys.size());
499 
500  n2.UL = n1.UR;
501  n2.UR = UR;
502  n2.BL = n1.BR;
503  n2.BR = cv::Point2i(UR.x,UL.y+halfY);
504  n2.vKeys.reserve(vKeys.size());
505 
506  n3.UL = n1.BL;
507  n3.UR = n1.BR;
508  n3.BL = BL;
509  n3.BR = cv::Point2i(n1.BR.x,BL.y);
510  n3.vKeys.reserve(vKeys.size());
511 
512  n4.UL = n3.UR;
513  n4.UR = n2.BR;
514  n4.BL = n3.BR;
515  n4.BR = BR;
516  n4.vKeys.reserve(vKeys.size());
517 
518  //Associate points to childs
519  for(size_t i=0;i<vKeys.size();i++)
520  {
521  const cv::KeyPoint &kp = vKeys[i];
522  if(kp.pt.x<n1.UR.x)
523  {
524  if(kp.pt.y<n1.BR.y)
525  n1.vKeys.push_back(kp);
526  else
527  n3.vKeys.push_back(kp);
528  }
529  else if(kp.pt.y<n1.BR.y)
530  n2.vKeys.push_back(kp);
531  else
532  n4.vKeys.push_back(kp);
533  }
534 
535  if(n1.vKeys.size()==1)
536  n1.bNoMore = true;
537  if(n2.vKeys.size()==1)
538  n2.bNoMore = true;
539  if(n3.vKeys.size()==1)
540  n3.bNoMore = true;
541  if(n4.vKeys.size()==1)
542  n4.bNoMore = true;
543 
544 }
545 
546 vector<cv::KeyPoint> ORBextractor::DistributeOctTree(const vector<cv::KeyPoint>& vToDistributeKeys, const int &minX,
547  const int &maxX, const int &minY, const int &maxY, const int &N, const int &level)
548 {
549  // Compute how many initial nodes
550  const int nIni = round(static_cast<float>(maxX-minX)/(maxY-minY));
551 
552  const float hX = static_cast<float>(maxX-minX)/nIni;
553 
554  list<ExtractorNode> lNodes;
555 
556  vector<ExtractorNode*> vpIniNodes;
557  vpIniNodes.resize(nIni);
558 
559  for(int i=0; i<nIni; i++)
560  {
561  ExtractorNode ni;
562  ni.UL = cv::Point2i(hX*static_cast<float>(i),0);
563  ni.UR = cv::Point2i(hX*static_cast<float>(i+1),0);
564  ni.BL = cv::Point2i(ni.UL.x,maxY-minY);
565  ni.BR = cv::Point2i(ni.UR.x,maxY-minY);
566  ni.vKeys.reserve(vToDistributeKeys.size());
567 
568  lNodes.push_back(ni);
569  vpIniNodes[i] = &lNodes.back();
570  }
571 
572  //Associate points to childs
573  for(size_t i=0;i<vToDistributeKeys.size();i++)
574  {
575  const cv::KeyPoint &kp = vToDistributeKeys[i];
576  vpIniNodes[kp.pt.x/hX]->vKeys.push_back(kp);
577  }
578 
579  list<ExtractorNode>::iterator lit = lNodes.begin();
580 
581  while(lit!=lNodes.end())
582  {
583  if(lit->vKeys.size()==1)
584  {
585  lit->bNoMore=true;
586  lit++;
587  }
588  else if(lit->vKeys.empty())
589  lit = lNodes.erase(lit);
590  else
591  lit++;
592  }
593 
594  bool bFinish = false;
595 
596  int iteration = 0;
597 
598  vector<pair<int,ExtractorNode*> > vSizeAndPointerToNode;
599  vSizeAndPointerToNode.reserve(lNodes.size()*4);
600 
601  while(!bFinish)
602  {
603  iteration++;
604 
605  int prevSize = lNodes.size();
606 
607  lit = lNodes.begin();
608 
609  int nToExpand = 0;
610 
611  vSizeAndPointerToNode.clear();
612 
613  while(lit!=lNodes.end())
614  {
615  if(lit->bNoMore)
616  {
617  // If node only contains one point do not subdivide and continue
