一种基于双背景模型的遗留物检测方法

摘要

Most abandoned object detection algorithms proposed are either computationally intensive or weak in tracking abandoned object with occlusion. For this, an abandoned objects detection method under static cameras is presented based on dual backgrounds, which are updated with different strategies and frequencies. The background model is driven by accumulative average values instead of complex filters and hence it is simple and of lower computational cost. Besides, the algorithm is able to track abandoned objects under occlusion. Experimental results under the benchmark datasets on PET 2006 show that the proposed method has a good performance.%针对一般遗留物检测算法运算量大和难以适应遮挡情况的问题,提出了一种静止单摄像机条件下快速有效的遗留物检测算法.算法建立了两个基于累积均值更新法的背景模型,分别称之为纯背景模型和脏背景模型.通过两个背景的差别得到静止目标块,并对静止目标块进行跟踪,当静止目标停留超过设定的时间即判定其为遗留物并触发报警.由于算法避免了使用复杂度数学概率背景模型,大大减低了背景更新的计算复杂度,使算法能满足视频监控系统实时处理的要求.同时,算法在静止目标跟踪模块中增加了碰撞帧数计数使遮挡情况下的遗留物跟踪得到更好的效果.在PETS2006数据集提供的多个视频序列实验中,该算法显示了良好的性能.