Finger vein recognition method based on multi-scale local feature fusion

Abstract

The invention discloses a finger vein recognition method based on multi-scale local feature fusion. According to the method, texture direction response values are extracted by using a multi-scale direction template, local vein direction features on a first-order gradient are obtained through comparison, then an MLBP operator is calculated by taking the direction response values as a substrate to obtain local texture detail features on a second-order gradient, and finally fusion is carried out through an optimal weight mode. According to the multi-gradient feature fusion mode, the image information is fully utilized, the stability of the features is enhanced, the structure of the finger vein can be well highlighted, and the grasp of the features to the global information can be enhanced ina multi-scale mode calculation mode.

Description

technical field [0001] The invention belongs to the technical field of biological feature identification and information security, in particular to the technical field of finger vein identification. Background technique [0002] Biometric technology is a convenient, safe and reliable identification method at present. Compared with other current biometric technologies such as fingerprint recognition, face, DNA (gene recognition), palmprint, iris, voiceprint and other finger vein recognition, it has fast speed, internal features, high security level, living body recognition, high precision, etc. Advantage. However, there are still some problems in the current finger vein recognition technology, such as insufficient utilization of information, weak anti-noise performance, and unstable features. [0003] The local binary pattern operator (Local Binary Pattern, LBP) is an algorithm based on local features, which reflects the changes in local textures by comparing the size of ne...

AI Technical Summary

Problems solved by technology

However, there are still some problems in the current finger vein recognition technology, such as insufficient utilization of information, weak anti-noise performance, unstable features, etc.

[0003] The local binary pattern operator (Local Binary Pattern, LBP) is an algorithm based on local features, which reflects the changes in local textures by comparing the size of neighboring pixel values, and is not affected by the linear transformation of the entire image. The calculation is simple and efficient, but the calculation scale is small, and the extracted finger vein features are defective in robustness; the Multi-block Mean Neighbors-based Binary pattern (MMNBP) proposed by Fu Hua et al. ) further enhances the robustness of the finger vein features through the operation of block and averaging, but the grasp of the global information is insufficient, and the vein structure is not prominent enough

ZHOU Y et al. proposed the Neighborhood Matching Random Transform (NMRT) based on the palm veins. This method extracts the main vein direction feature by comparing the linear pixel sum of the local area. The description of the vein direction is accurate, but it lacks the The description of local texture detail features, and compared with palm veins, finger veins have thinner vein width, more susceptible to noise, intricate texture, and rich local texture detail features

Moreover, the above local feature extraction algorithms are all based on pixel point extraction features, which are in the first-order gradient, and have certain defects in feature stability.

[0004] Finger vein image vein blurring and dust noise interference are difficult problems in finger vein image acquisition, and the current finger vein recognition algorithms based on local features are still in the first-order gradient calculation, local texture feature extraction is difficult, and feature stability is insufficient. Moreover, the current local feature algorithm has a single description of local texture features, and the amount of information is not fully utilized.

Images