-
Notifications
You must be signed in to change notification settings - Fork 0
/
Tracker.cpp
2829 lines (2383 loc) · 92 KB
/
Tracker.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#include "stdafx.h"
#include <vector>
#include <fstream>
#include <sstream>
#include <iostream>
#include <algorithm>
#include <Eigen/Core>
#include <Eigen/LU>
#include <chrono>
#include<cmath>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include "Tracker.h"
//定义外部变量——在其它地方已被定义，在这里要使用这个变量
extern std::vector<char*> frame_files;
extern std::vector<char*> frame_files_thermal;
extern std::string sequence_name1;
Tracker::Tracker(Config &config) :
/// Constant Setting
SEQUENCE_NAME(config.sequence_name),
SEQUENCE_PATH(config.sequence_path),
/*****************自己加的热红外*******************/
SEQUENCE_PATH_THERMAL(config.sequence_path_thermal),
/*****************自己加的热红外*******************/
RESULT_PATH(config.result_path),
IMAGE_TYPE(config.image_type),
IMAGE_TYPE_THERMAL(config.image_type_thermal),//新加入的thermal
INIT_FRAME(config.init_frame),
END_FRAME(config.end_frame),
NUM_FRAME(END_FRAME - INIT_FRAME + 1),
//NUM_FRAME(1),
NUM_CHANNEL(config.num_channel + 1),
PATCH_DIM(NUM_CHANNEL*CHANNEL_DIM),
OBJECT_DIM(NUM_PATCH*PATCH_DIM),
BBOX_DIM(1 * PATCH_DIM), // 整个目标的维度，即patch的维度（32维度）,目标变成整个bounding box
INIT_BOX(config.init_bbox)
{
/// Variable Setting
patch_w = config.patch_width;
patch_h = config.patch_height;
scale_w = config.scale_width;
scale_h = config.scale_height;
search_r = config.search_radius;
bbox_w = config.bbox_width; //全局的
bbox_h = config.bbox_height;
is_scale = false;
char image_name[100];
image = cv::imread(frame_files[0], IMAGE_TYPE);
cv::resize(image, image, cv::Size(), 1 / scale_w, 1 / scale_h);
cv::copyMakeBorder(image, image, patch_h, patch_h, patch_w, patch_w, cv::BORDER_CONSTANT, cv::Scalar());
/***********自己添加的表示热红外图像********/
image_thermal = cv::imread(frame_files_thermal[0], IMAGE_TYPE_THERMAL);
if (SEQUENCE_NAME == "occBike")
{
cv::copyMakeBorder(image_thermal, image_thermal, 0, 8, 0, 16, cv::BORDER_CONSTANT, cv::Scalar());
}
cv::resize(image_thermal, image_thermal, cv::Size(), 1 / scale_w, 1 / scale_h);
cv::copyMakeBorder(image_thermal, image_thermal, patch_h, patch_h, patch_w, patch_w, cv::BORDER_CONSTANT, cv::Scalar());
/***********自己添加的表示热红外图像********/
border_bbox.set(0, 0, image.cols - 1, image.rows - 1);
image_bbox.set(patch_w, patch_h, border_bbox.w - 2 * patch_w, border_bbox.h - 2 * patch_h);
object_bbox.x = std::round(config.init_bbox.x / scale_w) + patch_w;
object_bbox.y = std::round(config.init_bbox.y / scale_h) + patch_h;
object_bbox.w = std::round(config.init_bbox.w / scale_w);
object_bbox.h = std::round(config.init_bbox.h / scale_h);
if (object_bbox.x < 0)
object_bbox.x = 0;
if (object_bbox.x + object_bbox.w > image_bbox.x + image_bbox.w)
object_bbox.x = image_bbox.x + image_bbox.w - object_bbox.w;
if (object_bbox.y < 0)
object_bbox.y = 0;
if (object_bbox.y + object_bbox.h > image_bbox.y + image_bbox.h)
object_bbox.y = image_bbox.y + image_bbox.h - object_bbox.h;
result_box.resize(NUM_FRAME, Rect());
/********自己添加的 结果包围盒我没有用同一个*********/
result_box_thermal.resize(NUM_FRAME, Rect());//是否可以考虑使用同一个？
/********自己添加的 结果包围盒我没有用同一个*********/
}
Tracker::~Tracker() {};
void Tracker::display_result(int t, std::string window_name, std::string window_name_thermal)
{
int frame_id = t + INIT_FRAME;
//char image_name[100];
// sprintf_s(image_name, 100, "%05d.jpg", frame_id);
//image = cv::imread(SEQUENCE_PATH + image_name, cv::IMREAD_COLOR);
image = cv::imread(frame_files[frame_id-1], IMAGE_TYPE);
cv::rectangle(image,
cv::Rect((int)result_box[t].x, (int)result_box[t].y, (int)result_box[t].w, (int)result_box[t].h),
CV_RGB(255, 255, 0),
2);
// sprintf_s(image_name, 100, "#%d", frame_id);
// cv::putText(image, image_name, cvPoint(0, 60), 2, 2, CV_RGB(255, 255, 0), 3, 0);
cv::imshow(window_name, image);
//char buf[100];
//sprintf_s(buf, "d:\\imgs\\%04d.jpg", frame_id);
//cv::imwrite(buf, image);
