基于嵌入式YOLO网络的电力绝缘子自爆缺陷检测

doi:10.19595/j.cnki.1000-6753.tces.242110

Abstract
Figure/Table
References (0)
Related Citation (7)

Download: PDF (5725 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract

Insulators are critical insulating electrical components in power systems, and their proper insulating condition is essential to ensure the safe and stable operation of the power grid. To address the challenges of large model parameters and high computational complexity in existing insulator self-explosion fault detection models, which hinder their efficient deployment on resource-constrained embedded devices, this paper proposed a fault detection method for power insulators based on an embedded YOLO network.
First, a lightweight YOLO insulator defective target detection network based on depth-separable convolution, GhostNet network and SimSPPF spatial pyramid pooling weights was designed, which significantly reduces the number of parameters and computational complexity of the model while maintaining the detection accuracy. Meanwhile, the redundant parts in the network structure were effectively trimmed by the weight pruning method based on the L1 paradigm, which further compresses the model size. To compensate for the degradation in detection performance due to pruning, the features of the high-precision teacher model were migrated to the lightweight student model using knowledge distillation to improve the detection accuracy and performance of the pruned model.
The proposed lightweight YOLOv5s model was applied to a self-constructed insulator self-explosion defect detection dataset, achieving a significant improvement in detection speed and a notable reduction in model complexity, with only a slight decrease in overall average precision.When applied to another independent test set, the model also demonstrated strong generalization ability in detection. The model was designed to balance detection speed and accuracy, ensuring that it meets the memory and computational requirements of embedded devices. Additionally, the optimized model was deployed on an embedded device powered by the Rexchip RK3588, leveraging its computational performance to accelerate the inference process of the optimized model. Test results demonstrated that the deployed model could accurately detect insulator self-explosion defects at a real-time inference speed of 85 frames per second, which is 1.8 times faster than the original model. This significant improvement in efficiency demonstrates the effectiveness of the optimized model in detecting insulator self-explosion defects.
The following conclusions can be drawn from the experiments: (1) The designed lightweight insulator self-explosion defect detection model meets the memory requirements for embedded deployment and the computational performance requirements for real-time self-explosion defect detection, with minimal loss in detection performance; (2) Knowledge distillation technology compensates for the detection performance degradation caused by model lightweighting, achieving a good balance between detection speed and accuracy, while It also demonstrates strong generalization ability and robustness on the independent test set.(3) The model was deployed on the RK3588 embedded device, fully leveraging its computational performance to accelerate the efficient inference of the model. The lightweight model enables efficient detection of insulator self-explosion defects on embedded devices, contributing to improved insulator inspection efficiency in practical applications.
The dataset will be expanded in the future to include additional defect types, thereby enhancing the model's capacity to classify insulators with diverse defect types. Concurrently, the embedded device implemented in this study will be integrated with drones to enable real-time, on-site intelligent insulator inspection applications at the edge.

Key words： Insulator detection lightweight pruning knowledge distillation embedded

Received: 26 November 2024

PACS:	TM216
	TP183

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Li Peng
	Su Yunlong
	Ning Hao
	Meng Qingwei
	Wei Qiang

Cite this article:

Li Peng,Su Yunlong,Ning Hao等. Self-Explosion Defect Detection of Power Insulators Based on Embedded YOLO Network[J]. Transactions of China Electrotechnical Society, 0, (): 2110-.

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.242110 OR https://dgjsxb.ces-transaction.com/EN/Y0/V/I/2110

