A Novel Optical Design Solution for Computer Vision-Based Automated Defect Detection in Textile Fabric Production.

EasyChair Preprint no. 11649

Abstract

The automated defect detection system on industrial manufacturing lines in today's diverse world of consumer goods is a necessary requirement. Quality control is performed at various stages of a large-scale production process, including raw material and pre-production material inspections, in-process quality checks during production, and quality control of packaging, labeling before market releasing. In many industries, such as the garment industry, the quality of input materials significantly influences product quality, material utilization ratios, and ultimately the profitability of manufacturers. Fabric is one of the most critical input materials in the garment industry. However, during fabric production processes such as weaving, dyeing, and packaging, numerous factors can affect the quality of the raw fabric. Various fabric surface defects may occur, including yarn loss, yarn breakage, single yarn or area shrinkage, uneven dyeing, inconsistent color distribution, mold spots, and fabric thread breakage. These defects directly impact the final product and need to be eliminated during the classification process before entering production. Using manual labor to inspect each fabric roll with high accuracy becomes impractical in many cases due to several factors: experience, visual acuity, fabric roll speed, and psychological factors affecting operators' mental health from observing a monotonous surface for an extended period. All of these factors lead to the necessity of an error detection system on surfaces such as fabric. In this research, we introduce an approach to an optical system aimed at observing and detecting deviations in the fiber structure using images captured from a monochrome camera and a lighting system designed based on the actual structure of several types of fabrics used as research objects.

Keyphrases: Automatic Optical Inspection, computer vision, illumination, machine vision, optics

@Booklet{EasyChair:11649,
  author = {Sy Hieu Dau and Nguyen An Khang Le and Minh Thuan Tran and Phuc Dang},
  title = {A Novel Optical Design Solution for Computer Vision-Based Automated Defect Detection in Textile Fabric Production.},
  howpublished = {EasyChair Preprint no. 11649},

  year = {EasyChair, 2024}}