image_thermal = cv::imread(frame_files_thermal[frame_id-1], IMAGE_TYPE_THERMAL);
cv::rectangle(image_thermal,
cv::Rect((int)result_box[t].x, (int)result_box[t].y, (int)result_box[t].w, (int)result_box[t].h),
CV_RGB(255, 255, 0),
2);
//cv::imshow(window_name_thermal, image_thermal);
cv::waitKey(1);
}
void Tracker::run()
{
std::cout << "[Sequence] " << SEQUENCE_NAME << '\n';
std::cout << "-------------------------------------------- \n";
/// Initialize
std::cout << "Start initialization\n";
initialize();
display_result(0, sequence_name1, sequence_name1);
//cv::waitKey();
std::cout << "Complete initialization \n";
std::cout << "-------------------------------------------- \n";
/// track
std::cout << "Start tracking\n";
//initialize the scale model
initial_scale_model();
auto t0 = std::chrono::high_resolution_clock::now();
for (int t=1; t<NUM_FRAME; ++t)
{
int frame_id = t + INIT_FRAME;
//std::cout << frame_id << std::endl;
//if (frame_id == 43)
//{
// std::cout << frame_id << std::endl;
//}
track(frame_id);
display_result(t, sequence_name1, sequence_name1);
frame_idx = frame_id;
}
auto t1 = std::chrono::high_resolution_clock::now();
auto fps = (double)(NUM_FRAME-1.0) / std::chrono::duration_cast<std::chrono::seconds>(t1-t0).count();
std::cout << "FPS: " << fps << "\n";
std::cout << "Complete tracking \n";
std::cout << "-------------------------------------------- \n\n";
}
void Tracker::save(std::string cur_path)
{
//std::ofstream ofs(cur_path + "\\" + RESULT_PATH + "\\bbox\\" + SEQUENCE_NAME + "_ours.txt");// RGB-T
std::ofstream ofs(RESULT_PATH + "\\bbox\\" + "Ours_"+SEQUENCE_NAME + ".txt");
for (int i=0; i<result_box.size(); ++i)
{
char temp[100];
sprintf_s(temp, 100, "%.2lf %.2lf %.2lf %.2f %.2lf %.2lf %.2lf %.2f",
result_box[i].x + 1, result_box[i].y + 1,
result_box[i].x + 1 + result_box[i].w, result_box[i].y + 1,
result_box[i].x + 1 + result_box[i].w, result_box[i].y + 1+result_box[i].h,
result_box[i].x + 1 , result_box[i].y + 1 + result_box[i].h); // RGB-T
//sprintf_s(temp, 100, "%.2lf,%.2lf,%.2lf,%.2f",
// result_box[i].x + 1, result_box[i].y + 1, result_box[i].x + result_box[i].w + 1, result_box[i].y + result_box[i].h + 1); //RGB-D
ofs << temp << '\n';
}
ofs.close();
std::cout << "finish write result_box!" << std::endl;
//std::ofstream ofs1(RESULT_PATH + "\\query\\" + SEQUENCE_NAME + "_query.txt");
//for (int i = 0; i < NUM_FRAME; i++)
//{
// ofs1 << query_each_seq[i].transpose() << "\n";
//}
//ofs1.close();
//std::cout << "finish write query!" << std::endl;
//std::ofstream ofs2(RESULT_PATH + "\\Q\\" + SEQUENCE_NAME + "_q.txt");
//for (int i = 0; i < NUM_FRAME; i++)
//{
// ofs2 << Q_each_seq[i].transpose() << "\n";
//}
//ofs2.close();
//std::cout << "finish write Q!" << std::endl;
//std::ofstream ofs3(RESULT_PATH + "\\weight\\" + SEQUENCE_NAME + "_weight.txt");
////std::cout << "weight_each_seq: " << weight_each_seq.size() << std::endl;
//for (int i = 0; i < NUM_FRAME; i++)
//{
// ofs3 << weight_each_seq[i].transpose() << " ";
// ofs3 << "\n";
//}
//ofs3.close();
//std::cout << "finish write weight!" << std::endl;
for (int i = 0; i < MAXITER; i++)
{
iter_errs[i] /= result_box.size();
}
std::ofstream ofs4(RESULT_PATH + "\\iter_errs\\" + SEQUENCE_NAME + "_iter_errs.txt");
for (int i = 0; i < MAXITER; i++)
{
ofs4 << iter_errs[i] << '\n';
}
ofs4.close();
std::cout << "finish write iter_errs!" << std::endl;
////std::ofstream ofs5(cur_path + "\\" + RESULT_PATH + "\\W\\" + SEQUENCE_NAME + "_W_v.txt");
////std::ofstream ofs5_1(cur_path + "\\" + RESULT_PATH + "\\W\\" + SEQUENCE_NAME + "_W_thermal.txt");
////for (int i = 0; i < NUM_FRAME; i++)
////{
//// ofs5 << W_each_seq[i] << '\n';
//// ofs5_1 << W_each_seq_thermal[i] << '\n';
////}
////ofs5.close();
////ofs5_1.close();
////std::cout << "finish write W!" << std::endl;
//std::ofstream ofs6(RESULT_PATH + "\\S\\" + SEQUENCE_NAME + "_S.txt");
//for (int i = 0; i < NUM_FRAME; i++)
//{
// ofs6 << S_each_seq[i] << '\n';
//}
//ofs6.close();
//std::cout << "finish write S!" << std::endl;
