Surface Defect Detection of Preform Based on Improved YOLOv5

This paper proposes a lightweight detection model based on machine vision, YOLOv5-GC, to improve the efficiency and accuracy of detecting classifying surface defects in preforming materials. During this process, clear images entire are difficult obtain due stickiness, high reflectivity, black resin thermosetting plain woven prepreg. To address challenge, we built vision platform equipped with linescan camera high-intensity linear light source that captures material during process. solve problem defect case extremely small imbalanced samples, adopt transfer learning approach YOLOv5 neural network for recognition introduce coordinate attention Ghost Bottleneck module speed. Experimental results demonstrate proposed achieves rapid high-precision identification materials, outperforming other state-of-the-art methods. work provides promising solution contributing improvement composite quality.