车道线检测技术是自动驾驶领域的研究热点.通过对车道线图像基本特征的分析,研究了一种基于车道线识别的稳健方法.在初始检测部分,图像预处理中采用大津法来适应不同的光照条件,并根据车道线特征进行车道线初始检测.然后,用Hough变换检测车道线.在车道线跟踪部分,用Kalman滤波器预测参数,动态建立感兴趣区域,在此区域搜索车道线.并根据一定的失效判决模块,验证跟踪结果.最后,根据交通工具与车道线之间的距离来判决偏离情况.实验证明,该算法具有较高的实时性和稳健性.%Traffic lane detection technology has attracted increasing attention in field of autonomous driving. After the characteristic of the images of the traffic lane was analyzed, a robust approach was studied based on lane recognition, and the adaptive threshold based on Otsu algorithm was used for different illumination demands during image preprocessing . Lane was initially detected based on lane recognition. Lanes were then detected by Hough transformation. In lane tracking, the region of interest was established with parameters through the Kalman predictor. Then lanes were searched in the region. A judgement module that was responsible for estimating tracking results was introduced. Finally, whether the vehicle is depart from the lane depends on the distance between the lane and the vehicle. Experiment results indicate that the algorithm has good robustness and efficiency.
展开▼
