Human-face identification method of local-keep mapping based on statistic non-relative and orthogoual characteristics

Abstract

A face recognition method based on a statistically uncorrelated and orthogonal feature based on a part-preserving mapping in the field of image processing technology. The present invention first performs principal component analysis on the input training sample image to obtain the projection matrix of principal component analysis; then establishes a connection graph to obtain the similarity between any two nodes, and determines the adjacency points of all nodes according to the nearest neighbor principle , calculate the similarity matrix of the input data; then apply this similarity matrix to the local preservation mapping method, add two constraints of statistical irrelevance and orthogonality, use an iterative algorithm, according to the eigenvalue problem and the projection of the principal component analysis The uncorrelated and orthogonal projection matrix is ​​obtained from the matrix, and the training projection coefficient matrix and the test projection coefficient matrix are obtained; finally, the minimum distance method is used for identification. The invention has the minimum redundancy, is beneficial to realize the reconstruction of original data, and can improve the recognition performance when applied to face recognition.

Description

technical field [0001] The invention relates to a method in the technical field of image processing, in particular to a face recognition method based on a statistically irrelevant and orthogonal characteristic of a part-preserving mapping. It can be used in various civil and military systems such as video surveillance systems, automatic door guard systems, and military target tracking and identification systems. Background technique [0002] Face recognition technology has become a hot research topic today. This technology has been successfully applied in identification, man-machine interface, automatic teller machine, video surveillance and other fields. At present, the difficulty of face recognition is mainly that the recognition accuracy is relatively low when the face changes in light, skin color, expression, posture, etc. [0003] As one of the key links of face recognition, the feature extraction method is to map the original high-dimensional data to a low-dimensiona...

AI Technical Summary

Problems solved by technology

However, the basis vectors of the projection matrix of the locality-preserving mapping method are statistically correlated and non-orthogonal, so the extracted features contain redundancy, overlapping information will cause the actual distribution of features to be distorted, and non-orthogonal features are not conducive to reconstruction of the original data, these two shortcomings seriously affect the performance of locality-preserving mapping algorithms

So far, no one has proposed a locality-preserving mapping method that satisfies both the statistically uncorrelated and orthogonal properties

Images