Research On Identification And Location Of Typical Defects In Aerial Transmission Lines Based On Deep Learning

As energy channel for electric power,transmission line is an important part of the power system.With the development of drones and deep learning technology,it has become possible to use drones to take aerial photographs of transmission lines and use deep learning technology to automatically identify and locate its typical defects.Based on this,building an autonomous drone flight and an intelligent identification management system for aerial images can greatly improve the efficiency of transmission line inspections and reduce labor costs.This thesis applies deep learning technology to the identification and location of typical defects in aerial images of overhead transmission lines.The main research contents include:(1)Research and analyze the importance of defect identification and location of overhead transmission lines,summarize the application status of unmanned aerial vehicle technology in transmission line inspections,the research status of defects identification and location algorithms for overhead transmiss ion lines,and the research of deep learning algorithms in the field of computer vision application status.(2)Explain the basic principles and structure of convolutional neural networks in the field of deep learning,compare and analyze various target re cognition and positioning models,and choose YOLO V3 as the main model for recognition and positioning according to the characteristics of the identification and positioning of defects in overhead transmission lines.(3)Construct a typical defect data set of overhead transmission lines with 8defect categories and 3965 images,and manually label the data set.The data set is expanded through image enhancement,and the effectiveness of image enhancement in improving the generalization ability of the algorithm is verified through comparative experiments.(4)Aiming at the problem of poor recognition of the YOLO V3 model,combined with the characteristics of the defects of the overhead transmission line,the original YOLO V3 algorithm model is improved.The K-means clustering algorithm is used to optimize the preset anchor frame settings,the attention mechanism is introduced and the loss function is corrected based on GIo U.The improved YOLO V3 model has achieved significant improvements in performance paramet ers and recognition effects.(5)Develop autonomous flight modules and defect recognition management modules for UAVs to verify the effectiveness and practicability of the algorithms.