Research Of Intelligent Video Surveillance Key Technology

Video monitoring system in the people's daily life becomes moreand more popular and provides people safe and convenient life. As thetechnological development, in order to further improve the intelligentmonitoring system, a variety of intelligent technology for the monitoringsystem become popular. Moving target detection and tracking technologyis one of the basic and key research in intelligent monitoring system.The intelligent monitoring technology has been studied by thefollowing three aspects:(Ⅰ) Studying moving targets detection technology of video sequencesin the static background. Several common target motion detectionalgorithms have been proposed: optical flow, background, framedifference method, edge detection method. To analyze frame differencemethod and the background difference method, and improving threedifference methods, Integrating multiple targets getting from the originalthree frame difference methods, in order to add more details to themoving target. The experimental results show that the improvementmethod is indeed able to obtain more information of the moving target.Then, to solve the shortcomings of the "hole" phenomenon with the frame difference method, it is proposed the improvement methods withintegrating frame difference and background difference method inmovement target tracking, and using the improved Surendra algorithmwith OTSU dynamic threshold method to update background, thussolved the problems of target profile missing and target internal cavitation.Experiments results show that the improved method of detection is better.(Ⅱ) Studying moving targets tracking technology of video sequencesin the static background. Focusing on the CamShift algorithm, which hasbeen applied to track moving and improved in many ways, it can be usedfor multi-objective and multi-tone tracking and in order to improve therobustness the CamShift target tracking algorithm in the circumstances ofcolors similar. This paper combines the mean shift segmentation andregion growing method and uses hue rang, saturation and value colorhistogram to describe the target object. Through tracking the main colorof the target, improves the robustness of multi-tone target tracking. Formulti-target tracking, using separate track for each target object. Finally,experiment results show that the robustness and practicality ofImprovement methods.(Ⅲ) This paper has studied and optimized the monitoring systemconstructing technology. Focused on the algorithm optimization andassembly-level optimization of DCT transform and quantizationalgorithm, to analyze the experimental results, the encoding speed of H.264encorder has been improved, which can prove the effectiveness ofthe optimization algorithm.