The Technology Of Human Face Feature Extraction And Recognition

Face recognition has become an active area of research in the domain of pattern recognition and computer vision. There are broad applications in the fields of national security, military affairs etc. Many scholar research on it all the times. For the particularity of the face image, face recognition is such a difficult problem in the domain of pattern recognition. Using the image of face to recognize the face is still faced with great difficult. Also, there is much work to do if we want to make this technology be completely available.This paper is mainly dedicated to the research of the main methods about the human face feature extraction and recognition. Some representative theories and methods of face recognition are learned and understood seriously. Some advanced technology and tendency of developing on this domain are comprehended deeply. The research topics and the main contributions are as follows:(1) Some methods of face image preprocessing and face detection are discussed. The concepts of Haar rectangle features and "integral image" are introduced. And based on the basic principle of Adaboost algorithm and Haar rectangle features, the cascade classfiers for face detection is constructed. Regarding the Intel's software OpenCV as the foundation function library, the Microsoft's Visual C++ as the tools of opening up, the paper accomplish the face detection and the performance of the classifiers testing successfully. The excellent characteristic of the cascade classifiers is analyzed by comparing with other face detection algorithms.(2) After analysis, summarize and compare with various classical methods about face recognition, this paper summarize the advantage of current methods and the technology difficulties. The merits and demerits of the Eigenface method are regarded as the research emphasis which is analyzed particularly.(3) After studying and analyzing the mathematical principle of the Weighted PCA and the Supervised PCA(S-PCA) thoroughly, a new method of face recognition which is based on the S-PCA is proposed. Compared with the Eigenface method, the most advantage of this method is that it not only can make use of the"energy compaction"property of PCA, but also can make use of the classified information of the training set with Laplacian matrix. The experiments on the ORL face database and the face database of camera sampling improve that the method of S-PCA gains better recognition rate than PCA' s.(4) This paper takes the DCT (Discrete Cosine Transform) as the preprocessing method of data dimensionality reduction based on the two properties of DCT, its "energy compaction" and distance preserving for signal processing and image compression. A new method for face recognition based on the"DCT+S-PCA"algorithm is proposed. Experimental results show that the method is effective and correct. And it has been analyzed and compared with the method for face recognition based on the PCA,S-PCA and"Resize+S-PCA"algorithm respectively.(5) Implemented a simple face recognition system based on DCT+S-PCA algorithm.