中文摘要：

现有基于视频帧的车道背景建模方法建模过程较复杂,且易受光照、遮挡等因素的影响。提出一种基于多帧统计的视频车道背景建模方法,通过对多帧视频帧自主统计分析,首先建立无车的全背景图像;然后再次对多帧视频帧进行运动对象的位置统计,最终在全背景图像上获取完整的车道背景图像。该方法能有效确定视频中的背景区域,特别是能明确车道背景区域。算法思想简单,容易实现。实验结果表明该算法具有计算量小、车道检测完整、对光照的变化具有一定的自适应能力等特点。

英文摘要：

Existing video frames-based background modelling methods for traffic lane are very complicated in processing and are easily affected by the factors of illumination and occlusion,etc.This paper proposes a background modelling method for lane which is based on multi-frame statistics.With independent statistical analysis on multiple video frames,the full background image without cars is created first,and then the statistical analysis on position of the moving objects in multi-frame video frames is conducted so as to get the complete image of lane background from full background image at last.The method can effectively determine the background region in video,especially the background region of traffic lane.The algorithm is simple and easy to be realised.Experimental result shows that this algorithm has the characteristics of low computation load,complete detection on traffic lane,as well as certain self-adaptive capability to illumination variation,etc.