An Insulator Detection Model Using Bidirectional Feature Fusion Structure Based On YOLO X
Weicheng Shi, Xiaoqin Lyu, Lei Han · 2022 IEEE 17th Conference on Industrial Electronics and Applications (ICIEA) · 2022
Transmission line detection is the key to ensure the safe operation of the power system. How to identify the power equipment and further detect the faults is an important topic. Firstly, based on the You Only Look Once X(YOLO X), the Bidirectional Feature Fusion(BFF) structure is proposed to replace the Feature Pyramid Networks(FPN) and Path Aggregation Network(PAN) of the model, making multi-scale feature fusion more effective. Then, a transmission line insulator fault dataset is used to verify the effectiveness and accuracy of the proposed model. The results show that the mean average precision(mAP) of this model reaches 96.56% (compared to YOLO X increased by 0.54%), and the detection speed reaches 17 FPS on NVIDIA GeForce RTX 3080 Laptop.