Research On Video Based Real-time Multiple Face Detection, Recognition And Tracking Methods

Along with the increasingly serious of international situation, it is increasing thatdemanding for security in field of immigration clearance, criminal investigation,anti-terrorism and military installations. Consequently, to recognize person identity speedyand exactly in large-scale crowd has become an important way to protect the public socialsecurity, guarantee national harmony and reinforce pre-warning capability in public safety.This paper designs a real time automatic face detection, tracking and recognition systemfor video, which can detect, track faces and recognize human identity within the scope ofvideo. The system consists of three parts, which are multi-face detection, multi-facetracking and identity recognition.For face detection, this paper presents an face detection algorithm calledAdaboost-ASM face detection algorithm, combines Adaboost face detection algorithmwith the adaptive shape model (ASM), to solve the problem that Adaboost face detectionalgorithm recognizes the non-face region and complex region as a face region mistakenly.Finally, the Adaboost-ASM algorithm achieves the real time face detection and eliminatesthe non-face region at same time of the human face, and the face region are eliminated.For face tracking, this paper analyses the theory of Camshift algorithm in deeply.Aims at the problem of face number changing and faces interlacing, puts forward aenhanced multi-thread Camshift tracking algorithm, called MT-Camshift trackingalgorithm.For face recognition, this paper extract feature by2D-Gabor wavelet transform andfocuses on the analysis of merit and demerit of PCA method and Fisher linear discriminantmethod, combining PCA and Fisher linear discriminant analysis method, which is called asFisherface method, to reduce dimensions of human face feature and to find the projectionthat is help for the feature classification. Finally, calculates projection of face to recognizeand projection of sample face by cosine distance formula to identity faces. Under Visual Studio2008IDE, this paper developes a real-time multi-face autodetecting, tracking and recognizing system and makes a large number of experiment forverification. Experiments show that, the system achieves accuracy detecting andrecognizing result and makes a powerful real-time track which is a professional solutionfor human faces real-time detecting, tracking and recognizing.