Detection Method for Power Workers' Protection Rope Compliance Based on Improved YOLOv8

To ensure the safety of high-altitude operations, it is crucial to monitor power workers wearing two safety ropes. This study proposes an improved method based on drone image acquisition, applied in target detection. The method combines the CAA module in PKINet and the RepNCSPELAN module in YOLOv9 and is applied to YOLOv8n. Compared to the original and mainstream models, this method shows significant performance improvement on the safety rope dataset, with mAP@50 increasing from 94.1% to 99.3% and accuracy from 52.7% to 73.7% under the mAP@50–95 evaluation standard. By employing the reparameterization technique of the RepNCSP module and CSP design, the method enhances feature representation capability, reduces model parameters by approximately one-third, from 30 MB to 23 MB, and decreases computational complexity from 8.1GFLOPs to 6.4GFLOPs. This approach improves detection accuracy, and reduces computational requirements, making it more suitable for deployment on computation-constrained drone platforms, thus providing an efficient and practical safety protection solution for drone inspections.