Mobile terminal identity card image occlusion judgment method

Abstract

The invention belongs to the technical field of identity card identification, particularly relates to a mobile terminal identity card image occlusion judgment method. According to the invention, a light-weight identity card detection network is used for positioning an identity card area. The method comprises the following steps: preparing identity card shooting pictures of various scenes, markingthe position of an identity card by using a horizontal rectangular frame to form a training sample, adding a part of pictures without the identity card into the training sample to serve as negative samples, and outputting coordinate information of an identity card external rectangular frame in the input pictures after network training; and judging whether the identity card is shielded or not by using a lightweight classification network, and properly expanding the external rectangular area of the identity card obtained in the first step, cutting and sending the external rectangular area into the network for classification, and finally obtaining a result whether the identity card is shielded or not. According to the method, the occlusion judgment of the identity card image is completed by using the two lightweight network models, and the method has the characteristics of exquisite model, high speed, high accuracy and high robustness.

Description

technical field [0001] The invention relates to the technical field of ID card identification, in particular to a mobile terminal ID card image occlusion judgment method. Background technique [0002] In the identification process of ID card information, it often happens that the ID card is blocked by objects or the image is incomplete due to improper shooting methods of the user, which affects the recognition accuracy. Adding the ID card blocking judgment module can remind the user to adjust the position of the ID card in time To avoid object occlusion, thereby improving the accuracy of the recognition algorithm. [0003] One of the existing methods is to use traditional machine learning methods to judge ID card occlusion. For example, use traditional image detection methods such as contour extraction and morphological transformation to locate the ID card, and then manually design features and use support vector machines to locate the ID card. However, traditional machine ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a mobile terminal ID card image occlusion judgment method to solve the problem of how to realize the characteristics of high accuracy, small model and fast speed proposed in the above background technology, which is suitable for deployment and operation on the mobile terminal.

Images