Improving Contactless Fingerprint Recognition with Robust 3D Feature Extraction and Graph Embedding

TL;DR

This work addresses the modality gap in contactless fingerprint recognition by introducing a 3D feature extraction framework that recovers depth, 3D minutiae, and orientation from a single 2D image, followed by a 3D graph embedding-based matching network. By modeling fingerprints in 3D and applying a graph-based comparison (with an EdgeConv/TRM-inspired architecture) and pose-correction, the method achieves robust matching across large pose variations. A monocular depth estimator is trained to generalize across datasets and to enable 3D unwarping, while fusion with a strong 2D matcher (VeriFinger) further boosts performance. Experiments on UWA, PolyU 3D+, CFPose, and ZJU demonstrate state-of-the-art results in depth estimation, minutiae extraction, and cross-pose matching, highlighting the practical impact for real-world, contactless biometric systems.

Abstract

Contactless fingerprint has gained lots of attention in recent fingerprint studies. However, most existing contactless fingerprint algorithms treat contactless fingerprints as 2D plain fingerprints, and still utilize traditional contact-based 2D fingerprints recognition methods. This recognition approach lacks consideration of the modality difference between contactless and contact fingerprints, especially the intrinsic 3D features in contactless fingerprints. This paper proposes a novel contactless fingerprint recognition algorithm that captures the revealed 3D feature of contactless fingerprints rather than the plain 2D feature. The proposed method first recovers 3D features from the input contactless fingerprint, including the 3D shape model and 3D fingerprint feature (minutiae, orientation, etc.). Then, a novel 3D graph matching method is proposed according to the extracted 3D feature. Additionally, the proposed method is able to perform robust 3D feature extractions on various contactless fingerprints across multiple finger poses. The results of the experiments on contactless fingerprint databases show that the proposed method successfully improves the matching accuracy of contactless fingerprints. Exceptionally, our method performs stably across multiple poses of contactless fingerprints due to 3D embeddings, which is a great advantage compared to 2D-based previous contactless fingerprint recognition algorithms.

Figures (10)

• Figure 1: Compared with previous methods that only uses 2D fingerprint feature, our method extracts 3D features of contactless fingerprints and encodes them via a 3D graph neural network for fingerprint matching.
• Figure 2: The pipeline of the proposed method. Our method first extracts 3D features (3D shape, 3D minutiae feature, deep features, etc.) from the input contactless 2D fingerprint, then uses the extracted 3D features to conduct matching both in 2D space with unwarping and in 3D space by extracting fixed length graph embedding.
• Figure 3: The detailed pipeline of the proposed 3D feature network, which extracts 3D features from the input contactless 2D fingerprint. The network outputs a 128-dimensional deep feature along with the depth and minutiae feature for the following graph embedding. Each decoder (period, orientation, gradient, minutiae) contains 2 inputs and 1 output: (a) The deepest features of Lite-Mono Encoder as input. (b) The middle-level features of Lite - Mono Encoder that are used as input. (d) Final output. And (c) is the intermediate output only in period and orientation decoder.
• Figure 4: Illustration of calculating 3D minutiae. (a) represents the orientation of 2D/3D minutiae in a spherical coordinate. (b) shows a 3D minutia and the surface normal $\vec{n}$ perpendicular to it.
• Figure 5: The architecture of our 3D graph matching network. The specific architecture of EdgeConv can be found in wang2019dynamic.