Copied image recognition method and device based on deep learning, equipment and medium

Abstract

The invention discloses a copied image recognition method and device based on deep learning, equipment and a medium. The method comprises the steps of obtaining a to-be-identified image; carrying outgraying processing on the to-be-identified image to obtain a gray level image, and starting a first thread and a second thread; carrying out standardization processing on the gray level image througha first thread, and obtaining a direction gradient histogram feature of the gray level image; Preprocessing the grayscale image through a second thread, and obtaining a local binary pattern feature ofthe grayscale image; And generating an array to be identified according to the directional gradient histogram feature and the local binary pattern feature of the grayscale image, inputting the arrayto be identified into a reshot image identification model, and receiving an identification result output by the reshot image identification model, the identification result comprising that the image is a reshot image and that the image is a non-reshot image. According to the method, the feature extraction speed and efficiency are improved, and meanwhile, the image recognition speed, recognition accuracy and recognition efficiency are also improved.

Description

technical field [0001] The present invention relates to the field of deep learning, in particular to a method, device, equipment and medium for remake image recognition based on deep learning. Background technique [0002] With the development of the credit society, more and more application scenarios (such as: application scenarios involving finance, insurance, and security) need to verify user identities through document recognition and face recognition. However, in order to simplify the identification process and reduce the purchase cost of special equipment, the existing identity verification system verifies the identity of the user by obtaining the image of the face or certificate, so there is a situation where the identity verification system is deceived by using a re-taken photo to complete the identity verification. [0003] At present, in order to prevent this defect, the acquired image is identified through the local binary pattern (LBP, Local Binary Pattern) featu...

AI Technical Summary

Problems solved by technology

However, in order to simplify the identification process and reduce the purchase cost of special equipment, the existing identity verification system verifies the identity of the user by obtaining the image of the face or the certificate, so there is a situation where the identity verification system is completed by using a remake of the photo to deceive the identity verification system.

[0003] At present, in order to prevent this defect, the acquired image is identified through the local binary pattern (LBP, Local Binary Pattern) feature of the picture and the classification algorithm of the support vector machine (SVM, Support Vector Machine), and it is judged whether the image is a remake. If the obtained image is a remake, the identity verification cannot be passed

However, the remake recognition accuracy is around 80%, and the processing speed is slow; for some specific objects, such as the Mac screen (a computer screen introduced by Apple), the remake recognition accuracy is even lower

Images