opencv-Hough transform-circle detection

principle

If the center of the circle is at the origin, then x^2 + y^2 = r^2
If the center of the circle is not at the origin, then (x-a)^2 + (y-b) ^ 2 = r^2

Converted to polar coordinates, the formula is as follows
Insert picture description here
(a, b) are the coordinates of the center of the circle.
If we don’t know the coordinates of the center of the circle, then we can reverse it according to the point on the circle and the known radius
.

Find three points at random as shown in the figure below , then the same radius Next, the three circles will intersect at the center of the circle.
Through the addition of rgb, the pixel value of the intersection point will be very large, which looks the brightest, and this point is the origin.

API

void HoughCircles(InputArray   image,
	OutputArray  circles,
	int          method,
	double       dp,
	double       minDist,
	double       param1 = 100,
	double       param2 = 100,
	int          minRadius = 0,
	int          maxRadius = 0
	)

The meaning is as follows

//第一个参数，InputArray类型的image，输入图像，即源图像，需为8位的灰度单通道图像。
//第二个参数，InputArray类型的circles，经过调用HoughCircles函数后此参数存储了检测到的圆的输出矢量，每个矢量由包含了3个元素的浮点矢量(x, y, radius)表示。
//第三个参数，int类型的method，即使用的检测方法，目前OpenCV中就霍夫梯度法一种可以使用，它的标识符为CV_HOUGH_GRADIENT，在此参数处填这个标识符即可。
//第四个参数，double类型的dp，用来检测圆心的累加器图像的分辨率于输入图像之比的倒数，且此参数允许创建一个比输入图像分辨率低的累加器。上述文字不好理解的话，来看例子吧。例如，如果dp= 1时，累加器和输入图像具有相同的分辨率。如果dp=2，累加器便有输入图像一半那么大的宽度和高度。
//第五个参数，double类型的minDist，为霍夫变换检测到的圆的圆心之间的最小距离，即让我们的算法能明显区分的两个不同圆之间的最小距离。这个参数如果太小的话，多个相邻的圆可能被错误地检测成了一个重合的圆。反之，这个参数设置太大的话，某些圆就不能被检测出来了。
//第六个参数，double类型的param1，有默认值100。它是第三个参数method设置的检测方法的对应的参数。对当前唯一的方法霍夫梯度法CV_HOUGH_GRADIENT，它表示传递给canny边缘检测算子的高阈值，而低阈值为高阈值的一半。
//第七个参数，double类型的param2，也有默认值100。它是第三个参数method设置的检测方法的对应的参数。对当前唯一的方法霍夫梯度法CV_HOUGH_GRADIENT，它表示在检测阶段圆心的累加器阈值。它越小的话，就可以检测到更多根本不存在的圆，而它越大的话，能通过检测的圆就更加接近完美的圆形了。
//第八个参数，int类型的minRadius,有默认值0，表示圆半径的最小值。
//第九个参数，int类型的maxRadius,也有默认值0，表示圆半径的最大值。
 
//需要注意的是，使用此函数可以很容易地检测出圆的圆心，但是它可能找不到合适的圆半径

Code writing

Because Hough circle detection is more sensitive to noise, the median filter of the image must be done first.

Based on efficiency considerations, the Hough transform circle detection implemented in Opencv is based on the implementation of the image gradient, which is divided into two steps:

1. Detect edges and find possible centers
1. Calculate the optimal radius from the center of the candidate circle based on the first step

#define INPUT_TITLE "input image"          // 宏定义，原始图片
#define OUTPUT_TITLE "HoughCircle picture" // 宏定义，经过霍夫圆变换后的图片

#include<iostream>
#include<opencv2\opencv.hpp>
#include<opencv2\core.hpp>

using namespace std;
using namespace cv;

int main() {
    
    
	//img是最后经过霍夫圆变换的图片，src是通过imread读取的图片，change_src是src需要经过的变换的图片。
	Mat img, src, change_src;
	src = imread("C:/Users/86176/Pictures/picx/shuiyin.png");
	if (!src.data)
	{
    
    
		cout << "ERROR : could not load image.";
		return -1;
	}
	namedWindow(INPUT_TITLE, CV_WINDOW_AUTOSIZE);
	namedWindow(OUTPUT_TITLE, CV_WINDOW_AUTOSIZE);
	imshow(INPUT_TITLE, src);

	//中值滤波，去除椒盐噪声。并转为单通道。
	medianBlur(src, change_src, 3);
	cvtColor(change_src, change_src, CV_BGR2GRAY);

	//霍夫圆变换
	vector<Vec3f> pcircles;
	HoughCircles(change_src, pcircles, CV_HOUGH_GRADIENT, 1, 10, 100, 30, 5, 50);

	//
	src.copyTo(img);
	cout << "pcircles.size() = " << pcircles.size() << endl;//为了了解HoughCircles中第二个参数的大小。
															//将圆画在原图片上。
	for (int i = 0; i < pcircles.size(); i++)
	{
    
    
		Vec3f cc = pcircles[i];
		circle(img, Point(cc[0], cc[1]), cc[2], Scalar(0, 0, 255), 2, LINE_AA);
		circle(img, Point(cc[0], cc[1]), 2, Scalar(255, 0, 4), 2, LINE_AA);
	}

	imshow(OUTPUT_TITLE, img);

	waitKey(0);
	return 0;
}

The effect diagram is as follows. The three circles generated by pressing and holding shift are all recognized. As for the largest one, because the radius of the maximum recognized circle is limited to 50: if you change 50 to 150, it becomes the back That's it.

HoughCircles(change_src, pcircles, CV_HOUGH_GRADIENT, 1, 10, 100, 30, 5, 50);

Insert picture description here