融合ViBe与帧差法的交叉路口多车辆检测方法

Multi Vehicle Detection Based on ViBe and Frame Difference Method

摘要:

前景检测的精确程度是交叉路口车辆检测的重要因素,传统的基于背景建模的前景检测方法存在拖影现象,并且通常难以分辨出无牌车辆,针对上述问题,提出一种融合ViBe与帧差法的前景检测算法,并在此基础上结合车牌检测算法来检测场景当中的车辆;首先,结合帧差法和ViBe算法对背景像素点的判定结果,采用不同更新因子更新背景模型,其次,使用一种多条件过滤车牌检测算法定位运动区域当中车牌,最后,以检测到的车牌中心为锚点,定位出最终车牌区域;实验结果表明,该前景检测算法可有效应对交叉路口场景下的前景检测的拖影现象,同时车辆检测算法可以准确检测出进入场景时的车辆,并分辨出无牌车辆。

Abstract:

The accuracy of foreground detection is an important factor to the intersection of vehicle detection. There’s ghost phenomenon in traditional foreground detection method based on background modeling, and it’s difficult to distinguish unlicensed vehicles. Aiming at the above problems, put forward a foreground detection method based on ViBe and frame difference, and combining with license plate detection result to detect vehicles in scene. Firstly, using different factor to update background model depend on the result of frame difference method and the ViBe algorithm. Secondly, a multi condition filter is applied to detect license plate in foreground. Finally, take the center of license plate as the anchor to locate vehicle area. The experimental results show that multi vehicle detection based on ViBe and frame difference method can effectively deal with the ghost phenomenon in intersection scene, at the same time the vehicle detection algorithm can accurately detect vehicles enters the scene, and identify unlicensed vehicles.