打印文稿识别技术研究与设计

Research and Design of Print Document Recognition Technology

近年来由于现代数字电子技术的飞速发展,各类合同、文档、传单的真实性有待于考证;而且也迫切需要准确掌握各类打印文档的来源。针对打印文档的真伪性问题,提出在对单个字符图像进行灰度级压缩和二值去噪的基础上,利用统计法中的灰度共生矩阵来提取字符图像的特征值;并以这些特征值建立相应的样本数据库,借助神经网络工具箱对特征值进行分类和识别。试验对试验条件的控制较为严格,样本采集为5种全新打印机打印的文档各取5张,字符图像的提取采用同一扫描仪,并且控制工作时间和打印文档的数目一致。实验结果表明,观察提取的特征数据发现,不同打印文档的特征数据之间有明显差别,而且特征数据的正确性和可靠性明显提高。试验中对5种打印机全部正确识别。

In recent years,due to the rapid development of modern digital electronic technology,types of contracts,documents,leaflets authenticity to be verified,but also an urgent need for accurate information on the source of all kinds of printed documents.The authenticity of the document for printing problem,a single characterbased grayscale image compression and de-noising based on binary to extract characteristic value of the character images using statistical method GLCM be can used,and these eigenvalues establish the appropriate sample database,using neural network toolbox eigenvalues classification and identification.The trial of the test conditions more stringent controls,sample collection for the five kinds of new printer to print a document from each of five extracted character images using the same scanner,and control the number of working hours in line and print documents.Experimental results show that the extracted feature data observation found a significant difference between the different characteristics of the print document data,and the accuracy and reliability of the feature data has improved significantly.The test correctly identified all five kinds of printers.