618  lit++;
619  continue;
620  }
621  else
622  {
623  // If more than one point, subdivide
624  ExtractorNode n1,n2,n3,n4;
625  lit->DivideNode(n1,n2,n3,n4);
626 
627  // Add childs if they contain points
628  if(n1.vKeys.size()>0)
629  {
630  lNodes.push_front(n1);
631  if(n1.vKeys.size()>1)
632  {
633  nToExpand++;
634  vSizeAndPointerToNode.push_back(make_pair(n1.vKeys.size(),&lNodes.front()));
635  lNodes.front().lit = lNodes.begin();
636  }
637  }
638  if(n2.vKeys.size()>0)
639  {
640  lNodes.push_front(n2);
641  if(n2.vKeys.size()>1)
642  {
643  nToExpand++;
644  vSizeAndPointerToNode.push_back(make_pair(n2.vKeys.size(),&lNodes.front()));
645  lNodes.front().lit = lNodes.begin();
646  }
647  }
648  if(n3.vKeys.size()>0)
649  {
650  lNodes.push_front(n3);
651  if(n3.vKeys.size()>1)
652  {
653  nToExpand++;
654  vSizeAndPointerToNode.push_back(make_pair(n3.vKeys.size(),&lNodes.front()));
655  lNodes.front().lit = lNodes.begin();
656  }
657  }
658  if(n4.vKeys.size()>0)
659  {
660  lNodes.push_front(n4);
661  if(n4.vKeys.size()>1)
662  {
663  nToExpand++;
664  vSizeAndPointerToNode.push_back(make_pair(n4.vKeys.size(),&lNodes.front()));
665  lNodes.front().lit = lNodes.begin();
666  }
667  }
668 
669  lit=lNodes.erase(lit);
670  continue;
671  }
672  }
673 
674  // Finish if there are more nodes than required features
675  // or all nodes contain just one point
676  if((int)lNodes.size()>=N || (int)lNodes.size()==prevSize)
677  {
678  bFinish = true;
679  }
680  else if(((int)lNodes.size()+nToExpand*3)>N)
681  {
682 
683  while(!bFinish)
684  {
685 
686  prevSize = lNodes.size();
687 
688  vector<pair<int,ExtractorNode*> > vPrevSizeAndPointerToNode = vSizeAndPointerToNode;
689  vSizeAndPointerToNode.clear();
690 
691  sort(vPrevSizeAndPointerToNode.begin(),vPrevSizeAndPointerToNode.end());
692  for(int j=vPrevSizeAndPointerToNode.size()-1;j>=0;j--)
693  {
694  ExtractorNode n1,n2,n3,n4;
695  vPrevSizeAndPointerToNode[j].second->DivideNode(n1,n2,n3,n4);
696 
697  // Add childs if they contain points
698  if(n1.vKeys.size()>0)
699  {
700  lNodes.push_front(n1);
701  if(n1.vKeys.size()>1)
702  {
703  vSizeAndPointerToNode.push_back(make_pair(n1.vKeys.size(),&lNodes.front()));
704  lNodes.front().lit = lNodes.begin();
705  }
706  }
707  if(n2.vKeys.size()>0)
708  {
709  lNodes.push_front(n2);
710  if(n2.vKeys.size()>1)
711  {
712  vSizeAndPointerToNode.push_back(make_pair(n2.vKeys.size(),&lNodes.front()));
713  lNodes.front().lit = lNodes.begin();
714  }
715  }
716  if(n3.vKeys.size()>0)
717  {
718  lNodes.push_front(n3);
719  if(n3.vKeys.size()>1)
720  {
721  vSizeAndPointerToNode.push_back(make_pair(n3.vKeys.size(),&lNodes.front()));
722  lNodes.front().lit = lNodes.begin();
723  }
724  }
725  if(n4.vKeys.size()>0)
726  {
727  lNodes.push_front(n4);
728  if(n4.vKeys.size()>1)
729  {
730  vSizeAndPointerToNode.push_back(make_pair(n4.vKeys.size(),&lNodes.front()));
