摘要
利用车载机器视觉系统识别车道线以确定运动车辆相对道路的横向位置,是目前高速公路等规格化道路上实现车道保持的主要手段。通过理论计算确定采用车道线线性模型,能够满足识别的精度要求,同时提高了系统的实时性。卡尔曼滤波器动态确定感兴趣小窗口的大小和位置,实现车道线自动跟踪,使得图像预处理和车道线的Hough变换识别只在小窗口内进行,降低了计算成本。现场跑车试验结果表明,一帧道路图像的预处理和车道线识别与跟踪时间小于30 ms,且系统对道路上其他运动车辆等干扰具有较强的鲁棒性。
Lane detection and tracking with machine vision system is widely used to recognize the relative location of moving vehicle on the structural road. In the recognition, the linear road model can give a real time result and is proved to be precise by theoretic calculation. Pretreatment of the picture and Hough transform （HT） for lane detection are both carried out in region of interest（ROI） composed of two dynamic windows. The size and location of the windows are decided by Kalman filter. Practical experiments indicate that the time of detection for one frame of picture is less than 30 ms, and the tracking is robust against the disturbance of other cars.
出处
《工业仪表与自动化装置》
2009年第5期103-106,共4页
Industrial Instrumentation & Automation
基金
江苏省汽车工程重点实验室开放基金资助项目(QC200603)
江苏省交通科学研究计划项目(06C04)
