基于扩展卡尔曼滤波器的车道线检测算法  

Lane detection algorithm based on extended Kalman filter

作　　者：彭红[1] 肖进胜[1,2] 程显[1] 李必军[2] 宋晓[3] 

机构地区：[1]武汉大学电子信息学院,湖北武汉430072 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [3]中国电子科技集团公司第10研究所,四川成都610036

出　　处：《光电子．激光》2015年第3期567-574,共8页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(91120002;61471272)资助项目

摘　　要：提出一种将道路结构模型信息与扩展卡尔曼滤波器(EKF,extended Kalman filter)相结合的车道线检测算法。基于扫描线的自适应边缘检测算子进行边缘点的检测,针对车道模型建立了适合算法的自定义参数空间,进行边缘点的投票,提取出候选车道线,解决了传统Hough变换中处理速度慢的问题。根据道路几何学和车辆动力学建立新的车道模型,增加了车道信息待估计的参数,并利用车道线的特征约束排除干扰线得到车道线的内边界,结合EKF对车道线边界点坐标参数进行跟踪估计,以保证算法的稳定性与鲁棒性。实验结果表明,本文算法能够处理绝大多数的复杂车道情况,在实时性、鲁棒性和检测率上都取得很好的效果。Aiming at the problem of low lane detection accuracy and robustness,which is interrupted by strong light,shadows,non lane road markings and so on,this paper presents a new algorithm combined with information obtained from roadside structures model and extended Kalman filter（EKF）.It detects edge points using adaptive edge detection operator based on scan lines.In view of the roadside model,custom parameter space which is adapted to the algorithm in this paper is established.It can extract all candidate lanes by voting the edge points.The custom parameter space solves problem of slow processing speed in the traditional Hough transform.According to the roadway geometry and vehicle dynamics,a new driveway model is set up.The parameters of driveway information to be evaluated are increased.At the same time,we get the lane inner boundary by using the various constraints of the lane to eliminate interference lines.Tracking and estimating coordinate parameters of lane edge points are following that,combined with the extended kalman filter,which can ensure the stability and robustness of the algorithm.The results indicate that the introduced algorithm is robust in various complex road scenes and gets good results in time complexity robustness and performance of lane detection.

关 键 词：智能交通 车道线检测 车道模型 扩展卡尔曼滤波器(EKF) 

分 类 号：U491.116[交通运输工程—交通运输规划与管理]