731  lNodes.front().lit = lNodes.begin();
732  }
733  }
734 
735  lNodes.erase(vPrevSizeAndPointerToNode[j].second->lit);
736 
737  if((int)lNodes.size()>=N)
738  break;
739  }
740 
741  if((int)lNodes.size()>=N || (int)lNodes.size()==prevSize)
742  bFinish = true;
743 
744  }
745  }
746  }
747 
748  // Retain the best point in each node
749  vector<cv::KeyPoint> vResultKeys;
750  vResultKeys.reserve(nfeatures);
751  for(list<ExtractorNode>::iterator lit=lNodes.begin(); lit!=lNodes.end(); lit++)
752  {
753  vector<cv::KeyPoint> &vNodeKeys = lit->vKeys;
754  cv::KeyPoint* pKP = &vNodeKeys[0];
755  float maxResponse = pKP->response;
756 
757  for(size_t k=1;k<vNodeKeys.size();k++)
758  {
759  if(vNodeKeys[k].response>maxResponse)
760  {
761  pKP = &vNodeKeys[k];
762  maxResponse = vNodeKeys[k].response;
763  }
764  }
765 
766  vResultKeys.push_back(*pKP);
767  }
768 
769  return vResultKeys;
770 }
771 
772 void ORBextractor::ComputeKeyPointsOctTree(vector<vector<KeyPoint> >& allKeypoints)
773 {
774  allKeypoints.resize(nlevels);
775 
776  const float W = 30;
777 
778  for (int level = 0; level < nlevels; ++level)
779  {
780  const int minBorderX = edgeThreshold-3;
781  const int minBorderY = minBorderX;
782  const int maxBorderX = mvImagePyramid[level].cols-edgeThreshold+3;
783  const int maxBorderY = mvImagePyramid[level].rows-edgeThreshold+3;
784 
785  vector<cv::KeyPoint> vToDistributeKeys;
786  vToDistributeKeys.reserve(nfeatures*10);
787 
788  const float width = (maxBorderX-minBorderX);
789  const float height = (maxBorderY-minBorderY);
790 
791  const int nCols = width/W;
792  const int nRows = height/W;
793  const int wCell = ceil(width/nCols);
794  const int hCell = ceil(height/nRows);
795 
796  for(int i=0; i<nRows; i++)
797  {
798  const float iniY =minBorderY+i*hCell;
799  float maxY = iniY+hCell+6;
800 
801  if(iniY>=maxBorderY-3)
802  continue;
803  if(maxY>maxBorderY)
804  maxY = maxBorderY;
805 
806  for(int j=0; j<nCols; j++)
807  {
808  const float iniX =minBorderX+j*wCell;
809  float maxX = iniX+wCell+6;
810  if(iniX>=maxBorderX-6)
811  continue;
812  if(maxX>maxBorderX)
813  maxX = maxBorderX;
814 
815  vector<cv::KeyPoint> vKeysCell;
816  FAST(mvImagePyramid[level].rowRange(iniY,maxY).colRange(iniX,maxX),
817  vKeysCell,iniThFAST,true);
818 
819  if(vKeysCell.empty())
820  {
821  FAST(mvImagePyramid[level].rowRange(iniY,maxY).colRange(iniX,maxX),
822  vKeysCell,minThFAST,true);
823  }
824 
825  if(!vKeysCell.empty())
826  {
827  for(vector<cv::KeyPoint>::iterator vit=vKeysCell.begin(); vit!=vKeysCell.end();vit++)
828  {
829  (*vit).pt.x+=j*wCell;
830  (*vit).pt.y+=i*hCell;
831  vToDistributeKeys.push_back(*vit);
832  }
833  }
834 
835  }
836  }
837 
838  vector<KeyPoint> & keypoints = allKeypoints[level];
839  keypoints.reserve(nfeatures);
840 
