Research On Key Techniques Of Intelligent Video Surveillance

Intelligent Video Surveillance technology is an emerging research orientation in the field of computer vision.The main goal is to realize the high-level semantic analysis of the surveillance video by integrating computer vision technology, image/video processing technology and artificial intelligence technology.The analysis results can be used to control the video surveillance system itself.Thus,Intelligent Video Surveillance technology has a broad prospects in public security,intelligent transportation and other areas of application,which made the technology becomes a hot research by domestic and foreign experts and scholars.Video Surveillance System in the actual scenes are often very complex and volatile, and the basic of the description,understanding and follow-up to deal with in the intelligent video surveillance system is accurately detecting and tracking the moving targets.This paper is committed to the relevant models and algorithms of human detection and tracking under the specific surveillance scenes.After researching three classic algorithms for moving objects detection,and giving the analysis of different background models,a target detection algorithm based on shadow detection and Gaussian Mixture Model is proposed.What's more,the paper has established objects model based on the result of detection of moving targets,which can provide the accurate and effective input of the follow-up tracking.The algorithm has been simulated on the Matlab software platform.Experiments show the algorithm in dealing with multi-modal background is effective and can suppress the shadow of the movement perfectly.To conquer the drawback of selecting the target manually in the existing Particle Filter algorithm,an automatic modeling of the Particle Filter algorithm is proposed.To automatic model the target detected in the above-mentioned sections,which provides the input model of the Particle Filter algorithm based on color,realizes the automatic modeling and target tracking.The paper has simulated the Particle Filter tracking algorithm on the Matlab software platform,which use the PETS2006(Performance Evaluation of Tracking and Surveillance) common video database as the input data. Experiments show the tracking algorithm in automatic modeling target is effective and has good robustness on tracking non-rigid target,such as human.To overcome the shortcoming of single visual cue in complex environments,a tracking algorithm based on adaptive cue fusion mechanism is proposed.The color cue and texture cue are utilized to represent the target and democratic integration is applied to fusion these two cues,thus realizing the tracking algorithm on-line adjusting the weight of two cues and realizing the robust tracking.