Research On Defect Detection Of Electrical Insulators Based On Deep Learning

As an important component of transmission lines,insulators play the function of power insulation and supporting cables in high-voltage transmission lines,and their health status directly affects the safe operation of the entire transmission line.Therefore,regular inspection and maintenance of insulators are extremely important to ensure the safe and stable operation of high-voltage transmission lines.However,power line insulator defect recognition usually has problems such as complex background,small defect target size and inconspicuous defect features.Traditional recognition methods based on image processing,image analysis and pattern classification have many limitations in solving the above problems.In recent years,automatic detection of power line insulator defects using deep learning algorithms is gradually replacing the traditional manual inspection and identification methods,and is gradually becoming an emerging automatic inspection method for power line insulators.The main research work of this paper is as follows.(1)Dataset production and expansion: Firstly,insulators and insulator defects were labelled by the image annotation tool Label Img,and then the labelled PASCAL VOC format dataset was converted into a trainable label type for the YOLO network.Finally,the Imgaug library in Python is used to implement the affine transformation,colour transformation and noise filling of the images to enhance the dataset in order to improve the generalisation capability of the model,in view of the small number of insulator defect samples and the uneven sample of fault-free insulators and insulator defects.(2)Insulator defect offline identification system design: A better insulator defect detection model is obtained by training a PPYOLO network under the Paddle Paddle framework,and then the model is locally serviced for offline detection of insulator defects collected by UAVs.(3)Insulator defect online recognition system design: an improved lightweight YOLOv4-tiny based insulator defect detection algorithm is proposed.Firstly,the CBL(Conv-BN-Leaky Relu)module of the backbone feature extraction network is replaced with the Mobilevit module to enhance the feature extraction capability of the backbone network.Secondly,the attention mechanism Coordinate Attention(CA)was introduced to improve the network’s ability to focus more on information about the location of defects.Finally,the EIOU(Efficient Intersection Over Union)loss function is used to replace the original CIOU loss function to improve the convergence speed of the network.To verify the effectiveness of the proposed algorithm,the improved algorithm is compared with the mainstream Faster-RCNN algorithm,SSD algorithm,YOLOv3 algorithm and YOLOv4-tiny algorithm.The experimental results show that the proposed algorithm outperforms the above algorithms in terms of detection accuracy.Compared with the conventional YOLOv4-tiny algorithm,the proposed improved algorithm improves 1.64% in mean accuracy(m AP);0.32% in mean accuracy(AP)for missing insulator defects;and 4.96% in mean accuracy(AP)for broken insulator defects.The improved network model was finally deployed on the Nvidia Jetson AGX to implement a real-time online insulator defect detection function.