Real-Time Mechanical Intrusion and Avian Nest Detection in Transmission Corridors Using Enhanced YOLOv8

The reliable operation of power grids faces significant threats from both avian nests and mechanical intrusions within transmission corridors, impacting safety and stability. To address these challenges, this paper presents a real-time detection system for identifying both bird nests and mechanical intrusions like cranes and excavators on power transmission lines, leveraging an enhanced YOLOv8 model. First, the model's backbone is enhanced using the EFC2F module to improve feature extraction capabilities relevant to both types of disturbances. Second, a small-object detection head is added to increase the accuracy of detecting smaller avian nests and intrusion-related objects. Finally, an improved ECBAM attention module is incorporated before the detection head. This module separately strengthens location and category features, refining the model's ability to distinguish between targets and the surrounding environment. These improvements allow the model to maintain a stable inference speed while achieving enhanced accuracy in identifying bird nests and potential mechanical intrusions under diverse and challenging conditions. The model efficiently identifies target objects under normal conditions and performs well in challenging scenarios, including long-distance shooting, varying lighting, different angles, occlusions, and concealment. Experimental results demonstrate that the improved YOLOv8 model achieves significant improvements across multiple key metrics, validating the effectiveness of these enhancements for both avian nest and mechanical intrusion detection.