A New Approach to the Unsupervised Detection and Classification of the Spliced Yarn Joint

This paper presents an automatic vision-based system for unsupervised detection and classification of spliced yarn joints. In the splice detection process, a competitive learning method based on an LBG algorithm is used. In the splice classification process, a dynamic time warping (DTW) algorithm is used to classify the extracted splice joint into one of three categories, based on the degree of similarity between the spliced joint and the nonspliced remaining part of the same yarn. The use of DTW in the classification makes the proposed method adaptable to different types of yarns. Consequently, this method might be universally applicable for the classification of all spliced yarn joints. The proposed method has been evaluated using three types of experiments, yielding a promising result.