Skip to content

霍夫线变换

在本教程中,你将学到:

  • 如何使用 OpenCV 函数 HoughLines() 和 HoughLinesP() 来检测图像中的直线。
  1. 霍夫线变换是一种用于检测直线的变换。
  2. 在应用该变换之前,最好先进行边缘检测预处理。

Hough_Lines_Tutorial_Theory_0.jpg Hough_Lines_Tutorial_Theory_1.jpg Hough_Lines_Tutorial_Theory_2.jpg

  1. 如你所知,图像空间中的一条直线可以用两个变量来表示。例如:

    1. 在笛卡尔坐标系中:参数为 (m,b)(m,b)。
    2. 在极坐标系中:参数为 (r,θ)(r,\theta)

    对于霍夫变换,我们将在极坐标系中表达直线。因此,直线方程可以写为:

    y=(−cos⁡θsin⁡θ)x+(rsin⁡θ)y = \left ( -\dfrac{\cos \theta}{\sin \theta} \right ) x + \left ( \dfrac{r}{\sin \theta} \right )

    整理各项可得:r=xcos⁡θ+ysin⁡θr = x \cos \theta + y \sin \theta

  2. 一般来说,对于每个点 (x0,y0)(x_{0}, y_{0}),我们可以把经过该点的一族直线定义为:

    rθ=x0⋅cos⁡θ+y0⋅sin⁡θr_{\theta} = x_{0} \cdot \cos \theta + y_{0} \cdot \sin \theta

    也就是说,每一对 (rθ,θ)(r_{\theta},\theta) 都代表一条经过 (x0,y0)(x_{0}, y_{0}) 的直线。

  3. 如果对给定的 (x0,y0)(x_{0}, y_{0}) 画出经过它的那一族直线,我们会得到一条正弦曲线。例如,对于 x0=8x_{0} = 8、y0=6y_{0} = 6,我们得到如下曲线(在 θ\theta - rr 平面上):

    我们只考虑满足 r>0r > 0 且 0<θ<2π0< \theta < 2 \pi 的点。

  4. 我们可以对图像中的所有点执行同样的操作。如果两个不同点的曲线在 θ\theta - rr 平面中相交,就意味着这两个点属于同一条直线。例如,继续上面的例子,再为两个点 x1=4x_{1} = 4、y1=9y_{1} = 9 和 x2=12x_{2} = 12、y2=3y_{2} = 3 画出曲线,我们得到:

    三条曲线相交于同一个点 (0.925,9.6)(0.925, 9.6),这些坐标就是 (x0,y0)(x_{0}, y_{0})、(x1,y1)(x_{1}, y_{1}) 和 (x2,y2)(x_{2}, y_{2}) 所在直线的参数(θ,r\theta, r)。

  5. 上面这一切意味着什么?它意味着,一般来说,一条直线可以通过找出曲线之间的交点数量来被检测出来。相交的曲线越多,说明该交点所代表的直线上的点就越多。一般来说,我们可以定义一个阈值,表示检测一条直线所需的最少交点数量。

  6. 这正是霍夫线变换所做的事情。它跟踪图像中每个点的曲线之间的交点。如果交点数量超过某个阈值,就把它判定为一条直线,其参数就是交点的 (θ,rθ)(\theta, r_{\theta})。

OpenCV 实现了三种霍夫线变换:

a. 标准霍夫变换(Standard Hough Transform)

  • 基本就是我们上一节讲解的内容。它给出的结果是一个由 (θ,rθ)(\theta, r_{\theta}) 对组成的向量。
  • 在 OpenCV 中由函数 HoughLines() 实现。

b. 概率霍夫线变换(Probabilistic Hough Line Transform)

  • 霍夫线变换的一种更高效的实现。它输出检测到的直线的两个端点 (x0,y0,x1,y1)(x_{0}, y_{0}, x_{1}, y_{1})。
  • 在 OpenCV 中由函数 HoughLinesP() 实现。

c. 加权霍夫变换(Weighted Hough Transform)

