A Circular Local Binary Pattern and Convolutional Neural Network Approach to Mutilated Nigerian banknotes Recognition

The error and stress associated with manually counting large sums of money can be reduced through the use of electronic means. However, existing electronic currency note counting machines are unable to recognize currency note denominations. This has led to the study of bank currency note recognition systems. Recent improvement in machine learning techniques have led to their application to banknote recognition. Most banknote recognition systems have examined relatively clean banknotes, which is not the case in reality. In this study, an approach to the recognition of mutilated banknotes is proposed for Nigeria currency. The banknotes were preprocessed and passed to Circular Local Binary Pattern for features Extraction. After this, feature selection and classification were performed with Convolutional Neural Network (CNN). The system showed an accuracy of 100% for both the custom CNN model and the pre-trained CNN models.