Adaptively Weighted Sub-pattern PCA And Its Application In Face Recognition

Feature extraction is a critical problem in the research fields of pattern recognition. It is a process that extracts a new set of features of interest from the original data through some mapping. This thesis mainly focuses on Principal Component Analysis (PCA) based methods. Based on the previous research work, adaptively weighted sub-pattern PCA (Aw-SpPCA) is proposed and successfully applied to the task of face recognition, and then a native face recognition system based on Microsoft .Net technology is constructed. PCA is a popular and effective feature extraction method, and regarded as one of the industrial benchmarks in the application of face recognition. This method operates directly on the whole face image. Because only global information about face image can be extracted while local information is neglected, PCA is not very effective under different facial expression and illumination condition. In subpattern-based principal component analysis (SpPCA), an original whole large dimensional pattern denoted by a vector is partitioned into a set of equally-sized sub-patterns in a non-overlapping way and then local information on each sub-pattern is extracted. Although partially having overcome the limitations of PCA, SpPCA does not concern different contributions made by different sub-patterns, in other words, it endows equal importance to different parts of a pattern in classification. As a result, the global vector more likely contains redundant or even useless local information. The proposed Aw-SpPCA can adaptively compute the contributions of each part and then endows them to a classification task in order to enhance the robustness to both expression and illumination variations. Experiments on three standard face databases show that the proposed method is competitive. Furthermore, a native face recognition system based on Microsoft .Net technology on PC platform is constructed. This system consists of image capturing and preprocessing procedure, information storage procedure, face recognition procedure, etc.