基于朴素贝叶斯分类器的纸病离线静态辨识方法研究

Paper Defects Offline Static Identification Based on Naive Bayes Classifier

中文关键词:  朴素贝叶斯分类器  条件概率  后验概率  纸病离线静态辨识

Key Words:naive Bayes Classifier  conditional probability  posterior probability  paper defects offline static identification

中文摘要:

      针对当前纸病处理方法通用性弱、鲁棒性差的问题，研究了一种使用概率精确判别纸病类别的方法。该方法通过训练样本获得各类纸病特征量的条件概率分布，利用朴素贝叶斯分类器原理求得某一纸病特征向量属于各种纸病的后验概率，进而通过比较各后验概率的大小进行纸病辨识，这可满足纸病辨识的静态性能要求，同时，利用朴素贝叶斯分类器具有最小错误率的特点，保证纸病辨识精度。实验结果表明，该方法具有很强的通用性，能够有效、快速地对纸病进行辨识。

Abstract:

      Considering the problems of weak algorithm versatility, poor robustness of current paper defects identification methods. This paper proposed a method of using probability value to identify the category of the paper defects. This method obtained the prior probability distribution and conditional probability value of paper defects through training samples. These values were used to posterior probability calculation which could determine the type of paper defects, to simplify the computation and meet the real-time requirements of paper defect identification process, meanwhile the application of Bayesian Classifier with the smallest error rate characteristic could guarantee the accuracy requirement of the identification. Experiments results showed that this method could effectively and quickly identify defects on paper with strong versatility．