  • 使用边缘强度值,而不是标准霍夫变换中的二值 0 或 1。
  • 在 OpenCV 中由函数 HoughLines() 配合 use_edgeval=true 实现。
  • 示例见 samples/cpp/tutorial_code/ImgTrans/HoughLines_Demo.cpp。
  • 加载一幅图像;
  • 应用一次标准霍夫线变换和一次概率霍夫线变换;
  • 在三个窗口中显示原图和检测到的直线。

C++:

我们将要讲解的示例代码可以从 这里 下载。一个稍复杂一点的版本(同时展示标准与概率霍夫变换,并带有用于调整阈值的 Trackbar)可以在 这里 找到。

/**
* @file houghlines.cpp
* @brief This program demonstrates line finding with the Hough transform
*/
#include "opencv2/imgcodecs.hpp"
#include "opencv2/highgui.hpp"
#include "opencv2/imgproc.hpp"
using namespace cv;
using namespace std;
int main(int argc, char** argv)
{
// Declare the output variables
Mat dst, cdst, cdstP;
const char* default_file = "sudoku.png";
const char* filename = argc >=2 ? argv[1] : default_file;
// Loads an image
Mat src = imread( samples::findFile( filename ), IMREAD_GRAYSCALE );
// Check if image is loaded fine
if(src.empty()){
printf(" Error opening image\n");
printf(" Program Arguments: [image_name -- default %s] \n", default_file);
return -1;
}
// Edge detection
Canny(src, dst, 50, 200, 3);
// Copy edges to the images that will display the results in BGR
cvtColor(dst, cdst, COLOR_GRAY2BGR);
cdstP = cdst.clone();
// Standard Hough Line Transform
vector<Vec2f> lines; // will hold the results of the detection
HoughLines(dst, lines, 1, CV_PI/180, 150, 0, 0 ); // runs the actual detection
// Draw the lines
for( size_t i = 0; i < lines.size(); i++ )
{
float rho = lines[i][0], theta = lines[i][1];
Point pt1, pt2;
double a = cos(theta), b = sin(theta);
double x0 = a*rho, y0 = b*rho;
pt1.x = cvRound(x0 + 1000*(-b));
pt1.y = cvRound(y0 + 1000*(a));
pt2.x = cvRound(x0 - 1000*(-b));
pt2.y = cvRound(y0 - 1000*(a));
line( cdst, pt1, pt2, Scalar(0,0,255), 3, LINE_AA);
}
// Probabilistic Line Transform
vector<Vec4i> linesP; // will hold the results of the detection
HoughLinesP(dst, linesP, 1, CV_PI/180, 50, 50, 10 ); // runs the actual detection
// Draw the lines
for( size_t i = 0; i < linesP.size(); i++ )
{
Vec4i l = linesP[i];
line( cdstP, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0,0,255), 3, LINE_AA);
}
// Show results
imshow("Source", src);
imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst);
imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP);
// Wait and Exit
waitKey();
return 0;
}

Java:

