一种用于车辆检测的选择性背景更新方法

A selective background updating method for vehicle detection

针对传统背景更新方法实时性差,不能及时正确地处理背景局部突变的问题,文章提出一种用于车辆检测的选择性背景更新方法。该方法采用改进的对称差分法和背景帧差相融合检测车辆运动区域,并使每一个运动目标成为独立的连通域;采用基于两轮扫描法的种子填充技术,解决由于车辆表面与路面灰度接近而产生的运动目标"孔洞"问题;在此基础上对传统的选择性背景更新方法进行改进,对运动目标区域和非运动目标区域采用不同的更新方法进行更新。实验结果表明,所提出的改进方法具有良好的实时性,能够有效解决背景局部突变的问题,提高了车辆检测的准确性。

To solve the problems that traditional background updating methods have poor real-time performance and can not timely and correctly handle the local background mutation,this paper presents a selective background updating method for vehicle detection.The methods of improved symmetric difference and background difference are integrated to detect vehicle movement areas,and to make each of moving target become an independent connected domain.The seed filling technology based on twice scanning method is used to solve the problem of ＂hole＂ in moving targets caused by the similar gray scale of vehicle and road surfaces.The traditional selective background updating method is improved,and different updating methods are used to update the moving target area and the non-moving target area.Experimental results show that the improved method has good real-time performance and can effectively solve the problem of local background mutation,thus improving the accuracy of vehicle detection.