---

Preface

Corner detection （Corner Detection） It is a method used to obtain image features in computer vision system , It is widely used in motion detection 、 Image matching 、 Video tracking 、 In the fields of 3D modeling and target recognition , The page is called feature point detection .
A corner is usually defined as the intersection of two edges , More strictly speaking , Two local points of a region with different directions should have different neighborhood angles . And in practice , Most of the so-called corner detection methods detect image points with specific features , Not just “ Corner point ”. These feature points have specific coordinates in the image , And has some mathematical characteristics , Such as local maximum or minimum grayscale 、 Some gradient features, etc .

---

1、Harris Corner detection

In the field of image processing and computer vision , Point of interest (interest points), Also known as key points (key points)、 Characteristic point (feature points). It is widely used to solve object recognition 、 Image recognition 、 Image matching 、 Visual tracking 、 A series of problems such as 3D reconstruction . We don't look at the whole picture anymore , Instead, choose some special points , Then they are analyzed locally and purposefully . If enough of these spots can be detected , At the same time, their discrimination is particularly high , And it can accurately locate stable features , Then this method has use value .

Image feature types can be divided into the following three types ：

• edge
• Corner point （ Key points of interest ）
• speckle （Blobs）（ Region of interest ）

Corner detection algorithm

Current image processing neighborhood , Corner detection algorithms can be divided into the following three categories

• Corner detection based on gray image
• Corner detection based on binary image
• Corner detection based on contour curve

void cornerHarris( 
InputArray src, // The input image , Source image , fill  Mat  Class , And must be single channel  8  Bit or floating point image .
OutputArray dst, // The operation result after function call exists here , That is, this parameter is used to store  Harris  Output result of corner detection , The same size and type as the source image .
int blockSize,  // Represents the size of the neighborhood 
int ksize,      // Express  Sobel( The aperture size of the operator .
double k,       //Harris  Parameters 
int borderType = BORDER_DEFAULT  // Boundary mode of image pixels 
 );

C++ Code test example 1

#include <opencv2/opencv.hpp> 
using namespace cv;

static int test()
{
    
	// Load the image in grayscale mode and display 
	Mat srcImage = imread("1.jpg", 0);
	imshow(" The original picture ", srcImage);

	// Conduct Harris Corner detection to find out corner points 
	Mat cornerStrength;
	cornerHarris(srcImage, cornerStrength, 2, 3, 0.01);

	// Threshold operation on gray image , Get the binary diagram and display  
	Mat harrisCorner;
	threshold(cornerStrength, harrisCorner, 0.00001, 255, THRESH_BINARY);
	imshow(" Binary rendering after corner detection ", harrisCorner);

	waitKey(0);
	return 0;
}


int main()
{
    
	test();
	system("pause");
	return 0;
}

result

C++ Code test example 2

#include <opencv2/opencv.hpp> 
using namespace cv;
using namespace std;

string  window_name1 = "【 Program window 1】";      // Window Title Definition  
string  window_name2 = "【 Program window 2】";      // Window Title Definition  

Mat g_srcImage, g_srcImage1, g_grayImage;
int thresh = 30; // Current threshold 
int max_thresh = 175; // Maximum threshold 

void on_CornerHarris(int, void*);// Callback function 

static int test()
{
    
	//【0】 change console The font color 
	system("color 3F");

	//【1】 Load the original drawing and clone it 
	g_srcImage = imread("1.jpg", 1);
	if (!g_srcImage.data) {
     cout <<" Error reading picture , Please make sure whether there is... In the directory imread The picture specified by the function does not exist ~！ \n"; return false; }
	imshow(" The original picture ", g_srcImage);
	g_srcImage1 = g_srcImage.clone();

	//【2】 Save a grayscale image 
	cvtColor(g_srcImage1, g_grayImage, COLOR_BGR2GRAY);

	//【3】 Create windows and scroll bars 
	namedWindow(window_name1, WINDOW_AUTOSIZE);
	createTrackbar(" threshold : ", window_name1, &thresh, max_thresh, on_CornerHarris);

	//【4】 Call the callback function once , To initialize 
	on_CornerHarris(0, 0);

	waitKey(0);
	return(0);
}

int main()
{
    
	test();
	system("pause");
	return 0;
}

void on_CornerHarris(int, void*)
{
    
	//----------------------【1】 Define some local variables ---------------
	Mat dstImage;// Goal map 
	Mat normImage;// Normalized graph 
	Mat scaledImage;// Graph of 8-bit unsigned integer after linear transformation 

