Dynamic Trajectory for Visual Perception of Abnormal Behavior

Computer vision and intelligent analysis technology has been extensively applied to video surveillance and security system. Motion trajectory feature is used to detect abnormal motion behavior. Because the trajectory mainly reflects the spatial change of the moving object in the foreground, there are some limitations to reflect the time information. In this paper, a time encoding method for particle trajectory is proposed, which is based on the information of temporal and spatial changes of particle motion. Based on the trajectory of motion, it describes how the object moves and its position changes in space, reflecting the spatial distribution of motion mode and energy. Based on the dynamic particle trajectory, the video motion evolution pattern is presented in the form of visual behavior template, and the performance of visual features is enhanced by pseudo color coding. Compared with the spatial motion trajectory, the proposed time encoding trajectory method can effectively improve the recognition performance.