[1] 何宁辉, 王世杰, 刘军福, 等. 基于深度学习的航拍图像绝缘子缺失检测方法研究[J]. 电力系统保护与控制, 2021, 49(12): 132-140.
He Ninghui, Wang Shijie, Liu Junfu, et al.Research on infrared image missing insulator detection method based on deep learning[J]. Power System Protection and Control, 2021, 49(12): 132-140.
[2] 刘传洋, 吴一全, 刘景景. 无人机航拍图像中绝缘子缺陷检测的深度学习方法研究进展[J/OL]. 电工技术学报, 2024: 1-21. (2024-09-04). http://kns.cnki.net/KCMS/detail/detail.aspx? filename=DGJS20240903001&dbname=CJFD&dbcode=CJFQ.
Liu Chuanyang, Wu Yiquan, Liu Jingjing. Research progress of deep learning method for insulator defect detection in aerial images of unmanned aerial vehicles[J/OL]. China Industrial Economics, 2024: 1-21. (2024-09-04). http://kns.cnki.net/KCMS/detail/detail.aspx? filename=DGJS20240903001&dbname=CJFD&dbcode=CJFQ.
[3] 李斌, 屈璐瑶, 朱新山, 等. 基于多尺度特征融合的绝缘子缺陷检测[J]. 电工技术学报, 2023, 38(1): 60-70.
Li Bin, Qu Luyao, Zhu Xinshan, et al.Insulator defect detection based on multi-scale feature fusion[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 60-70.
[4] 刘思言, 王博, 高昆仑, 等. 基于R-FCN的航拍巡检图像目标检测方法[J]. 电力系统自动化, 2019, 43(13): 162-168.
Liu Siyan, Wang Bo, Gao Kunlun, et al.Object detection method for aerial inspection image based on region-based fully convolutional network[J]. Automation of Electric Power Systems, 2019, 43(13): 162-168.
[5] Wang C Y, Bochkovskiy A, Liao H M.YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, BC, Canada, 2023: 7464-7475.
[6] Girshick R, Donahue J, Darrell T, et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 2014: 580-587.
[7] 徐建军, 黄立达, 闫丽梅, 等. 基于层次多任务深度学习的绝缘子自爆缺陷检测[J]. 电工技术学报, 2021, 36(7): 1407-1415.
Xu Jianjun, Huang Lida, Yan Limei, et al.Insulator self-explosion defect detection based on hierarchical multi-task deep learning[J]. Transactions of China Electrotechnical Society, 2021, 36(7): 1407-1415.
[8] 苟军年, 杜愫愫, 刘力. 基于改进掩膜区域卷积神经网络的输电线路绝缘子自爆检测[J]. 电工技术学报, 2023, 38(1): 47-59.
Gou Junnian, Du Susu, Liu Li.Transmission line insulator self-explosion detection based on improved mask region-convolutional neural network[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 47-59.
[9] 罗潇, 於锋, 彭勇. 基于深度学习的无人机电网巡检缺陷检测研究[J]. 电力系统保护与控制, 2022, 50(10): 132-139.
Luo Xiao, Yu Feng, Peng Yong.UAV power grid inspection defect detection based on deep learning[J]. Power System Protection and Control, 2022, 50(10): 132-139.
[10] 张焕龙, 齐企业, 张杰, 等. 基于改进YOLOv5的输电线路鸟巢检测方法研究[J]. 电力系统保护与控制, 2023, 51(2): 151-159.
Zhang Huanlong, Qi Qiye, Zhang Jie, et al.Bird nest detection method for transmission lines based on improved YOLOv5[J]. Power System Protection and Control, 2023, 51(2): 151-159.
[11] 王道累, 张正刚, 张世恒, 等. 基于密集连接网络的航拍绝缘子旋转目标精准定位方法[J]. 电力系统保护与控制, 2024, 52(1): 35-43.
Wang Daolei, Zhang Zhenggang, Zhang Shiheng, et al.Accurate positioning method of insulator rotating target in aerial photography based on dense connection network[J]. Power System Protection and Control, 2024, 52(1): 35-43.
[12] 贾晓芬, 于业齐, 郭永存, 等. 航拍绝缘子自爆缺陷的轻量化检测方法[J]. 高电压技术, 2023, 49(1): 294-300.
Jia Xiaofen, Yu Yeqi, Guo Yongcun, et al.Lightweight detection method of self-explosion defect of aerial photo insulator[J]. High Voltage Engineering, 2023, 49(1): 294-300.
[13] Zha Mingfeng, Qian Wenbin, Yi Wenlong, et al.A lightweight YOLOv4-based forestry pest detection method using coordinate attention and feature fusion[J]. Entropy, 2021, 23(12): 1587.
