A plain cloth defect detection method based on the combination of Laws texture and single-class SVM

The invention provides a plain cloth defect detection method based on the combination of Laws texture and single-class SVM. The plain cloth defect detection method adopts a linear array camera to collect an image of the cloth in motion, and preprocesses the collected cloth image. The nonlinear training data is mapped to the high dimensional feature space by constructing a single classification SVM, and the optimal separation function is constructed in the high dimensional feature space. The sample images of standard fabric without defects were input to single-class SVM, and the feature is extracted by LawsTexture Feature Extraction method, and the extracted feature is trained. Cloth images collected on site are preprocessed and input into single-class SVM to judge whether there are defectsor not. If there are defects, the local binary threshold segmentation and morphological processing are carried out on the judged image, and the detection results are obtained at last. The invention improves the detection eff