摘要
为满足车辆行驶时能对各种车道线(实线、虚线、直道、大弯道)准确识别,提出一种基于Meanshift原理和RANSAC(Random Sample Consensus)算法的车道识别方法;该方法首先利用改进的最大熵阈值分割方法和图像灰度概率密度特征对左右车道线目标进行初定位,动态地建立车道线ROI(Region of Interests),然后运用Meanshift算法对左右车道线进行精确定位,最后利用RANSAC算法对各搜索框中候选车道线的重心进行筛选,并采用最小二乘法对左右车道线进行拟合;实验结果表明,该方法可以识别各种车道线型,并具有较好的鲁棒性;车道检测平均时间为80ms/f,车道跟踪平均时间为40ms/f。
In order to meet vehicle request for various lane markers （full line, dotted line, straight lane, curved line） recognition, a lane recognition method based on Meanshift principle and RANSAC algorithm is proposed. Firstly, lane marker objects are initially detected with applying improved maximum entropy image segmentation technology and intensity probability density of image. Then, the ROI is built dynamically and the left and right lane markers are detected with using Meanshift principle. With the candidate barycenter of lane marker in each detecting window, RANSAC algorithm and least square method are used to fit curved lane. Experimental results show that this method can recognize various lane markers and has better robustness. The load detectiong speed is about 80ms/f and load tracking speed is about 40ms/f.
出处
《计算机测量与控制》
北大核心
2013年第5期1344-1347,共4页
Computer Measurement &Control
基金
国家自然基金面上项目(61175075
51075137
91120306)
国家863项目(2012AA112312)
江苏省汽车工程重点实验室开放基金项目(QC201002)
资助中央高校基本科研业务费