841  keypoints = DistributeOctTree(vToDistributeKeys, minBorderX, maxBorderX,
842  minBorderY, maxBorderY,mnFeaturesPerLevel[level], level);
843 
844  const int scaledPatchSize = patchSize*mvScaleFactor[level];
845 
846  // Add border to coordinates and scale information
847  const int nkps = keypoints.size();
848  for(int i=0; i<nkps ; i++)
849  {
850  keypoints[i].pt.x+=minBorderX;
851  keypoints[i].pt.y+=minBorderY;
852  keypoints[i].octave=level;
853  keypoints[i].size = scaledPatchSize;
854  }
855  }
856 
857  // compute orientations
858  for (int level = 0; level < nlevels; ++level)
859  computeOrientation(mvImagePyramid[level], allKeypoints[level], umax, halfPatchSize);
860 }
861 
862 void ORBextractor::ComputeKeyPointsOld(std::vector<std::vector<KeyPoint> > &allKeypoints)
863 {
864  allKeypoints.resize(nlevels);
865 
866  float imageRatio = (float)mvImagePyramid[0].cols/mvImagePyramid[0].rows;
867 
868  for (int level = 0; level < nlevels; ++level)
869  {
870  const int nDesiredFeatures = mnFeaturesPerLevel[level];
871 
872  const int levelCols = sqrt((float)nDesiredFeatures/(5*imageRatio));
873  const int levelRows = imageRatio*levelCols;
874 
875  const int minBorderX = edgeThreshold;
876  const int minBorderY = minBorderX;
877  const int maxBorderX = mvImagePyramid[level].cols-edgeThreshold;
878  const int maxBorderY = mvImagePyramid[level].rows-edgeThreshold;
879 
880  const int W = maxBorderX - minBorderX;
881  const int H = maxBorderY - minBorderY;
882  const int cellW = ceil((float)W/levelCols);
883  const int cellH = ceil((float)H/levelRows);
884 
885  const int nCells = levelRows*levelCols;
886  const int nfeaturesCell = ceil((float)nDesiredFeatures/nCells);
887 
888  vector<vector<vector<KeyPoint> > > cellKeyPoints(levelRows, vector<vector<KeyPoint> >(levelCols));
889 
890  vector<vector<int> > nToRetain(levelRows,vector<int>(levelCols,0));
891  vector<vector<int> > nTotal(levelRows,vector<int>(levelCols,0));
892  vector<vector<bool> > bNoMore(levelRows,vector<bool>(levelCols,false));
893  vector<int> iniXCol(levelCols);
894  vector<int> iniYRow(levelRows);
895  int nNoMore = 0;
896  int nToDistribute = 0;
897 
898 
899  float hY = cellH + 6;
900 
901  for(int i=0; i<levelRows; i++)
902  {
903  const float iniY = minBorderY + i*cellH - 3;
904  iniYRow[i] = iniY;
905 
906  if(i == levelRows-1)
907  {
908  hY = maxBorderY+3-iniY;
909  if(hY<=0)
910  continue;
911  }
912 
913  float hX = cellW + 6;
914 
915  for(int j=0; j<levelCols; j++)
916  {
917  float iniX;
918 
919  if(i==0)
920  {
921  iniX = minBorderX + j*cellW - 3;
922  iniXCol[j] = iniX;
923  }
924  else
925  {
926  iniX = iniXCol[j];
927  }
928 
929 
930  if(j == levelCols-1)
931  {
932  hX = maxBorderX+3-iniX;
933  if(hX<=0)
934  continue;
935  }
936 
937 
938  Mat cellImage = mvImagePyramid[level].rowRange(iniY,iniY+hY).colRange(iniX,iniX+hX);
939 
940  cellKeyPoints[i][j].reserve(nfeaturesCell*5);
