一种监控视频结构化的时域分割算法

A Structured Temporal Segmentation Algorithm for Surveillance Video

当前,监控视频数据量呈现爆炸式增长,快速、准确的从中提取有效信息成为研究热点。运动是监控视频的有效特征,有效信息往往蕴含于监控视频的运动片段之中。因此,本文首先针对监控视频的特点,参考普通视频的层次划分,定义了监控视频的层次结构;其次,按照所定义的监控视频层次结构,从像素差异、边缘特征、分块直方图三个维度进行监控视频时域分割;最后,对三种分割方法进行了性能分析与对比。本文所定义的监控视频层次结构有利于快速锁定有效信息,从时域分割出运动片段可大大降低监控视频数据量。

At present,the amount of surveillance video data is exploding,and how to extract effective infor-mation quickly and accurately is becoming a research hotspot.The salient feature of surveillance video is mo-tion,and usually,motion segments contain valid information.Therefore,in this paper,first of all,the hierarchical structure of surveillance video is defined according to the characteristics of surveillance video and the hierar-chical structure of ordinary video.Secondly,according to the defined surveillance video hierarchical structure,the surveillance video is segmented in time-domain from three dimensions:pixel difference,edge feature,and block histogram.Finally,the performance analysis and comparison of the three time-domain segmentation methods are carried out.The surveillance video hierarchy defined in this paper is useful for quickly locking valid information,and the time-domain motion segments can greatly reduce the amount of surveillance video data.