Research On Surveillance Video Segmentation And Summarization

Nowadays, the equipment of civil, commercial and military surveillance is growing up rapidly due to the increasing demand of security. Millions of surveillance cameras record video around the clock, producing huge video archives, finding critical events is like finding a needle in a haystack, making the archive almost worthless. In order to browse the video data quickly and make full use of the surveillance video data, it is very important to do research on surveillance video segmentation and summarization.This paper first introduces the current researches of video summarization, and then summaries the traditional video summarization technologies. According to the summary of traditional video shot change detection algorithms, we propose a new surveillance video sequence segmentation method based on moving object detection. This method detects moving objects by adaptive background subtraction technique with a three-frame differencing first, and the sum of foreground pixels is calculated as difference feature. Last, a sliding window based adaptive threshold is proposed to determine the segment boundaries. The experimental results show the algorithm has the advantanges of low complexity and high performance. The paper also put forward a surveillance video summarization method by key-frame based on trajectory extraction.according the features of surveillance video. Trajectory is extracted from each moving object, together with the last frame of corresponding video segment to represent the moving history of each object intuitively. The proposed video summarization algorithm can save browsing time and storage space obviously, and has large amount of information.In order to test the effectiveness of proposed algorithms, we develop a video summarization system by Microsoft Visual C++ 6.0 and OpenCV. It contains several function modules such as video playing, video segmentation, trajectory extraction and summarization displaying. The system can satisfy the users'basic requirements and has practical value for future research.