People Counting And Crowd Density Analysis In Video Surveillance

As the economy rises rapidly, large quantities of video monitoring systems have been widely applied in stations, intersections of roads, public squares, banks, supermarkets, and many other public areas. Yet, it remains an urgent problem how to implement the supervision and management over people in these areas efficiently based on gathered video information. Obviously, the analysis and comprehension of the astonishing amount of gathered video data which is required to complete the work has already surpassed the limit of human labor. Therefore, target detecting, tracing, recognizing and many other technologies have been used to calculate the flow and density of crowds in order to secure the its safety in public areas, the implementations of these techniques have been turning an increasingly hot research spot in field of intelligent video monitoring.In this paper, two differently typical applications are studied, which put forward their own solution. The main contents include two aspects: The first one, in connection with the situation of detecting pedestrian in the gangways, the pedestrian detection tracking and virtual loop are used to count the number of pedestrians by monitoring the scenes. In the method of targets detection and tracking, firstly, the method of background subtraction is used to extract target, secondly, analyzing the targets extracted from before and after images can achieve counting tracking pedestrian. In the method of virtual loop, using the outline of pedestrian in the virtual loop achieves statistics.The second one, in connection with the situation of large crowd scenes, the method of the combination of image pixel and texture is used to estimate the number of persons in the scenes. Firstly, the number of pedestrian are estimated by image pixel in the scenes, secondly, the scenes are sorted based on the method of the combination of image pixel and texture, then the SVM algorithm is used for the crowd density analysis.According to the studying every part, it turned out that the methods are effective in the solving of people counting in the really video surveillance.