Pedestrian Behavior Analysis Method Based On Video

With deep development of intelligent video surveillance, people want to enjoy more convenience brought by intelligent monitoring system, even wants them to replace human beings to do more things. Community’s attention about public safety continues increasing, and area to be monitored is also expanding, then the demand for intelligent monitoring system is increasing rapidly. If the analysis of video content could be down by a computer, on which a lot of useless content are ignored and do necessary responds, there will save a lot of human resources and storage space in the disc. Pedestrian is one of the most important elements in the video scene, therefore an analysis of pedestrian behavior in the video scene of great significance to the understanding of the security state.This paper studies the behavior recognition process about pedestrians, based on the features of pedestrian movement. On this basis, we studied abnormal behavior recognition on the characteristics of pedestrians’movement paths. First, we declared the process of the behavior recognitions based on surveillance video is target detection, target tracking, behavior recognition and understanding of advanced behavior. we did much analysis on some target detection and tracking ways, pointed the advantages and disadvantages of these ways, and the applicability of the process on pedestrian behavior analysis. Then, we analysis a variety of features in the images including human, spatial and temporal characteristics is chosen as the parameter on human behavior expression for the good applicability. Then we made the recognition process using the Hidden Markov Model, and then the experimental procedure process is given. On this basis, we explored the initial stage of the process on video semantic understanding. The video is selected, in which pedestrian walking motion are contained, for abnormal walking recognition using the existing linear fitting method. Then we designed a pedestrian wandering scene detection experiment using the accumulated value of speed direction changes. At last we completed a adjustable wandering and abnormal walking test in different monitoring environments.In this study, a thought are given on pedestrian behavior recognitions and semantic understanding, and a primary exploration are did on the semantic understanding. Some positive effect will produced on subject of this research in the future.