Textile defect classification using discriminative wavelet frames

Abstract

The classification of defects is highly demanded for automated inspection of textile products. In this paper, a new method for textile defect classification is proposed by using discriminative wavelet frames. Multiscale texture properties of textile image are characterized by its wavelet frames representation. For a better description of the latent structure of textile image, wavelet frames adapted to textile are generated rather than using standard ones. Based on discriminative feature extraction (DFE) method, the wavelet frames and the back-end classifier are simultaneously designed with the common objective of minimizing classification errors. The proposed method has been evaluated on the classification of 466 defect samples containing eight classes of textile defects, and 434 nondefect samples. In comparison with standard wavelet frames, the designed discriminative wavelet frames has been shown to largely improve the classification performance, where 95.8% classification accuracy was achieved.