基于路径搜索的道路检测算法

Road detection algorithm based on path search

针对有车道线和无车道线的结构化道路检测方法,提出一种基于路径搜索的道路检测算法。首先选择感兴趣区域减少计算量,采用像素分割算法提取车道线起始搜索点并确定搜索框,基于梯度、像素、纹理变化进行特征融合,根据特征变化大小从起始搜索框出发进行搜索,最后对选取点进行曲线拟合形成车道线。该算法比基于Hough变换的道路检测算法具有鲁棒性更强、在弯曲道路线检测效果更好的特点,在有、无车道线的结构化路面均表现良好。

Aiming at the structured road detection method with and without lane lines,a road detection algorithm based on route search is proposed.Firstly,the interested area was selected to reduce calculation amount.The pixel segmentation algorithm was used to extract the starting search point of the lane line and determine the search box.The feature fusion was carried out according to the gradient,pixel and texture changes.The search was carried out from the starting search box according to the size of the feature change.Finally the lane line was formed by the curve fitting of selected points.Compared with the road detection based on Hough transform,the algorithm has the characteristics of more robustness and better performance in curved road line.The algorithm shows good results in structured road surfaces with and without lane lines.