//std::ofstream ofs7(RESULT_PATH + "\\R\\" + SEQUENCE_NAME + "_R.txt");
//for (int i = 0; i < NUM_FRAME; i++)
//{
// ofs7 << R_each_seq[i] << '\n';
//}
//ofs7.close();
//std::cout << "finish write R!" << std::endl << std::endl << std::endl;
query_each_seq.clear();
Q_each_seq.clear();
W_each_seq.clear();
W_each_seq_thermal.clear();
S_each_seq.clear();
R_each_seq.clear();
//patch_weight.clear();
}
void Tracker::initialize()
{
iter_errs.resize(MAXITER);
mu_errs.resize(MAXITER);
std::cout << "- initialize feature map\n";
update_feature_map(INIT_FRAME);
std::cout << "- initialize patch weight\n";
initialize_mask();
initialize_seed();
update_patch_weight();
std::cout << "- initialize classifier\n";
update_classifier(true);
update_result_box(INIT_FRAME);
}
void Tracker::initial_scale_model()
{// caculate the global variable scaleFactors、min_scale_factor、max_scale_factor
int nScales = numScales;
scaleFactors.resize(nScales, 0.0);
double scale_step = Scale_step;
for (int numi = 1; numi <= nScales; numi++)
{
scaleFactors[numi - 1] = pow(scale_step, ceil(nScales / 2.0) - numi);
//cout << "scale_model_temp = " << scale_model_temp << endl;
}
if (feature_bbox.w > feature_bbox.h) {
min_scale_factor = pow(scale_step, ceil(log(5 / feature_bbox.h) / log(scale_step)));
}
else {
min_scale_factor = pow(scale_step, ceil(log(5 / feature_bbox.w) / log(scale_step)));
}
if (image_bbox.w / object_bbox.w > image_bbox.h / object_bbox.h) {
max_scale_factor = pow(scale_step, floor(log(image_bbox.h / object_bbox.h) / log(scale_step)));
}
else {
max_scale_factor = pow(scale_step, floor(log(image_bbox.w / object_bbox.w) / log(scale_step)));
}
min_scale_factor *= 6;
double PI = 3.1415926, tmp = 0.0;
scale_window.resize(nScales, 0.0);
scale_window[0] = 0.00001;//hann()窗正常求值应为0，但0作用特征后输入到分类器做运算结果异常
for (int N = 1; N < nScales; N++)
{
tmp = 0.5 - 0.5*cos(2 * PI * N / (nScales - 1));
if (N == nScales / 2)
tmp = 1.0;
scale_window[N] = tmp;
//scale_window[N] = (scale_window[N] + 2) / 3;//可提高对尺度敏感性
//cout << N << ':' << scale_window[N] << endl;
}
}
void Tracker::update_scale()
{
// 等比例缩放scale_w,scale_h,search_r；但是patch_w、patch_h不变
if (best_bbox.w < best_bbox.h)
{
scale_w = best_bbox.w / 32.0;
patch_w = std::round(best_bbox.w / (8.0*scale_w));
patch_h = std::round(best_bbox.h / (8.0*scale_w));
scale_h = best_bbox.h / (8.0*patch_h);
}
else
{
scale_h = best_bbox.h / 32.0;
patch_h = std::round(best_bbox.h / (8.0*scale_h));
patch_w = std::round(best_bbox.w / (8.0*scale_h));
scale_w = best_bbox.w / (8.0*patch_w);
}
search_r = sqrt(best_bbox.w*best_bbox.h / (scale_w*scale_h));//搜索区域
}
void Tracker::track(int frame_id)
{
//clock_t t0 = clock();
update_feature_map(frame_id);
//clock_t t1 = clock();
//std::cout << "feature: " << t1 - t0 << std::endl;
bool is_updated = update_object_box(frame_id);
//std::cout << is_updated << std::endl;
update_result_box(frame_id); //在此之后修改新的scale_w、 scale_h 和 search_r.
if (is_updated)
{
if (is_scale)
{
update_scale(); //实时判断并更新原始图片的缩放尺度 scale_w/scale_h/search_r
//update object_bbox resize(image) image_bbox feature_bbox border_bbox
//char image_name[100];
//sprintf_s(image_name, 100, "%04d.jpg", frame_id);
//image = cv::imread(SEQUENCE_PATH + image_name, IMAGE_TYPE);
image = cv::imread(frame_files[frame_id - 1], IMAGE_TYPE);
cv::resize(image, image, cv::Size(), 1 / scale_w, 1 / scale_h);
cv::copyMakeBorder(image, image, patch_h, patch_h, patch_w, patch_w, cv::BORDER_CONSTANT, cv::Scalar());
border_bbox.set(0, 0, image.cols - 1, image.rows - 1);
image_bbox.set(patch_w, patch_h, border_bbox.w - 2 * patch_w, border_bbox.h - 2 * patch_h);
object_bbox.x = std::round(best_bbox.x / scale_w) + patch_w;
object_bbox.y = std::round(best_bbox.y / scale_h) + patch_h;
object_bbox.w = std::round(best_bbox.w / scale_w);
object_bbox.h = std::round(best_bbox.h / scale_h);
if (object_bbox.x < 0)
object_bbox.x = 0;
if (object_bbox.x + object_bbox.w > image_bbox.x + image_bbox.w)
object_bbox.x = image_bbox.x + image_bbox.w - object_bbox.w;
if (object_bbox.y < 0)
