Video based abnormal behavior detection
Bao Hong, Shi Yu-fang, Bo Xu · 2011
In this paper, problems of breakpoint produced during motion region detection, feature selection for human behavior recognition as well as classification and identification of human behavior were studied. Problem of breakpoints was solved by means of combining background subtraction with frame difference method. Shape feature is selected by experiment as the identification indicators of human motor behavior. Refinement of the shape characteristics was made. The mass center locus and its x, y components were proved to have high recognition performance for human behavior identification and classification. On this basis, a video-based human behavior detection system was designed.