一种基于HGPC的交通监控视频抖动异常检测方法

A Detection Based on HGPC for Traffic Surveillance Video Shaking

在交通监控视频中,不可避免的会出现导致监控系统性能下降的视频图像抖动的异常现象。为检测此类异常现象,提出了一种基于分级的灰度投影相关系数法(Hierarchical Gray Project Correlation,HGPC)。首先,使用全局的灰度投影相关系数法对视频进行粗检测;然后,为提高检测精度,采用局部灰度投影相关系数法进行细检测。该算法在海信网络科技有限公司交通监控贵阳视频库上进行检测,精度达到91.4%,高于传统的测试方法,且检测时间仅为47ms,满足实时性要求,证明此算法有效。

In the traffic surveillance videos, the video shaking inevitably arises. The video shaking make monitor system＇s performance degradation. A hierarchical gray project the correlation algorithm （HGPC） is proposed to solve the problem of the traffic surveillance video shaking. First, global hierarchical gray project correlation algorithm （GHGPC） is used to detect the video roughly. Then, the local hierarchical gray project correlation algorithm （LHGPC） is used to deal with the results from the rough matching in order to further improve the detection accuracy. Hisense network technology co. , LTD. Guiyang intelligent traffic monitoring video library is detectd based on HGPC algo posed algorithm is 91.4%, which is superior to the previous metho only 47ms, which meets the real-time requirement. Therefore, HGP to dispose the video shaking.