Bird’s Nest Detection Algorithm for Transmission Line Towers Based on Deep Learning

Abstract

The transmission line’s ability to operate safely will be substantially compromised by the presence of the nest of birds atop the tower of either a transmission line. During recent decades, A popular area of study now is the efficient location and detection of bird nests on transmission line towers. As deep learning technology has advanced and hardware processing power has increased, using deep learning technology can be gained by utilizing deep learning technology for bird’s nest target recognition. This research suggests a technique for detecting bird nests on transmission system towers through YOLOv5. The method consists of input, backbone, neck, and prediction. By introducing pre-training weights and various data enhancement methods, After multiple network training sessions, it is feasible to find and recognize the bird’s nest on the power system’ tower. The outcomes reveal that the YOLOv5 model’s mean average precision (mAP) of the bird’s nest on the top of power tower can reach 85.6%, and the model is only 14.4 MB in size. This work suggested a YOLOv5-based method for determining bird nests on transmission line towers that can meet the demands of high precision and low volume while also showing promise for future development.