-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsurfmatch.cpp
158 lines (147 loc) · 5.34 KB
/
surfmatch.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
#include "pch.h"
#include <iostream>
#include <opencv2/cudaarithm.hpp>
#include <opencv2/cudafilters.hpp>
#include <opencv2/xfeatures2d/cuda.hpp>
#include <opencv2/cudafeatures2d.hpp>
#include "surfmatch.h"
using namespace std;
using namespace cv::xfeatures2d;
/****** surf 匹配 vector<Point2f>********/
int scale = 3; //显示缩放
void surf_match(const cv::Mat& img1, const cv::Mat& img2, cv::Mat& uvmark) {
cv::cuda::GpuMat gmat1(3648, 5472, CV_8UC1);//创建一个加载图片的空gpumat
cv::cuda::GpuMat gmat2(3648, 5472, CV_8UC1);
cv::cuda::GpuMat gmat01;
cv::cuda::GpuMat gmat02;
gmat01.upload(img1);
gmat02.upload(img2);
//ROI 块赋值
/*gmat01(cv::Rect(3087, 513, 1193, 1130)).copyTo(gmat1(cv::Rect(3087, 513, 1193, 1130)));
gmat02(cv::Rect(2187, 493, 1311, 1130)).copyTo(gmat2(cv::Rect(2187, 493, 1311, 1130)));*/
gmat01.copyTo(gmat1);
gmat02.copyTo(gmat2);
/*gmat1.download(imgdemo);
cv::namedWindow("img1", 0);
cv::resizeWindow("img1", (int)(img1.size().width / 10), (int)(img1.size().height / 10));
cv::imshow("img1", img1);
cv::namedWindow("demo", 0);
cv::resizeWindow("demo", (int)(img1.size().width / 10), (int)(img1.size().height / 10));
cv::imshow("demo", imgdemo);
cv::waitKey(0);
cv::destroyAllWindows();*/
/*下面这个函数的原型是：
explicit SURF_CUDA(double
_hessianThreshold, //SURF海森特征点阈值
int _nOctaves=4, //尺度金字塔个数
int _nOctaveLayers=2, //每一个尺度金字塔层数
bool _extended=false, //如果true那么得到的描述子是128维，否则是64维
float _keypointsRatio=0.01f,
bool _upright = false
);
要理解这几个参数涉及SURF的原理*/
cv::cuda::SURF_CUDA surf(10, 4, 3);
/*分配下面几个GpuMat存储keypoint和相应的descriptor*/
cv::cuda::GpuMat keypt1, keypt2;
cv::cuda::GpuMat desc1, desc2;
/*检测特征点*/
surf(gmat1, cv::cuda::GpuMat(), keypt1, desc1);
surf(gmat2, cv::cuda::GpuMat(), keypt2, desc2);
/*匹配，下面的匹配部分和CPU的match没有太多区别,这里新建一个Brute-Force Matcher，一对descriptor的L2距离小于0.1则认为匹配*/
auto matcher = cv::cuda::DescriptorMatcher::createBFMatcher(cv::NORM_L2);
//vector< vector< DMatch> > match_vec;
vector<cv::DMatch> match_vec;
matcher->match(desc1, desc2, match_vec);
/*int count = 0;
for (auto & d : match_vec) {
if (d.distance < 0.1)
count++;
}*/
// downloading results Gpu -> Cpu
vector< cv::KeyPoint> keypoints1, keypoints2;
vector< float> descriptors1, descriptors2;
surf.downloadKeypoints(keypt1, keypoints1);
surf.downloadKeypoints(keypt2, keypoints2);
surf.downloadDescriptors(desc1, descriptors1);
surf.downloadDescriptors(desc2, descriptors2);
cout << "surf done" << endl;
int ptcount = (int)match_vec.size();
cv::Mat p1(ptcount, 2, CV_32F);
cv::Mat p2(ptcount, 2, CV_32F);
//change keypoint to mat
cv::Point2f pt;
vector<cv::Point2f> p01, p02;
for (int i = 0; i < ptcount; i++)
{
pt = keypoints1[match_vec[i].queryIdx].pt;
p01.push_back(pt);
p1.at<float>(i, 0) = pt.x;
p1.at<float>(i, 1) = pt.y;
pt = keypoints2[match_vec[i].trainIdx].pt;
p02.push_back(pt);
p2.at<float>(i, 0) = pt.x;
p2.at<float>(i, 1) = pt.y;
}
//use RANSAC to calculate F
cv::Mat fundamental;
vector <uchar> RANSACStatus;
fundamental = findFundamentalMat(p1, p2, RANSACStatus, cv::FM_RANSAC, 10, 0.99);
//下面的代码为RANSAC优化后的特征点匹配效果
//calculate the number of outliner
int outlinerCount = 0;
for (int i = 0; i < ptcount; i++)
{
if (RANSACStatus[i] == 0)
outlinerCount++;
}
//calculate inLiner
vector<cv::Point2f> inliner1, inliner2;
vector<cv::DMatch> inlierMatches;
int inlinerCount = ptcount - outlinerCount;
inliner1.resize(inlinerCount);
inliner2.resize(inlinerCount);
inlierMatches.resize(inlinerCount);
int inlinerMatchesCount = 0;
for (int i = 0; i < ptcount; i++)
{
if (RANSACStatus[i] != 0){
double diss = abs(p1.at<float>(i, 1) - p2.at<float>(i, 1));
//cout <<"diss is: "<< diss << endl; // 横坐标小于100，基本处于一条水平线的才是正确匹配点
if (diss < 100)
{
inliner1[inlinerMatchesCount].x = p1.at<float>(i, 0);
inliner1[inlinerMatchesCount].y = p1.at<float>(i, 1);
inliner2[inlinerMatchesCount].x = p2.at<float>(i, 0);
inliner2[inlinerMatchesCount].y = p2.at<float>(i, 1);
inlierMatches[inlinerMatchesCount].queryIdx = inlinerMatchesCount;
inlierMatches[inlinerMatchesCount].trainIdx = inlinerMatchesCount;
inlinerMatchesCount++;
}
}
}
inliner1.resize(inlinerMatchesCount);
inliner2.resize(inlinerMatchesCount);
inlierMatches.resize(inlinerMatchesCount);
vector<cv::KeyPoint> key1(inlinerMatchesCount);
vector<cv::KeyPoint> key2(inlinerMatchesCount);
cv::KeyPoint::convert(inliner1, key1);
cv::KeyPoint::convert(inliner2, key2);
cv::Mat out;
cv::Mat uvtrans = cv::Mat_<cv::Point2f>(inliner1.size(), 2);
//vector<Point2f> p001, p002;
for (int i = 0; i < inliner1.size(); i++)
{
uvtrans.at<cv::Point2f>(i, 0) = inliner1[i];
//p001.push_back(inliner1[i]);
uvtrans.at<cv::Point2f>(i, 1) = inliner2[i];
//p002.push_back(inliner2[i]);
}
uvmark = uvtrans;
//cout << uvmark << endl;
drawMatches(img1, key1, img2, key2, inlierMatches, out);
cv::namedWindow("good match result", 0);
cv::resizeWindow("good match result", (int)(out.size().width / scale), (int)(out.size().height / scale));
cv::imshow("good match result", out);
cv::waitKey(0);
cv::destroyAllWindows();
}