基于计算机视觉的自动驾驶汽车车道检测

Detection of Self-Driving Car Lane Based on Computer Vision

提出一种基于车道线宽度的边缘检测方法。首先将道路图像转换为鸟瞰图,在此基础上进行图像灰度化处理,并选择图像中包含车道线信息的区域作为感兴趣区域;利用车道线有固定宽度的特征,检测感兴趣区域内的图像像素值跳变的间隔宽度,并与车道线的固定宽度相比较,得到车道像素点;通过抛物线车道模型,利用随机抽样一致性算法,拟合图像上的像素点,得到实际的车道线;利用Matlab及其自动驾驶工具进行仿真试验。试验证明:上述车道检测跟踪方法能对车道线进行有效跟踪,为自动驾驶汽车的横向控制提供了导航信息。

In this paper,An edge detection method based on lane line width is proposed.Firstly,the road image is transformed into aerial view,then,the image is grayscale processed,lastly,the region containing lane information in the image is selected as the region of interes,the interval width of image pixel value jump in the area of interest is detected by using the feature of lane line with fixed width,and the lane pixel point is obtained by comparing with the fixed width of lane line,and the actual lane line is obtained by parabolic lane model,using random sampling consistency algorithm to fit the point on the image,and the simulation test is carried out by using Matlab and its automatic driving tool,The test results show that the above lane detection and tracking method can effectively track the lane line and provide navigation information for the lateral control of autonomous vehicles.