Finger vein recognition method and system based on triple loss and lightweight network

Abstract

The invention discloses a finger vein recognition method and system based on triple loss and a lightweight network. The method comprises the following steps: preprocessing a finger vein sample image in a database; constructing a small lightweight network model structure, selecting a triple (A, P, N) in the training set, and carrying out network training; taking the training set as character information of a to-be-input finger vein database, reading images in the training set and inputting the images into the model to obtain an embedins vector library; inputting a certain sample image into the network model, and if the similarity between the sample image and the sample image in the training set is greater than a threshold value, determining that the person is not in the database; otherwise, selecting a distance minimum value in a range conforming to a threshold value, and identifying the distance minimum value as a person with the highest similarity. According to the method, triple loss is introduced to be combined with a lightweight network and a training model, spatial mapping of vein features is achieved, the situation that a network model needs to be trained multiple times when a database dynamically changes is avoided, and finally the precision and efficiency of finger vein recognition are effectively improved.

Description

technical field [0001] The invention relates to the technical field of digital image processing, in particular to a finger vein recognition method and system based on a triple loss and a lightweight network. Background technique [0002] In recent years, the problem of finger vein recognition has been a popular direction in the field of image processing. The process includes four parts: image acquisition, preprocessing, feature extraction, and recognition and classification, among which feature extraction is the most core part. In the traditional field of image recognition, features are basically extracted manually, which is not only time-consuming and laborious, but also cannot guarantee the effectiveness of features, and has poor robustness. Using deep learning for feature extraction avoids these problems. [0003] Deep learning is a collection of algorithms that use various machine learning algorithms on multi-layer neural networks to solve various problems such as images...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a finger vein recognition method based on triple loss and lightweight network, which solves the problem of sensitivity to noise and poor robustness in existing recognition metho

Images