The Design And Implementation Of Online Pursuit System Based On Face Detection And Recognition Technology

It's difficult to adopt the typical recognition method to environment of video surveillance directly as it designed to be based on high definition still image,largesample learning and multi-iteration recognition method,and we may face problems,like low detection rate,high latency and insufficient compute capability.In this thesis,a proposed detection and recognition algorithm could only require low definition video and low compute capability.The contributions of the thesis are:1.A proposed algorithm fusing gradient algorithm to facial feature template could find the facial features accurately in low definition environment.Followed an escalating verification mechanism can effectively reduce fault rate and increase productivity,eventually,we increased the recognition rate in low definition environment.2.A proposed algorithm fusing morphology template method of layered face to histogram analysis,could make the face detection efficiency and accurate in low definition environment.3.About the problem of complex background,it's common to subtract video frame with background to get the foreground objects,this thesis is based on this algorithm,furthermore,we try to use morphology algorithm on the result to get rid of noises without losing any important information.The implemented system was deployed to Chongqing Jiangbei Public Security Bureau,and the result of detection and recognition is obvious good in real production.