直线模型下的车道线跟踪与车道偏离检测  

Lane Tracking and Departure Detection Based on Linear Model

作　　者：余厚云[1,2] 张为公[1] 

机构地区：[1]东南大学仪器科学与工程学院,江苏南京210096 [2]南京航空航天大学机电学院,江苏南京210016

出　　处：《自动化仪表》2009年第11期1-3,7,共4页Process Automation Instrumentation

基　　金：江苏省汽车工程重点实验室开放基金资助项目(编号:QC200603);江苏省交通科学研究计划资助项目(编号:06C04)

摘　　要：为了解决规格化道路上车道线跟踪及车道偏离检测的问题,利用Kalman滤波器来动态确定感兴趣小窗口的大小和位置。首先,在小窗口内采用Hough变换方法进行车道线识别;同时,根据摄像机的成像几何性质,推导出车道偏离程度与道路图像中车道线斜率之间的函数关系,从而简化了摄像机标定过程。现场试验表明,完成一帧道路图像的预处理及车道线识别的所需时间小于30 ms,车辆直行情况下的车道偏离率相对测量误差小于5%,试验结果验证了该方法的实时性和正确性。In order to implement lane tracking and departure detection on standardized road, the size and location of the small window in region of interest （ ROI ） are determined dynamically by adopting Kalman filter. Firstly, the line of lane is identified in small window with Hough trans- form （HT）, then in accordance with the geometric characteristics of the images from video camera, the functional relationship between lane drift degree and the slope of lane line in image is derived, thus the calibration for the camera can be simple. The test on site shows that the time period for pre-processing one frame of lane image and identifying lane line is less than 30 ms, the relative measuring error for straight running vehicle is less than 5%. The real-time performance and correction of the method have been verified by the test results.

关 键 词：直线模型 车道线 HOUGH变换 KALMAN滤波 针孔模型 车道偏离 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]