Detection Of Fake Bank Currency With Machine Learning Algorithms

Abstract

The rapid advancement in printing and scanning technologies has led to an increase in counterfeit currency production, posing a significant challenge for financial institutions and economies. Existing banknote authentication systems, though effective, are often prohibitively expensive, limiting their accessibility. This thesis presents a cost-effective, accurate, and reliable approach for detecting counterfeit banknotes using machine learning and image processing techniques. The proposed system extracts and analyzes key currency features, such as micro-printing, watermarks, and ultraviolet (UV) lines, by leveraging Optical Character Recognition (OCR), Face Recognition, and the Canny Edge Detection along with the Hough Transformation Algorithm implemented in MATLAB. The system compares extracted features of suspected banknotes with genuine currency templates to determine authenticity. The model's efficiency and accuracy were tested using the Bangladeshi 1000 Taka note, ensuring its practical applicability. The proposed solution emphasizes affordability, scalability, and ease of deployment, making it suitable for both large financial institutions and smaller businesses. Through rigorous experimentation, the system demonstrated high reliability and precision, providing a promising tool for counterfeit currency detection.