YOLOv8 Improved Transmission Line Foreign Body Detection Technology Research

In order to improve the detection performance of the transmission line, the YOLOv8 network is improved. First, use the repvitBlock and EMA attention mechanism to construct the C2F-RE improvement of the deep C2F module in the backbone network. EMA module replace the PANET Concat connection to achieve multi-scale feature fusion; finally, use the Inner-Wiou V3 as the boundary box loss function, and the EMA-Slide Loss as a classification loss function to enhance the generalization of the model. The average accuracy rate of the YOLOv8 model on the improved YOLOV8 model on bird nests, balloons, kites, and garbage detection was made through the data of Nanwang electric dataset data 95.3%, the Map@0.5 value detected by this method is 2.3%higher than the YOLOv8 is directly used.