交叉注意力与质量优化结合的内外指纹融合  

Internal and External Fingerprint Fusion Combining Cross Attention and Quality Optimization

作　　者：石梦菠 王海霞[1] SHI Mengbo;WANG Haixia(School of Computer and Science Technology,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学计算机与科学技术学院,杭州310023

出　　处：《小型微型计算机系统》2024年第8期1914-1920,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61976189)资助.

摘　　要：指纹作为一种应用广泛的生物特征,使用不同传感器采集的内部和外部指纹图像各有优缺点.本文提出了一种基于交叉注意力和质量优化内外指纹融合网络,旨在充分利用两者之间的优势,最大化提高指纹图像质量.该网络利用交叉注意力模块让两个分支交互作用,并关注彼此的细节信息,从而提高特征表征能力.同时,设计了基于质量指标OCL的损失函数,以保留高质量区域、减少低质量区域,实现质量优化.实验结果表明,本文融合方法在视觉效果、质量评分和匹配性能以及细节点提取等方面均具有优越性,等错误率达到了1.44%,并大大降低了时间性能.这项工作为今后残缺指纹的识别提供了参考和见解.Fingerprints,as a widely used biometric feature,have their own advantages and disadvantages in terms of the quality of internally and externally captured fingerprint images using different sensors.In this paper,a cross-attention and quality-optimized internal-external fingerprint fusion network is proposed to fully utilize the advantages of both,and maximize the quality of fingerprint images.The network uses a cross-attention module to allow the two branches to interact and focus on each other's detail information,thereby improving feature representation ability.At the same time,a loss function based on the quality indicator OCL is designed to preserve high-quality regions and reduce low-quality regions for quality optimization.Experimental results show that the proposed fusion method has superior performance in visual effect,quality evaluation,matching performance,and fine feature extraction,with an equal error rate of 1.44%,and greatly reduces time performance.This work provides reference and insight for future recognition of mutilated fingerprints.

关 键 词：内部指纹 外部指纹 交叉注意力 质量指标OCL 指纹融合网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]