基于PCA的纸病特征再提取算法研究

Research on Re-extraction Algorithm of Paper Defect Characteristics Based on PCA

中文关键词:  纸病特征  特征维数  主成分分析  检测算法  运算量

Key Words:paper defect features  feature dimension  principal component analysis  detection algorithm  computation amount

中文摘要:

      针对传统纸病检测中相似纸病辨识准确率低及纸病提取特征维数高致使纸病辨识过程时间较长的问题，提出一种基于主成分分析（PCA）的纸病特征再提取算法。该算法以多种纸病的图像为研究对象，对可能存在相关关系的高维原始纸病特征量进行PCA降维处理并去除相关成分，形成相互独立且更具代表性的纸病新特征，在减少数据处理量的同时使纸病辨识准确率明显提高。实验表明，PCA算法可显著提高纸病辨识准确率并可大幅缩短算法平均运行时间。

Abstract:

      Because of the low accuracy in identification of similar paper defects in traditional paper defect detection and the slow running speed of the system caused by high feature dimension extraction， a PCA-based paper defect feature re-extraction algorithm was proposed. This method took various paper defect images as the research object， PCA was adoped to deal with the dimension reduction of high-dimensional original features that may have correlations and remove their related components so as to form new defect features which were indepen-dent and more representative， so that the data processing amount was reduced. At the same time， the identification accuracy of paper defects could be significantly improved. Experiments showed that the algorithm could significantly improve the accuracy of paper defect identification and the average running time of the system was greatly shortened．