Transmission line route tree target detection technology based on improved YOLOv5s

Aiming at the problem of low target detection accuracy in unmanned aerial vehicle (UAV) inspection, this paper proposes a target detection algorithm based on improved YOLOv5s for transmission line routes and trees. Firstly, a squeeze-excitation (SE) module is introduced into the backbone network of YOLOv5s, which selectively emphasizes feature information using global information, suppresses non critical features, and recalibrates features. Secondly, the traditional pyramid feature fusion structure of YOLOv5s is replaced with a bifrustum feature fusion (BFF) structure to enhance the extraction ability of validity information and reduce the influence of interference information on recognition accuracy. By constructing an inspection dataset for experimental verification, the results showed that the improved YOLOv5s achieved a MAP of 91.26%, and the detection accuracy of lines and trees reached 94.5% and 89.37%, respectively. The improved model improves detection accuracy while balancing speed, demonstrating the effectiveness of the algorithm.