941 
942  FAST(cellImage,cellKeyPoints[i][j],iniThFAST,true);
943 
944  if(cellKeyPoints[i][j].size()<=3)
945  {
946  cellKeyPoints[i][j].clear();
947 
948  FAST(cellImage,cellKeyPoints[i][j],minThFAST,true);
949  }
950 
951 
952  const int nKeys = cellKeyPoints[i][j].size();
953  nTotal[i][j] = nKeys;
954 
955  if(nKeys>nfeaturesCell)
956  {
957  nToRetain[i][j] = nfeaturesCell;
958  bNoMore[i][j] = false;
959  }
960  else
961  {
962  nToRetain[i][j] = nKeys;
963  nToDistribute += nfeaturesCell-nKeys;
964  bNoMore[i][j] = true;
965  nNoMore++;
966  }
967 
968  }
969  }
970 
971 
972  // Retain by score
973 
974  while(nToDistribute>0 && nNoMore<nCells)
975  {
976  int nNewFeaturesCell = nfeaturesCell + ceil((float)nToDistribute/(nCells-nNoMore));
977  nToDistribute = 0;
978 
979  for(int i=0; i<levelRows; i++)
980  {
981  for(int j=0; j<levelCols; j++)
982  {
983  if(!bNoMore[i][j])
984  {
985  if(nTotal[i][j]>nNewFeaturesCell)
986  {
987  nToRetain[i][j] = nNewFeaturesCell;
988  bNoMore[i][j] = false;
989  }
990  else
991  {
992  nToRetain[i][j] = nTotal[i][j];
993  nToDistribute += nNewFeaturesCell-nTotal[i][j];
994  bNoMore[i][j] = true;
995  nNoMore++;
996  }
997  }
998  }
999  }
1000  }
1001 
1002  vector<KeyPoint> & keypoints = allKeypoints[level];
1003  keypoints.reserve(nDesiredFeatures*2);
1004 
1005  const int scaledPatchSize = patchSize*mvScaleFactor[level];
1006 
1007  // Retain by score and transform coordinates
1008  for(int i=0; i<levelRows; i++)
1009  {
1010  for(int j=0; j<levelCols; j++)
1011  {
1012  vector<KeyPoint> &keysCell = cellKeyPoints[i][j];
1013  KeyPointsFilter::retainBest(keysCell,nToRetain[i][j]);
1014  if((int)keysCell.size()>nToRetain[i][j])
1015  keysCell.resize(nToRetain[i][j]);
1016 
1017 
1018  for(size_t k=0, kend=keysCell.size(); k<kend; k++)
1019  {
1020  keysCell[k].pt.x+=iniXCol[j];
1021  keysCell[k].pt.y+=iniYRow[i];
1022  keysCell[k].octave=level;
1023  keysCell[k].size = scaledPatchSize;
1024  keypoints.push_back(keysCell[k]);
1025  }
1026  }
1027  }
1028 
1029  if((int)keypoints.size()>nDesiredFeatures)
1030  {
1031  KeyPointsFilter::retainBest(keypoints,nDesiredFeatures);
1032  keypoints.resize(nDesiredFeatures);
1033  }
1034  }
1035 
1036  // and compute orientations
1037  for (int level = 0; level < nlevels; ++level)
1038  computeOrientation(mvImagePyramid[level], allKeypoints[level], umax, halfPatchSize);
1039 }
1040 
1041 static void computeDescriptors(const Mat& image, vector<KeyPoint>& keypoints, Mat& descriptors,
1042  const vector<Point>& pattern)
1043 {
1044  descriptors = Mat::zeros((int)keypoints.size(), 32, CV_8UC1);
1045 
1046  for (size_t i = 0; i < keypoints.size(); i++)
1047  computeOrbDescriptor(keypoints[i], image, &pattern[0], descriptors.ptr((int)i));
1048 }
1049 
1050 void ORBextractor::operator()( InputArray _image, InputArray _mask, vector<KeyPoint>& _keypoints,
