Prediction of Cotton Yarn’s Characteristics by Image Processing and ANN

Abstract

Machine learning and computer vision were employed in quality assessment in the textile field for more objectivity and less expense. The estimation of yarn various parameters is of great importance for the producers and customers in order to achieve optimal quality. This research utilized image processing and artificial neural networks in order to evaluate yarn tenacity, elongation%, coefficient of mass variation%, and yarn imperfections for ring-spun and compact cotton yarns. Cotton yarn samples were collected from two mills: ring spinning and compact spinning mills. The images were taken and image analysis was employed then feature vectors were defined as the inputs of the backpropagation neural networks. Two systems were built; each one contained three modules for the estimation of the different yarn’s properties. Using the multilayer network structure proved to improve the performance of the networks leading to better parameters’ modeling. Yarn properties estimation for different yarn types was achieved using a moderately priced method.