Improved YOLOv8n for Foreign-Object Detection in Power Transmission Lines

Effective and accurate detection of foreign objects in transmission lines plays a crucial role in achieving intelligent power inspection. However, in the real world, detecting objects that are too far or too small can lead to inaccurate object detection tasks. Therefore, this article proposes an improved model based on YOLOv8n to improve detection performance. We introduce attention mechanism into the YOLOv8n network and add a small object detection module to improve detection accuracy. Considering the requirements of detection tasks for detection speed and accuracy, after comparing the three attention mechanisms of CBAM, ECA, and GAM, we chose the backbone network formed by the fusion of YOLOv8n and ECA attention mechanism, and added a small object detection module in the head section. The results show that compared to the unimproved YOLOv8n model, this method can effectively improve detection accuracy and still perform excellently in detection speed and robustness.