Knowledge-based Face Detection And Face Recognition System

Automated face recognition is an important part of Biometrics. It is one of the most active and challenging tasks for computer vision and pattern recognition. This dissertation mainly studies approaches to real time frontal face detection and recognition. And an automated face recognition simulation system is established in the end. The essential contents can be listed as follows:In order to detect the frontal upright face in real time, the integral image is applied to the face detection method based on knowledge. The integral image is used to calculate sub-block's statistic value, and it degrades the computationalcomplexity from O(n5) to O(n3) , thus the speed of detection is improvedremarkably. Then organ-based blocking scheme in the face detection method based on knowledge is introduced to the detection method based on support vector machine (SVM), it reduces the dimension of face character vectors and improves the detection precision.A hidden Markov model method (HMM) is used to recognize face in our automated recognition system. The element of HMM is introduced firstly. Then the recognition method based on HMM, the model of face and the training scheme are expatiated.Finally, an automated face recognition simulation system is established which mainly includes video input module, face detection module and face recognition module. In the online recognition experiments, an method of automated constructing face database is proposed. This method which is based upon real time detection of face could segment faces from background images automatically, therefore improve the constructing efficiency and uniform the standard of face images. The system is tested in various environments of complex background in room and results show that the system has reached expected goals in time complexity and recognition rate.