Minutiae-Anchored Local Dense Representation for Fingerprint Matching

2025-07-21

AI Summary

To address the insufficient robustness and accuracy of fingerprint matching under challenging acquisition conditions—including pressure distortion, plain impression, partial capture, touchless imaging, and latent prints—this paper proposes a minutiae-anchored local dense representation method. Our approach innovatively couples deep semantic features extracted from orientation-normalized local image patches with their spatial coordinates, constructing a 3D tensor descriptor that explicitly encodes spatial correspondence. Additionally, a foreground segmentation mask is incorporated to constrain matching to biometrically relevant regions. This design enables multi-level, fine-grained dense feature representation and mask-guided efficient matching. Evaluated on diverse multi-source fingerprint datasets, the method achieves state-of-the-art (SOTA) identification accuracy while maintaining high computational efficiency, making it suitable for large-scale real-time fingerprint recognition applications.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Search and Optimization: Local Search
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Fingerprint matching under diverse capture conditions remains a fundamental challenge in biometric recognition. To achieve robust and accurate performance in such scenarios, we propose DMD, a minutiae-anchored local dense representation which captures both fine-grained ridge textures and discriminative minutiae features in a spatially structured manner. Specifically, descriptors are extracted from local patches centered and oriented on each detected minutia, forming a three-dimensional tensor, where two dimensions represent spatial locations on the fingerprint plane and the third encodes semantic features. This representation explicitly captures abstract features of local image patches, enabling a multi-level, fine-grained description that aggregates information from multiple minutiae and their surrounding ridge structures. Furthermore, thanks to its strong spatial correspondence with the patch image, DMD allows for the use of foreground segmentation masks to identify valid descriptor regions. During matching, comparisons are then restricted to overlapping foreground areas, improving efficiency and robustness. Extensive experiments on rolled, plain, parital, contactless, and latent fingerprint datasets demonstrate the effectiveness and generalizability of the proposed method. It achieves state-of-the-art accuracy across multiple benchmarks while maintaining high computational efficiency, showing strong potential for large-scale fingerprint recognition. Corresponding code is available at https://github.com/Yu-Yy/DMD.

Problem

• Robust fingerprint matching under diverse capture conditions
• Integrating minutiae and ridge textures in structured representation
• Improving efficiency and accuracy in large-scale fingerprint recognition

Innovation

• Minutiae-anchored local dense representation
• Three-dimensional tensor for spatial features
• Foreground segmentation for efficient matching