1051  OutputArray _descriptors)
1052 {
1053  if(_image.empty())
1054  return;
1055 
1056  Mat image = _image.getMat();
1057  assert(image.type() == CV_8UC1 );
1058 
1059  // Pre-compute the scale pyramid
1060  ComputePyramid(image);
1061 
1062  vector < vector<KeyPoint> > allKeypoints;
1063  ComputeKeyPointsOctTree(allKeypoints);
1064  //ComputeKeyPointsOld(allKeypoints);
1065 
1066  Mat descriptors;
1067 
1068  int nkeypoints = 0;
1069  for (int level = 0; level < nlevels; ++level)
1070  nkeypoints += (int)allKeypoints[level].size();
1071  if( nkeypoints == 0 )
1072  _descriptors.release();
1073  else
1074  {
1075  _descriptors.create(nkeypoints, 32, CV_8U);
1076  descriptors = _descriptors.getMat();
1077  }
1078 
1079  _keypoints.clear();
1080  _keypoints.reserve(nkeypoints);
1081 
1082  int offset = 0;
1083  for (int level = 0; level < nlevels; ++level)
1084  {
1085  vector<KeyPoint>& keypoints = allKeypoints[level];
1086  int nkeypointsLevel = (int)keypoints.size();
1087 
1088  if(nkeypointsLevel==0)
1089  continue;
1090 
1091  // preprocess the resized image
1092  Mat workingMat = mvImagePyramid[level].clone();
1093  GaussianBlur(workingMat, workingMat, Size(7, 7), 2, 2, BORDER_REFLECT_101);
1094 
1095  // Compute the descriptors
1096  Mat desc = descriptors.rowRange(offset, offset + nkeypointsLevel);
1097  computeDescriptors(workingMat, keypoints, desc, pattern);
1098 
1099  offset += nkeypointsLevel;
1100 
1101  // Scale keypoint coordinates
1102  if (level != 0)
1103  {
1104  float scale = mvScaleFactor[level]; //getScale(level, firstLevel, scaleFactor);
1105  for (vector<KeyPoint>::iterator keypoint = keypoints.begin(),
1106  keypointEnd = keypoints.end(); keypoint != keypointEnd; ++keypoint)
1107  keypoint->pt *= scale;
1108  }
1109  // And add the keypoints to the output
1110  _keypoints.insert(_keypoints.end(), keypoints.begin(), keypoints.end());
1111  }
1112 }
1113 
1114 void ORBextractor::ComputePyramid(cv::Mat image)
1115 {
1116  for (int level = 0; level < nlevels; ++level)
1117  {
1118  float scale = mvInvScaleFactor[level];
1119  Size sz(cvRound((float)image.cols*scale), cvRound((float)image.rows*scale));
1120  Size wholeSize(sz.width + edgeThreshold*2, sz.height + edgeThreshold*2);
1121  Mat temp(wholeSize, image.type()), masktemp;
1122  mvImagePyramid[level] = temp(Rect(edgeThreshold, edgeThreshold, sz.width, sz.height));
1123 
1124  // Compute the resized image
1125  if( level != 0 )
1126  {
1127  resize(mvImagePyramid[level-1], mvImagePyramid[level], sz, 0, 0, INTER_LINEAR);
1128 
1129  copyMakeBorder(mvImagePyramid[level], temp, edgeThreshold, edgeThreshold, edgeThreshold, edgeThreshold,
1130  BORDER_REFLECT_101+BORDER_ISOLATED);
1131  }
1132  else
1133  {
1134  copyMakeBorder(image, temp, edgeThreshold, edgeThreshold, edgeThreshold, edgeThreshold,
1135  BORDER_REFLECT_101);
1136  }
1137  }
1138 
1139 }
1140 