object_bbox.y = 0;
if (object_bbox.y + object_bbox.h > image_bbox.y + image_bbox.h)
object_bbox.y = image_bbox.y + image_bbox.h - object_bbox.h;
//2016-09-28
//Update patch_mask[64] & expand_patch_mask[100]
initialize_mask();
//由于跟踪目标的尺度发生变化，分类器更新之前，必须先更新scale/object_bbox等参数
update_feature_map(frame_id);
}
//clock_t t0 = clock();
update_patch_weight();
//clock_t t1 = clock();
//std::cout << "weight: " << t1 - t0 << std::endl;
//t0 = clock();
update_classifier();
//t1 = clock();
//std::cout << "update: " << t1 - t0 << std::endl;
}
/*update_result_box(frame_id);*/
}
void Tracker::update_result_box(int frame_id)
{
int t = frame_id - INIT_FRAME;
if (is_scale)
{
result_box[t].x = best_bbox.x;
result_box[t].y = best_bbox.y;
result_box[t].w = best_bbox.w;
result_box[t].h = best_bbox.h;
//由于跟踪目标尺度发生变化，所以进入下一帧前 object_bbox 等也要更新，但前提是先计算新的scale_w,scale_h,search_r等尺度参数
}
else
{
result_box[t].x = (object_bbox.x - patch_w)*scale_w;
result_box[t].y = (object_bbox.y - patch_h)*scale_h;
result_box[t].w = object_bbox.w*scale_w;
result_box[t].h = object_bbox.h*scale_h;
}
// std::cout << "is_Scale : " << is_scale << " best_bbox: " << best_bbox.x << " " << best_bbox.y << " " << best_bbox.w << " " << best_bbox.h << std::endl;
// std::cout << "is_Scale : " << is_scale << " result_bbox: " << result_box[t].x + 1 << " " << result_box[t].y + 1 << " " << result_box[t].w << " " << result_box[t].h << std::endl;
}
void Tracker::initialize_mask()
{
patch_mask.clear();
std::vector<Rect> patch = extract_patch(object_bbox);
for (int i=0; i<patch.size(); ++i)
patch_mask.push_back(Rect(patch[i].x-object_bbox.x,
patch[i].y-object_bbox.y,
patch[i].w,
patch[i].h));
expanded_patch_mask.clear();
std::vector<Rect> expanded_patch = extract_expanded_patch(object_bbox);
for (int i=0; i<expanded_patch.size(); ++i)
expanded_patch_mask.push_back(Rect(expanded_patch[i].x-object_bbox.x,
expanded_patch[i].y-object_bbox.y,
expanded_patch[i].w,
expanded_patch[i].h));
patch_mask_thermal.clear();
std::vector<Rect> patch_thermal = extract_patch(object_bbox);
for (int i = 0; i<patch_thermal.size(); ++i)
patch_mask_thermal.push_back(Rect(patch_thermal[i].x - object_bbox.x,
patch_thermal[i].y - object_bbox.y,
patch_thermal[i].w,
patch_thermal[i].h));
expanded_patch_mask_thermal.clear();
std::vector<Rect> expanded_patch_thermal = extract_expanded_patch(object_bbox);
for (int i = 0; i < expanded_patch_thermal.size(); ++i)
expanded_patch_mask_thermal.push_back(Rect(expanded_patch_thermal[i].x - object_bbox.x,
expanded_patch_thermal[i].y - object_bbox.y,
expanded_patch_thermal[i].w,
expanded_patch_thermal[i].h));
//bbox_mask.clear();
//bbox_mask.push_back(Rect(object_bbox.x, object_bbox.y, bbox_w, bbox_h)); //object_bbox.w object_bbox.h
bbox_mask = Rect(object_bbox.x, object_bbox.y, object_bbox.w, object_bbox.h); //object_bbox.w object_bbox.h bbox_w, bbox_h
}
void Tracker::initialize_seed()
{
std::vector<Rect> expanded_patch = extract_expanded_patch(object_bbox, expanded_patch_mask);
patch_weight_v.resize(NUM_PATCH, 0.0);
patch_weight_i.resize(NUM_PATCH, 0.0);
}
Eigen::MatrixXd Tracker::getMarkMatrix(Eigen::MatrixXd scoreMap, int h, int w)
{
// get the peaks in the scoreMap
double maxTmp = 0;
bool flag_tmp;
Eigen::MatrixXd markMatrix = Eigen::MatrixXd::Zero(h, w);
for (int i = 0; i < scoreMap.cols(); i++)
{
for (int j = 0; j < scoreMap.rows(); j++)
{
if (i == 0 && j == 0) //左上角
{
double neighPoint1 = scoreMap(i, j + 1); //右
double neighPoint2 = scoreMap(i + 1, j); //下
double neighPoint3 = scoreMap(i + 1, j + 1); // 右下
flag_tmp = scoreMap(i, j) > neighPoint1 && scoreMap(i, j) > neighPoint2 && scoreMap(i, j) > neighPoint3;
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (i == 0 && j == w - 1) //右上角
{
double neighPoint1 = scoreMap(i, j - 1); //左
double neighPoint2 = scoreMap(i + 1, j - 1); // 左下
double neighPoint3 = scoreMap(i + 1, j); //下
flag_tmp = scoreMap(i, j) > neighPoint1 && scoreMap(i, j) > neighPoint2 && scoreMap(i, j) > neighPoint3;
