Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fixed-length Dense Descriptor for Efficient Fingerprint Matching

Zhiyu Pan, Yongjie Duan, Jianjiang Feng, Jie Zhou

TL;DR

This work tackles efficient fingerprint matching with robustness to partials and background noise by introducing Fixed-length Dense Descriptor (FDD), a 3D fixed-length descriptor that aligns with the fingerprint space. The method uses a dual-branch network (minutia and texture) with a 2D sinusoidal positional embedding and a pose-aligned pipeline to produce dense descriptors, then matches fingerprints via region-overlapping cosine similarity. Experiments on diverse datasets show that FDD outperforms previous fixed-length descriptors, especially for partial, cross-modal, and noisy impressions, and even binarized or low-dimensional FDD variants remain effective; fusion with minutiae-based methods further improves accuracy. The results suggest FDD is a practical, efficient alternative for large-scale fingerprint recognition, enabling fast similarity searches while maintaining robustness to pose and background variability.

Abstract

In fingerprint matching, fixed-length descriptors generally offer greater efficiency compared to minutiae set, but the recognition accuracy is not as good as that of the latter. Although much progress has been made in deep learning based fixed-length descriptors recently, they often fall short when dealing with incomplete or partial fingerprints, diverse fingerprint poses, and significant background noise. In this paper, we propose a three-dimensional representation called Fixed-length Dense Descriptor (FDD) for efficient fingerprint matching. FDD features great spatial properties, enabling it to capture the spatial relationships of the original fingerprints, thereby enhancing interpretability and robustness. Our experiments on various fingerprint datasets reveal that FDD outperforms other fixed-length descriptors, especially in matching fingerprints of different areas, cross-modal fingerprint matching, and fingerprint matching with background noise.

Fixed-length Dense Descriptor for Efficient Fingerprint Matching

TL;DR

This work tackles efficient fingerprint matching with robustness to partials and background noise by introducing Fixed-length Dense Descriptor (FDD), a 3D fixed-length descriptor that aligns with the fingerprint space. The method uses a dual-branch network (minutia and texture) with a 2D sinusoidal positional embedding and a pose-aligned pipeline to produce dense descriptors, then matches fingerprints via region-overlapping cosine similarity. Experiments on diverse datasets show that FDD outperforms previous fixed-length descriptors, especially for partial, cross-modal, and noisy impressions, and even binarized or low-dimensional FDD variants remain effective; fusion with minutiae-based methods further improves accuracy. The results suggest FDD is a practical, efficient alternative for large-scale fingerprint recognition, enabling fast similarity searches while maintaining robustness to pose and background variability.

Abstract

In fingerprint matching, fixed-length descriptors generally offer greater efficiency compared to minutiae set, but the recognition accuracy is not as good as that of the latter. Although much progress has been made in deep learning based fixed-length descriptors recently, they often fall short when dealing with incomplete or partial fingerprints, diverse fingerprint poses, and significant background noise. In this paper, we propose a three-dimensional representation called Fixed-length Dense Descriptor (FDD) for efficient fingerprint matching. FDD features great spatial properties, enabling it to capture the spatial relationships of the original fingerprints, thereby enhancing interpretability and robustness. Our experiments on various fingerprint datasets reveal that FDD outperforms other fixed-length descriptors, especially in matching fingerprints of different areas, cross-modal fingerprint matching, and fingerprint matching with background noise.
Paper Structure (15 sections, 6 equations, 5 figures, 5 tables)

This paper contains 15 sections, 6 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Network Design
  5. Fingerprint Matching
  6. Objective Functions
  7. Classification Loss
  8. Auxiliary Loss
  9. Local Similarity Loss
  10. Implementation Details
  11. Experiments
  12. Datasets
  13. Compared Methods
  14. Results
  15. Conclusion

Figures (5)

  • Figure 1: The illustration of Fixed-length Dense Descriptors (FDD) of a rolled fingerprint and a mated plain fingerprint. Two corresponding regions and their features are marked.
  • Figure 2: Overview of fingerprint matching using the proposed FDD. First, fingerprint pose alignment is performed according to Duan et al.duan2023estimating, followed by the extraction of fixed-length dense descriptors with the descriptor extraction network, which are then used for matching.
  • Figure 3: The overall network architecture of FDD consists of the minutia branch and the texture branch, which extract minutia maps and masks, respectively, along with their corresponding descriptors. "Desc." is an abbreviation for "Descriptor".
  • Figure 4: Image examples from different fingerprint datasets (a) NIST SD4, (b) NIST SD14, (c) DPF, (d) FVC2002 DB3A, (e) FVC2004 DB1A, (f) FVC2006 DB1A, (g) N2N Plain, (h) PolyU CL2CB, (i) NIST SD27, (j) THU Contact10K. Images are input at 500 ppi, and they have been rescaled here to ensure optimal visualization.
  • Figure 5: Examples of the FDD extracted from genuine pairs of (a) NIST SD4, (b) DPF, (c) PolyU CL2CB, and (d) NIST SD27. The fingerprint images shown in the figure have been aligned based on their estimated poses.