基于多特征的打印文件层级分类研究

打印文件鉴别是一种广泛应用于安全领域的取证技术,因此对其检测的准确率和速度均有较高要求。考虑到单个特征的信息不全,基于多特征融合的方法来提高准确率,同时使用基于Ada Boost的SVM级联分类器进行分类判定。研究过程包括数据采集、图像预处理、GMM和LBP特征提取、特征融合、单个SVM分类器分类以及基于Ada Boost的层级SVM分类器分类。通过对4 000张图片集提取GMM和LBP特征,然后进行特征融合,输入分类器分类,结果表明,该方法能够在一定程度上提高鉴别的准确率和速度,具有良好的可扩展性。...

Full description

Saved in:
Bibliographic Details
Published in电子技术应用 Vol. 42; no. 3; pp. 113 - 115
Main Author 周晶晶 陈庆虎 彭文花 鄢煜尘
Format Journal Article
LanguageChinese
Published 武汉大学电子信息学院,湖北武汉,430072 2016
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:打印文件鉴别是一种广泛应用于安全领域的取证技术,因此对其检测的准确率和速度均有较高要求。考虑到单个特征的信息不全,基于多特征融合的方法来提高准确率,同时使用基于Ada Boost的SVM级联分类器进行分类判定。研究过程包括数据采集、图像预处理、GMM和LBP特征提取、特征融合、单个SVM分类器分类以及基于Ada Boost的层级SVM分类器分类。通过对4 000张图片集提取GMM和LBP特征,然后进行特征融合,输入分类器分类,结果表明,该方法能够在一定程度上提高鉴别的准确率和速度,具有良好的可扩展性。
Bibliography:gaussian mixture model; local binary pattern; feature fusion; cascade classifier; print document identification
Zhou Jingjing,Chen Qinghu,Peng Wenhua,Yan Yuchen (School of Electronic Information, Wuhan University, Wuhan 430072, China)
11-2305/TN
Printed document identification is a kind of technology which is widely used in the security field. So it needs higher ac-curacy and speed. Considering the information of a single feature is not complete, this paper improves the accuracy based on mul-ti- feature. The research process includes data acquisition, image preprocessing, GMM and LBP feature extraction, feature fusion,SVM classifier and cascade detector. Through extracting the GMM and LBP features from 4 000 image sets, the feature fusion is performed, and the results show that the proposed method can improve the accuracy and speed of the identification in a certain ex-tent.
ISSN:0258-7998
DOI:10.16157/j.issn.0258-7998.2016.03.032