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (i == h - 1 && j == 0) //左下角
{
double neighPoint1 = scoreMap(i - 1, j); //上
double neighPoint2 = scoreMap(i, j + 1); // 右
double neighPoint3 = scoreMap(i - 1, j + 1); //右上
flag_tmp = scoreMap(i, j) > neighPoint1 && scoreMap(i, j) > neighPoint2 && scoreMap(i, j) > neighPoint3;
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (i == h - 1 && j == w - 1) //右下角
{
double neighPoint1 = scoreMap(i - 1, j); //上
double neighPoint2 = scoreMap(i, j - 1); //左
double neighPoint3 = scoreMap(i - 1, j - 1); //左上
flag_tmp = scoreMap(i, j) > neighPoint1 && scoreMap(i, j) > neighPoint2 && scoreMap(i, j) > neighPoint3;
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (i == 0 && j != 0 && j != w - 1) //上边框
{
double neighPoint1 = scoreMap(i, j - 1);//左
double neighPoint2 = scoreMap(i + 1, j - 1); //左下
double neighPoint3 = scoreMap(i + 1, j); //下
double neighPoint4 = scoreMap(i + 1, j + 1); //右下
double neighPoint5 = scoreMap(i, j + 1);//右
flag_tmp = (scoreMap(i, j) > neighPoint1) && (scoreMap(i, j) > neighPoint2) && (scoreMap(i, j) > neighPoint3) && (scoreMap(i, j) > neighPoint4) && (scoreMap(i, j) > neighPoint5);
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (i == h - 1 && j != 0 && j != w - 1) //下边框
{
double neighPoint1 = scoreMap(i, j - 1);//左
double neighPoint2 = scoreMap(i - 1, j - 1); //左上
double neighPoint3 = scoreMap(i - 1, j); //上
double neighPoint4 = scoreMap(i - 1, j + 1); //右上
double neighPoint5 = scoreMap(i, j + 1);//右
flag_tmp = (scoreMap(i, j) > neighPoint1) && (scoreMap(i, j) > neighPoint2) && (scoreMap(i, j) > neighPoint3) && (scoreMap(i, j) > neighPoint4) && (scoreMap(i, j) > neighPoint5);
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (j == 0 && i != 0 && i != h - 1)//左边框
{
double neighPoint1 = scoreMap(i - 1, j);//上
double neighPoint2 = scoreMap(i - 1, j + 1); //右上
double neighPoint3 = scoreMap(i, j + 1); //右
double neighPoint4 = scoreMap(i + 1, j + 1); //右下
double neighPoint5 = scoreMap(i + 1, j);//下
flag_tmp = (scoreMap(i, j) > neighPoint1) && (scoreMap(i, j) > neighPoint2) && (scoreMap(i, j) > neighPoint3) && (scoreMap(i, j) > neighPoint4) && (scoreMap(i, j) > neighPoint5);
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else if (j == w - 1 && i != 0 && i != h - 1) //右边框
{
double neighPoint1 = scoreMap(i - 1, j);//上
double neighPoint2 = scoreMap(i - 1, j - 1); //左上
double neighPoint3 = scoreMap(i, j - 1); //左
double neighPoint4 = scoreMap(i + 1, j - 1); //左下
double neighPoint5 = scoreMap(i + 1, j);//下
flag_tmp = (scoreMap(i, j) > neighPoint1) && (scoreMap(i, j) > neighPoint2) && (scoreMap(i, j) > neighPoint3) && (scoreMap(i, j) > neighPoint4) && (scoreMap(i, j) > neighPoint5);
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
else //中间
{
double neighPoint1 = scoreMap(i, j - 1);//左
double neighPoint2 = scoreMap(i - 1, j - 1); //左上
double neighPoint3 = scoreMap(i - 1, j); //上
double neighPoint4 = scoreMap(i - 1, j + 1); //右上
double neighPoint5 = scoreMap(i, j + 1);//右
double neighPoint6 = scoreMap(i + 1, j + 1); //右下
double neighPoint7 = scoreMap(i + 1, j); //下
double neighPoint8 = scoreMap(i + 1, j - 1);//左下
flag_tmp = (scoreMap(i, j) > neighPoint1) && (scoreMap(i, j) > neighPoint2) && (scoreMap(i, j) > neighPoint3) && (scoreMap(i, j) > neighPoint4) && (scoreMap(i, j) > neighPoint5) && (scoreMap(i, j) > neighPoint6) && (scoreMap(i, j) > neighPoint7) && (scoreMap(i, j) > neighPoint8);
markMatrix(i, j) = flag_tmp ? 1 : 0;
}
}
}
return markMatrix;
}
bool Tracker::isOverBoundry(int index_x, int index_y, int x_shift, int y_shift)
{
bool flag = false; //默认未超出边界
int x = index_x - search_r + x_shift;
int y = index_y - search_r + y_shift;
if (x < -32 || x > 32 || y < -32 || y > 32)
{
flag = true;
}
return flag;
}
bool Tracker::isTruePeak(Eigen::MatrixXd scoreMap, double peak, int index_x, int index_y)
{
int winSize = 5;
int xShift[5] = { -2, -1, 0, 1, 2 };
int yShift[5] = { -2, -1, 0, 1, 2 };
//int x_new, y_new;
//double maxPeak = 0;
double neighPoint;
bool flag = false; // 判断是否越界
