Available at: https://digitalcommons.calpoly.edu/theses/3310
Date of Award
6-2026
Degree Name
MS in Electrical Engineering
Department/Program
Electrical Engineering
College
College of Engineering
Advisor
Jane Zhang
Advisor Department
Electrical Engineering
Advisor College
College of Engineering
Abstract
This thesis addresses the global challenge of counterfeit paper currency by proposing a classical computer vision framework for distinguishing genuine United States Dollar (USD) bills from counterfeit ones using image data. In contrast to existing approaches that rely on a large number of easily reproducible visual features, this work prioritizes the detection of a single, robust security feature: the ultraviolet (UV) reactive security strip embedded in genuine USD bills of denominations $5 and above. By focusing on a feature that is inherently difficult to replicate, the proposed method reduces the likelihood of counterfeit bills being misclassified as genuine.
The system operates in two stages. First, the denomination of a bill is determined independently of its authenticity using a combination of relatively insecure visual features, including the shape of corner numerals and the portrait face. Template Matching with Sum of Absolute Differences (SAD) is used on the sample corner numeral, and the Viola-Jones algorithm with Eigenface-based principal component analysis is used to compare the sample bill’s portrait face. Metrics extracted from these two features are combined through a k-Nearest Neighbors classifier to accurately classify sample bill denomination. In the second stage, authenticity is assessed by detecting the presence of the UV security strip in images captured with ultraviolet backlight illumination. Using the known denomination to constrain the search region, the strip is identified through edge detection in L*a*b* color space and extraction of vertical line structures via the Hough Transform.
The proposed approach is evaluated on a dataset of 240 bill samples, consisting of both genuine and counterfeit $5, $10, and $20 notes. The system achieves 100% accuracy in denomination classification and an overall authenticity classification error rate of approximately 5.4%, with only a small portion attributed to false negatives. Misclassifications are primarily caused by background patterns containing vertical lines producing false detections as well as weak fluorescence of genuine strips reducing visibility under UV light.
Overall, this thesis demonstrates that classical computer vision techniques can achieve reliable counterfeit detection in data-constrained environments by emphasizing high-integrity security features. The modular design further allows for the integration of additional features, offering a scalable and efficient framework for future enhancements.