A binary encoding and recognition method of facial features based on two-layer self-organizing neural network

Abstract

The invention discloses a binary coding and recognition method of facial features based on two-layer self-organizing neural network. Two-layer self-organizing neural network constitutes a visual dictionary; based on the visual dictionary and two-layer SOM architecture, the face image is encoded, and the face features are projected into the deep semantic space; the encoded face image is pooled to reduce the encoding dimension and generate binary The encoding describes the face; finally, the image encoding distance is calculated by the AND operation to classify the face image. The invention uses the SOM to simulate the human visual perception mechanism to form a binary feature information extraction scheme, which is conducive to rapid face recognition, and has strong robustness to visual occlusion and light changes in the case of face recognition.

Description

technical field [0001] The invention belongs to the technical field of face recognition design, in particular to a binary coding and recognition method of face features based on a two-layer self-organizing neural network. Background technique [0002] Because of its easy acquisition, strong non-invasiveness, and easy feature extraction, facial features have great advantages in biometric recognition. The research on face recognition technology has always been a research hotspot in computer vision. After decades of development, face recognition technology has made great progress in recognition accuracy, environmental impact, etc., and has been widely used in identity recognition, access control, and judicial confirmation. However, in the actual application of face recognition, most of the application scenarios have insufficient training samples, such as single-sample recognition, which brings great problems to practical applications. [0003] Single-sample face recognition wi...

AI Technical Summary

Problems solved by technology

For example, Kohonen et al. (T.Kohonen, Self-Organization and Associative Memory, 2nd edition, Berlin: Springer-Verlag, 1988) used SOM in the application of face recall, and it still has high performance in the presence of noise and occlusion. Accuracy, Liu et al. (N.Liu, J.J.Wang, Y.H.Gong, "Deep Self-Organizing Map for Visual Classification," International Joint Conference on Neural Networks (IJCNN), 12-17July 2015) using the mechanism of human multi-layer visual perception Building a multi-layer SOM network has achieved high accuracy in character recognition, but no effective method has been proposed to apply SOM to single-sample face recognition

Images