	//---------------------【2】 initialization ----------------------------
	// Set to zero the two pictures currently to be displayed , That is, clear their values the last time this function was called 
	dstImage = Mat::zeros(g_srcImage.size(), CV_32FC1);
	g_srcImage1 = g_srcImage.clone();

	//-----------------------【3】 Formal testing ------------------------
	// Corner detection 
	cornerHarris(g_grayImage, dstImage, 2, 3, 0.04, BORDER_DEFAULT);

	//  Normalization and transformation 
	normalize(dstImage, normImage, 0, 255, NORM_MINMAX, CV_32FC1, Mat());
	convertScaleAbs(normImage, scaledImage);// The normalized graph is linearly transformed into 8 Bit unsigned integer  

	//-----------------【4】 Drawing ---------------------------
	//  What will be detected , And the corner points that meet the threshold conditions are drawn 
	for (int j = 0; j < normImage.rows; j++)
	{
    
		for (int i = 0; i < normImage.cols; i++)
		{
    
			if ((int)normImage.at<float>(j, i) > thresh + 80)
			{
    
				circle(g_srcImage1, Point(i, j), 5, Scalar(10, 10, 255), 2, 8, 0);
				circle(scaledImage, Point(i, j), 5, Scalar(0, 10, 255), 2, 8, 0);
			}
		}
	}
	//------------------【4】 Show the end result ---------------------
	imshow(window_name1, g_srcImage1);
	imshow(window_name2, scaledImage);
}

result

2、Shi-Tomasi Corner detection

Besides using Harris Carry out corner detection , We can also usually use Shi-Tomasi Methods corner detection .Shi-Tomasi The algorithm is Harris Algorithm improvement , The original definition of this algorithm is matrix M The determinant value of is the same as M Trace subtraction , Then compare the difference with the preset threshold . later Shi and Tomasi An improved method is proposed , If the smaller of the two eigenvalues is greater than the minimum threshold , You will get strong corners .

C++ Code example

#include <opencv2/opencv.hpp> 
using namespace cv;
using namespace std;

string window_name = "【Shi-Tomasi Corner detection 】";

Mat g_srcImage, g_grayImage;
int g_maxCornerNumber = 33;
int g_maxTrackbarNumber = 500;
RNG g_rng(12345);// Initialize random number generator 

void on_GoodFeaturesToTrack(int, void*)
{
    
	//【1】 For variables less than or equal to 1 Time processing 
	if (g_maxCornerNumber <= 1) {
     g_maxCornerNumber = 1; }

	//【2】Shi-Tomasi Algorithm （goodFeaturesToTrack function ） Parameter preparation for 
	vector<Point2f> corners;
	double qualityLevel = 0.01;// The minimum eigenvalue acceptable for corner detection 
	double minDistance = 10;// Minimum distance between corners 
	int blockSize = 3;// The neighborhood range specified when calculating the derivative autocorrelation matrix 
	double k = 0.04;// Weight factor 
	Mat copy = g_srcImage.clone();	// Copy the source image to a temporary variable , As a region of interest 

	//【3】 Conduct Shi-Tomasi Corner detection 
	goodFeaturesToTrack(g_grayImage,// The input image 
		corners,                     // The output vector of the detected corner 
		g_maxCornerNumber,           // The maximum number of corners 
		qualityLevel,                // The minimum eigenvalue acceptable for corner detection 
		minDistance,                 // Minimum distance between corners 
		Mat(),                       // Region of interest 
		blockSize,                   // The neighborhood range specified when calculating the derivative autocorrelation matrix 
		false,                       // Don't use Harris Corner detection 
		k);                         // Weight factor 


	//【4】 Output text messages 
	cout << "\t> The number of corners detected this time is ：" << corners.size() << endl;

	//【5】 Draw the detected corner 
	int r = 4;
	for (int i = 0; i < corners.size(); i++)
	{
    
		// Draw corners in random colors 
		circle(copy, corners[i], r, Scalar(g_rng.uniform(0, 255), g_rng.uniform(0, 255),
			g_rng.uniform(0, 255)), -1, 8, 0);
	}

	//【6】 Show （ to update ） window 
	imshow(window_name, copy);
}

static void test()
{
    
	//【0】 change console The font color 
	system("color 2F");

	//【1】 Load the source image and convert it to a grayscale image 
	g_srcImage = imread("1.jpg", 1);
	cvtColor(g_srcImage, g_grayImage, COLOR_BGR2GRAY);

