Recognizing Banknote Fitness with a Visible Light One Dimensional Line Image Sensor

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Accurate banknote fitness classification is crucial for currency recirculation.
• Previous research primarily utilized visible, infrared, and ultraviolet light sensors.
• Limited research exists on Indian paper currency fitness classification.

Purpose of the Study:

• To propose a novel method for classifying the fitness of Indian banknotes using visible light.
• To address the challenge of classifying banknote soiling, a key factor in determining fitness.
• To develop an efficient system for recirculating clean Indian Rupees.

Main Methods:

• Utilized a one-dimensional line image sensor with visible light for banknote imaging.
• Extracted features from specific regions of interest (ROIs) with minimal texture.
• Employed 1-level discrete wavelet transformation (DWT) for feature extraction.
• Selected optimal DWT features using linear regression analysis against densitometer data.
• Classified banknote fitness using a support vector machine (SVM) with selected features.

Main Results:

• The proposed method demonstrated superior performance in classifying banknote fitness compared to existing approaches.
• Accurate identification of soiling on Indian banknotes was achieved.
• The feature selection process effectively identified discriminative DWT features.

Conclusions:

• The developed method offers an effective and novel approach for Indian banknote fitness classification.
• Visible light imaging combined with DWT and SVM provides a robust solution for currency sorting.
• This research contributes to improving the efficiency of Indian currency management and recirculation.