Bird activities seriously affect the safe operation of power transmission lines. This paper proposes a lightweight object detection model based on improved YOLOv4 to recognize typical bird species threatening transmission line security. A dataset composed of 3000 images about 10 bird species that easily cause transmission line outages was constructed. An improved YOLOv4 model was established by replacing the feature extraction network with GhostNet. The focus layer was added in GhostNet, and the standard convolution in the path aggregation network (PANet) was replaced with the depthwise separable convolution, thus to greatly reduce the amount of model parameters. After model training, the improved YOLOv4 was applied to detect bird targets in 300 test sample images. The results indicate that the mean average precision (mAP) reaches 97.55%, and frames per second (FPS) is 43, which is much faster than YOLOv4. The detection accuracy and efficiency of the proposed model were compared to the existing models such as SSD, YOLOv4, etc. This study can be applied for bird recognition and therefore contribute to achieve differentiated prevention of bird-related outages.

Transmission line,bird-related outages,object detection,bird recognition,YOLOv4