我们将要讲解的示例代码可以从 这里 下载。

/**
* @file HoughLines.java
* @brief This program demonstrates line finding with the Hough transform
*/
import org.opencv.core.*;
import org.opencv.core.Point;
import org.opencv.highgui.HighGui;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
class HoughLinesRun {
public void run(String[] args) {
// Declare the output variables
Mat dst = new Mat(), cdst = new Mat(), cdstP;
String default_file = "../../../../data/sudoku.png";
String filename = ((args.length > 0) ? args[0] : default_file);
// Load an image
Mat src = Imgcodecs.imread(filename, Imgcodecs.IMREAD_GRAYSCALE);
// Check if image is loaded fine
if( src.empty() ) {
System.out.println("Error opening image!");
System.out.println("Program Arguments: [image_name -- default "
+ default_file +"] \n");
System.exit(-1);
}
// Edge detection
Imgproc.Canny(src, dst, 50, 200, 3, false);
// Copy edges to the images that will display the results in BGR
Imgproc.cvtColor(dst, cdst, Imgproc.COLOR_GRAY2BGR);
cdstP = cdst.clone();
// Standard Hough Line Transform
Mat lines = new Mat(); // will hold the results of the detection
Imgproc.HoughLines(dst, lines, 1, Math.PI/180, 150); // runs the actual detection
// Draw the lines
for (int x = 0; x < lines.rows(); x++) {
double rho = lines.get(x, 0)[0],
theta = lines.get(x, 0)[1];
double a = Math.cos(theta), b = Math.sin(theta);
double x0 = a*rho, y0 = b*rho;
Point pt1 = new Point(Math.round(x0 + 1000*(-b)), Math.round(y0 + 1000*(a)));
Point pt2 = new Point(Math.round(x0 - 1000*(-b)), Math.round(y0 - 1000*(a)));
Imgproc.line(cdst, pt1, pt2, new Scalar(0, 0, 255), 3, Imgproc.LINE_AA, 0);
}
// Probabilistic Line Transform
Mat linesP = new Mat(); // will hold the results of the detection
Imgproc.HoughLinesP(dst, linesP, 1, Math.PI/180, 50, 50, 10); // runs the actual detection
// Draw the lines
for (int x = 0; x < linesP.rows(); x++) {
double[] l = linesP.get(x, 0);
Imgproc.line(cdstP, new Point(l[0], l[1]), new Point(l[2], l[3]), new Scalar(0, 0, 255), 3, Imgproc.LINE_AA, 0);
}
// Show results
HighGui.imshow("Source", src);
HighGui.imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst);
HighGui.imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP);
// Wait and Exit
HighGui.waitKey();
System.exit(0);
}
}
public class HoughLines {
public static void main(String[] args) {
// Load the native library.
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
new HoughLinesRun().run(args);
}
}

Python:

我们将要讲解的示例代码可以从 这里 下载。

"""
@file hough_lines.py
@brief This program demonstrates line finding with the Hough transform
"""
import sys
import math
import cv2 as cv
import numpy as np
def main(argv):
default_file = 'sudoku.png'
filename = argv[0] if len(argv) > 0 else default_file
# Loads an image
src = cv.imread(cv.samples.findFile(filename), cv.IMREAD_GRAYSCALE)
# Check if image is loaded fine
if src is None:
print ('Error opening image!')
print ('Usage: hough_lines.py [image_name -- default ' + default_file + '] \n')
return -1
# Edge detection
dst = cv.Canny(src, 50, 200, None, 3)
# Copy edges to the images that will display the results in BGR
cdst = cv.cvtColor(dst, cv.COLOR_GRAY2BGR)
cdstP = np.copy(cdst)
# Standard Hough Line Transform
lines = cv.HoughLines(dst, 1, np.pi / 180, 150, None, 0, 0)
# Draw the lines
if lines is not None:
for i in range(0, len(lines)):
rho = lines[i][0][0]
theta = lines[i][0][1]
a = math.cos(theta)
b = math.sin(theta)
x0 = a * rho
y0 = b * rho
pt1 = (int(x0 + 1000*(-b)), int(y0 + 1000*(a)))
pt2 = (int(x0 - 1000*(-b)), int(y0 - 1000*(a)))
cv.line(cdst, pt1, pt2, (0,0,255), 3, cv.LINE_AA)
# Probabilistic Line Transform
linesP = cv.HoughLinesP(dst, 1, np.pi / 180, 50, None, 50, 10)
# Draw the lines
if linesP is not None:
for i in range(0, len(linesP)):
l = linesP[i][0]
cv.line(cdstP, (l[0], l[1]), (l[2], l[3]), (0,0,255), 3, cv.LINE_AA)
# Show results
cv.imshow("Source", src)
cv.imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst)
cv.imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP)
# Wait and Exit
cv.waitKey()
return 0
if __name__ == "__main__":
main(sys.argv[1:])

