Neural network modeling of fabric pilling evaluation

In quality control of modern textile industry, it is necessary to develop some objective methods to evaluate cloth appearances, such as pilling, wrinkling and seam puckering. This paper presents a new method to evaluate the fabric pilling based on artificial intelligence techniques (image analysis and neural network). In this work, we first digitalizc the fabric appearance using digital camera, after that, we extract the essential features of pills on fabric surface using image analysis techniques, which could characterize the fabric appearances objectively. At last, we apply neural network models to establish the correlationship between these essential features and final rating grade of fabric pilling. The experimental results in this paper show that good correlation can be achieved between objective and subjective evaluation data. It proves that artificial intelligent system is able to simulate the functions of human eyes and brain. The aim of this paper is to develop an artificial expert system for the quality control of fabric appearances, so as to liberate human being from the tedious, rcpcatablc, time- consuming subjective judging works. It will help us to build up an objective evaluation and quick response management system for modern textile industry.