Plate surface defect recognition method based on wavelet moment invariant and locality preserving projection

A feature extraction method based on wavelet moment invariant and locality preserving projection (LPP) was presented and applied to the automatic recognition of plate surface defects. 3-level wavelet decomposition was performed on the surface images, details of the plate surface images were decomposed into components on several scales, and then the noise scattered in detail components of all the scales was reduced by wavelet shrinkage. Moment invariants were extracted from amplitude spectra of all the components, and then the feature vector composed by all the moment invariants was reduced from 77-demension to 8-dimension via LPP. At last, an AdaBoost classifier based on decision trees was constructed to classify the samples. Experimental results demonstrated that the feature extraction method presented in this paper was applicable to the classification of plate surface defects, and the classification rate was 91.60%.