opencv3 环境安装及实验surf特征的描述符匹配文章整理

OpenCV3.1 SIFT使用

OpenCV3对OpenCV的模块进行了调整，将开发中与nofree模块放在 了OpenCV_contrib中（包含SIFT），gitHub上的官方项目分成了两个，opencv 与 opencv_contrib。所以，要使用sift接口需在opencv3.1基础上，再安装opencv_contrib。本文主要记录如何安装opencv_contrib，配置Xcode，sift接口的用法。 
环境：OSX + Xcode + OpenCV3.1

• OpenCV31 SIFT使用
  • install opencv_contrib
  • configuration Xcode
    • pro_name Build Setting Search Paths
    • pro_name Build Setting Other Linker Flags
  • sample of sift
  • References

install opencv_contrib

• download contrib source code https://github.com/Itseez/opencv_contrib, follow README.md to install

$ cd <opencv_build_directory>$ cmake -DOPENCV_EXTRA_MODULES_PATH=<opencv_contrib>/modules <opencv_source_directory>$ make -j5$ sudo make install
1
2
3
4
1
2
3
4

Where <opencv_build_directory> and <opencv_source_directory> is directory in opencv3.1 install tutorial

configuration Xcode

like How to develop OpenCV with Xcode

pro_name Build Setting > Search Paths

• /usr/local/lib
• /usr/local/include

pro_name Build Setting >Other Linker Flags

• -lopencv_stitching -lopencv_superres -lopencv_videostab -lopencv_aruco -lopencv_bgsegm -lopencv_bioinspired -lopencv_ccalib -lopencv_dnn -lopencv_dpm -lopencv_fuzzy -lopencv_line_descriptor -lopencv_optflow -lopencv_plot -lopencv_reg -lopencv_saliency -lopencv_stereo -lopencv_structured_light -lopencv_rgbd -lopencv_surface_matching -lopencv_tracking -lopencv_datasets -lopencv_text -lopencv_face -lopencv_xfeatures2d -lopencv_shape -lopencv_video -lopencv_ximgproc -lopencv_calib3d -lopencv_features2d -lopencv_flann -lopencv_xobjdetect -lopencv_objdetect -lopencv_ml -lopencv_xphoto -lippicv -lopencv_highgui -lopencv_videoio -lopencv_imgcodecs -lopencv_photo -lopencv_imgproc -lopencv_core

sample of sift

• sample in (souce_dir)/samples/cpp/tutorial_code/xfeatures2D/LATCH_match.cpp or bellow

  #include "opencv2/xfeatures2d.hpp"// // now, you can no more create an instance on the 'stack', like in the tutorial// (yea, noticed for a fix/pr).// you will have to use cv::Ptr all the way down://cv::Ptr<Feature2D> f2d = xfeatures2d::SIFT::create();//cv::Ptr<Feature2D> f2d = xfeatures2d::SURF::create();//cv::Ptr<Feature2D> f2d = ORB::create();// you get the picture, i hope..//-- Step 1: Detect the keypoints:std::vector<KeyPoint> keypoints_1, keypoints_2;    f2d->detect( img_1, keypoints_1 );f2d->detect( img_2, keypoints_2 );//-- Step 2: Calculate descriptors (feature vectors)    Mat descriptors_1, descriptors_2;    f2d->compute( img_1, keypoints_1, descriptors_1 );f2d->compute( img_2, keypoints_2, descriptors_2 );//-- Step 3: Matching descriptor vectors using BFMatcher :BFMatcher matcher;std::vector< DMatch > matches;matcher.match( descriptors_1, descriptors_2, matches );
  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
  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

References

https://github.com/Itseez/opencv_contrib 
https://github.com/Itseez/opencv 
http://blog.csdn.net/lijiang1991/article/details/50756065 
http://docs.opencv.org/3.1.0/d5/d3c/classcv_1_1xfeatures2d_1_1SIFT.html#gsc.tab=0

一个实验代码：：

不多说什么了，直接上代码吧：

[java] view plain copy

 

