基于熵最大化边缘提取的直线型车道标识线识别及跟踪方法  

Linear Lane Mark Identification and Track Method Based on Entropy Maximization Edge Extraction

作　　者：贾阳[1] 王荣本[2] 余天洪[2] 金立生[2] 

机构地区：[1]哈尔滨工业大学航天学院,哈尔滨150001 [2]吉林大学交通学院,长春130022

出　　处：《吉林大学学报（工学版）》2005年第4期420-425,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家自然科学基金资助项目(50175046);中国博士后科学基金资助项目(2004036397)

摘　　要：为了得到较为理想的车道标识线边缘图像以便于更好地对车道标识线进行识别,提出了一种基于熵最大化的边缘提取的车道标识线识别及跟踪方法。该方法结合了基于一维熵最大化分割方法和基于二维直方图的熵最大化分割方法,利用图像变窗口处理技术实现对道路图像中车道标识线边缘的分割,然后利用基于统计的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。To obatain more ideal edge image for lane mark idenfitication, a lane mark identification and track method based on entropy maximization edge extraction was proposed. This method combined both one-dimensional and two-dimensional histogram entropy maximization partition methods, and image window variation technology was employed to partition the lane mark edge of the road image, then the lane mark parameters were acquired by the statistical method, finally the lane mark real-time track was performed by estabishing trapezoid areas of interest. Experimental results show that the proposed method is characterized by realiability, good robustness and real-timeness.

关 键 词：交通运输系统工程 一维熵 二维熵 边缘分割 车道标识识别 车道标识跟踪 

分 类 号：U495[交通运输工程—交通运输规划与管理]