Shi-Tomasi角点检测

矫情吗；* 2022-02-21 19:41 269阅读 0赞

[官网教程][Link 1]

## 理论介绍 ##

*  在[Harris角点检测][Harris]的介绍时候得到协方差矩阵 M M M,它的特征值： λ 1 , λ 2 \\lambda\_1 ,\\lambda\_2 λ1,λ2

Harris打分函数： R = d e t ( M ) − k ( T r a c e ( M ) ) 2 R=det(M)-k(Trace(M))^2 R=det(M)−k(Trace(M))2  
 d e t ( M ) = λ 1 λ 2 det(M)=\\lambda\_1\\lambda\_2 det(M)=λ1λ2  
 T r a c e ( M ) = λ 1 + λ 2 Trace(M)=\\lambda\_1+\\lambda\_2 Trace(M)=λ1\+λ2

*  但是Shi-Tomasi使用的打分函数为： R = m i n ( λ 1 ， λ 2 ) R=min(\\lambda\_1，\\lambda\_2) R=min(λ1，λ2)

当R>阈值便认为是角点，

--------------------

## 例子 ##

#include <iostream>
    #include <vector>
    #include <opencv2/opencv.hpp>
     
    using namespace cv;
    using namespace std;
     
    RNG rng(122345);
    int maxCorners = 23;
    int maxTrackbar = 100;
    
    
    Mat img, img_gray;
    const char* source_window = "corners_detector";
    
    
    void goodFeaturesTraack_Demo(int, void*);
    
    int main()
    { 
    
        img = imread("../res/image.png",cv::IMREAD_COLOR);
        if(img.empty())
        { 
    	cout << "can't load the image" << endl;
        }
    
        cv::cvtColor(img, img_gray, cv::COLOR_BGR2GRAY);
    
        namedWindow(source_window);
        cv::createTrackbar("Max maxCorners:", source_window, &maxCorners, maxTrackbar, goodFeaturesTraack_Demo);
    
        imshow(source_window, img);
        goodFeaturesTraack_Demo(0, 0);
    
    
        waitKey();
        return 0;
    }
    
    
    
    
    void goodFeaturesTraack_Demo(int, void*)
    { 
        maxCorners = MAX(maxCorners, 1);
        vector<Point2f> corners;
        double qualityLevel = 0.01;
        double minDistance = 10;
        int blockSize = 3, gradientSize = 3;
        bool useHarriDetector = false;
        double k = 0.04;
    
        Mat copy = img.clone();
        cv::goodFeaturesToTrack(
    	    img_gray,
    	    corners,
    	    maxCorners,
    	    qualityLevel,
    	    minDistance,
    	    Mat(),
    	    blockSize,
    	    gradientSize,
    	    useHarriDetector,
    	    k);
    
        cout << "Number of corners detected" << corners.size() << endl;
        int radius = 4;
        for(int i=0; i<corners.size(); i++)
        { 
    	circle(copy, corners[i], radius, Scalar(rng.uniform(0,255), rng.uniform(0,255),rng.uniform(0,255)), FILLED);
        }
        imshow(source_window, copy);
    }

结果：

![在这里插入图片描述][20190411114413307.png_pic_center]

--------------------

## OpenCV API ##

[void cv::goodFeaturesToTrack ][void cv_goodFeaturesToTrack]

( 	
    	InputArray  	image,  //8-bit or floating-point 32-bit, single-channel image. 
    		OutputArray  	corners,   // vector of detected corners : vector\<Point2f>
    		int  	maxCorners, // 最多检测角点，如果实际检测超过maxCorners,返回maxCorners个最突出的，<=0意味着返回全部
    		double  	qualityLevel,// 0-1之间取值，参数意味着最小能接受的角点质量(阈值)，阈值＝qulityLevel*best corner quality measure（最小特征值）
    		                       //角点的quality measure < 阈值，都将忽略，
    		                       //比如：最好角点quality measure =1500, qualityLevel=0.01, 则角点quality measure<15将被忽略
    		double  	minDistance,//　返回的角点之间最小的欧式距离
    		InputArray  	mask = noArray(),// 　默认整张图，或者兴趣区域(提供相同大小的CV_8UC1)，返回只有兴趣域内的角点
    		int  	blockSize = 3,//　计算每个像素的导数协方差矩阵的临域大小
    		bool  	useHarrisDetector = false,// 使用cornerHarris或者cornerMinEigenVal，两种检测算法，本教程使用后者
    		double  	k = 0.04 //　对于Harris detector的参数k,一般取0.04-0.06
    	)

[Link 1]: https://docs.opencv.org/master/d8/dd8/tutorial_good_features_to_track.html
[Harris]: https://blog.csdn.net/zzyczzyc/article/details/88360360
[20190411114413307.png_pic_center]: /images/20220221/80b8e2eccb1842749b9eb5d1f7d0dac0.png
[void cv_goodFeaturesToTrack]: https://docs.opencv.org/master/dd/d1a/group__imgproc__feature.html#ga1d6bb77486c8f92d79c8793ad995d541