Forgery Detection and Value Identification of Euro Banknotes

Arcangelo Bruna, Giovanni Maria Farinella, Giuseppe Claudio Guarnera, Sebastiano Battiato

TL;DR

This paper presents a low-cost, hardware-software system for detecting counterfeit Euro banknotes and identifying their value using near-infrared (NIR) imaging. By leveraging unique NIR features of Euro banknotes, the system achieves robust authentication and value recognition, even under varying lighting conditions and with slight misalignments, offering a practical solution for consumer-grade validation devices.

What Problem Does This Paper Address?

The paper tackles the persistent challenge of counterfeit currency, specifically focusing on Euro banknotes. While sophisticated security features are embedded in banknotes, counterfeiters continuously develop new methods to bypass detection. Existing validation systems, especially for consumer use, often rely on expensive mechanical parts or visible light imaging, which can be easily fooled by high-quality forgeries. Furthermore, there's a growing need for accessible solutions for visually impaired individuals to identify both the value and authenticity of banknotes. This work aims to bridge this gap by developing a cost-effective, non-mechanical system that utilizes the infrared spectrum to overcome the limitations of visible light detection.

Methods

The core of the proposed system lies in its use of near-infrared (NIR) imaging. Unlike visible light, certain inks and security features on Euro banknotes exhibit distinct reflective and absorptive properties in the NIR spectrum, making them more discriminative for detecting forgeries. The system comprises:

  • Hardware Prototype: A low-cost setup featuring IR LEDs for illumination, an IR camera, an analog-to-digital converter, and an ARM-based microprocessor. Crucially, it avoids mechanical parts, contributing to its low cost and durability.
  • Calibration Phase: Addresses non-uniformity in IR LED illumination by creating a "brightness map" from a white sheet image. This map is used to normalize subsequent banknote images, ensuring consistent processing.
  • Training Phase: This is where the system learns to distinguish genuine from counterfeit banknotes and recognize their values. It involves:
  • Feature Extraction: Identifying key regions (dark and bright areas) in genuine banknotes under IR light that are characteristic of their authenticity. These are defined as $F_g$ (dark areas) and $T_g$ (bright areas).
  • Counterfeit Analysis: Analyzing the IR response of counterfeit banknotes ($F_{c,j}$ and $T_{c,j}$) to identify deviations from genuine patterns.
  • Thresholding: Determining optimal thresholds ($percD^*$ and $percB^*$) based on pixel intensity distributions in specific regions to classify banknotes as genuine or counterfeit.
  • Value Recognition: Learning specific "patches" (template regions) and their locations that are unique to each banknote denomination. These patches are used to match against input images, with a wider search area to accommodate minor misalignments.
  • Use Module: Once trained, the system takes an input banknote image, applies calibration, and then uses the learned thresholds and patches to determine genuineness and identify the banknote's value.

Datasets

The paper utilized a dataset of genuine and counterfeit Euro banknotes provided by the Bank of Italy. This dataset included:

  • A training set of 1,000 images.
  • A test set of 1,750 images.

These images were acquired under various environmental lighting conditions and with varying degrees of banknote misalignment to ensure the system's robustness. The authors note that obtaining real counterfeit banknotes, especially for IR analysis, is challenging, and their dataset was crucial for testing.

Key Insights and Ideas

  • NIR as a Discriminative Spectrum: The most significant insight is the effectiveness of the near-infrared spectrum for Euro banknote authentication. Certain security features and inks, invisible or indistinguishable in visible light, reveal crucial differences under IR illumination, making it a powerful tool against sophisticated forgeries.
  • Low-Cost, Mechanical-Free Design: The emphasis on a hardware design without mechanical parts is a key innovation. This significantly reduces manufacturing costs, maintenance, and potential wear-and-tear, making the system more accessible for consumer applications.
  • Robustness to Real-World Conditions: The system's design incorporates mechanisms (calibration, wider search areas for patches) to handle variations in lighting, illumination non-uniformity, and slight misalignments, which are common in practical usage scenarios.
  • Dual Functionality: The system effectively combines counterfeit detection and banknote value recognition within a single, integrated framework, offering a comprehensive solution.

Key Results

  • Counterfeit Detection: The system achieved 100% True Positive (correctly classifying counterfeit banknotes) and 95.7% True Negative (correctly classifying genuine banknotes) rates. The False Positive rate (genuine banknote incorrectly classified as counterfeit) was 4.3%, and the False Negative rate (counterfeit banknote incorrectly classified as genuine) was 0%.
  • Banknote Value Classification: The system demonstrated high accuracy across different denominations. For example, it achieved 88% accuracy for 5 Euro notes and 98% for 20 Euro notes, with most misclassifications occurring between similar denominations (e.g., 5 and 20 Euro).
  • Computational Speed: The system achieves a computational speed of approximately 1 second for validity checking and 2 seconds for both validity and value recognition, which is suitable for real-time applications.
  • Low Hardware Cost: The estimated cost of the hardware prototype is around $30 when produced at scale.

Limitations and Future Work

  • False Negatives in Specific Conditions: The paper notes that most false negatives (genuine banknotes misclassified as counterfeit) occurred for images acquired under direct sunlight or for very worn banknotes. Wear and tear on frequently used banknotes (like 5 and 10 Euro) was identified as a primary cause for misclassification.
  • Limited Comparison: A direct comparison with other state-of-the-art methods is difficult because many existing approaches do not address Euro banknotes specifically, focus only on recognition (not validation), or use different spectral ranges.
  • Future Work:
    • Increased Accuracy and Reduced Computation: Efforts will be made to further improve recognition accuracy and decrease computational time.
  • Alternative Feature Spaces and Classifiers: Exploring different feature spaces and classification algorithms suitable for constrained domains.
  • Multi-Currency Support: Extending the solution to recognize other currencies like US Dollars and UK Pounds.
  • Audio Output: Augmenting the prototype with audio signals to provide validity and value information for visually impaired and blind individuals.

Summary updated

Abstract

This paper describes both hardware and software components to detect counterfeits of Euro banknotes. The proposed system is also able to recognize the banknote values. Differently than other state-of-the-art methods, the proposed approach makes use of banknote images acquired with a near infrared camera to perform recognition and authentication. This allows one to build a system that can effectively deal with real forgeries, which are usually not detectable with visible light. The hardware does not use any mechanical parts, so the overall system is low-cost. The proposed solution is reliable for ambient light and banknote positioning. Users should simply lean the banknote to be analyzed on a flat glass, and the system detects forgery, as well as recognizes the banknote value. The effectiveness of the proposed solution has been properly tested on a dataset composed by genuine and fake Euro banknotes provided by Italy's central bank.
BibTeX
@inproceedings{bruna2013forgery,
  author = {Arcangelo Bruna and Giovanni Maria Farinella and Giuseppe Claudio Guarnera and Sebastiano Battiato},
  title = {{Forgery Detection and Value Identification of Euro Banknotes}},
  booktitle = {Sensors},
  year = {2013},
  url = {https://openreview.net/forum?id=rnPElTEAj6},
}
Figure