BCVerifyNet: Optical Character Recognition-Based Number Verification and Transformer Efficient Net-Based Handwritten Character and Signature Verification in Bank Cheques

The developed model initially gathered the bank cheque images from various benchmark data and then preprocessed them to remove the unwanted blurs from the images. Here, filtering and contrast enhancement approaches are suggested for image preprocessing. Then, the preprocessed images are segmented with the help of a Multi-scale TransUnet[Formula: see text] Transformer-based Efficient Network (MSTUnet[Formula: see text] to increase the efficiency of the bank cheque system of verification. After segmentation, the IFSC code, account number, cheque no, Amount, and Signature are separately obtained. The cheque is verified using the Optical Character Recognition (OCR) method to find the numbers. The handwritten characters are recognised using a Transformer-based efficient network (TUnet) and verify the handwritten characters. The same TUnet effectively obtains the signature features. Then, the similarities of the signature features are checked with the signatures from the database and then it is verified whether the particular cheque is genuine or forged. The experimental result is compared with the traditional bank cheque verification systems regarding various measures. According to the overall analysis, the developed MSTUnet[Formula: see text] model attains 96% accuracy in dataset 1 and 94% accuracy in dataset 2, which is more effective than Unet, Unet3[Formula: see text], Resunet and TransUnet. Hence, the proposed model achieves the desired result with accurate and better performance than other traditional techniques.