A novel classification approach of weld defects based on dual-parameters optimization of PCA and LDA

摘要

To improve the classification accuracy of defects, a novel algorithm has been developed based on dual-parameters optimization of the principal component analysis (PCA) and the linear discriminant analysis (LDA). The original defect images are transformed to eigen-defects by PCA which contains all features of these defects. Then, LDA is used to classify eigen-defects. The optimal parameters of PCA and LDA are given when the PCA-LDA model gets the maximum value of classification accuracy. For estimating the actual classification accuracy of the proposed method in a concrete system, Bootstrap evaluation method is introduced. The experimental result demonstrates that the accuracy of this method is 91.12%, which promotes the accuracy by 0.37%, 3.61% and 8.72% comparing with PCA-SVM, SVM and MLP-ANN.