1. #include <iostream>  
2. #include <stdio.h>  
3. #include "opencv2/core.hpp"  
4. #include "opencv2/core/utility.hpp"  
5. #include "opencv2/core/ocl.hpp"  
6. #include "opencv2/imgcodecs.hpp"  
7. #include "opencv2/highgui.hpp"  
8. #include "opencv2/features2d.hpp"  
9. #include "opencv2/calib3d.hpp"  
10. #include "opencv2/imgproc.hpp"  
11. #include"opencv2/flann.hpp"  
12. #include"opencv2/xfeatures2d.hpp"  
13. #include"opencv2/ml.hpp"  
14.   
15. using namespace cv;  
16. using namespace std;  
17. using namespace cv::xfeatures2d;  
18. using namespace cv::ml;  
19.   
20. int main()  
21. {  
22.     Mat a = imread("box.png", IMREAD_GRAYSCALE);    //读取灰度图像  
23.     Mat b = imread("box_in_scene.png", IMREAD_GRAYSCALE);  
24.   
25.     Ptr<SURF> surf;      //创建方式和2中的不一样  
26.     surf = SURF::create(800);  
27.   
28.     BFMatcher matcher;  
29.     Mat c, d;  
30.     vector<KeyPoint>key1, key2;  
31.     vector<DMatch> matches;  
32.   
33.     surf->detectAndCompute(a, Mat(), key1, c);  
34.     surf->detectAndCompute(b, Mat(), key2, d);  
35.   
36.     matcher.match(c, d, matches);       //匹配  
37.   
38.     sort(matches.begin(), matches.end());  //筛选匹配点  
39.     vector< DMatch > good_matches;                
40.     int ptsPairs = std::min(50, (int)(matches.size() * 0.15));  
41.     cout << ptsPairs << endl;  
42.     for (int i = 0; i < ptsPairs; i++)  
43.     {  
44.         good_matches.push_back(matches[i]);  
45.     }  
46.     Mat outimg;  
47.     drawMatches(a, key1, b, key2, good_matches, outimg, Scalar::all(-1), Scalar::all(-1),vector<char>(), DrawMatchesFlags::NOT_DRAW_SINGLE_POINTS);  //绘制匹配点  
48.   
49.     std::vector<Point2f> obj;  
50.     std::vector<Point2f> scene;  
51.   
52.     for (size_t i = 0; i < good_matches.size(); i++)  
53.     {  
54.         obj.push_back(key1[good_matches[i].queryIdx].pt);  
55.         scene.push_back(key2[good_matches[i].trainIdx].pt);  
56.     }  
57.   
58.     std::vector<Point2f> obj_corners(4);  
59.     obj_corners[0] = Point(0, 0);  
60.     obj_corners[1] = Point(a.cols, 0);  
61.     obj_corners[2] = Point(a.cols, a.rows);  
62.     obj_corners[3] = Point(0, a.rows);  
63.     std::vector<Point2f> scene_corners(4);  
64.   
65.     Mat H = findHomography(obj, scene, RANSAC);      //寻找匹配的图像  
66.     perspectiveTransform(obj_corners, scene_corners, H);  
67.   
68.     line(outimg,scene_corners[0] + Point2f((float)a.cols, 0), scene_corners[1] + Point2f((float)a.cols, 0),Scalar(0, 255, 0), 2, LINE_AA);       //绘制  
69.     line(outimg,scene_corners[1] + Point2f((float)a.cols, 0), scene_corners[2] + Point2f((float)a.cols, 0),Scalar(0, 255, 0), 2, LINE_AA);  
70.     line(outimg,scene_corners[2] + Point2f((float)a.cols, 0), scene_corners[3] + Point2f((float)a.cols, 0),Scalar(0, 255, 0), 2, LINE_AA);  
71.     line(outimg,scene_corners[3] + Point2f((float)a.cols, 0), scene_corners[0] + Point2f((float)a.cols, 0),Scalar(0, 255, 0), 2, LINE_AA);  
72.     imshow("aaaa",outimg);  
73.     cvWaitKey(0);  
74. }  

运行图： 

[java] view plain copy

 

