Font Size: a A A

Research On Printer Source Identification Based On Deep Learning

Posted on:2023-07-18Degree:MasterType:Thesis
Country:ChinaCandidate:R TianFull Text:PDF
GTID:2568307022957029Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the popularity of printers and the decline in printing costs,printed documents have become an important carrier of information everywhere in people’s lives.However,while printers bring convenience to people,they also provide opportunities for criminals to take advantage of,and the number of cases related to printed documents has been increasing year by year.For the problem that characters in Chinese documents are difficult to be extracted quickly and completely,this paper proposes an improved character extraction method,which is improved on the basis of projection segmentation method,because the method based on projection segmentation can extract characters quickly,but will produce oversegmentation,this paper improves the method for this situation,firstly,the image tilt angle is calculated using Hough transform,and then the image tilt angle is calculated according to the image tilt angle.After that,the binary image is obtained by Canny edge detection of the characters,and the possible noise points are removed,and morphological processing is added to make the regions of the characters connected with each other,and finally the projection segmentation is performed.This provides an effective data set for printer source recognition methods.For the traditional printer source recognition technology,which is complicated to extract features manually and relies on a single character,this paper adopts a printer source recognition method based on convolutional neural network,which can better extract the features of the character image.At the same time,unlike traditional printer source recognition methods which are often based on a single character,this paper extends the object of study to mixed Chinese characters,and proposes a corresponding channel attention model to improve the recognition rate of mixed Chinese characters in response to the decreasing accuracy of mixed Chinese characters,so that the method is not limited to the recognition of single characters in printed documents,making this paper’s method have better application The application value of this paper is improved.The experimental results show that the printer source recognition method proposed in this paper can better achieve the recognition of Chinese printed documents.Compared with the traditional printer source identification method,it is free from the reliance on a specific single character,and can better achieve the traceability of Chinese print documents,so as to determine the printer source of Chinese print documents.
Keywords/Search Tags:Printer source recognition, Character Extraction, convolutional neural networks, channel attention
PDF Full Text Request
Related items