1141 } //namespace rtabmap
d
std::vector< float > mvInvScaleFactor
Definition: ORBextractor.h:120
static void computeDescriptors(const Mat &image, std::vector< KeyPoint > &keypoints, Mat &descriptors, const std::vector< Point > &pattern, int dsize, int WTA_K)
Definition: Orb.cpp:716
std::vector< float > mvInvLevelSigma2
Definition: ORBextractor.h:122
GLM_FUNC_DECL genType round(genType const &x)
static void computeOrbDescriptor(const KeyPoint &kpt, const Mat &img, const Point *pattern, uchar *desc)
std::vector< cv::KeyPoint > DistributeOctTree(const std::vector< cv::KeyPoint > &vToDistributeKeys, const int &minX, const int &maxX, const int &minY, const int &maxY, const int &nFeatures, const int &level)
f
GLM_FUNC_DECL vecType< T, P > sqrt(vecType< T, P > const &x)
void operator()(cv::InputArray image, cv::InputArray mask, std::vector< cv::KeyPoint > &keypoints, cv::OutputArray descriptors)
unsigned char uchar
Definition: matrix.h:41
std::vector< float > mvScaleFactor
Definition: ORBextractor.h:119
GLM_FUNC_DECL detail::tmat4x4< T, P > scale(detail::tmat4x4< T, P > const &m, detail::tvec3< T, P > const &v)
std::vector< cv::Point > pattern
Definition: ORBextractor.h:103
GLM_FUNC_DECL T angle(detail::tquat< T, P > const &x)
static int bit_pattern_31_[256 *4]
Definition: Features2d.h:41
void DivideNode(ExtractorNode &n1, ExtractorNode &n2, ExtractorNode &n3, ExtractorNode &n4)
GLM_FUNC_DECL genType step(genType const &edge, genType const &x)
GLM_FUNC_DECL genType cos(genType const &angle)
void ComputeKeyPointsOctTree(std::vector< std::vector< cv::KeyPoint > > &allKeypoints)
GLM_FUNC_DECL genType sin(genType const &angle)
const float factorPI
void ComputeKeyPointsOld(std::vector< std::vector< cv::KeyPoint > > &allKeypoints)
std::vector< cv::Mat > mvImagePyramid
Definition: ORBextractor.h:93
void ComputePyramid(cv::Mat image)
std::vector< int > umax
Definition: ORBextractor.h:117
static void computeOrbDescriptor(const KeyPoint &kpt, const Mat &img, const Point *pattern, uchar *desc, int dsize, int WTA_K)
Definition: Orb.cpp:138
GLM_FUNC_DECL genType max(genType const &x, genType const &y)
GLM_FUNC_DECL genType pow(genType const &base, genType const &exponent)
static bool computeOrientation(Eigen::Vector3f A, Eigen::Vector3f E, Eigen::Quaternionf &orientation)
static float IC_Angle(const Mat &image, const int half_k, Point2f pt, const std::vector< int > &u_max)
Definition: Orb.cpp:106
std::vector< int > mnFeaturesPerLevel
Definition: ORBextractor.h:115
GLM_FUNC_DECL genType ceil(genType const &x)
static float IC_Angle(const Mat &image, Point2f pt, const vector< int > &u_max, int halfPatchSize)
Definition: ORBextractor.cc:83
std::vector< cv::KeyPoint > vKeys
Definition: ORBextractor.h:47
std::vector< float > mvLevelSigma2
Definition: ORBextractor.h:121
#define GET_VALUE(idx)


rtabmap
Author(s): Mathieu Labbe
autogenerated on Mon Dec 14 2020 03:34:59