Automatic Detection and Classification of Defects in Knitted Fabrics

Abstract

High technological equipments are currently used in the knitted fabric production process. However, repetitive tasks are still being performed by human operators in steps of this process, thus reducing productivity and product quality. The inspection of fabric rolls which occurs after the knitting process corresponds to an example. As this activity is still performed by a human inspector, mistakes are common because of human inability to maintain attention during long working hours. In this work, the development of an automated inspection system for knitted fabric in raw and flat state is presented. In order to automatically inspect the fabric, new methods of detection and classification using computer vision are proposed. The system was evaluated in a series of tests using a large image database created from rolls supplied by the industry. Excellent rates on defect detection and classification were achieved, demonstrating the effectiveness of the proposed approach.