A Defect Recognition Method Based on Dual-Parameter Optimization KPCA and DMSVM

摘要

In order to improve the accuracy of defect classification for automated radiographic NDT (Nondestructive Testing), a dual-parameter optimization KPCA (kernel principal component analysis) method for defect recognition is proposed. Firstly, 25 kinds of defect shape-location characteristics and 4 kinds of defect gray characteristics are presented to describe the defect. Secondly, dual-parameter optimization KPCA is applied to transform the defect features to a high dimensional space, and the nonlinear characteristics are obtained in high dimensional space. Finally, DMSVM (direct multiclass support vector machine) is used to classify the defects. A case study is provided to illustrate our work. The experiment shows that the method based on dual-parameter optimization KPCA improves the accuracy of defects recognition to 91.87%, which is 2.67% higher than the traditional DMSVM method.