一种改进的ViBe运动目标检测算法

An improved ViBe moving object detection algorithm

为抑制传统ViBe算法在检测运动目标时产生的"鬼影",提高监控视频运动目标检测的准确性,提出了一种改进的ViBe运动目标检测算法。该算法采用连续相邻的多帧图像序列代替传统ViBe算法中第一帧图像,构建背景模型,从根源上解决传统ViBe算法在运动目标检测中存在的"鬼影"问题。利用Canny边缘检测算子和形态学运算相结合的方式准确完整的提取运动区域,降低算法的复杂度且减少运动区域的提取时间。提出一种背景模型更新策略判定条件,提高背景模型的质量,消除高频扰动对运动目标检测的影响,从而实时保证运动目标检测的准确性和鲁棒性。实验结果表明:经过4种算法的对比,改进的ViBe算法能够有效的抑制鬼影,且在高频扰动的情况下能较好的适应动态背景,显著提高运动目标检测的准确性。

In order to restrain the ghosting phenomenon generated by the traditional ViBe algorithm in detecting moving objects,and to improve the accuracy of moving object detection in surveillance video,an improved ViBe moving object detection algorithm is proposed in the paper. In the presented algorithm,continuous multi-frame image sequence is used to replace the first frame image in the traditional ViBe algorithm in order to construct the background model,which is to solve the problem of ghosting phenomenon in moving object detection based on traditional ViBe algorithm fundamentally. By using Canny operator edge detection and morphological operations,the motion region is extracted accurately and completely. The complexity of the algorithm and the extraction time of the moving region are reduced. This paper presents a background model update strategy decision conditions,which can improve the quality of background model and eliminate high frequent disturbance of the moving target detection. It can ensure the accuracy and robustness of moving object detection in real time. The experimental results show that through the comparison of four algorithms,the improved ViBe algorithm can restrain ghost phenomenon effectively. Under the condition of high frequency disturbance,it can better adapt to the dynamic background,and significantly improve the accuracy of moving object detection.