Patterned fabric defect detection system using near infrared imaging

Patterned fabric defect detection based on inspection with visual light source suffers from two main problems; the fabric pattern itself, which complicates the defect detection process and the undesirable effect of surrounding illumination. These problems lead to reducing the detection success rates due to underdetection and misdetection. In this paper, a computer vision system that can detect fabric defects in patterned fabrics is proposed. The proposed system utilizes near-infrared imaging to overcome visual light source imaging drawbacks. It employs the non-extensive standard deviation filtering and minimum error thresholding method to detect defects. In addition to the simplicity of the proposed algorithm, it produces high accuracy rate that reaches 97%. The proposed algorithm can also be extended to defect detection on plain fabrics.