Surveillance video summarization based on moving object detection and trajectory extraction

Abstract

With the ever growing surveillance data, it is important and urgent to effectively and efficiently manage and utilize the huge amount of surveillance video repositories. In order to browse the video data quickly and make full use of the surveillance video data, we propose a fully automatic and computationally efficient framework for analysis and summarization of surveillance videos with the techniques of moving object detection and trajectory extraction. The video is first partitioned into segments based on moving object detection, then trajectory is extracted from each moving object, and then keyframes are selected, together with the trajectories to represent the video segment. The experimental results indicate the practical applicability of the proposed method.