Statistical Hough transform lane detection method based on gradient constraint

Abstract

A statistical Hough transform lane detection method based on gradient constraint solves the problem that the standard Hough transform depends on the edge detection results in lane detection, and utilizes a Gaussian kernel function to carry out modeling on each pixel in an image without edge detection. An initial data set of the statistical Hough transform is restrained through a gradient threshold method, and finally, a continuous probability density function of lane parameters is obtained. Under the highway environment, the method can detect the lane quickly and accurately, and has very strong robustness.

Description

technical field [0001] The invention relates to the field of lane detection, in particular to a gradient-constrained statistical Hough transform lane detection method. Background technique [0002] With the growth of the number of cars, the rate of accidents on highways has skyrocketed, most of which are caused by drivers' ignorance of their surroundings or visual disturbances. Vision-based lane detection is a key technology for safe driving assistance, and can be used in many aspects such as lane departure warning, adaptive cruise control, and automatic driving. [1] . A lot of research on lane detection has been done at home and abroad. [2-9] . my country's highway technical standards stipulate that the minimum flat curve radius of the expressway is 650m, and the radius of curvature of the lane line is 650m. The lane image within 40m from the viewing plane in front of the vehicle can be approximated as a straight line, and this approximation can be established in most ca...

AI Technical Summary

Problems solved by technology

Although the standard Hough transform method is simple and has good results on discontinuous lane lines, it must be applied to the edge image. The quality of the edge detection result directly affects the final detection result. Even if there is a good edge image, individual A limited number of edge points will also cause the parameter histogram in the standard Hough transform to be sparse, which will bring trouble to the final search for the parameter peak. [2]

Images