bool isPeak = false; // 判断是否是真正的peak
for (int i = 0; i < winSize; i++)
{
for (int j = 0; j < winSize; j++)
{
// 判断flag
flag = isOverBoundry(index_x, index_y, xShift[i], yShift[j]);
if (!flag)
{
neighPoint = scoreMap(index_x + xShift[i], index_y + yShift[j]);
// 只要有一个邻域值比该值大，就移除此极值点
if (peak <= neighPoint)
isPeak = false;
else
isPeak = true;
}
}
}
return isPeak;
}
bool Tracker::update_object_box(int frame_id)
{
bool is_updated = false;
Rect sample(object_bbox);
Rect best_sample(object_bbox);
double best_score = -DBL_MAX;
//多峰检测
/*double maxPeak = 0;
int maxPeak_x;
int maxPeak_y;
double secPeak = 0;
Eigen::MatrixXd scoreMap = Eigen::MatrixXd::Zero(2 * search_r + 1, 2 * search_r + 1);
Eigen::MatrixXd markB = Eigen::MatrixXd::Zero(2 * search_r + 1, 2 * search_r + 1);
Eigen::MatrixXd markB_new = Eigen::MatrixXd::Zero(2 * search_r + 1, 2 * search_r + 1);
Eigen::MatrixXd optScoreMap_first = Eigen::MatrixXd::Zero(2 * search_r + 1, 2 * search_r + 1);
Eigen::MatrixXd optScoreMap_second = Eigen::MatrixXd::Zero(2 * search_r + 1, 2 * search_r + 1);
bool isPeak;*/
for (int iy = -search_r; iy <= search_r; ++iy)
{
for (int ix = -search_r; ix <= search_r; ++ix)
{
sample.x = (int)object_bbox.x + ix;
sample.y = (int)object_bbox.y + iy;
if (!sample.is_inside(image_bbox))
continue;
Eigen::VectorXd sample_feature = extract_test_feature(sample);
double score = (1 - OMEGA)*classifier.test(sample_feature) + OMEGA*classifier0.test(sample_feature);
if (score > best_score)
{
best_score = score;
best_sample.set(sample);
}
//scoreMap(iy + search_r, ix + search_r) = score;
}
}
//markB = getMarkMatrix(scoreMap, 2 * search_r + 1, 2 * search_r + 1);
////std::ofstream ofs_score("D:\\scoreMap.txt");
////ofs_score << scoreMap;
////ofs_score.close();
////int temp = 0;
//// 找所有局部极值点 并确定其是否为局部值
//for (int i = 0; i < scoreMap.rows(); i++)
//{
// for (int j = 0; j < scoreMap.cols(); j++)
// {
// optScoreMap_first(i, j) = scoreMap(i, j) * markB(i, j);
// optScoreMap_second(i, j) = optScoreMap_first(i, j);
// if (optScoreMap_first(i, j) >= maxPeak)
// {
// maxPeak = optScoreMap_first(i, j);
// maxPeak_x = j;//i;
// maxPeak_y = i;//j;
// }
// if (optScoreMap_first(i, j) != 0)
// {
// //temp++;
// isPeak = isTruePeak(scoreMap, optScoreMap_first(i, j), i, j);
// //std::cout << temp << " Peak: " << isPeak << "\n";
// if (!isPeak)
// {
// optScoreMap_second(i, j) = 0;
// }
// }
// }
//}
//// 找第二大值
//for (int i = 0; i < scoreMap.rows(); i++)
//{
// for (int j = 0; j < scoreMap.cols(); j++)
// {
// if (optScoreMap_second(i, j) != maxPeak)
// {
// secPeak = secPeak > optScoreMap_second(i, j) ? secPeak : optScoreMap_second(i, j);
// }
// }
//}
//// 判断是否为全局峰值
////if ((maxPeak - secPeak) > PeakThreshold)
////std::cout << "maxPeak: " << maxPeak << " secPeak: " << secPeak << " ratio: " << secPeak / maxPeak << "\n";
//
Eigen::VectorXd best_sample_feature = extract_test_feature(best_sample);
double validation_score = classifier.validation_test(best_sample_feature);
//std::cout << "validation_score: " << validation_score << std::endl;
if (validation_score > THETA)
//if (validation_score > THETA && !((secPeak * 1.0 / maxPeak) > PeakThreshold))
{
object_temp_bbox.x = (best_sample.x - patch_w)*scale_w;
object_temp_bbox.y = (best_sample.y - patch_h)*scale_h;
object_temp_bbox.w = best_sample.w*scale_w;
object_temp_bbox.h = best_sample.h*scale_h;
//std::cout << "best:" << best_bbox.x << " " << best_bbox.y << " " << best_bbox.w << " " << best_bbox.h << " " << std::endl;
//先恢复到原图大小，在原图分辨率上做尺度变化
if (frame_id % interval == 0)
scale_estimation(frame_id, object_temp_bbox);
double max_scale_score = 0;
int num = 0, max_scale_score_num = 0;
double w_temp = 0, h_temp = 0;
double min_w_temp = 0, min_h_temp = 0;
double rate_temp;
if (validation_score_s.size())
{
max_scale_score = *validation_score_s.begin();
std::vector<double>::iterator iter;
for (iter = validation_score_s.begin(); iter != validation_score_s.end(); iter++)
{
if (max_scale_score < *iter && num != (numScales - 1) / 2)
