Automated System for Indian Banknote Recognition using Image Processing and Deep Learning

This work proposes to develop an automated system for Indian banknote recognition using image processing and deep learning techniques which will be invariant of orientation and independent of the sides or faces of the notes. The images of currency notes are preprocessed to compensate for rotation and then passed through a technique to detect whether they are Indian banknotes or not. If found to be Indian banknote, then the denomination is classified in two different ways: (i) extracting colour and texture features to form feature vectors and the test samples are classified using k-NN based on those features, (ii) images of Indian banknotes are directly fed to a Convolutional Neural Network (CNN) for classification.