基于多层码本模型的遗弃行李检测算法

提出一种在公共场所智能检测遗弃行李的新算法框架。首先使用多层码本模型检测场景中的稳定前景区域作为可疑区域,之后结合梯度方向直方图特征设计一种消除光照影响的机制,最后使用同样基于梯度方向直方图的快速行人检测算法消除行人的影响。实验结果表明,该方法计算复杂度低,在一些标准数据库上能有效检测遗弃行李,并对复杂场景有较好的鲁棒性。提出的算法框架可以应用到公共场所的监控系统中。

Abandoned luggage detection method based on multi-layer codebook model

This paper proposed a novel framework for detecting abandoned luggage in public places intelligently. Firstly created a multi-layer codebook model to detect the stable foreground of the scene as the suspicion regions. Then designed a new mechanism combined with HOG features to eliminate the influence of illumination change. At last, applied a human detector also based on HOG to eliminate the false alarms caused by pedestrians. Experiments demonstrate the algorithm is with low computing complexity and effective on standard databases. The method is also robust in complex scenes. The algorithm framework proposed in this paper can be applied in the monitoring system of public places.