Research on Object Detection of Suspension Clamps Based on Improved YOLOv8

Aiming at the problems of partially obscuring the target features of the overhanging wire clip detection and being susceptible to environmental interference in the high altitude, a transmission line overhanging wire clip detection network based on YOLOv8 is proposed. In order to adapt to the requirements of the deployment of small embedded devices for UAVs, the C2f module in the Neck layer is improved to a lightweight C2f RVBCA module to reduce the model size and increase the operation rate. The proposed network is compared with other mainstream YOLO series target detection algorithms by constructing and expanding the data set for training, and the results show that the proposed network performs better in detecting the overhanging wire clips on transmission lines. Compared with the benchmark model, the number of model parameters and computational effort of our model are reduced by 15.6% and 16.9%, respectively, and the mAP50 is improved by 2 percentage points on the homemade dataset, and our model outperforms the benchmark model in tests in multiple environments.