Mobile terminal account opening data bank information automatic extraction method based on machine learning

Abstract

The invention discloses a mobile terminal account opening data bank information automatic extraction method based on machine learning, and the method comprises the steps of recognizing the type of anaccount opening data image through a trained image classification model; identifying the direction of the account opening data picture; performing image operation according to the identification result, and performing rotation forward processing on the non-forward account opening data picture; identifying a text area in the account opening data picture; for an account opening data picture of whichthe type is an account opening license, predicting position information of a text region by using the trained table regression device model; verifying the predicted position information, and intercepting a picture of the target text area; and performing OCR character extraction on the intercepted picture to obtain text information of the target character area. Automatic identification and extraction of the bank information in the account opening data are realized at the mobile terminal so that the back-end service pressure can be greatly reduced, the cost can be saved, the model size can be greatly reduced and the method is more suitable for mobile terminal application.

Description

technical field [0001] The present invention relates to the use of machine learning tools and methods to solve the problem of target image extraction, especially the extraction of specific text regions, especially the classification and identification of account opening data based on the image classifier model and form regression model trained by supervised machine learning. The method for automatically identifying the account opening bank account and the name of the account opening bank of the account opening license type is specifically a method for automatically extracting bank information from mobile terminal account opening data based on machine learning. Background technique [0002] When registering on a B2B platform (such as an online bidding platform), it is generally necessary for the registered company to provide some qualification information, usually requiring the information of the opening bank. When registering on the mobile phone terminal, the user needs to t...

AI Technical Summary

Problems solved by technology

Although the aforementioned patent applications are all oriented to mobile terminals, they essentially use neural network algorithms to identify image features to achieve the purpose of target detection. They belong to the application of image target detection models, which need to store images obtained during training. Features and weights lead to a large model size, and it is impossible to integrate multiple models in the mobile application package, resulting in poor user experience

Images