|
In this paper, the face recognition system in the main part of the technology are studied , including image preprocessing, face detection and positioning, image normalization , feature extraction, face recognition algorithm . The image preprocessing includes a normalized scale of brightness , contrast, etc. of the original image preprocessing and extracted face image , the process of calibration of the rotation angle of illumination invariance processing . Mainly discussed in face detection and positioning of face detection based on neural network and knowledge modeling technology, and a face location algorithm based on the output characteristics of the face detector , and achieved good results . In the feature extraction Yang proposed IMPCA method made ??a promotion. Face recognition algorithm part of the focus of this study , in this section , we first proposed the concept of contribution to the matrix , according to the contribution of the different sizes of each mode -dimensional feature recognition effect , determine a contribution to the matrix , and then use the pretreatment eigenvectors (Mode ) . Then using subspace eigenvectors obtained pretreatment classification, through experiments made ??on the Manchester face database show that , to identify the effect of greatly improved after pretreatment . Secondly, the intersection of the sub- space caused by the mode misrecognized implicitly through the introduction of nuclear methods to achieve a transformation from a low-dimensional space to a high dimensional space , and purpose to eliminate intersect or overlap between subspaces , thereby further improving the recognition rate , and finally through the experiment confirmed the feasibility of the proposed method .
|