Authors: Assistant Professor P.Premchand Goud, Jakkula Supriya
Abstract: Even though digital banking services are widely used, bank cheques are still crucial for financial transactions. Traditional methods of verifying checks, on the other hand, rely heavily on human intervention, which may be laborious, error-prone, and even fraudulent. This study proposes a framework for automated bank cheque verification that uses deep learning and image processing to overcome these obstacles and increase the trustworthiness and efficiency of cheque authentication. By use of picture preprocessing and segmentation, the suggested method retrieves crucial information from checks, such as the account number, legal amount, courtesy amount, IFSC code, and the signature of the account holder. While a Convolutional Neural Network (CNN) is used for handwritten numerical character identification, Optical Character identification (OCR) is used for machine-printed text recognition. Signature verification and classification are handled by Support Vector Machines (SVMs), while feature extraction is handled by Scale Invariant Feature Transforms (SIFTs). By automating the verification process, the suggested architecture drastically cuts down on human participation while still adhering to the CTS-2010 criteria for cheque verification. Results from experiments show that the suggested technique works, with CNN model accuracy reaching 99.14% for handwritten digit recognition, OCR accuracy reaching 97.7% for machine-printed text recognition, and SIFT-SVM accuracy reaching 98.10% for signature verification. Based on these results, the suggested framework is a safe, efficient, and accurate way to verify bank checks automatically, which speeds up processing and improves the authenticity of the checks.