C++:

const char* default_file = "sudoku.png";
const char* filename = argc >=2 ? argv[1] : default_file;
// Loads an image
Mat src = imread( samples::findFile( filename ), IMREAD_GRAYSCALE );
// Check if image is loaded fine
if(src.empty()){
printf(" Error opening image\n");
printf(" Program Arguments: [image_name -- default %s] \n", default_file);
return -1;
}

Java:

String default_file = "../../../../data/sudoku.png";
String filename = ((args.length > 0) ? args[0] : default_file);
// Load an image
Mat src = Imgcodecs.imread(filename, Imgcodecs.IMREAD_GRAYSCALE);
// Check if image is loaded fine
if( src.empty() ) {
System.out.println("Error opening image!");
System.out.println("Program Arguments: [image_name -- default "
+ default_file +"] \n");
System.exit(-1);
}

Python:

default_file = 'sudoku.png'
filename = argv[0] if len(argv) > 0 else default_file
# Loads an image
src = cv.imread(cv.samples.findFile(filename), cv.IMREAD_GRAYSCALE)
# Check if image is loaded fine
if src is None:
print ('Error opening image!')
print ('Usage: hough_lines.py [image_name -- default ' + default_file + '] \n')
return -1

使用 Canny 检测器检测图像的边缘

Section titled “使用 Canny 检测器检测图像的边缘”

C++:

// Edge detection
Canny(src, dst, 50, 200, 3);

Java:

// Edge detection
Imgproc.Canny(src, dst, 50, 200, 3, false);

Python:

# Edge detection
dst = cv.Canny(src, 50, 200, None, 3)

现在我们将应用霍夫线变换。我们会讲解如何使用 OpenCV 为此提供的两个函数。

首先,应用变换:

C++:

// Standard Hough Line Transform
vector<Vec2f> lines; // will hold the results of the detection
HoughLines(dst, lines, 1, CV_PI/180, 150, 0, 0 ); // runs the actual detection

Java:

// Standard Hough Line Transform
Mat lines = new Mat(); // will hold the results of the detection
Imgproc.HoughLines(dst, lines, 1, Math.PI/180, 150); // runs the actual detection

Python:

# Standard Hough Line Transform
lines = cv.HoughLines(dst, 1, np.pi / 180, 150, None, 0, 0)

其参数含义如下:

  • dst:边缘检测器的输出。它应该是一幅灰度图(尽管实际上是一幅二值图)。
  • lines:一个向量,用来存储检测到的直线的参数 (r,θ)(r,\theta)。
  • rho:参数 rr 的分辨率,以像素为单位。我们使用 1 像素。
  • theta:参数 θ\theta 的分辨率,以弧度为单位。我们使用 1 度(即 CV_PI/180)。
  • threshold:检测一条直线所需的最少交点数量。
  • srn 和 stn:默认为 0 的参数。更多信息请查阅 OpenCV 参考文档。

然后通过绘制直线来显示结果。

C++:

// Draw the lines
for( size_t i = 0; i < lines.size(); i++ )
{
float rho = lines[i][0], theta = lines[i][1];
Point pt1, pt2;
double a = cos(theta), b = sin(theta);
double x0 = a*rho, y0 = b*rho;
pt1.x = cvRound(x0 + 1000*(-b));
pt1.y = cvRound(y0 + 1000*(a));
pt2.x = cvRound(x0 - 1000*(-b));
pt2.y = cvRound(y0 - 1000*(a));
line( cdst, pt1, pt2, Scalar(0,0,255), 3, LINE_AA);
}

Java:

