Finger vein image recognition method based on fusion of multi-scale coding and VGG model

Abstract

The invention proposes a finger vein image recognition method based on fusion of multi-scale coding and a VGG model, and the method comprises the steps of carrying out the extraction of a region of interest (ROI) of a finger vein image in a database, obtaining a finger vein ROI image, and carrying out the preprocessing of the finger vein ROI image; performing feature coding on the preprocessed finger vein ROI image by adopting a multi-scale local binary pattern (LBP) coding operator to obtain a feature value of a central pixel point in the finger vein ROI image; reconstructing to obtain a multi-scale coding convolution filter; performing parallel fusion on the multi-scale coding images obtained after reconstruction through a memory; establishing an improved VGG convolutional neural network model; taking the fused feature image as input, combining the fused feature image with a VGG model, and performing feature extraction on the preprocessed finger vein ROI image to obtain a feature vector; and according to the feature vector, performing similarity calculation on the to-be-matched finger vein ROI image by adopting a feature matching method.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a finger vein image recognition method based on fusion of multi-scale coding and a VGG model. Background technique [0002] Biometric technology is a technology for human identity authentication through human biological characteristics or behavioral characteristics. Among them, behavioral characteristics include gait, voice, signature, etc. Human biological characteristics mainly include two categories, external biological characteristics: such as face shape, palm shape, fingerprints, iris, etc.; internal biological characteristics: such as palm veins, finger veins and dorsal hand veins Wait. At present, common identification methods such as fingerprints, voice, signatures, faces, etc. are easy to forge and fragile, and the detection methods of iris, DNA, etc. are complicated and unfriendly. Finger veins are located under the skin and are usually collected using near-...

AI Technical Summary

Problems solved by technology

[0004] Due to the excessive number of learning parameters of the existing convolutional neural network model, the learning cost of the network model is high, and there may be a problem of overfitting

At present, people usually directly reduce the number of parameters of the neural network structure to reduce the learning cost, but this may cause a decrease in network performance.

Images