Bank card number recognition method based on deep learning

Abstract

The invention discloses a bank card number recognition method based on deep learning. Sigma transformation is carried out through the pixel horizontal arrangement degree of the card to correct the inclined picture to be in the horizontal direction; a sliding window capable of detecting a target is selected, then improved VggNet is used as a basic discrimination model, the size of input data is restretched to 64 * 64, an SGD optimization method is adopted to train the model, and recognition of card number numbers is achieved. Taking the value of 0.25 * the length of the sliding window as the size of the moving pace to translate from top to bottom; the position of the window is judged according to the judgment threshold value of each position, a small sliding block with the size capable of accommodating the length of a character is arranged in the sliding window, and a judgment result is generated through detection from left to right so as to achieve positioning and recognition of the bank card number. According to the method, sliding block processing is carried out depending on a discrimination threshold value, the improved VggNet model method has a high recognition rate, and compared with common plane card number recognition, the method has a higher recognition rate for concave-convex card character positioning.

Description

technical field [0001] The invention relates to the technical field of character recognition, in particular to a bank card number recognition method based on deep learning. Background technique [0002] With the rise of the mobile Internet economy, banks have launched their own mobile banking APPs to facilitate online transfers, shopping, investment and other economic activities for users. However, in the process of binding the card, the user often feels inconvenient because the bank card number has many digits and needs to be checked multiple times. Instead, the bank card automatic identification technology is used to take a photo of the front of the bank card and identify the bank card number, and then Help customers to automatically fill in the bank card number, which will greatly facilitate customers to bind cards [0003] Although the size of the bank card is basically uniform, the background of the bank card is varied, the card number has different methods such as fla...

AI Technical Summary

Problems solved by technology

[0004] The present invention aims at the defects in the existing bank card number recognition technology that it is not easy to locate and recognize the concave and convex card numbers, and provides a bank card number recognition method based on deep learning, which is a sliding window based on deep learning that uses detection and positioning The design method of the bank card number recognition system realized by means of positioning, and training and recognition through the improved VggNet model framework

Images