Deep Learning-Based Banknote Fitness Classification Using the Reflection Images by a Visible-Light One-Dimensional...

Area of Science:

• Computer Science
• Artificial Intelligence
• Image Processing

Background:

• Banknote fitness classification is crucial for automatic currency sorting to determine recirculation suitability.
• Existing methods often require pre-classification of banknote denomination and input direction.
• Visible-light reflection images have been used, but limitations persist.

Purpose of the Study:

• To develop a deep learning-based method for banknote fitness classification that is independent of denomination and input direction.
• To improve the accuracy and efficiency of automatic paper currency sorting systems.

Main Methods:

• Utilized a convolutional neural network (CNN) for image analysis.
• Employed visible-light one-dimensional line image sensors to capture banknote reflection images.
• Trained and tested the model on diverse banknote image databases (KRW, INR, USD).

Main Results:

• The proposed deep learning method achieved higher classification accuracy compared to existing methods.
• Demonstrated effectiveness across multiple currencies (KRW, INR, USD) with varying fitness levels.
• Successfully classified banknote fitness irrespective of denomination and input orientation.

Conclusions:

• The deep learning-based approach offers a robust solution for automated banknote fitness classification.
• This method overcomes the limitations of pre-classification, enhancing sorting system capabilities.
• The findings have significant implications for improving the efficiency and accuracy of currency processing worldwide.