1. //-------------读取模板------------  
2.     cv::Mat img_object = imread("/storage/emulated/0/ApplePearFace/imgTemp.jpg");  
3.     //-------------图像处理---------  
4.     cv::Mat img_scene(yimage);  
5.   
6.     /* 
7.     // 检测surf特征点 
8.     int minHessian = 400; 
9.     OrbDescriptorExtractor detector(minHessian); 
10.  
11.     std::vector<KeyPoint> keypoints_1, keypoints_2; 
12.     detector.detect(img_1, keypoints_1); 
13.     detector.detect(img_2, keypoints_2); 
14.     //-- Step 2: Calculate descriptors (feature vectors) 
15.     OrbDescriptorExtractor extractor; 
16.     Mat descriptors_1, descriptors_2; 
17.     extractor.compute(img_1, keypoints_1, descriptors_1); 
18.     extractor.compute(img_2, keypoints_2, descriptors_2); 
19.  
20.     //-- Step 3: Matching descriptor vectors with a brute force matcher 
21.     BFMatcher matcher(NORM_L2); 
22.     std::vector< DMatch > matches; 
23.     matcher.match(descriptors_1, descriptors_2, matches); 
24.  
25.     //-- Draw matches 
26.     Mat img_matches; 
27.     drawMatches(img_1, keypoints_1, img_2, keypoints_2, matches, img_matches);*/  
28.   
29.     // 读取数据  
30.     //cv::Mat img_object = cv::imread("doll01.jpg");  
31.     //cv::Mat img_scene = cv::imread("doll012.jpg");  
32.     if (!img_object.data || !img_scene.data) {  
33.         cout << "Error reading images." << endl;  
34.         return 0;  
35.     }  
36.   
37.     // 构建特征检测器和描述子提取器  
38.     cv::Ptr<cv::FeatureDetector> detector = cv::FeatureDetector::create("ORB");  
39.     cv::Ptr<cv::DescriptorExtractor> descriptor = cv::DescriptorExtractor::create("ORB");  
40.   
41.     // 检测特征点  
42.     vector<cv::KeyPoint> kp_object, kp_scene;  
43.     detector->detect(img_object, kp_object);  
44.     detector->detect(img_scene, kp_scene);  
45.   
46.     // 计算描述子  
47.     cv::Mat desp_object, desp_scene;  
48.     descriptor->compute(img_object, kp_object, desp_object);  
49.     descriptor->compute(img_scene, kp_scene, desp_scene);  
50.   
51.     /* 
52.     if (desp_object.type() != CV_32F) { 
53.     desp_object.convertTo(desp_object, CV_32F); 
54.     } 
55.  
56.     if (desp_scene.type() != CV_32F) { 
57.     desp_scene.convertTo(desp_scene, CV_32F); 
58.     } 
59.     */  
60.   
61.     // 匹配描述子  
62.     vector<cv::DMatch> matches;  
63.     cv::FlannBasedMatcher matcher(new cv::flann::LshIndexParams(20, 10, 2));  
64.     matcher.match(desp_object, desp_scene, matches);  
65.     //cout << "Find total " << matches.size() << " matches." << endl;  
66.   
67.     // 筛选匹配  
68.     //double min_dist = 100000;  
69.     //for (int i = 0; i < matches.size(); i++) {  
70.     //  float a = matches[i].distance;  
71.     //  if (a < min_dist) {  
72.     //      min_dist = matches[i].distance;  
73.     //  }  
74.     //}  
75.   
76.     //vector<cv::DMatch> good_matches;  
77.     //for (int i = 0; i < matches.size(); i++) {  
78.     //  
79.     //  if (matches[i].distance < 3 * min_dist) {  
80.     //      good_matches.push_back(matches[i]);  
81.     //  }  
82.     //}  
83.   
84.     // 显示匹配  
85.     //cout << "Good matches=" << matches.size() << endl;  
86.     cv::Mat img_matches;  
87.     cv::drawMatches(img_object, kp_object, img_scene, kp_scene, matches, img_matches);  
88.   
89.     // 定位目标  
90.     cv::vector<cv::Point2f> obj_points;  
91.     cv::vector<cv::Point2f> scene;  
92.   
93.     for (int i = 0; i < matches.size(); i++) {  
94.         obj_points.push_back(kp_object[matches[i].queryIdx].pt);  
95.         scene.push_back(kp_scene[matches[i].trainIdx].pt);  
96.     }  
97.     cv::Mat H = cv::findHomography(obj_points, scene, CV_RANSAC);  
98.   
99.   
100.     cv::vector<cv::Point2f> obj_corners(4);  
101.     cv::vector<cv::Point2f> scene_corners(4);  
102.     obj_corners[0] = cv::Point(0, 0);  
103.     obj_corners[1] = cv::Point(img_object.cols, 0);  
104.     obj_corners[2] = cv::Point(img_object.cols, img_object.rows);  
105.     obj_corners[3] = cv::Point(0, img_object.rows);  
106.   
107.     cv::perspectiveTransform(obj_corners, scene_corners, H);  
108.   
109.     cv::line(img_matches, scene_corners[0] + cv::Point2f(img_object.cols, 0), scene_corners[1] + cv::Point2f(img_object.cols, 0), cv::Scalar(0, 255, 0), 4);  
110.     cv::line(img_matches, scene_corners[1] + cv::Point2f(img_object.cols, 0), scene_corners[2] + cv::Point2f(img_object.cols, 0), cv::Scalar(0, 255, 0), 4);  
111.     cv::line(img_matches, scene_corners[2] + cv::Point2f(img_object.cols, 0), scene_corners[3] + cv::Point2f(img_object.cols, 0), cv::Scalar(0, 255, 0), 4);  
112.     cv::line(img_matches, scene_corners[3] + cv::Point2f(img_object.cols, 0), scene_corners[0] + cv::Point2f(img_object.cols, 0), cv::Scalar(0, 255, 0), 4);  
113.   
114.   
115.     cv::Mat dstSize;  
116.     cv::resize(img_matches, dstSize, Size(2 * h, w));