	//【2】 Create windows and sliders , And initialize the display and callback functions 
	namedWindow(window_name, WINDOW_AUTOSIZE);
	createTrackbar(" Maximum number of corners ", window_name, &g_maxCornerNumber, g_maxTrackbarNumber, on_GoodFeaturesToTrack);
	imshow(window_name, g_srcImage);
	on_GoodFeaturesToTrack(0, 0);
	waitKey(0);
}

int main()
{
    
	test();
	system("pause");
	return 0;
}

result

3、 Sub pixel corner detection

Background Overview

If the purpose of image processing is not to extract feature points for recognition, but to carry out geometric measurement , This usually requires higher accuracy , The function goodFeaturesToTrack（） Only simple pixel coordinate values can be provided , in other words , Sometimes you need real coordinate values instead of integer coordinate values .
The position of sub-pixel corner detection is in the camera calibration stage 、 Track and reconstruct the trajectory of the camera , Or when reconstructing the three-dimensional structure of the tracking target , Is a basic measurement .

C++ Code test example

#include <opencv2/opencv.hpp> 
#include <iostream>
using namespace cv;
using namespace std;


string window_name = "【 Sub pixel corner detection 】";       // Window Title Definition 
Mat g_srcImage, g_grayImage;
int g_maxCornerNumber = 33;
int g_maxTrackbarNumber = 500;
RNG g_rng(12345);// Initialize random number generator 


void on_GoodFeaturesToTrack(int, void*)
{
    
	//【1】 For variables less than or equal to 1 Time processing 
	if (g_maxCornerNumber <= 1) {
     g_maxCornerNumber = 1; }

	//【2】Shi-Tomasi Algorithm （goodFeaturesToTrack function ） Parameter preparation for 
	vector<Point2f> corners;
	double qualityLevel = 0.01;// The minimum eigenvalue acceptable for corner detection 
	double minDistance = 10;// Minimum distance between corners 
	int blockSize = 3;// The neighborhood range specified when calculating the derivative autocorrelation matrix 
	double k = 0.04;// Weight factor 
	Mat copy = g_srcImage.clone();	// Copy the source image to a temporary variable , As a region of interest 

	//【3】 Conduct Shi-Tomasi Corner detection 
	goodFeaturesToTrack(g_grayImage,// The input image 
		corners,// The output vector of the detected corner 
		g_maxCornerNumber,// The maximum number of corners 
		qualityLevel,// The minimum eigenvalue acceptable for corner detection 
		minDistance,// Minimum distance between corners 
		Mat(),// Region of interest 
		blockSize,// The neighborhood range specified when calculating the derivative autocorrelation matrix 
		false,// Don't use Harris Corner detection 
		k);// Weight factor 

	//【4】 Output text messages 
	cout << "\n\t>------------- The number of corners detected this time is ：" << corners.size() << endl;

	//【5】 Draw the detected corner 
	int r = 4;
	for (unsigned int i = 0; i < corners.size(); i++)
	{
    
		// Draw corners in random colors 
		circle(copy, corners[i], r, Scalar(g_rng.uniform(0, 255), g_rng.uniform(0, 255),
			g_rng.uniform(0, 255)), -1, 8, 0);
	}

	//【6】 Show （ to update ） window 
	imshow(window_name, copy);

	//【7】 Parameter setting of sub-pixel corner detection 
	Size winSize = Size(5, 5);
	Size zeroZone = Size(-1, -1);
	TermCriteria criteria = TermCriteria(TermCriteria::EPS + TermCriteria::MAX_ITER, 40, 0.001);

	//【8】 Calculate the sub-pixel corner position 
	cornerSubPix(g_grayImage, corners, winSize, zeroZone, criteria);

	//【9】 Output corner information 
	for (int i = 0; i < corners.size(); i++)
	{
    
		cout << " \t>> Exact corner coordinates [" << i << "] (" << corners[i].x << "," << corners[i].y << ")" << endl;
	}
}


static void test()
{
    
	//【0】 change console The font color 
	system("color 2F");

	//【1】 Load the source image and convert it to a grayscale image 
	g_srcImage = imread("1.jpg", 1);
	cvtColor(g_srcImage, g_grayImage, COLOR_BGR2GRAY);

	//【2】 Create windows and sliders , And initialize the display and callback functions 
	namedWindow(window_name, WINDOW_AUTOSIZE);
	createTrackbar(" Maximum number of corners ", window_name, &g_maxCornerNumber, g_maxTrackbarNumber, on_GoodFeaturesToTrack);
	imshow(window_name, g_srcImage);
	on_GoodFeaturesToTrack(0, 0);

	waitKey(0);
}

int main()
{
    
	test();
	system("pause");
	return 0;
}

result