// Draw the lines
for (int x = 0; x < lines.rows(); x++) {
double rho = lines.get(x, 0)[0],
theta = lines.get(x, 0)[1];
double a = Math.cos(theta), b = Math.sin(theta);
double x0 = a*rho, y0 = b*rho;
Point pt1 = new Point(Math.round(x0 + 1000*(-b)), Math.round(y0 + 1000*(a)));
Point pt2 = new Point(Math.round(x0 - 1000*(-b)), Math.round(y0 - 1000*(a)));
Imgproc.line(cdst, pt1, pt2, new Scalar(0, 0, 255), 3, Imgproc.LINE_AA, 0);
}

Python:

# Draw the lines
if lines is not None:
for i in range(0, len(lines)):
rho = lines[i][0][0]
theta = lines[i][0][1]
a = math.cos(theta)
b = math.sin(theta)
x0 = a * rho
y0 = b * rho
pt1 = (int(x0 + 1000*(-b)), int(y0 + 1000*(a)))
pt2 = (int(x0 - 1000*(-b)), int(y0 - 1000*(a)))
cv.line(cdst, pt1, pt2, (0,0,255), 3, cv.LINE_AA)

首先,应用变换:

C++:

// Probabilistic Line Transform
vector<Vec4i> linesP; // will hold the results of the detection
HoughLinesP(dst, linesP, 1, CV_PI/180, 50, 50, 10 ); // runs the actual detection

Java:

// Probabilistic Line Transform
Mat linesP = new Mat(); // will hold the results of the detection
Imgproc.HoughLinesP(dst, linesP, 1, Math.PI/180, 50, 50, 10); // runs the actual detection

Python:

# Probabilistic Line Transform
linesP = cv.HoughLinesP(dst, 1, np.pi / 180, 50, None, 50, 10)

其参数含义如下:

  • dst:边缘检测器的输出。它应该是一幅灰度图(尽管实际上是一幅二值图)。
  • lines:一个向量,用来存储检测到的直线的参数 (xstart,ystart,xend,yend)(x_{start}, y_{start}, x_{end}, y_{end})。
  • rho:参数 rr 的分辨率,以像素为单位。我们使用 1 像素。
  • theta:参数 θ\theta 的分辨率,以弧度为单位。我们使用 1 度(即 CV_PI/180)。
  • threshold:检测一条直线所需的最少交点数量。
  • minLineLength:构成一条直线所需的最少点数。点数少于此值的直线会被忽略。
  • maxLineGap:两个点之间被视为同一条直线的最大间隔。

然后通过绘制直线来显示结果。

C++:

// Draw the lines
for( size_t i = 0; i < linesP.size(); i++ )
{
Vec4i l = linesP[i];
line( cdstP, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(0,0,255), 3, LINE_AA);
}

Java:

// Draw the lines
for (int x = 0; x < linesP.rows(); x++) {
double[] l = linesP.get(x, 0);
Imgproc.line(cdstP, new Point(l[0], l[1]), new Point(l[2], l[3]), new Scalar(0, 0, 255), 3, Imgproc.LINE_AA, 0);
}

Python:

# Draw the lines
if linesP is not None:
for i in range(0, len(linesP)):
l = linesP[i][0]
cv.line(cdstP, (l[0], l[1]), (l[2], l[3]), (0,0,255), 3, cv.LINE_AA)

C++:

// Show results
imshow("Source", src);
imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst);
imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP);

Java:

// Show results
HighGui.imshow("Source", src);
HighGui.imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst);
HighGui.imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP);

Python:

# Show results
cv.imshow("Source", src)
cv.imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst)
cv.imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP)

C++:

// Wait and Exit
waitKey();
return 0;

Java:

// Wait and Exit
HighGui.waitKey();
System.exit(0);

Python:

# Wait and Exit
cv.waitKey()
return 0

hough_lines_result1.png hough_lines_result2.png

使用一幅输入图像,例如这幅数独图像。 使用标准霍夫线变换,我们得到如下结果:

使用概率霍夫线变换,我们得到如下结果:

你可能会观察到,当改变阈值时,检测到的直线数量会发生变化。原因显而易见:阈值设得越高,检测到的直线就越少(因为需要更多的点才能判定一条直线)。