基于视频序列的车道线检测和跟踪

Video-based lane detection and tracking algorithm

车道线检测是无人驾驶系统以及一系列辅助驾驶系统的关键技术环节。传统的车道线检测方法容易受到其他交通标志线干扰,提出了一种新的车道线检测与跟踪方法。该方法首先使用自适应阈值算法提取特征,通过ROI二次设置以及跟踪区域规划,逐步减小感兴趣区域,最后将感兴趣区域内的特征点从图像坐标系转换到世界坐标系下,以最小二乘方法进行曲线拟合。在高速公路及城区道路等多种工况下的实验表明,提出的方法能够正确实时的识别出车道线,有效的消除了其他交通标线的干扰。

The lane detection is a critical technology for unmanned driving system and a series of driving assistance systems. Traditional lane detection algorithms are easily disturbed by the other traffic lines on the road. A new method for lane detection and tracking is put forward in this paper. First, the adaptive image threshold algorithm is adopted for lane features extraction. By means of setting the ROI twice and planning the tracking region, the ROI decreases gradually. At last,the feature points in the ROI are transformed from the image coordinate system to the real-world coordinate system before they are used to accomplish curve fitting through least square method. Experiments under various conditions,such as highway and urban environments, show that the method proposed in this paper is able to accurately recognize the lane marks in real time and effectively eliminate the disturbance of the other traffic line markings.