A Non-Invasive System for On-line Surface Defect Detection on Special-shaped Steel towards Real Production Lines

The non-invasive system based on machine vision has been widely used in the field of surface defect detection for flat steel products. However, due to the steel features of multiple surfaces, various specifications and changeable producing speeds, on-line surface defect detection for the special-shaped steel products is still a great challenge. In view of the above problems, a non-invasive system, which is suitable for the online detection of special-shaped steel, is designed and introduced in this paper. Specifically, the proposed system is composed of two subsystems working in coordination: an image acquisition system and an image processing system. The image acquisition system, which adopts a specifically developed image acquisition device as the core, can obtain high-quality full-surface images of steel with different specifications. The image processing system can detect surface defects by embedding the latest YOLOv5 algorithm. The results show a satisfactory performance in both detection speed and precision, which confirm that the proposed system can be reliably employed in real production lines of special-shaped steel.