{
max_scale_score = *iter;
max_scale_score_num = num;
}
num++;
}
}
//std::cout << frame_id << '/' << NUM_FRAME << ": " << validation_score << ',' << max_scale_score << std::endl;
if (!validation_score_s.empty())
{
validation_score = validation_score_s[int((numScales - 1) / 2)];//由于特征金字塔里倍数为1.0时的分类器得分（因为新的方式里特征提取后做归一化时与原先归一化结果不同）
}
//if (frame_id % interval == 0)
// std::cout << frame_id << '/' << NUM_FRAME << ": " << validation_score << ',' << max_scale_score << std::endl;
if (max_scale_score - validation_score > IsScale_Threshold)
{
// std::cout << "isScale: True" << std::endl;
is_scale = true;
//best_bbox = best_sample_s.at(max_scale_score_num);
//访问容器里的元素，若使用at()，需执行范围检查，如果参数无效，at()就会抛出一个std::out_of_range异常
best_bbox = best_sample_s[max_scale_score_num];
//std::cout << "best:" << best_bbox.x << " " << best_bbox.y << " " << best_bbox.w << " " << best_bbox.h << " " << std::endl;
currentScaleFactor *= scaleFactors[max_scale_score_num];
if (currentScaleFactor < min_scale_factor)
{
currentScaleFactor = min_scale_factor;
is_scale = false;
object_bbox.set(best_sample); // non-scale variety
}
else
if (currentScaleFactor > max_scale_factor)
{
currentScaleFactor = max_scale_factor;
is_scale = false;
object_bbox.set(best_sample); // non-scale variety
}
}
//cout << best_bbox.x << ' ' << best_bbox.y << ' ' << best_bbox.w << ' ' << best_bbox.h << endl;
//然后，比较 best_sample 和 best_sample_s 分类器得分
else
{
is_scale = false;
object_bbox.set(best_sample); // non-scale variety
/*is_updated = true;*/
}
//进入下一帧之前，清空 validation_score_s 和 best_sample_s 容器里的内容
validation_score_s.clear();
best_sample_s.clear();
is_updated = true;
}
//std::cout << is_updated << std::endl;
return is_updated;
}
Eigen::VectorXd Tracker::extract_test_feature_s(cv::Mat &expand_roi, cv::Mat &expand_roi_thermal)
{
//Eigen::VectorXd feature(Eigen::VectorXd::Zero(2 * OBJECT_DIM));//新加入thermal特征，特征维度*2
Eigen::VectorXd feature(Eigen::VectorXd::Zero(2 * OBJECT_DIM + 2 * BBOX_DIM));//新加入thermal特征，特征维度*2
for (int i = 0; i<NUM_PATCH; ++i)
{
int x_min = 0 + patch_mask[i].x;
int y_min = 0 + patch_mask[i].y;
feature.segment(i*PATCH_DIM * 2, PATCH_DIM) = patch_weight_v[i] * feature_map_s[expand_roi.cols*y_min + x_min];
feature.segment(i*PATCH_DIM * 2 + PATCH_DIM, PATCH_DIM) = patch_weight_i[i] * feature_map_thermal_s[expand_roi_thermal.cols*y_min + x_min];//新加入thermal特征
}
Rect r_bbox(object_bbox.x, object_bbox.y, object_bbox.w, object_bbox.h);
feature.segment(OBJECT_DIM * 2, BBOX_DIM) = global_ration*feature_map_bbox_s[expand_roi.cols*(expand_roi.rows-1) + expand_roi.cols - 1];
feature.segment(OBJECT_DIM * 2 + BBOX_DIM, BBOX_DIM) = global_ration*feature_map_bbox_thermal_s[expand_roi_thermal.cols * (expand_roi_thermal.rows - 1) + expand_roi_thermal.cols - 1];
//feature.segment(OBJECT_DIM * 2, BBOX_DIM) = global_ration*feature_map_bbox_s[expand_roi.cols*bbox_mask.y + bbox_mask.x];
//feature.segment(OBJECT_DIM * 2 + BBOX_DIM, BBOX_DIM) = global_ration*feature_map_bbox_thermal_s[expand_roi_thermal.cols * bbox_mask.y + bbox_mask.x];
feature.normalize();// 向量归一化？？？ 与在搜索窗口里固定尺度对当前位置提取的特征（可能）不一样！！！
return feature;
}
void Tracker::compute_color_histogram_map_s(cv::Mat &expand_roi1, cv::Mat &expand_roi_thermal)
{
double bin_size = 32.0;
int num_color_channel = NUM_CHANNEL - 1;
for (int i = 0; i<num_color_channel; ++i)
{
cv::Mat tmp(expand_roi1.rows, expand_roi1.cols, CV_8UC1);
tmp.setTo(0);
for (int j = 0; j<CHANNEL_DIM; ++j)
{
for (int y = 0; y<expand_roi1.rows; ++y)
{
const uchar* src = image_channel_s[i].ptr(y);
uchar* dst = tmp.ptr(y);
for (int x = 0; x<expand_roi1.cols; ++x)
{
int bin = (int)((double)(*src) / bin_size);
*dst = (bin == j) ? 1 : 0;
++src;
++dst;
}
}
cv::integral(tmp, integ_hist_s[i*CHANNEL_DIM + j]);
}
/************** 新加入的热红外thermal **********/
cv::Mat tmp_thermal(expand_roi_thermal.rows, expand_roi_thermal.cols, CV_8UC1);
tmp_thermal.setTo(0);
for (int j = 0; j<CHANNEL_DIM; ++j)
{
for (int y = 0; y<expand_roi_thermal.rows; ++y)
{
const uchar* src_thermal = image_channel_thermal_s[i].ptr(y);
