| With the development of artificial intelligence, human-computer interaction, computer vision has became a research area of concern gradually. Face-recognition and head pose estimation is an important target for researchers. In recent years, the representation-based face-recognition techniques focus mainly on constraint conditions and dictionary learning. Few researcher study which sample data features influence the performance of representation-based classification algorithms. In order to solve this problem, we define the structure-scatter degree, which represents the structure features of training sample sets. Experimental results show that sets with a higher structure scatter more likely allows a classification algorithm to obtain a higher recognition rate. Further, the block contribution degree(DBC) of a training sample set is defined to evaluate whether a sample set is suitable for block-based sparse-representation classification algorithms. Experimental results indicate that if the DBC approaches to zero, the block technique is unlikely to improve the performance of a representation-based classification algorithm. On the basis, the self-adaptive optimization method was proposed to generate an optimal block size, an overlapping degree, and a block-weighting scheme.Face recognition is an important part of the face detection in natural environment. Active Appearance Model(AAM) is a typical face detection method which is one of the effective methods of head pose estimation. According to the establish and established fitting process of the AAM which is sensitive to initial settings promblem. HOG-LBP technique is used to extract feature histogram from a tested face image, the parameters of a model best matching with the tested face image are further selected for initial values according to feature histogram, and Adaboost is used for face detection. Experimental results show that the proposed method can improve the robustness and velocity of AAM significantly. |
PDF Full Text Request