Research On Pedestrian Detection And Behavior Analysis In Video Surveillance

In modern society, the video surveillance system can be seen everywhere. Video surveillance systems can make people’s lives easier, safer, and improve the working efficiency. However, at present, most of the current video surveillance systems only remain in the original video recording stage. Although some systems contain motion detection functions, but it is the human that mostly complete the actual monitoring tasks. As video monitoring scale expands gradually, the limitations of video surveillance systems which rely on artificial way in real-time monitoring are becoming more and more serious. Neither we have enough screens for video displaying, nor it is possible to arrange enough persons staring at the screen all day long, and it is facing massive video data retrieval. So to using the computer vision technology related to realize intelligent video surveillance system of demand more and more urgent. As a result, the demand for video surveillance systems using computer vision technology becomes more and more urgent.The pedestrian detection and behavior analysis is very important in intelligent video detection. Researching on pedestrian detection and behavior analysis has very important significance to the development of intelligent video surveillance. In this paper, we study the problem of pedestrian detection and behavior analysis in video surveillance. Particularly, we construct an object detector in view of computer vision, and it can search objects in video and can detect them. From the image sequence, we must find the objects concerned, namely the pedestrian, and we must have the ability to mark different parts of the body to distinguish behavior. More exactly it is to achieve the two goals in object detection:determining the pedestrian and their behavior in image.In this article, we introduce the pedestrian detection and behavior analysis separately. First, we detect pedestrian in the image, and then analysis its behavior. For pedestrian detection algorithm applied to image feature extraction method, this paper adopts the method of Histograms of Oriented Gradient (HOG) descriptors to determine whether the image contains the pedestrian. After getting the physical characteristics, SVM classifier that is often used in the human body detection algorithm is applied to experimental objects classification. In the step of behavior analysis, Active-Basis is introduced into the system to extract the different parts of the human body.In our laboratory, we constructed a pedestrian detection and behavior analysis system, to verify that the algorithm is valid, and to improve the algorithm in order to make the algorithm more practical in applications. Our aim is that in a real-time video surveillance, the detection of pedestrians is real-time and it has the ability to conduct simple classification. At the same time, there is no significant delay to ensure smooth monitoring screen, with a higher accuracy rate.