uchar* dst_thermal = tmp_thermal.ptr(y);
for (int x = 0; x<expand_roi_thermal.cols; ++x)
{
int bin_thermal = (int)((double)(*src_thermal) / bin_size);
*dst_thermal = (bin_thermal == j) ? 1 : 0;
++src_thermal;
++dst_thermal;
}
}
cv::integral(tmp_thermal, integ_hist_thermal_s[i*CHANNEL_DIM + j]);
}
/************** 新加入的热红外thermal **********/
}
}
void Tracker::compute_gradient_histogram_map_s(cv::Mat &expand_roi1, cv::Mat &expand_roi_thermal)
{
float bin_size = 22.5;
float radian_to_degree = 180.0 / CV_PI;
cv::Mat gray_image(expand_roi1.rows, expand_roi1.cols, CV_8UC1);
if (IMAGE_TYPE == cv::IMREAD_COLOR)
cv::cvtColor(expand_roi1, gray_image, CV_BGR2GRAY);
else
expand_roi1.copyTo(gray_image);
cv::Mat x_sobel, y_sobel;
cv::Sobel(gray_image, x_sobel, CV_32FC1, 1, 0);
cv::Sobel(gray_image, y_sobel, CV_32FC1, 0, 1);
std::vector<cv::Mat> bins;
for (int i = 0; i<CHANNEL_DIM; ++i)
bins.push_back(cv::Mat::zeros(expand_roi1.rows, expand_roi1.cols, CV_32FC1));
for (int y = 0; y<expand_roi1.rows; ++y)
{
float* x_sobel_row_ptr = (float*)(x_sobel.row(y).data);
float* y_sobel_row_ptr = (float*)(y_sobel.row(y).data);
std::vector<float*> bins_row_ptrs(CHANNEL_DIM, nullptr);
for (int i = 0; i<CHANNEL_DIM; ++i)
bins_row_ptrs[i] = (float*)(bins[i].row(y).data);
for (int x = 0; x<expand_roi1.cols; ++x)
{
if (x_sobel_row_ptr[x] == 0)
x_sobel_row_ptr[x] += 0.00001;
float orientation = atan(y_sobel_row_ptr[x] / x_sobel_row_ptr[x])*radian_to_degree + 90;
float magnitude = sqrt(x_sobel_row_ptr[x] * x_sobel_row_ptr[x] + y_sobel_row_ptr[x] * y_sobel_row_ptr[x]);
for (int i = 1; i<CHANNEL_DIM; ++i)
{
if (orientation <= bin_size*i)
{
bins_row_ptrs[i - 1][x] = magnitude;
break;
}
}
}
}
/************** 新加入的thermal **************/
cv::Mat gray_image_thermal(expand_roi_thermal.rows, expand_roi_thermal.cols, CV_8UC1);
if (IMAGE_TYPE == cv::IMREAD_COLOR)
cv::cvtColor(expand_roi_thermal, gray_image_thermal, CV_BGR2GRAY);
else
expand_roi_thermal.copyTo(gray_image_thermal);
cv::Mat x_sobel_thermal, y_sobel_thermal;
cv::Sobel(gray_image_thermal, x_sobel_thermal, CV_32FC1, 1, 0);
cv::Sobel(gray_image_thermal, y_sobel_thermal, CV_32FC1, 0, 1);
std::vector<cv::Mat> bins_thermal;
for (int i = 0; i<CHANNEL_DIM; ++i)
bins_thermal.push_back(cv::Mat::zeros(expand_roi_thermal.rows, expand_roi_thermal.cols, CV_32FC1));
for (int y = 0; y<expand_roi_thermal.rows; ++y)
{
float* x_sobel_row_ptr_thermal = (float*)(x_sobel_thermal.row(y).data);
float* y_sobel_row_ptr_thermal = (float*)(y_sobel_thermal.row(y).data);
std::vector<float*> bins_row_ptrs_thermal(CHANNEL_DIM, nullptr);
for (int i = 0; i<CHANNEL_DIM; ++i)
bins_row_ptrs_thermal[i] = (float*)(bins_thermal[i].row(y).data);
for (int x = 0; x<expand_roi_thermal.cols; ++x)
{
if (x_sobel_row_ptr_thermal[x] == 0)
x_sobel_row_ptr_thermal[x] += 0.00001;
float orientation = atan(y_sobel_row_ptr_thermal[x] / x_sobel_row_ptr_thermal[x])*radian_to_degree + 90;
float magnitude = sqrt(x_sobel_row_ptr_thermal[x] * x_sobel_row_ptr_thermal[x] + y_sobel_row_ptr_thermal[x] * y_sobel_row_ptr_thermal[x]);
for (int i = 1; i<CHANNEL_DIM; ++i)
{
if (orientation <= bin_size*i)
{
bins_row_ptrs_thermal[i - 1][x] = magnitude;
break;
}
}
}
}
/************** 新加入的thermal **************/
int color_dim = PATCH_DIM - CHANNEL_DIM;
for (int i = 0; i<CHANNEL_DIM; ++i)
{
cv::integral(bins[i], integ_hist_s[color_dim + i]);
cv::integral(bins_thermal[i], integ_hist_thermal_s[color_dim + i]);
}
}
Eigen::VectorXd Tracker::extract_patch_feature_s(int x_min, int y_min, int x_max, int y_max)
{
Eigen::VectorXd feature(PATCH_DIM);
int color_dim = PATCH_DIM - CHANNEL_DIM;
double patch_area = patch_w*patch_h;
for (int i = 0; i<color_dim; ++i)
{
double sum = integ_hist_s[i].at<int>(y_min, x_min)
+ integ_hist_s[i].at<int>(y_max, x_max)
- integ_hist_s[i].at<int>(y_max, x_min)
- integ_hist_s[i].at<int>(y_min, x_max);
//if (frame_idx == 97)
// cout << "sum = " << sum << ';' << x_min << ',' << y_min << ',' << x_max << ',' << y_max << endl;