Fuzzy Fingerprint Recognition

QingE Wu, Lifen Ding, Jincheng Li, Xiaoliang Qian, Weidong YangResearch 2015 · DOI

TL;DR

This paper introduces a novel fuzzy fingerprint recognition system that enhances accuracy and speed. It achieves this through a new preprocessing technique to isolate the region of interest and a wavelet packet transform-based feature extraction method, leading to improved anti-noise capabilities compared to existing approaches.

What Problem Does This Paper Address?

Fingerprint recognition is a crucial technology for information security due to the unique and stable nature of fingerprints. However, existing methods often struggle with issues like noise, rotation, and translation in fingerprint images, which can degrade accuracy and speed. The paper aims to address these limitations by proposing a more robust and efficient fingerprint recognition system.

Methods

The proposed system comprises three main stages:

  • Fingerprint Image Preprocessing:
    • Region of Interest (ROI) Localization: A novel method is presented to quickly and conveniently identify the ROI. This involves searching for sharp gray-level mutations along specific directions (OP and OQ) to define the boundaries of the region containing rich fingerprint information. This approach aims to eliminate the influence of noise, rotation, and translation.
  • Comparison with Existing Algorithms: The paper compares its ROI extraction algorithm with others like covariance, Canny, and Fuzzy C-Means (FCM), demonstrating superior performance in preserving fingerprint details and reducing background noise. A combined approach with the Canny algorithm is also explored, with one variant (Fig. 5c) being highlighted as ideal for preserving original image information.
  • Feature Extraction:
    • Wavelet Packet Transform: The study utilizes wavelet packet decomposition to extract energy features from the preprocessed fingerprint image. This method decomposes the image into different frequency bands and directions, allowing for the capture of texture features at various scales.
  • Energy Feature Calculation: The energy of the image at different nodes of the wavelet packet decomposition is calculated and normalized to create characteristic vectors. These vectors represent the fingerprint's texture features.
  • Matching and Recognition:
    • Minimum Distance Principle: A new recognition method is proposed that compares the characteristic vectors of an unknown fingerprint with those stored in a database. The minimum difference between these vectors is used to determine the identity of the fingerprint.
  • Hierarchical Recognition: To improve efficiency, a two-stage hierarchical recognition approach is introduced. The first stage uses a computationally less complex decomposition coefficient feature for initial filtering, creating a candidate set. The second stage then uses the energy feature for more precise matching within this smaller set.

Datasets

The paper states that the main fingerprint dataset used is from a biometric identification research center, constructed by an image institute. This database consists of 100 different fingers, with 6 fingerprint images sampled from each finger at different times. The paper does not mention the creation of any new benchmark datasets.

Key Insights and Ideas

  • ROI Localization is Crucial: The paper emphasizes that effectively isolating the Region of Interest is critical for improving subsequent feature extraction and matching accuracy. Their proposed method appears to be more effective than existing ones in this regard.
  • Wavelet Packet Transform for Texture: The use of wavelet packet decomposition for extracting energy features is a key technique. This allows for a multi-scale and multi-directional analysis of fingerprint texture, which is more informative than simpler feature extraction methods.
  • Hierarchical Matching for Speed: The introduction of a hierarchical matching strategy is a practical approach to balance accuracy and speed. By using a less complex feature for initial filtering, the system can quickly narrow down potential matches before applying a more detailed analysis.
  • Fuzzy Logic Integration (Implied): While the title mentions "Fuzzy Fingerprint Recognition," the detailed methods described primarily focus on wavelet packet transforms and minimum distance principles. The "fuzzy" aspect might be implicitly present in the gray-level mutation detection or the overall approach to handling uncertainty, but it's not explicitly detailed as a core algorithmic component in the provided text.

Key Results

  • Improved Accuracy: The proposed recognition method is reported to be more accurate than several existing methods, including feature fusion and pattern entropy, LDSSS, and a system method with eight-neighbors deviation.
  • Faster Recognition Speed: The paper claims the proposed method is faster than existing better recognition methods.
  • Stronger Anti-Noise Ability: The system demonstrates improved robustness against noise.
  • Quantitative Results: The paper presents a comparison of correct recognition rates in Figure 7. For a sample size of 500, the proposed method achieved an average correct recognition rate of 97.95%, outperforming Pattern Entropy (96.94%), LDSSS (95.97%), and System Method (95.00%). The recognition rate increases with the number of samples and tends to level off.

Limitations and Future Work

  • Acquisition Device Requirements: The proposed ROI extraction algorithm requires a high-level acquisition device to obtain high-quality fingerprint images and necessitates building a large database of high-quality fingerprints for experiments.
  • Future Study Questions: The authors identify several areas for future research:
  • Choosing or constructing a proper wavelet packet basis function.
  • Determining the optimal number of decomposition levels for the best recognition effect.
  • Investigating how to combine fingerprint and its features for better and faster distinguishing effects.

Summary updated

Abstract

In order to effectively carry out the fingerprint image recognition, this study proposes a fingerprint preprocessing algorithm and feature extraction technology. In the fingerprint image preprocessing, this study gives the location method of region of interest. In terms of feature extraction, this study extracts the decomposition energy features of fingerprint. In terms of matching recognition, this study presents a new recognition method for fingerprint and uses the method to carry out the matching and recognition of fingerprint. In the simulation, the presented fingerprint recognition method is compared with several existing major fingerprinting methods. The comparison results show that the proposed recognition method is more accurate and faster than some existing better recognition methods. Its recognition precision is higher, the recognition speed is faster and the anti-noise ability is stronger. These researches in this study on fingerprinting promote an improving of the accuracy and speed of recognition and provide a new way of thinking for target recognition, which has an important theoretical references and practical significance.
BibTeX
@inproceedings{wu2015fuzzy,
  author = {QingE Wu and Lifen Ding and Jincheng Li and Xiaoliang Qian and Weidong Yang},
  title = {{Fuzzy Fingerprint Recognition}},
  booktitle = {Research},
  year = {2015},
  doi = {10.19026/rjit.6.2166},
}
Figure