Bank Cheque Validation Using Image Processing

Abstract

Bank cheques, as documents issued by banks can be used as a form of bills capable of monetary exchange, allowing a payee a certain sum of money from the account of drawer. However, due to many fraudulent practices and a need of faster cheque clearance, there had been advances in the process of cheque clearance. Consequently, to aid the process of cheque validation this research work focuses on implementing image processing techniques such as OCR, ANN and Deep Learning to extract key parameters essential for cheque validation. These techniques can be used in sequential manner to automate the task of cheque validation. For extracting machine typographic information Optical Character Recognition is used. Whereas, for the handwritten characters we have used CNN trained using MNIST dataset. The accuracy achieved in handwritten character recognition is 99.14%. For testing purposes IDRBT cheque dataset is used comprising cheque leaflets of different banks.