Attention CNN-based document certificate type image tampering detection method

Abstract

The invention provides a document certificate type image tampering detection method based on attention CNN. The method comprises the following steps: adding a convolutional neural network of a multi-semantic attention mechanism into a framework of a model in a network structure, increasing the attention of tampering edges, and generating an attention graph, meanwhile, a maximum entropy Markov model is used for modeling correlation between adjacent regions in the attention map; the expansion convolution can capture multi-scale context information; the overall framework of the convolutional neural network introduced with the attention mechanism has two branches, the two branch networks have the same weight parameter and structure in a whole convolutional layer and a full connection layer, and in a training stage, a set of a plurality of loss functions is used for training; the obtained preliminary detection result is a binary image, and the binary image is subjected to corrosion expansion processing to obtain a final result. The method has the advantages that various tampering means can be handled, detection is convenient, and the accuracy is high.

Description

technical field [0001] The invention relates to the field of image tampering detection, in particular to a kind of tampering detection of document and certificate images based on unnatural images of attention convolutional neural network (CNN). Background technique [0002] The technology of detecting tampered and forged images is called image forensics. It has been developed and researched for more than ten years. The focus is on the tampering detection of natural images, and the related detection and research on unnatural images of documents and certificates is still very limited. With the rapid development of the Internet, unnatural images such as electronic invoices and certificates have begun to be widely used. The emergence and popularization of image editing tools, such as Adobe Photoshop, make the tampering of documents and certificates like selfie editing. Figure 1 As simple as that, due to the mature development of this kind of PS technology, it is difficult to di...

AI Technical Summary

Problems solved by technology

This method must ensure that the OCR recognition is accurate and the alignment of the reference characters must be precise, otherwise there will be a large error. When the distance between the character and the baseline is large, the recognition error will be large, so this method has great limitations.

Images