Face Recognition Method Based on Weighted Discriminative Sparse Constrained Nonnegative Matrix Factorization

Abstract

The invention discloses a face recognition method based on weighted discriminative sparse constrained non-negative matrix decomposition, which mainly solves the problem that the prior art is not robust to occluded faces and the recognition rate is low. The technical solution is: 1. Construct a non-negative weight matrix according to the occluded area of ​​the test image; 2. Introduce the weight matrix into the generalized KL divergence objective function, impose sparse constraints on the base matrix, and impose intra-class sums on the coefficient matrix. Divergence constraints to obtain weighted discriminative sparse constrained non-negative matrix decomposition objective function; 3. Solve the objective function, decompose the training data matrix to obtain the base matrix and coefficient matrix; 4. Project the test data matrix on the base matrix, Obtain the corresponding low-dimensional representation set as the final test data; 5. Use the coefficient matrix as the training data, use the nearest neighbor classifier to classify the test data, and output the result. The invention improves the face recognition effect in the case of occlusion, and can be used for identity recognition and information security.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a method for extracting and identifying salient features of a human face image, which can be used for identity identification and information security. Background technique [0002] With the rapid development of the Internet, identity authentication technology has a very important application status in many fields such as e-commerce, human-computer interaction, public safety and network transmission. Compared with the traditional identification technology that combines information encryption and other strategies to add discriminative information to samples, biometric identification technologies such as fingerprints, irises, and voice that use image processing and pattern recognition to identify personal identities are unique and reliable. , convenience and not easy to be stolen and so on. Compared with other biological features, face features have the advantages of dire...

AI Technical Summary

Problems solved by technology

DNMF makes good use of the category information of the data, but its robustness and adaptability to face occlusion are poor, especially for occlusions with large continuous areas in the face image

Images