PCA和贝叶斯分类技术在焊缝缺陷识别中的应用  

Application of PCA-Bayesian classification technology to recognition of weld defects

作　　者：蔡晓龙[1] 穆向阳[1] 高炜欣[1] 李亮[1] 

机构地区：[1]西安石油大学陕西省钻机控制重点实验室,710065

出　　处：《焊接》2014年第3期31-35,70,共5页Welding & Joining

基　　金：陕西省自然科学基础研究计划项目(2010JQ8033)

摘　　要：以埋弧焊管焊缝的X射线检测图像为应用对象,针对图像中噪声点与微小缺陷易混淆的问题,提出将朴素贝叶斯与主成分分析法相结合,应用到焊缝图像的缺陷识别中的思路。首先,通过主成分分析进行特征向量的去冗余和正交化处理;其次,采用核密度估计的方法进行未知分布样本的条件概率密度函数估计。最后,利用贝叶斯原理实现缺陷类型的判别。试验表明,通过与主成分分析法的结合,朴素贝叶斯方法在焊缝缺陷识别的准确率上提高了5.5%,可有效地应用于焊缝检测图像的缺陷识别。To solve the problem of confusion in X-ray images noises and small defects,a method of combining the principal component analysis(PCA) with the Naive Bayesian classifier(NBC) was proposed in the weld seam defects recognition of submerged arc welding of pipe.Firstly,the PCA was adopted to ensure the features vector orthogonal and less redundancy.Then the kernel density estimation method was used to calculate the conditional probability density function of unknown sample distribution.At last,the Bayesian theory was applied to classify the type of weld defects.Experimental results demonstrated that the accuracy of the NBC combining with the PCA on weld defects recognition can be improved by 5.5,which demonstrate that the PCA-NBC method can be effectively applied to the defects recognition of X-ray images.

关 键 词：朴素贝叶斯 主成分分析 缺陷识别 

分 类 号：TG441.7[金属学及工艺—焊接]