一种多阈值Hough变换车道线检测算法

A Multi Threshold Hough Transform Lane Detection Algorithm

针对传统Hough变换虚线检测率不足的问题,提出一种多阈值Hough变换车道线检测算法。该算法在对图像进行灰度化处理、逆透视变换、二值化处理的基础上,在预设好的多条直线位置进行突变点检测,并对突变点进行分类、拟合、合并,最后进行Hough变换。3种实际路况的实验表明,该算法能够较准确、稳定地检测出车道线,平均识别率达到98.70%,高于传统Hough直线检测算法的平均识别率(86.84%),而且可通过计算车道线线段的长度和点的个数来判断虚线和实线。

To improve the low detection rate of traditional Hough transform dotted lines,a multi-threshold Hough transform lane line detection algorithm is proposed.The algorithm performs grayscale processing,inverse perspective transformation,and binarization processing on the image,then detects mutation points in multiple preset straight line positions,and then classifies,fits,merges and finally Hough transforms the mutation points.Experiments in three real road conditions show that the algorithm can detect lane lines accurately and stably,with an average recognition rate of 98.70%,which is higher than that of the traditional Hough line detection algorithm,and it can distinguish the dotted from solid lines by calculating the length and dot number of the lane line segment.