Research On Identification Algorithm Of Burst Defect Of Transmission Line Insulator String Based On Aerial Image

Overhead transmission line network is an important intermediate link between power generation and distribution in the power system,how to conduct inspections safely and effectively has attracted wide attention from many parties.The traditional manual line inspection method has been difficult to adapt to the construction speed and specification requirements of transmission lines,which brings huge challenges to the safety guarantee of power transmission.With the continuous reform and improvement of automated and semi-automated power line inspection technology,the proportion of large-scale mechanized transmission line inspections will become higher and higher in the future.Among them,the method of relying on drones for auxiliary detection has received more and more attention and research.Insulators,as an important component connecting power lines and transmission towers,are of great significance to the entire transmission network to ensure its safety and stability.Therefore,the use of image processing and other technologies to identify and detect defects in transmission line insulators is a research work with potential to tap.Traditional insulator detection technology relies heavily on the completeness of artificially designed features,and the detection speed and accuracy are poor in robustness,which is difficult to satisfy.Aiming at the above problems,this paper uses deep neural network and image processing technology to carry out the following research on the "burst" phenomenon of transmission line insulators:(1)For the aerial images obtained by the drone,the available information is screened,using data enhancement technology to assist in the production of data sets for the detection and positioning of insulators on overhead transmission lines.At the same time,analyze and compare the current deep learning technologies with outstanding detection results to determine the main detection process.(2)Aiming at the problem that the number of "burst" insulators in the data set is small,the environment is more complex,and the background contains a lot of redundant information,a new idea of insulator detection with a generative countermeasure network as the main model is proposed.Optimize the structure of the insulator and discriminator to improve the resolution and texture details of the output insulator image,and increase the amount of available data while completing the insulator detection task.(3)Aiming at the problem that the current insulator "burst" defect location algorithm requires more manual feature selection and low detection accuracy,use the extracted connected domain information to characterize the physical shape of the insulator disc,calculate the number of discs in combination with the square-line fitting technology,so as to classify and locate the insulator strings that have "burst" defects.The method proposed in this paper performs well on the task of detecting “burst”of insulators.The detection accuracy of the insulator detection task and the "burst" defect location task reached 92.4% and 95.6% respectively.Experimental results show that the proposed algorithm can effectively solve the problem of "burst" defect detection of insulators under complex backgrounds such as mountain forests.It provides a solid foundation for the subsequent application of insulator defect detection technology.