The Research And Application Of Key Technology For Video Intelligent Surveillance System

In the field of video intelligent surveillance,foreground detection , moving target trackingand target recognition are the key technology.In the foreground detection, the results ofmoving object detection,target tracking and target recognition play an import part in targetbehavior analysis and understanding of post-processing. Moving object detection, targettracking and target recognition are not only the basic and key work of video part for videointelligent surveillance,but also are the hot research topics of home and abroad scholars.In thispaper,I have done a lot of work in the foreground detection of complex of scene videosurveillance ,besides,it proposes the mothod of weighted background modeling based onstatistical classification.This paper has analyzed both parametric background modeling and no-parameterbackground modeling,in addition,which has analyzed the principles of conventional GaussianMixture,framed difference and mean shift.In this paper I also have proposed the method ofstatistical classification weighted background modeling for the problems of low runningspeed of Gaussian detection,unfavourable effect of framed difference and diffcultlyprocessing of complex scene for mean shift.This way needn't to set the background pixelsto guide line with Gauss for a period of time,which belongs to the algorithms ofnon-parameter background .This method need not to train.For many states of backgroundpixel,it uses many state categories to express.In addition,it updates the weights according tohistory matching cases.The background state is the state of computing the sum of the weightsabove a certain threshold.The experiments show that it gains excellent detection results andreal-time detection speed in the complex scene.This paper also analyses the method of the target tracking and recognition,it introduces thecurrent widely used principle of Camshift target tracking arithmetic.We adopt some methodsof the documents for the disadvantage of the conventional Camshiftarget trackingarithmetic.The experiments show that we can track many target,in addition,this method hasbetter robustness for partially sheltering and color interference.The analysis and recognition of the target trace mainly analyses the five features to recognize and classify the movingtarget. the five feature are the application scene of target trace,the choices of moving target'sheight to width ratio, dispersion, angle standard deviation and so on.In the end of the paper,it has designed the community video surveillance station,besides,ithas implemented a experiment platform of video intelligent surveillance system adopting thecritical technology stated in the passage.we can achieve the functions of part of the analysis ofabnormal behavior,the Identification between car and human and the detection of video targetby adopting this video intelligent surveillance system.