Multi-feature description and local decision weighting face identification method

Abstract

The invention discloses a multi-feature description and local decision weighting face identification method. The method comprises the following steps: 1, first of all, constructing a global complementation subspace by use of an independent composition analysis algorithm, and performing coarse classification on samples to be detected; 2, by use of a proposed unified local mean value mode, through combination with other two texture description algorithms, constructing a local complementation subspace and obtaining posterior probability values of samples which cannot be identified easily by the coarse classification; and 3, according to size of the posterior probability values, arranging grade fractions, and solving accurate decision results of the samples to be detected on local complementation subblocks. The method provided by the invention has the following advantages: 1, the method can effectively improve the face texture feature expression capability and enhances the accuracy of feature representation; and 2, the global and local complementation subspaces are constructed for describing human faces and only the samples which cannot be identified easily are projected to the local subspace for accurate classification, such that the problem of not high identification rate or long identification time of a conventional method is overcome.

Description

technical field [0001] The invention relates to a feature extraction method and classification discrimination, belonging to the field of pattern recognition, in particular to a face recognition method based on multi-feature description and local decision weighting. Background technique [0002] Face recognition is a hot research topic in recent years. The description and classification of face images are the two main steps of face recognition. According to the No Free Lunch (NFL) theorem, there is no single algorithm that can be superior to other algorithms in any case, so research on multi-feature and multi-classifier fusion of images has become the current mainstream development direction. [0003] Using global and local features to describe human faces is a commonly used method at present. The global feature mainly describes and expresses the attributes and information of the whole face, while the local feature mainly describes the information of the details of the facia...

AI Technical Summary

Problems solved by technology

In recent years, many scholars have proposed improvements to the LBP algorithm, which have improved the recognition ability of the algorithm to a certain extent, but they all use the local size relationship between neighboring points to describe the texture information, while ignoring the same square The overall relationship between the upward pixel and the gray value of the central pixel

When there are random noise points or there are lighting and edge changes, the performance will be greatly reduced

[0006] The Symmetric Local Graph Structure (SLGS) operator is a recently proposed texture description algorithm. It is an improvement of MFA Abdullah's LGS algorithm. It is no longer limited to the circular neighborhood and uses fewer pixels. points to describe the texture features, but it only considers the horizontal direction of the central pixel point, and does not consider the vertical direction, and the vertical direction also includes a lot of information about the texture

Images