A Carrier System for UAVs Connecting Machines and Stations for Autonomous Inspection of Power Lines

Based on the problems existing in the current power line inspection, such as the short effective operation distance and the need for the drone to return to the take-off point for landing after inspection, as well as the high rate of missed detection of small target defects in the massive inspection data, this paper proposes a drone aircraft carrier, which integrates multiple functions such as remote precise positioning of drones, multi-drone formation, intelligent collaborative inspection, and real-time intelligent data analysis. It innovatively adopts cutting-edge technologies such as drone tracking and positioning, relay endurance, and intelligent image recognition, significantly enhancing inspection efficiency and reducing line operation and maintenance costs. In terms of system target recognition, this paper improves the YOLOv4-tiny algorithm to obtain the YOLOv4-tiny-3l-spp algorithm based on the deep learning algorithm. During the construction of the sample set, the mixup algorithm is used to enhance the original data set, thereby enhancing the generalization ability of the trained neural network model and improving the accuracy of target recognition by the neural network model. Secondly, the SGDR algorithm is used to train the neural network model, continuously adjusting the learning rate during the training process to improve the performance of the neural network model. Finally, the SPP layer can fuse the features of different layers for different-sized image inputs, generating fixed-size image outputs and enhancing the recognition effect of the target. Based on the improved algorithm, models for missing nuts on screws, bird nests, insulator burst pieces, and offline construction machinery were trained, and the average accuracy of fault recognition for transmission lines was increased by 2.85%, with an average increase of 3.55% for small targets.