[14] 王道累, 张世恒, 袁斌霞, 等. 基于改进YOLOv5的轻量化玻璃绝缘子自爆缺陷检测研究[J]. 高电压技术, 2023, 49(10): 4382-4390.
Wang Daolei, Zhang Shiheng, Yuan Binxia, et al.Research on self-explosion defect detection of lightweight glass insulators based on improved YOLOv5[J]. High Voltage Engineering, 2023, 49(10): 4382-4390.
[15] 陈奎, 刘晓, 贾立娇, 等. 基于轻量化网络与增强多尺度特征融合的绝缘子缺陷检测[J]. 高电压技术, 2024, 50(3): 1289-1300.
Chen Kui, Liu Xiao, Jia Lijiao, et al.Insulator defect detection based on lightweight network and enhanced multi-scale feature fusion[J]. High Voltage Engineering, 2024, 50(3): 1289-1300.
[16] 姜香菊, 王瑞彤, 马彦鸿. 基于轻量级改进RT-DETR边缘部署算法的绝缘子缺陷检测[J/OL]. 电工技术学报, 2024: 1-14. (2024-04-26). http://kns.cnki.net/KCMS/detail/detail.aspx? filename=DGJS20240423004&dbname=CJFD&dbcode=CJFQ.
Jiang Xiangju, Wang Ruitong, Ma Yanhong. Insulator defect detection based on lightweight improved RT-DETR edge deployment algorithm[J/OL]. China Industrial Economics, 2024: 1-14. (2024-04-26). http://kns.cnki.net/KCMS/detail/detail.aspx? filename=DGJS20240423004&dbname=CJFD&dbcode=CJFQ.
[17] 阮顺领,王京,顾清华,等.面向边缘计算的矿区障碍检测模型研究[J/OL].煤炭科学技术,1-12[2024-09-23].
Ruan Shunling, Wang Jing, Gu Qinghua, et al.Research on mining area obstacle detection model for edge computing[J/OL]. Coal Science and Technology, 1-12[2024-09-23].
[18] 李登攀, 任晓明, 颜楠楠. 基于无人机航拍的绝缘子掉串实时检测研究[J]. 上海交通大学学报, 2022, 56(8): 994-1003.
Li Dengpan, Ren Xiaoming, Yan Nannan.Real-time detection of insulator drop string based on UAV aerial photography[J]. Journal of Shanghai Jiao Tong University, 2022, 56(8): 994-1003.
[19] 罗志聪,何陈涛,陈登捷,等.基于轻量化YOLOv8的百香果快速检测模型[J/OL].农业机械学报,1-12[2024-09-23].
Luo Zhicong, He Chentao, Chen Dengjie, et al.Passion fruit rapid detection model based on lightweight YOLOv8[J/OL]. Transactions of the Chinese Society for Agricultural Machinery, 1-12[2024-09-23].
[20] 王年涛, 王淑青, 黄剑锋, 等. 基于改进YOLOv5神经网络的绝缘子缺陷检测方法[J]. 激光杂志, 2022, 43(8): 60-65.
Wang Niantao, Wang Shuqing, Huang Jianfeng, et al.Insulator defect detection method based on improved YOLOv5 neural network[J]. Laser Journal, 2022, 43(8): 60-65.
[21] Kaiser L, Gomez A N, Chollet F. Depthwise separable convolutions for neural machine translation[EB/OL].2017: 1706.03059. https://arxiv.org/abs/1706.03059v2.
[22] Han Kai, Wang Yunhe, Tian Qi, et al.GhostNet: more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2020: 1580-1589.
[23] He Kaiming, Zhang Xiangyu, Ren Shaoqing, et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
[24] Li Hao, Kadav A, Durdanovic I, et al. Pruning filters for efficient ConvNets[EB/OL].2016: 1608.08710. https://arxiv.org/abs/1608.08710v3.
[25] Gou Jianping, Yu Baosheng, Maybank S J, et al.Knowledge distillation: a survey[J]. International Journal of Computer Vision, 2021, 129(6): 1789-1819.
[26] Tao Xian, Zhang Dapeng, Wang Zihao, et al.Detection of power line insulator defects using aerial images analyzed with convolutional neural networks[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2020, 50(4): 1486-1498.
[27] 第八届”泰迪杯” 数据挖掘挑战赛B题:电力巡检智能缺陷检测.[2024-04-19]. https://www.tipdm.org/u/cms/www/202003/0104085491vr.pdf
The 8th Teddy Cup Data Mining Challenge B: Intelligent Defect Detection for Power Inspection. [2024-04-19]. https://www.tipdm.org/u/cms/www/202003/0104085491vr.pdf