基于熵最大化图像分割的直线型车道标识识别及跟踪方法

为了解决在道路路面材料不一致或光照不均匀情况下的车道标识线的识别和跟踪问题,提出一种基于熵最大化的图像分割可变形模板的车道标识线识别及跟踪方法.该方法结合了图像变窗口处理技术和基于熵最大化分割方法来实现对道路图像的理想分割,然后利用可变形模板匹配方法得到车道标识线参数,最后采用建立梯形感兴趣区域的方法实现对车道标识线的实时跟踪.试验结果表明：该方法具有很好的可靠性、鲁棒性和实时性.

Linear Lane Identification and Tracking Method Study Based on Maximum Entropy Image Segmentation

To realize lane mark identification and tracking under such conditions as uneven road surface materials and different illumination etc. a lane mark identification and tracking method based on maximum entropy image segmentation is proposed. The authors use this method combined with image window variation technology and maximum entropy segmentation method to divide up the lane mark of road image, then to get the lane mark parameters by the match technique based on the bi-nomalized adjustable template, finally to track the lane mark in real-time by establishing trapezoid areas. Experimental results show that the proposed method is characterized by reliability, good robustness and real time.