| Organism feature recognition using the human body inherent characteristics for authentication can greatly improve the security, and reduce the possibility of forging and stealing identification card. The face recognition technology which has traits of non-contact recognition, strong initiative and friendly operation by using face image as the object, has gradually become a new and developing research hotspot in the region of biometric recognition.Face recognition includes face detection and face recognition, in which face detection is the precondition of face recognition. According to the continuity character of video sequence image, the paper considers the algorithm based on regional features to realize the face target detection, which has high detection rate. In order to satisfy the real-time requirement of system, the paper studies and proposes the improving AdaBoost algorithm to detect face target, which has the average detection time of 110ms. Then it uses MeanShift tracking algorithm integrated with Kalman filter to track the face target, which realize the only testing of target, and improve on the real-time system.Based on the face detection, this paper proposes the integrated algorithm combined Fisher face algorithm with support vector machine (SVM) classification methods for face recognition. First, use discrete wavelet transform and principal component analysis methods to compress the detection face image and reduce the dimension. Then adopt Fisher face method to extract the face feature. Finally, classify the extract features to realize the different face recognition by SVM.VS2008 and OpenCV based on software development platform is used to construct the detection and recognition system by the paper. Through the detection and recognition experiment for static image and dynamic video, the testing result shows that the average face detection time is 0.1 S/Frame, and the face recognition rate can reach 97.0% based on Yale face dataset and 97.5% based on ORL database. |
PDF Full Text Request