基于改进ViBe算法与三帧差法的运动检测算法

A novel moving objects detection algorithm based on improved ViBe and three-frame differencing algorithm

针对ViBe(visual background extractor)算法存在的鬼影和漏检问题,文章提出了一种基于改进ViBe算法和三帧差法的运动目标检测的方法。首先针对ViBe算法检测结果存在鬼影的问题提出一种改进的方法,改变ViBe算法中仅通过第1帧建立背景模型的方式;该文在前5帧中每个像素点的邻域随机选取4个像素点建立一种具有时间和空间信息的背景模型进行运动检测,并采用"或"类型三帧差法做"或"运算改善漏检的问题;然后对显著性检测结果做"与"运算去除过检点以提高运动目标检测的准确性;最后进行适当地后处理得到最终的检测结果。该算法能够去除噪声、抑制鬼影以及减少空洞点,实验结果表明能够快速并准确地检测出运动目标。

A moving objects detection algorithm based on improved visual background extractor(ViBe)and three-frame differencing algorithm is proposed to deal with the ghost and missed detection problems.Firstly,an improved ViBe algorithm is proposed to deal with the ghost problem,changing the method of setting up background model by the first frame of ViBe.In the improved algorithm,the background model is set up from the first five frames and four pixels in the neighborhood are selected randomly.The background model like this has spatial and temporal information.Then the missed detection problem is improved by the“OR”process with three-frame differencing algorithm.After that the“AND”process is taken with the salient object detection result to filter the over-detecting pixels.Finally,the final moving objects detection results are obtained after the proper post-processing.The proposed algorithm can remove noise,avoid ghost,decrease holes and detect moving objects quickly and accurately.