VAT invoice identification method, device, computer equipment and storage medium

Abstract

The present invention relates to a value-added tax invoice identification method, device, computer equipment and storage medium. The method includes obtaining value-added tax invoice data to be identified to obtain value-added tax invoice data to be identified; text identification is performed on the value-added tax invoice data to be identified using a recognition model , to get the recognition result; output the recognition result. The present invention establishes a recognition model, and uses the recognition model to perform text recognition on value-added tax invoices to be recognized, wherein the recognition module includes four sub-models, and the recognition module performs modeling based on the real value-added tax invoice data set captured by the mobile phone to form a classification sub-model , VAT electronic ordinary invoice detection sub-model, VAT general and special invoice detection sub-model, and character recognition sub-model to realize end-to-end fast and accurate identification of VAT invoices. The entire identification model is simple, and the accuracy of invoice identification is relatively stable. Model deployment and maintenance are convenient.

Description

Technical field [0001] The present invention relates to invoice recognition methods, and more particularly to a VAT invoice identification method, device, computer device, and storage medium. Background technique [0002] Informatization technology plays an increasingly important role in daily operation, production, and management of enterprises, and paper on paperless lines has become an unstoppable trend. VAT invoices as a credential of business transactions, is an important basis for verifying business exchanges and tax returns. In addition, VAT invoices are enrolled and verified in business systems is an extremely important but very cumbersome task in corporate financial management. If the invoice is not processed in time, it will lead to confusion in corporate financial management to further affect the normal operation of the company. Artificial entry and collection invoice information can expand space to a certain extent, but the data of the invoice document is repeatedly m...

AI Technical Summary

Problems solved by technology

[0003] Existing invoice recognition methods include detection and character recognition of key character areas of invoices, wherein the detection of key character areas of invoices generally uses template matching and CPTN text recognition network methods, and the methods used for character recognition include convolution Neural network and recurrent neural network, but the existing recognition method has a complex network model, and a large amount of invoice data needs to be collected during training, and the character area is marked, and the workload is very heavy; the model is deployed online, the operation is complicated, and the recognition speed is slow, resulting in The efficiency improvement of the invoice input system is not obvious; the accuracy of invoice recognition is unstable, and the robustness against noise is weak; the requirements for production environment deployment are high, and it is difficult to maintain in the later operation and maintenance process

[0004] Therefore, it is necessary to design a new method to solve the problems of complex models, unstable invoice recognition accuracy, difficult model deployment and maintenance

Images