基于多特征融合的车辆检测算法

摘要

提出了一种综合利用车辆阴影、车辆边缘和对称性3种特征进行车辆检测的算法:首先,进行车辆阴影检测,它提供了可能包含车辆的初始搜索区域;随后,进行的车辆边缘检测在排除部分虚警的同时,进一步缩小了车辆区域的范围;最后,通过对称性检测实现车辆的准确定位与检测.实验表明:该算法具有较低的误检率.%A method to detect vehicles based on multi-features: shadows, edges and symmetry is presented. Shadows are first used to mark areas that may have a vehicle,edges then are used to eliminate some areas without vehicles and shrink the areas, symmetry is used at last in order to locate the vehicles. Experiments show that this method can detect vehicles with low error rate.