Development of an Online Automated Fabric Inspection System

In this work, the development of a system for the automated inspection of mesh fabrics in real time through computer vision techniques is described. Here, a processing pipeline that includes a distortion removal procedure, a texture-based detector, a shape and an artificial neural network classifier is proposed. Different from other works found in the literature, the developed system was evaluated under real inspection conditions. The obtained results show that the proposed system not only achieves compatible defect detection and classification rates than the current state-of-the-art methods but also met functional and non-functional requirements observed in industry.