基于因子分析的打印文档鉴定方法

Identification Method of Printed Document Based on Factor Analysis

针对现有打印文档鉴定方法在检材和样本中无相同字符时基本失效的问题,提出一种基于因子分离的打印文档鉴定方法。该方法将字符图像看作是融入了字符形态结构的残缺纹理图像,并建立基于灰度共生矩阵特征的2个因子模型,该特征可划分为纹理因子和字符因子两部分。通过因子分离方法获得与字符无关的独立纹理特征,并利用最小距离分类器来实现打印文档的同机鉴定。选取35台激光打印机参与测试,鉴别正确率达94.29%,证明了该方法的有效性。

A printed document identification method based on factor separation is proposed, since the existing printed document i- dentification methods rose efficacy when there are no identical characters between test and training documents. In the proposed method, the character image is considered as an incomplete texture image integrated with character structure. A two-factor model based on gTay-level co-occurrence matrix feature is established, and the feature is decomposed into texture factor and character factor. Thus, character-independent texture feature is obtained, and minimum distance classifier is applied to implement printed document identification. Thirty-five laser printers are selected for testing, and the accuracy can reach to 94.29% , which proves the effectiveness of the proposed method.