融合频谱分析和KPCA-SVM的磁瓦内部缺陷检测方法

Method of the magnetic tile internal defect detection combining spectrum analysis with KPCA-SVM

针对磁瓦内部缺陷自动化检测的实际需求,文章提出了一种融合频谱分析、核主成分分析和支持向量机的磁瓦内部缺陷音频检测方法。通过频谱分析找到峰值频点所处频率段,并提取频率段幅值数据;利用核主成分分析对幅值数据进行特征提取和降维;构造支持向量机,使用所提取的特征进行磁瓦内部缺陷识别。试验证明,3类磁瓦缺陷识别率达到了100%,优于双谱切片方法。结果表明,提出的检测方法检测速度快、高效、准确,能够很好地应用于磁瓦内部缺陷检测。

Aiming at the actual demand of automatic detection of internal defects of magnetic tiles,a novel approach is proposed based on spectrum analysis,kernel principal component analysis(KPCA)and support vector machine(SVM).The frequency bands containing the peak points are found and the corresponding amplitude data was extracted.Then the feature was extracted and the dimension of feature is reduced by KPCA.The internal defect was identified by SVM based on the extracted feature.The experiment showed that the accuracy rate of three types of magnetic tiles achieved 100%,higher than the bispectrum method.The results demonstrate that the proposed method is of fast speed,high effective,accurate and can be well applied to magnetic tiles internal defect detection.