CAJ | 학술논문

针对打印文件检验在公共安全应用中的需求，提出一种基于灰度共生矩阵的打印文件鉴别方法。鉴别过程包括打印文件图像采集，图像预处理，特征提取和分类判别。用自制装置采集打印文件图像，经预处理后得到单个打印字符，利用灰度共生矩阵方法对不同打印文件中的相同字符进行纹理特征提取，并采用欧氏距离分类器鉴别出源打印机。选取45台激光打印机参与测试，将打印字符的图像灰度级压缩至64级再进行特征提取，鉴别正确率达93．58％，且大幅减少了算法耗时，证明了该方法的有效性。
침대타인문건검험재공공안전응용중적수구，제출일충기우회도공생구진적타인문건감별방법。감별과정포괄타인문건도상채집，도상예처리，특정제취화분류판별。용자제장치채집타인문건도상，경예처리후득도단개타인자부，이용회도공생구진방법대불동타인문건중적상동자부진행문리특정제취，병채용구씨거리분류기감별출원타인궤。선취45태격광타인궤삼여측시，장타인자부적도상회도급압축지64급재진행특정제취，감별정학솔체93．58％，차대폭감소료산법모시，증명료해방법적유효성。
Aiming at the requirement of the printed document examination in public security application, a prin-ted document examination method based on gray level co-occurrence matrix is proposed.The method is composed of printed document image acquisition, image preprocessing, feature extraction, discrimination and classification.By using self-designed device to acquire printed document image, single characters are obtained after preprocessing. Gray level co-occurrence matrix method is applied for texture feature extraction from the same characters of different printed documents, and the source printer is identified by Euclidean distance classifier.Forty-five laser printers are selected for testing.The grayscale of printed character image is compressed to 64 levels before feature extraction, which significantly reduces the time-consuming of algorithm.The identification accuracy of 93.58% is achieved, which demonstrates the effectiveness of this method.