Detection and recognition of bank card number based on dimension clustering and multi-scale prediction

Abstract

The invention relates to a method for detecting and identifying bank card number based on dimension clustering and multi-scale prediction. The method comprises the following steps: step S1, obtainingbank card data set and marking the obtained bank card data set; Step S2, training a card number positioning and identification model of the bank card according to the collected bank card data set, andobtaining a trained card number positioning and identification model of the bank card; Step S3: Inputing the bankcard image to be detected into the trained bankcard number positioning and identification model, and carry out card number positioning and identification on the bankcard image to be detected. The invention improves the recall rate of the detection of the bank card number and the accuracy rate of the identification classification, and can ensure the real-time property of the detection and the identification.

Description

technical field [0001] The invention relates to the fields of deep learning and computer vision, in particular to a bank card number detection and recognition method based on dimensional clustering and multi-scale prediction. Background technique [0002] Text occurring in natural scenes is an important source of information for us. For example, billboards, traffic signs, texts on various documents, etc. These texts contain clear semantic information, providing people with necessary instructions and reminders. If the detection and recognition of these texts can be realized, the understanding and analysis of the content of these scenes can be realized. With the advancement of human science and technology, it has become an inevitable trend to use machines to detect and understand text in scenes. With the widespread use of mobile devices with cameras and the continuous development of mobile payment, people have become accustomed to using mobile applications for payment, such ...

AI Technical Summary

Problems solved by technology

For example: (1) The card number cannot be separated from the background of the card surface, and the accuracy rate of using traditional character segmentation methods and character recognition methods is not high

(2) Since the number and organizational form of bank card numbers are different, using a fixed template to segment bank card numbers will lead to segmentation errors

(3) Using the traditional template matching algorithm for card number recognition, the card number and background adhesion cannot guarantee the correct rate

(4) If the face correction of the bank card fails, it is difficult to locate the card number of the non-standard bank card using the traditional character positioning method

[0004] The location and identification of bank card numbers in natural scenarios plays an important role in production and life, but its research progress is slow. The reasons can be attributed to three points: it is difficult to collect bank card data sets that belong to private data, there are many types of bank cards and The scene of bank card shooting is complicated

Images