Applying pattern recognition principles to grading textile yarns

This research reports on a simple and effective system for grading textile yams by pattern recognition theory. During the learning processes of textile yarn grading, yam property vectors can be transferred to the principal axis vector (PAV) Of the orthonormal function with Karhunen-Loeve expansion, in order to select features and then reduce dimensions. With the Bayes classifier or the minimum distance method, the decision function for grading textile yams can be obtained. Then, under the premise of not sacrificing the identification rate, the learning process can be repeated to search for the transfer matrix and the decision function with PAVS of the lowest identifiable dimension. This system is a quick and effective way to grade textile yarns by using PAVS of only one dimension, thus both simplifying the identification system and providing objective grading results.