智能汽车中基于视觉的道路检测与跟踪算法  

Research on Vision-Based Lane Detection and Tracking for Intelligent Vehicles

作　　者：刘富强[1] 田敏[1] 胡振程[2] 

机构地区：[1]同济大学信息与通信工程系,上海200092 [2]日本熊本大学计算机工程系,熊本8608555

出　　处：《同济大学学报（自然科学版）》2007年第11期1535-1541,共7页Journal of Tongji University:Natural Science

基　　金：科技部国际合作基金资助项目(2005DFA10100)

摘　　要：提出了一个基于视觉的道路环境识别算法.该算法能实时提取图像序列中道路信息,获取车辆在道路上的位置、姿态信息,预测前方道路状况,将道路环境信息反馈给智能汽车用于其控制车辆的主动安全.首先联合摄像机内参构建了道路的三维数学模型,突破了以往一些系统将道路看作一个平面的二维模型的局限,算法利用道路标志线的颜色突变特性有效地提取道路边界,并将扩展卡尔曼滤波器与道路模型结合对道路和车体的状态进行实时跟踪分析,获取实时的道路环境信息.实验证明该算法可在直道和弯道以及标志线间断或标志线周围有干扰的各种道路中稳定地工作,在光影条件不利或前方有车的情形下算法仍具有较高的鲁棒性,能够适应多变的道路环境,提供实时有效的信息数据.This paper presents a new algorithm to detect and track road boundary from the scene captured in front of the vehicle. The intelligent driving system can extract from the image series the realtime road information such as vehicle＇ s pose and its displacement and predicte the state of the coming road and feeding the information back to the control system as well. A 3D road model is established on the basis of the inner parameters of the camera. The road markings are extracted successfullg from the image according to their colour characteristics. The Extended Kalman Filter combined with the road model is introduced into tracking the lane frame by frame, which makes the needed information avail- able on real time. The algorithm adapts well to various road environment straight or curve. Experimental results show that even in a complex condition with a lot of disturbance caused by shadow, vehicles in front or other white markings on road, the algorithm also acts in a robust way.

关 键 词：智能汽车 道路跟踪 卡尔曼滤波器 图像标定 自适应跟踪 

分 类 号：U476[机械工程—车辆工程]