Design And Application Of Transmission Line Insulator Defect Detection System Based On Image Processing Technology

With the continuous expansion of power grid construction,UAV aerial patrolling has gradually replaced the traditional manual patrolling,and plays an increasingly important role in operation and maintenance of transmission line.To solve the problems of processing the large amount of picture data and to promote the efficiency of defect inspection in the transmission line operation and maintenance team,this paper improved the target detection algorithm based on deep learning,and successfully designed and developed the insulator image automatic naming software,defect detection system,fault management system and the web page.The paper mainly studies the following aspects:(1)Based on the different types of defects of insulator in transmission line and the daily operational mode,this paper proposes a new method to acquire defective images and provides the format requirements.To better process the large amount of image data in operation and maintenance work,this paper uses Qt Creator to write a batch naming system for UAV aerial images,which realizes the batch naming of transmission line images and the positioning of longitude and latitude of images.This system can improve the accuracy rate and working efficiency,and also lay a foundation for subsequent insulator image recognition.(2)After comparing the single-stage algorithms with the two-stage algorithms,this paper selects the YOLOv3 target detection algorithm to realize the target detection of insulator images,and uses the Res Net classification model to complete the classification of defects.This paper introduces the Darknet-53 network structure,activation function and loss function of YOLOv3 target detection algorithm,and improves the YOLOv3 algorithm.This paper puts forward a target detection algorithm based on image block to optimize the recognition results,which enhanced the feature recognition effect of long-distance aerial insulator pictures and improved the accuracy.(3)This paper uses translation,rotation,mirror,reverse,Gaussian blur,Gaussian noise to triple the original data of insulators’ defective image,and uses Label Img to annotate insulator data sets.This paper uses Tensor Flow framework for deep learning,comparing and analyzing reliability,accuracy,recall ratio,m AP values and FPS value results of various algorithms.This paper uses Python programming language,combined with Py Charm,Py QT5 and Open CV to complete the development of insulator defect detection system,and tests its application effects in practical level.(4)By combining VB6.0,ASP,Adobe Dreamweaver CS5 and database technology,this paper designs and develops a local fault management system and web page for transmission lines,which is convenient for operation and maintenance personnel to conduct statistics and query on the previous defect data in the later stage.The three kinds of software mentioned in this paper have been applied in the transmission team of a bureau of Guangdong Power Grid for practical use.They help the transmission line operation and maintenance personnel to achieve image processing and fault statistics.This software follows the future trend of digital power grid and can improve the daily production and operation efficiency of the transmission department.