Identity card quality inspection method and system based on deep learning

Abstract

The invention provides an identity card quality inspection method and system based on deep learning. The method comprises the steps: acquiring a first image containing an identity card image and a background image, preprocessing the first image to obtain a second image only containing the identity card image, inputting the second image into a set reflection quality inspection model for reflectionquality inspection to obtain a third image, and inputting the third image into a set fuzzy quality inspection model for fuzzy quality inspection to obtain a fourth image, wherein the set fuzzy qualityinspection model and the set reflective quality inspection model are both obtained according to the basic structure of SqueezeNet. According to the invention, the problem that a plurality of index thresholds need to be adjusted during identity card quality inspection in the traditional image processing technology and equalization is difficult to achieve on positive and negative sample recall indexes is solved, and a traditional image processing method for separating a reflective area from a non-reflective area by using brightness, gradient and color cast can not distinguish reflective areas (characters and head portraits) of an identity card from reflective areas of the non-critical area and can not identify Great Wall watermarks similar to reflective areas.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an ID card quality inspection method and system based on deep learning. Background technique [0002] With the development of Internet finance, on the one hand, the number of securities investment groups continues to expand, and the number of accounts opened in securities companies has increased significantly; Online account opening is a typical scenario. In the process of online account opening business, customers need to submit ID cards, account opening statement videos and other materials, and securities companies need to judge whether the quality of the ID cards, videos and other materials submitted by customers is qualified, such as whether the documents are reflective, blurred, blocked, missing corners, etc. , missing edges, whether there are reflections, blurs, occlusions in images and videos, and the account opening statement does not meet the requirements, etc....

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a method and system for quality inspection of ID cards based on deep learning, which solves the problem that traditional image processing technology needs to adjust multiple index thresholds in the quality inspection of ID cards, and it is difficult to achieve a balance in positive and negative sample recall indicators; and use brightness, The traditional image processing method of separating reflective and non-reflective areas by gradient and color shift cannot distinguish the key areas of the ID card, such as text, avatar, etc., from the non-key areas of reflection and cannot recognize the problem that the Great Wall watermark similar to reflection cannot be recognized

Images