摘要： 针对视频中噪声点难以消除和路面背景难以提取的问题，提出一种基于车辆平均轨迹的车道提取方法。对视频中的车辆进行跟踪，得到它们在每一帧中的位置，将这些位置中心点根据算法拟合成一条平均轨迹。从视频帧中选择出合适的一帧作为背景图，对背景图进行形态学处理以消除天空及两边景物等干扰较大的噪声点，并通过进一步处理得到二值化的背景图。将平均轨迹与二值化图进行匹配，提取出车道。实验结果表明，该方法适用于多种道路且不易受到噪声点的干扰，成功率较高并具有较强的灵活性和适应性。

关键词: 图像处理, 车辆跟踪, 平均轨迹, 形态学处理, 背景提取, 车道提取

Abstract: On account of the problem that noise points are various and the background is difficult to be extracted in the videos, the lane extraction method based on average vehicle trajectory is proposed. Vehicle positions are acquired in each frame by tracking and the center points of these positions is fit into an average trajectory. One suitable frame is selected from video frames as background and morphologic processing is done so as to eliminate the influential noise points such as the sky or scenery on both sides, and the binary background is obtained by further processing. The lane is extracted by matching the average trajectory with the binary background. Experimental results show that this method can apply to varieties of roads, and it has high success rate and great flexibility and adaptability.

Key words: image processing, vehicle tracking, average trajectory, morphologic processing, background extraction, lane extraction

中图分类号: 

• TP391.4