Face Recognition Method Based on Siamese Networks Under Non-Restricted Conditions

Face recognition is an important part of computer vision, and has a vital role to play in public safety. Because of the problems of low accuracy and low efficiency in the case of uneven image quality and occlusion due to sudden changes in light in face recognition, we design a Siamese Neural Network based on Local Binary Pattern (also called LBP) and Frequency Feature Perception. The network is based on the Siamese neural network and adopts the Uniform LBP algorithm and Frequency Feature perception to achieve face recognition under non-restricted conditions. the LBP algorithm can eliminate the effect of lighting on the image and provide vector-level input to the network model; the frequency feature perception divides the image features into low-frequency features and high-frequency features, and compresses the low-frequency features in the Siamese neural network to increase the network’s recognition efficiency while exchanging information with high-frequency features to retain their feature data while eliminating target noise data. This maintains the recognition rate of the network and improves the computational speed of the network. Simulation experiments are conducted on the standard face datasets CASIA-WebFace, Yale-B, and LFW, and compared with other network models. The experimental results show that the proposed SN-LF network structure in this paper can improve the recognition accuracy of the algorithm and obtain a better recognition accuracy.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.