Research And Development Of Human Face Detection Technology In Video Surveillance System

Due to Anti-terrorism and national security needs,Security Industry becomes the fastest-growing industries.With the rapid economic growth, the demand for security products from the major sectors gradually popularizes.On the other hand,China’s policy guidance,such as the proposed construction of“Safe Community”,the Olympic Games and World Expo,provides a good opportunity for the security industry.As a result,high-end,high value-added product platform is the top priority of the Security Industry.Among that,Face Detection in the Video is one of the key techniques of Video Surveillance System.Also,it has high value of research and good future of application.Under the backgroud of security video surveillance system,the paper works on research and development of the core technology and key issues of face detection technology.In this paper,it describes new color detection method and improves the face detection method based on AdaBoost algorithm.Based on the security video surveillance system needs,an intelligence video surveillance for real-time face detection framework is proposed. Combined with pix-based and block-based color detection method, two integration framework of pixel-based and block-based color detection method is proposed.The second integration framework of first block-based classifier then pix-based classifier,combined with color segmentation method,completes the study as a whole.Color detection integration framework reduces the area needed to search in face detection process so as to enhance the algorithm of real-time performance.Color segmentation further screens detected regional candidates of human face. It can remove the noise caused by the false face region so as to enhance the algorithm accuracy and to reduce false detection rate.Based on the research of AdaBoost face detection algorithm,it determines the use of simple features and improved classifier cascade method,and trains in machine training methods.Also it uses facial image database to obtain more efficient cascade classifier to do fast face detection.It can reduce the false detection rate and increase the speed of detection.Cascade classifier framework that added auxiliary decision functions proposed in this paper can balance detection rate and false detection rate better.Finally,aider research on integration framework of pixel-based and block-based color detection method and color segmentation,and the improved face detection method based on AdaBoost algorithm,a face detection framework is proposed.It has been integrated into the Shanghai Science and Technology project“Research and Applications of all-IP intelligent network security system technology”with China Security & Surveillance Technology,Inc.Face detection can obtain dynamic face target and mark.Then targeted face information can be recorded by alarm module.After judging the face detection target,the system is able to use high-speed camera for PTZ tracking and carry out high-definition close-up photography of suspicious goals.The practical application shows that the proposed framework of face detection can detect face in real time and has better detection rate and real-time rate.