改进的ViBe算法及其在交通视频处理中的应用

基于ViBe算法的运动目标检测具有很高的处理效率,但在复杂背景下也存在缺陷,比如不能有效抑制运动目标的残影和鬼影区域、易受噪音干扰,在交通视频的处理中更为明显。鉴于ViBe算法的这些不足,本文对ViBe原算法提出了改进,结合像素生命长度对鬼影或残影像素进行判断,利用二次更新策略提高消除速度。根据像素值变化引入最大类间方差法,将原有的固定像素阈值变为动态阈值,以提高交通视频中车辆目标检测的抗干扰能力。实验结果表明,在3组不同交通场景视频中,改进算法能够在较少帧数内消除目标的残影或者鬼影。对结果的量化分析也证明了改进算法由于应用了动态阈值,使目标检测的精准度和整体性都有所提高。

An improved ViBe algorithm and its application in traffic video processing

Moving object detection based on the ViBe algorithm has high processing efficiency,but also some defects under complicated background,for example,it is unable to effectively suppress the ghost shadows and residual shadows of moving objects,vulnerable to noise,which is more apparent in processing of traffic videos.Subject to such deficiencies,the original ViBe algorithm was improved in the paper,as in to make judgment on pixel of ghost shadows and residual shadows by using life length matrix of pixel,to apply second update strategy to improve shadow elimination speed.According to changes of pixel value,OTSU method was introduced to transform the original fixed pixel threshold value into dynamic threshold and to promote the anti-jamming capability of vehicle objects detection contained in traffic videos.The results show that among the three groups of traffic scene video,the improved algorithm is able to eliminate the ghost shadows and residual shadows of moving objects within few frames.The quantitative analysis on results also proved the application of dynamic threshold in such improved algorithm could promote the accuracy and overall performance of object detection in traffic videos.