深度融合的指纹识别算法误识率仿真数据

本项目开展高效安全的指纹识别算法研究。传统的结构特征点算法的优点是识别效果非常稳定，因为由纹理走势推测出的结构特征点具有非常高的稳定性和特异性，不会随着使用环境和手指干湿特性的变化而变化，传统的算法不容易被错误的图像攻击。然而，在当下传感器小型化的趋势下，传统的特征点算法碰到了图像小而无特征的尴尬境地，其识别效果随着指纹传感器采集面积的减小而急剧下降。项目提出一个优势互补的深度融合算法：利用传统纹理特征算法的结构特征稳定性降低甚至杜绝异物攻击，同时保留图像算法多特征点的特性以有效识别使小面积指纹。该算法实现了识别效果（高通过率，低误识率）和安全性的最优化。This project focuses on research into efficient and secure fingerprint recognition algorithms. The traditional structural minutiae-based fingerprint recognition algorithms boast a core advantage of highly robust recognition performance: the structural minutiae inferred from fingerprint ridge patterns possess extremely high stability and distinctiveness, remaining unaffected by changes in usage environments or the dry/wet condition of fingers. Thus, traditional minutiae-based algorithms are less susceptible to spoofing attacks via counterfeit fingerprint images. However, amid the ongoing trend of miniaturizing fingerprint sensors, traditional minutiae-based algorithms face a critical dilemma: small-area acquired fingerprint images often lack sufficient valid minutiae, resulting in a sharp decline in recognition performance as the sensor's acquisition area decreases. This project proposes a deep fusion algorithm that leverages complementary strengths: on one hand, utilizing the structural feature stability of traditional texture-based minutiae algorithms to reduce or even eliminate spoofing attacks; on the other hand, retaining the multi-minutiae extraction capability of image-based algorithms to effectively recognize small-area fingerprints. This algorithm achieves an optimal balance between recognition performance (high genuine acceptance rate, low false acceptance rate) and security.

搜集汇总

背景与挑战

背景概述

该数据集源自国家重点研发计划项目'高能效人机交互芯片技术'，提供深度融合指纹识别算法的误识率仿真数据。算法旨在通过融合传统结构特征点的稳定性与图像多特征点的优势，以应对小面积指纹采集挑战，从而优化识别效果，实现低误识率和高安全性。