forked from MasteringOpenCV/code
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathOFFeatureMatcher.cpp
executable file
·157 lines (140 loc) · 4.68 KB
/
OFFeatureMatcher.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
/*****************************************************************************
* ExploringSfMWithOpenCV
******************************************************************************
* by Roy Shilkrot, 5th Dec 2012
* http://www.morethantechnical.com/
******************************************************************************
* Ch4 of the book "Mastering OpenCV with Practical Computer Vision Projects"
* Copyright Packt Publishing 2012.
* http://www.packtpub.com/cool-projects-with-opencv/book
*****************************************************************************/
#include "OFFeatureMatcher.h"
#include <opencv2/video/video.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/gpu/gpu.hpp>
#include <opencv2/flann/flann.hpp>
#ifdef __SFM__DEBUG__
#include <opencv2/highgui/highgui.hpp>
#include <omp.h>
#include <sstream>
#endif
#include <set>
using namespace std;
using namespace cv;
//c'tor
OFFeatureMatcher::OFFeatureMatcher(
bool _use_gpu,
std::vector<cv::Mat>& imgs_,
std::vector<std::vector<cv::KeyPoint> >& imgpts_) :
AbstractFeatureMatcher(_use_gpu),imgpts(imgpts_), imgs(imgs_)
{
//detect keypoints for all images
FastFeatureDetector ffd;
// DenseFeatureDetector ffd;
ffd.detect(imgs, imgpts);
}
void OFFeatureMatcher::MatchFeatures(int idx_i, int idx_j, vector<DMatch>* matches) {
vector<Point2f> i_pts;
KeyPointsToPoints(imgpts[idx_i],i_pts);
vector<Point2f> j_pts(i_pts.size());
// making sure images are grayscale
Mat prevgray,gray;
if (imgs[idx_i].channels() == 3) {
cvtColor(imgs[idx_i],prevgray,CV_RGB2GRAY);
cvtColor(imgs[idx_j],gray,CV_RGB2GRAY);
} else {
prevgray = imgs[idx_i];
gray = imgs[idx_j];
}
vector<uchar> vstatus(i_pts.size()); vector<float> verror(i_pts.size());
if(use_gpu) {
gpu::GpuMat gpu_prevImg,gpu_nextImg,gpu_prevPts,gpu_nextPts,gpu_status,gpu_error;
gpu_prevImg.upload(prevgray);
gpu_nextImg.upload(gray);
gpu_prevPts.upload(Mat(i_pts).t());
gpu::PyrLKOpticalFlow gpu_of;
gpu_of.sparse(gpu_prevImg,gpu_nextImg,gpu_prevPts,gpu_nextPts,gpu_status,&gpu_error);
Mat j_pts_mat;
gpu_nextPts.download(j_pts_mat);
Mat(j_pts_mat.t()).copyTo(Mat(j_pts));
Mat vstatus_mat,verror_mat;
gpu_status.download(vstatus_mat);
gpu_error.download(verror_mat);
Mat(vstatus_mat.t()).copyTo(Mat(vstatus));
Mat(verror_mat.t()).copyTo(Mat(verror));
} else {
CV_PROFILE("OpticalFlow",calcOpticalFlowPyrLK(prevgray, gray, i_pts, j_pts, vstatus, verror);)
}
double thresh = 1.0;
vector<Point2f> to_find;
vector<int> to_find_back_idx;
for (unsigned int i=0; i<vstatus.size(); i++) {
if (vstatus[i] && verror[i] < 12.0) {
to_find_back_idx.push_back(i);
to_find.push_back(j_pts[i]);
} else {
vstatus[i] = 0;
}
}
std::set<int> found_in_imgpts_j;
Mat to_find_flat = Mat(to_find).reshape(1,to_find.size());
vector<Point2f> j_pts_to_find;
KeyPointsToPoints(imgpts[idx_j],j_pts_to_find);
Mat j_pts_flat = Mat(j_pts_to_find).reshape(1,j_pts_to_find.size());
vector<vector<DMatch> > knn_matches;
//FlannBasedMatcher matcher;
BFMatcher matcher(CV_L2);
CV_PROFILE("RadiusMatch",matcher.radiusMatch(to_find_flat,j_pts_flat,knn_matches,2.0f);)
CV_PROFILE("Prune",
for(int i=0;i<knn_matches.size();i++) {
DMatch _m;
if(knn_matches[i].size()==1) {
_m = knn_matches[i][0];
} else if(knn_matches[i].size()>1) {
if(knn_matches[i][0].distance / knn_matches[i][1].distance < 0.7) {
_m = knn_matches[i][0];
} else {
continue; // did not pass ratio test
}
} else {
continue; // no match
}
if (found_in_imgpts_j.find(_m.trainIdx) == found_in_imgpts_j.end()) { // prevent duplicates
_m.queryIdx = to_find_back_idx[_m.queryIdx]; //back to original indexing of points for <i_idx>
matches->push_back(_m);
found_in_imgpts_j.insert(_m.trainIdx);
}
}
)
cout << "pruned " << matches->size() << " / " << knn_matches.size() << " matches" << endl;
#if 0
#ifdef __SFM__DEBUG__
{
// draw flow field
Mat img_matches; cvtColor(imgs[idx_i],img_matches,CV_GRAY2BGR);
i_pts.clear(); j_pts.clear();
for(int i=0;i<matches->size();i++) {
//if (i%2 != 0) {
// continue;
// }
Point i_pt = imgpts[idx_i][(*matches)[i].queryIdx].pt;
Point j_pt = imgpts[idx_j][(*matches)[i].trainIdx].pt;
i_pts.push_back(i_pt);
j_pts.push_back(j_pt);
vstatus[i] = 1;
}
drawArrows(img_matches, i_pts, j_pts, vstatus, verror, Scalar(0,255));
stringstream ss;
ss << matches->size() << " matches";
// putText(img_matches,ss.str(),Point(10,20),CV_FONT_HERSHEY_PLAIN,1.0,Scalar(255),2);
ss.clear(); ss << "flow_field_" << omp_get_thread_num() << ".png";
imshow( ss.str(), img_matches );
int c = waitKey(0);
if (c=='s') {
imwrite(ss.str(), img_matches);
}
destroyWindow(ss.str());
}
#endif
#endif
}