Defect Classification for Silk Fabric Based on Four DFT Sector Features

Abstract

We cannot imagine a world without textile and textile industry. Vital role is played by textile industry in today's world of business. Quality inspection, reliability, durability and fabric with less defects are an important factors for good apparel organizations. Fabric defect classification holds an inimitable position in demand of worthy products. In this paper we have experimented to classify the fabric defects for silk material based on its structural failures. We have used the DFT sectorization process on TILDA textile images to extract features in order to classify the defects. The Feature Vector Database (FVDB) is generated by means of four DFT sectors. FVDB is used as input in WEKA for defect classification based on two test options i.e. 10-fold cross validation and full training set. It has been observed that the rate of classification for silk cloth declines in 10-fold cross validation as compared to full training set. All characterization calculations are analyzed dependent on their accuracy and Kappa statistics. It has been observed that the Random Forest is most efficient algorithm for defect classification for silk fabric due to its high rate of classification.