基于二维时空图像分析的车辆检测和分割方法

Vehicle Detection and Separation Based on 2D SpatioTemporal Image Analysis

使用二维时空图像分析法 ,能够较好地解决光照变化和阴影给车辆分割带来的困难 ,且该方法的计算量小 ,适用于低成本的实时交通监控系统 .通过道路上方安装的摄像机获得图像序列 ,构成全景图和外极面图两种二维时空图像 .对它们进行背景减除、图像差分和取阈值等处理 ,不仅可以统计车流量 ,而且可以估计车速和道路占用率 .由于算法考虑了日照阴影和晚间灯光对图像的影响 。

A successful visual traffic monitoring system must meet two critical requirements——robust and low cost. Those problems on the vehicle separation caused by illumination changes or shadows can be easily settled when using 2D spatio temporal image. It is also suitable for the real time operation with inexpensive system. Image sequence can be acquired through the camera mounted over the road, and then two 2 D ST images, a panoramic view image (PVI) and an epipolar plane image, are formed for each lane. The vehicle can be extracted and counted in the PVI. By analyzing these images, it can not only count the vehicles but also estimate the vehicle speeds and lane occupancy. Having taken the different light conditions into account, including shadows in daytime and lights at nights, the algorithm is robust and effective.