Research On Intelligent Video Surveillance System

Recently,with the continuous development of video coding technology,rapid increase of Internet bandwidth,and fast reducing of storage device cost,digital videos have become the major information carrier in all kinds of surveillance systems for its convenience.Video surveillance systems have become useful surveillance tools at all kinds of public places such as airports and banks.Traditional video surveillance systems are mainly based on operations by human operaters,who watch the surveillance videos and obtain the knowledge from the videos.It brings a big manual labor and it may also bring the possible mistakes because of people's carelessness or even the possibility of the collusion between security operaters and the criminals.So,the research on intelligent video surveillance systems based on digital videos is very valuable.Automatic or semi-automatic operations on surveillance videos using computers are performed in intelligent surveillance systems,to meet different surveillance requirements,replace or partly replace the operations of human's,reduce the mistakes and improve the surveillance quality.The application scope of video surveillance is very wide,and different techniques are needed on different applications and requirements.A specific intelligent surveillance system combines lots of techniques in areas as image processing,video processing,and pattern recognition,etc.In the first part of this thesis,key techniques of intelligent video surveillance systems are reviewed and analyzed.For recognition and segmentation of people and luggage,a method based on the complexity of detected edges is addressed.And for the problem of the difficulties of background model initialize and update in outdoor surveillance videos,an adaptive background modeling method is proposed.Two complete intelligent surveillance systems are addressed in the last part of this thesis for practical applications.Experimental results demonstrate the effectiveness and effeciency of the systems.