SPSS-YOLO: An Improved Steel Wire Rope Damage Detection Model Based on YOLOv8

Steel wire ropes, as essential components of industrial safety, require precise damage detection to ensure safe production. Addressing the high missed detection rates and poor environmental adaptability associated with traditional detection methods, we propose a novel high-accuracy and real-time approach, named SPSS-YOLO. This network is based on YOLOv8 and incorporates Swin-Transformer layers to enhance multi-scale feature extraction capabilities. It integrates the SimAM attention mechanism to optimize key feature focus, employs the SPPFCSPC module to improve multi-scale feature fusion efficiency, and utilizes the SIoU loss function to enhance bounding box localization precision. Experiments conducted on a self-built dataset (comprising 3,613 images with three types of defects: wear, broken wires, and protrusions) demonstrate that the improved model outperforms mainstream algorithms such as SSD and Faster R-CNN across metrics including accuracy, recall, and mAP50, while significantly reducing parameter count and computational load. Ultimately, SPSS-YOLO achieves an optimized balance between detection efficiency and accuracy, improving the detection precision of protrusion damage by 6.1% compared to YOLOv8, thus providing a new solution for real-time steel wire rope damage detection in industrial settings.