基于车道线宽度滤波和抽样行扫描的车道线检测

Lane detection based on lane marking width filtering and sampling line scan

车道线检测是汽车驾驶安全辅助系统的关键技术之一.针对现有检测算法存在的算法复杂度高、检测结果不稳定且容易受周围车辆和远方景物影响的问题,提出了一种基于车道线宽度滤波和抽样行扫描的应用于高速或快速公路的车道线快速检测算法.该算法以单目摄像机拍摄的视频为研究对象,首先对视频图像进行预处理生成二值图像,其次进行车道线宽度滤波去除非车道线噪声,然后采用抽样行扫描和特征点筛选的方法确定拟合特征点,最后采用最小二乘拟合的方法提取车道线.车道线宽度滤波和特征点筛选步骤保证算法的准确性,基于抽样行扫描的车道线识别方法保证算法实时性.实验结果表明,本算法在不同场景下平均准确率达96%以上,具有很好的鲁棒性和实用性.

Lane detection is a key technique of vehicle driving safety assistant system.The existing algorithms are easily to be interfered by surrounding vehicles and far distance scene.Aiming at the high complexity and instability of existing algorithms,the paper proposes a lane detection algorithm based on lane marking width filtering and sampling line scan for highway and thruway.The real-time videos are gotten by the monocular camera installed in the driving car.Firstly,image preprocessing is applied to produce binary image and a kind of filtering algorithm based on lane marking width is proposed to remove non-lane markings.Secondly,the sampling line scan and the sorting method are used to generate lane feature points.At last,lane marking is extracted from binary image using the least square fitting.The steps of lane marking width filtering and lane feature points sorting ensure the accuracy of proposed algorithm,while the lane marking identification method based on sampling line scan ensures the real-time performance of proposed algorithm.The experiment results show that the average accuracy is over 96% in different scenes and the proposed algorithm is robust and practical.