Research On Algorithm Of Indoor Moving Target Detection And Tracking In Intelligent Video Surveillance System

Intelligent video surveillance is playing a more and more important role in various aspects of life in the human society, especially the application in building security, traffic management, disaster relief expedition, is of great significance. Intelligent video surveillance uses analysis method to analyse the video image sequence automatically, realize the target detection and tracking in the dynamic scene, and accomplish the target recognition, subsequent behavior analysis, alarm etc.. The dynamic image of the object motion provides more abundant and valuable information for people, so the moving target is an important research content in the field of intelligent video surveillance. Key problems of moving target detection and tracking technology are accurate detection and foreground reconstruction, effective tracking. The main research work of this paper includes the following contents.In the area of target detection, this paper comparative analysised and verified several commonly used moving objects detection methods, and selected suitable morphological operations to denoise the detected foreground object; in the aspect of target tracking, the paper derived and discussed in detail on the tracking process based on Mean shift algorithm, Calman filtering algorithm and nearest neighbor association algorithm, used different video to realize different tracking algorithm, and realized the matching tracking of multiple moving objects by Calman filter.In addition, aimed at some key technical problems of moving target detection and tracking, this paper put forward the coupling color space conversion and Calman filter algorithm, to achieve foreground segmentation, details perfect, prediction and tracking, shadow detection and removal systematically in the entire process. The realization process of coupling algorithm is:using the adaptive background difference algorithm to separate foreground and background of video image in RGB space, and using Calman filter to predict the variation of position of moving target, the detected foreground object was locked in the minimum enclosing rectangle box to track, utilizing the value information difference with real object to eliminate the false target of error detection (shadow) in HSV space by threshold segmentation. After several video processing verification, it is proved that the calculation method in this paper is simple, easy to implement, and under circumstances of video environments of different complexity, error detection of different cause, error detection of different number, targets of different number, can real-time detect and track targets, and can accurately suppress shadow and error detection in the foreground.Finally, based on the basis of existing algorithms, the feasibility of algorithm is verified through a lot of experiments. The experimental results show that, the algorithm has good effect and is practical.