车型识别系统中利用阴影特点快速定位车体

A Fast Method Based on Shadow for Localization Vehicle in Vehicle Recognition System

阴影在车型识别系统中对于车体的准确定位产生很大的影响 ,为了能准确地定位车体 ,提出一种快速定位阴影与车体之间边界的算法即根据平行投影知识和成像原理的方法来定位阴影与车体之间的边界 ,同时利用该算法也能找出货车本身所具有的特征如货车的车头和车尾的定位 ,利用所寻找的特征点也可以进一步判断车所属的大致类型。利用该算法进行了大量的实验 ,从实验结果来看 ,该算法具有较好的稳定性和有效性。

It is a very important research for Vehicle Recognition System (for short VRS) in Intelligent Transportation System (for short ITS). However, it is influenced for the accurate of localization vehicle according ot the shadow.The method we provide for locating shadow is based on the parallel projection and imaging principle. We adopt vertical distance difference projection and horizontal distance difference projection for locating the boundary between the vehicle and the shadow, and the feature point in the contour of shadow, respectively. This method can also find the feature of the vans, such as vehiclehead and compartment. We can also apply these features to classify the different classification of the vehicle. Very promising experimental results are provided using real time vehicle image sequences. Experimental results have confirmed that this proposed localization method is efficient and reliable, and its computation cost can also satisfy the VRS.