An On-Line Fabric Classification Technique Using a Wavelet-Based Neural Network Approach

Abstract

A sewing system is described that classifies both the fabric type and number of plies encountered during apparel assembly, so that on-line adaptation of the sewing param eters to improve stitch formation and seam quality can occur. Needle penetration forces and presser foot forces are captured and decomposed using the wavelet transform. Salient features extracted using the wavelet transform of the needle penetration forces form the input to an artificial neural network, which classifies the fabric type and number of plies being sewn. A functionally linked wavelet neural network is trained on a moderate number of stitches for five fabrics, and can correctly classify both fabric type and number of plies being sewn with 97.6% accuracy. This network is intended for use with dedicated DSP hardware to classify fabrics on-line and control sewing parameters in real time.