Automatic Recognition of Financial Instruments Based on Anisotropic Partial Differential Equations

In this paper, anisotropic partial differential equations are used to conduct an indepth study and analysis of automatic recognition of financial bills. Firstly, it obtains the invoices of the group enterprise, uses scanning technology and related image recognition technology to capture, process, compress and slice the paper bill content, and then carries out data identification and verification of the image. It classifies the obtained electronic data information into bills, converts it into electronic information related to bills according to the corresponding categories of the bill template, and stores it in the bill table of the database to achieve the management operation of formatted electronic files. After categorizing the bills according to the electronic information of bills to match the business scenarios, financial journal vouchers can be generated according to the preconfigured voucher templates of the corresponding business scenarios, and the financial journal vouchers are converted into voucher messages using XML technology. Finally, we use agent technology to design middleware for heterogeneous financial systems to realize the function of communicating voucher messages to each other in different business systems. The system automatically extracts the key information of invoices through OCR technology and performs real-time verification and cyclical feedback to the verification results to the suppliers. The system has realized the intelligent management of the power company’s VAT invoices, thus greatly enhancing the efficiency of VAT invoice verification and settlement. The automatic tax invoice recognition system adopts a network structured tax invoice recognition model, which eliminates the cumbersome steps of character decomposition and character classification in traditional OCR character recognition. After several trials, it has obtained better experimental results in terms of recognition accuracy, with an accuracy rate of over 93% in the recognition of tax invoice data set.