Currency Recognition System Using Image Processing

Abstract

This paper presents a currency recognition system developed using image processing and machine learning techniques to address the challenges faced by visually impaired individuals and in situations involving worn or damaged currency notes. The system captures images of currency notes and preprocesses them using techniques such as noise removal, grayscale conversion, and image enhancement. Feature extraction is performed using methods like Scale-Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF), and ORB. The core of the system is built using the EfficientNet-B0 deep learning architecture, implemented with PyTorch and TensorFlow, and integrated with a user-friendly interface using Streamlit. The system aims to provide real-time, accurate classification of currency denominations, supporting multiple currencies and varying note conditions. Testing confirms successful functionality across unit, integration, and acceptance phases. The system offers a scalable, cost-effective, and portable alternative to traditional hardware-based solutions.