On-line Defect Detection and Classification of Latex Gloves

At present, the detection of latex glove is mainly done manually. In order to realize the online detection of latex glove, this paper presents a method of defect detection and classification of latex glove based on multi-feature extraction to achieve online detection and classification of latex glove. Firstly, the target pixel and the background pixel were segmented using the OTSU method. Secondly, the image was pre-processed using median filtering, edge extraction and morphological closed arithmetic processing, and then the defect was framed using the minimum circumscribed rectangular frame selection method. Finally, the characteristics of three kinds of defects were proposed by analysis. Feature vectors were extracted using texture features and HOG features. The classification model selects SVM support vector machine to train the classifier to realize the classification of defects. When training the classifier, the Gaussian radial basis function was used as the kernel function. In view of the characteristics of the types of defects in latex glove, first classify the defects in a “one-to-one” manner, and then implement multi-classification. The experimental results show that the online defect detection and classification of latex glove based on multi-feature extraction has achieved good results.