Photo background similarity clustering method based on convolutional neural network and computer

Abstract

The invention discloses a photo background similarity clustering method based on a convolutional neural network. The method comprises the following steps of preprocessing an original image based on aconvolutional neural network algorithm so as to correct a direction of an identification target in the original image; carrying out instance segmentation on foreground image features and background image features of the recognition target contained in the original image, and carrying out background extraction; performing background separation on the image subjected to instance segmentation; performing feature extraction on the separated background image to obtain a high-dimensional spatial feature map; and performing similarity clustering processing on the high-dimensional spatial feature map.The invention further provides a computer program system for implementing the method. According to the method, based on a pixel-level instance segmentation algorithm, foreground areas (portraits andidentity cards) in a real application scene are detected and removed, similarity comparison is carried out through background areas, and meanwhile, the recognition accuracy can be greatly improved byutilizing the convolutional neural network obtained through migration training.

Description

technical field [0001] The invention belongs to the technical field of image processing, and specifically relates to a convolutional neural network-based similarity clustering method for photo backgrounds and a computer. Background technique [0002] As a financial exchange between consumers and sellers, payment involves money-related links. As a third-party payment company, the first thing to do is to ensure the security of user accounts and payments. According to a multi-segment market research report, the total annual global fraud loss is about 500 million One hundred million U.S. dollars. Global losses on credit, debit, prepaid and private label payment cards totaled $16.31 billion last year alone. E-retailers and wholesalers lose more than 7.5 percent of their annual revenue to fraud, and insurance fraud (excluding health insurance) totals more than $40 billion annually. Research by DataVisor shows that large-scale financial risks have the phenomenon of "group siege"....

AI Technical Summary

Problems solved by technology

[0007] However, when using a single pHash algorithm to calculate the similarity of pictures, it often happens that the calculated similarity between pictures with completely different contents is very high, which is very likely to cause misjudgment, and the accuracy of the results cannot be guaranteed.

On the other hand, this method cannot perform similarity calculations for local features of pictures, and cannot solve problems in business scenarios

Images