Research On Insulator Detection And Defect Recognition Based On Deep Learning

Transmission line is an important part of the power system.inspection of transmission lines is a necessary work.Insulators are an important component of transmission lines.Since insulators are mostly outdoors and are exposed to complex natural environment for a long time,faults are prone to affect the safety and stability of transmission lines.Regular inspection of the working state of insulators is of great significance to ensure the safe operation of power system.crucial.With the application of UAV aerial photography inspection method,the background of the images collected by aerial photography is complex and changeable,and the amount of data is large,and the traditional manual screening method is difficult to meet the requirements of accurate and efficient detection.In recent years,deep learning technology,especially convolutional neural network research,has achieved remarkable results.The application of deep learning-based target detection technology to insulator detection is of great significance to improving the intelligence level of power inspection.In this paper,insulators are taken as the research object,and the problems of insulator detection and defect recognition in the current aerial images are studied.The main work is as follows:Aiming at the problem of insufficient insulator image data,which easily leads to overfitting during network training,data enhancement methods such as rotation transformation,flip transformation,brightness adjustment,and noise addition are used to expand the number of insulator images and create a data set.Aiming at the complex background of aerial images and the variable scale of insulators,an improved Faster R-CNN algorithm is proposed.Firstly,use the Res Net-50 network as the feature extraction network to improve the feature expression ability of the network,and introduce the feature pyramid structure to achieve deep and shallow feature fusion,and improve the adaptability of the network to the change of insulator scale;Secondly,the ROI Align method is used to improve the pooling of the region of interest to eliminate the positioning error caused by the loss of insulator feature information in the pooling process and improve the detection accuracy.Finally,the K-means clustering method is used to obtain a priori box suitable for the proportion of insulators.The experimental results show that the average precision of the improved algorithm has been improved,and the identification and positioning of insulators can be accurately completed.Aiming at the problem that the defect parts of insulators in aerial images are too small and the defect feature information is not obvious,a cascaded insulator defect detection method is proposed.After locating the insulator,the image of the insulator region is extracted by cropping,and YOLOv3 is used as the secondary network to complete the detection of defect parts.At the same time,Efficient Net,which combines the attention mechanism CBAM module,is designed as the backbone network of YOLOv3 to improve its detection of defect parts.feature extraction capability.The experimental results show that the cascading insulator defect detection method has certain advantages over other target detection methods in terms of detection accuracy,and can effectively identify the defect parts of insulators.