Challenge: Real-time detection of defects ≥ 1 mm in laminated textiles of different widths and shades, and defect labelling. The defects can occur on the mesh, in the mesh or underneath it. The maximum speed of the conveyor belt is 50m/min.
Solution: The system hardware comprises of four powerful Basler line cameras with adequate optics, customized line light sources, a label printer for defect labelling and two powerful computers for image analysis and inter-device communication.
A powerful custom-made algorithm successfully detects defects. The algorithm adapts dynamically to the width of the roll and the smaller variations in the illumination of the laminated textile that can occur despite controlled conditions. The defect detection is particularly difficult due to the non-homogeneity of the inspected surface (white mash on dark or light foam). As the algorithm detects a defect, it sends a signal to the label printer, which in turn applies a label on the defect area. Real-time synchronization of computers prevents application of several labels on the same area, and enables parallel and synchronized processing. The user interface displays the temporary images from all four cameras and the images of the last recognized defects. Saving configuration settings for different types of laminated textile enables a quick and simple transition when changing textile types.