Fabric Defect Detection and Classification Using Image Analysis

Abstract

Conventional image analysis hardware was used to image solid-shade, unpattemed, woven fabrics. Two different software approaches for detecting and classifying knot and slub defects were studied and compared. The approaches were based on either gray level statistics or morphological operations. The autocorrelation function was used for both methods to identify fabric structural repeat units, and statistical or morphological computations were based on these units. Plain weave and twill weave fabrics were used to compare the performance of each software approach.