基于机器视觉的车道偏离报警系统  

Lane departure warning system based on machine vision

作　　者：于兵[1] 张为公[1] 龚宗洋[1] 

机构地区：[1]东南大学仪器科学与工程学院,南京210096

出　　处：《东南大学学报（自然科学版）》2009年第5期928-932,共5页Journal of Southeast University：Natural Science Edition

基　　金：江苏省交通科学研究计划资助项目(06C04);苏州市科技发展计划资助项目(SG0723)

摘　　要：为了提高基于机器视觉车道偏离报警系统的可靠性和实用性,对基于视觉的车道偏离报警系统各个环节的优化做了研究.介绍了基于视觉的车道偏离报警系统的构成和工作原理,提出了各个环节的实现方法.通过选择直线车道数学模型和限定车道提取的感兴趣区域(ROI)以简化系统复杂度和提高检测精度.首先使用方向可调滤波器进行图像预处理,然后使用Kalman预测器和距离判别法得到车道线有效点集,最后采用抗干扰能力强的Hough变换得出车道线参数.研究并采纳了一种无需对摄像头标定的车道偏离决策方法,通过综合道路图像中2条车道线的斜率值来判断车辆偏离车道的程度.实验表明,该系统具有良好的车道识别能力以及准确的偏离决策能力,能够满足高速公路环境车道偏离报警要求.Several studies of optimization were carried out to improve the practicability and stability of lane departure warning system （LDWs）based on machine vision. The system configuration and work flow of LDWs are introduced and the realization of each segment is described. The linear lane model and region of interest （ROI） are adopted to simplify the system structure and improve the system accuracy firstly. Then, by means of preprocess of image based on steerable filter, tracking method based on Kalman prediction and gathering lane edge points solution through distance discrimination way, the lane is extracted robustly. Finally, the decision method about lane departure decision without calibrating the camera is studied. Experiments show that the developed system exhibits good performances in recognition reliability, and warning decision, which can satisfy the requirements of lane departure warning system in the structured highway environment.

关 键 词：车道偏离报警 方向可调滤波器 卡尔曼跟踪 偏离决策 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]