A finger vein recognition method based on an FCN full convolution network

Abstract

The invention discloses a finger vein recognition method based on an FCN full convolution network. The finger vein recognition method comprises the steps that collecting a finger vein image; preprocessing the finger vein image to obtain a preprocessed finger vein image; performing edge detection, least square linear fitting and direction correction on the preprocessed finger vein image, and extracting a median section ROI finger vein image with rickest finger veins on the basis of the edge detection, the least square linear fitting and the direction correction; extracting features from the extracted ROI finger vein image by adopting the trained FCN full convolution network, and classifying each pixel point to obtain a to-be-registered / recognized finger vein image with distinguished vein points and background points; and collecting a to-be-registered finger vein image, constructing a registered finger vein database, and carrying out retrieval and recognition on the to-be-recognized finger vein image in the finger vein database to obtain a matching recognition result. In the embodiment of the invention, a clearer finger vein image can be obtained, and the matching recognition efficiency is higher and more accurate.

Description

technical field [0001] The invention relates to the fields of biometric feature recognition technology, image recognition and deep learning, and in particular to a finger vein recognition method based on an FCN full convolution network. Background technique [0002] The research and application of identity authentication technology based on biometrics is becoming more and more extensive; the current society's demand for safer and more friendly identity authentication has put forward higher requirements for biometric technology; and finger veins are alive and unique. It will produce feature duplication and allow non-contact identification, so it has become a category that has attracted more attention in the field of biometrics. [0003] Most of the current finger vein recognition methods are based on the knowledge in the field, designing image processing, filtering, etc.; when the imaging quality of the acquisition equipment is low, it is difficult to artificially design feat...

AI Technical Summary

Problems solved by technology

[0004] At present, the problems to be solved are: the success rate of recognition depends on the imaging quality of finger vein images, it is difficult to separate the finger vein area from the background area, and it is difficult to extract accurate and robust features, etc.

Images