基于人工图像数据扩充的输电线路绝缘子识别

doi:10.16182/j.issn1004731x.joss.21-0646

摘要/Abstract

摘要：

深度学习方法在计算机视觉领域发展迅速，但依赖于海量训练数据。输电线路绝缘子自动识别任务中，航拍图像数量不足、多样性差等问题影响识别的准确性。提出人工绝缘子图像数据扩充方法，通过3D建模创建人工绝缘子图像，并构建导向反向补偿网络，对创建的人工图像进行补偿优化，用补偿后的人工图像扩充航拍绝缘子图像数据集。在多个典型卷积神经网络上进行绝缘子识别对比实验，结果显示：所提方法使绝缘子识别准确率平均提升2.1%，且网络相对轻量级，验证了所提方法的有效性和优势。

关键词: 人工图像, 数据扩充, 绝缘子, 导向反向传播, 卷积神经网络

Abstract:

Deep learning method has developed rapidly in the field of computer vision, but relies on a large quantities of training data. In the task of transmission line insulator automatic detection, problems such as insufficient number of aerial insulator images and poor diversity affect the accuracy of insulator recognition. An artificial insulator images data expansion method is proposed. Artificial insulator images are created by modeling software, and a compensation network is constructed. The artificial images are compensated and optimized by compensation network, and the aerial insulator image data set is expanded by the compensated artificial insulator images. The insulator recognition experiments are carried out on several typical convolutional neural networks. The results show that the proposed method improves the accuracy of insulator identification by an average of 2.1%, and the network is relatively lightweight, which verifies the effectiveness and advantages of the proposed method.

Key words: artificial images, data expansion, insulator, guided backpropagation, convolutional neural network

中图分类号:

王亚茹, 杨凯, 翟永杰, 郭聪彬, 赵文清, 苏杰. 基于人工图像数据扩充的输电线路绝缘子识别[J]. 系统仿真学报, 2022, 34(11): 2337-2347.

Yaru Wang, Kai Yang, Yongjie Zhai, Congbin Guo, Wenqing Zhao, Jie Su. Transmission Line Insulator Recognition Based on Artificial Images Data Expansion[J]. Journal of System Simulation, 2022, 34(11): 2337-2347.

图/表 10

图1

图2

图3

图4

图5

图6

图7

图8

表1

表2

参考文献 29

[1]	El-Mehalawy N, Awaad M, Eliyan T, et al. Electrical Properties of ZnO/Alumina Nano Composites for High Voltage Transmission Line Insulator[J]. Journal of Materials Science-Materials in Electronics (S0957-4522), 2018, 29(16): 13526-13533.
[2]	Asadpourahmadchali M, Niasati M, Alinejad-Beromi Y. Improving Tower Grounding vs. Insulation Level to Obtain the Desired Back-flashover Rate for HV Transmission Lines[J]. International Journal of Electrical Power and Energy Systems (S0142-0615), 2020, 123(3): 106171.
[3]	胡毅, 刘凯, 吴田, 等. 输电线路运行安全影响因素分析及防治措施[J]. 高电压技术, 2014, 40(11): 3491-3499.
	Hu Yi, Liu Kai, Wu Tian, et al. Analysis of Influential Factors on Operation Safety of Transmission Line and Countermeasures[J]. High Voltage Engineering, 2014, 40(11): 3491-3499
[4]	Nguyen V N, Jenssen R, Roverso D. Automatic Autonomous Vision-based Power Line Inspection: A Review of Current Status and the Potential Role of Deep Learning[J]. International Journal of Electrical Power & Energy Systems (S0142-0615), 2018, 99: 107-120.
[5]	邵瑰玮, 刘壮, 付晶, 等. 架空输电线路无人机巡检技术研究进展[J]. 高电压技术, 2020, 46(1): 14-22.
	Shao Guiwei, Liu Zhuang, Fu Jing, et al. Research Progress in Unmanned Aerial Vehicle Inspection Technology on Overhead Transmission Lines[J]. High Voltage Engineering, 2020, 46(1): 14-22.
[6]	Siddiqui Z A, Park U, Lee S W, et al. Robust Powerline Equipment Inspection System Based on a Convolutional Neural Network[J]. Sensors (S1424-8220), 2018, 18(11): 3837.
[7]	张晶晶, 韩军, 赵亚博, 等. 形状感知的绝缘子识别与缺陷诊断[J]. 中国图象图形学报, 2020, 19(8): 1194-1201.
	Zhang Jingjing, Han Jun, Zhao Yabo, et al. Insulator Recognition and Defects Detection Based on Shape Perceptual[J]. Journal of Image and Graphics, 2020, 19(8): 1194-1201.
[8]	Ge B, Hou C, Liu Y, et al. Anomaly Detection of Power Line Insulator from Aerial Imagery with Attribute Self-supervised Learning[J]. International Journal of Remote Sensing (S0143-1161), 2021(2): 1-21.
[9]	Govindaraju P, Muniraj C. Monitoring and Optimizing the State of Pollution of High Voltage Insulators Using Wireless Sensor Network Based Convolutional Neural Network[J]. Microprocessors and Microsystems (S0141-9331), 2020, 79: 103299.
[10]	Tao X, Zhang D, Wang Z, 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 (S2168-2216), 2018, 50(4): 1486-1498.
[11]	Ling Z, Qiu R C, Jin Z, et al. An Accurate and Real-time Self-blast Glass Insulator Location Method Based on Faster R-CNN and U-net with Aerial Images[J]. CSEE Journal of Power and Energy Systems (S2096-0042), 2019, 5(4): 474-482.
[12]	Zajac M, Zołna K, Rostamzadeh N, et al. Adversarial Framing for Image and Video Classification[C]//AAAI Conference on Artificial Intelligence. Honolulu, HI: AAAI, 2019, 33(1): 10077-10078.
[13]	Xi P, Tang Z, Fei Y, et al. Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT: IEEE Press, 2018: 2226-2234.
[14]	Goodfellow I J, Pouget-Abadie J, Mirza M, et al. Generative Adversarial Networks[J]. Advances in Neural Information Processing Systems (S1049-5258), 2014, 3: 2672-2680.
[15]	Jairo V, Chen Y Q, Wang J. FaultFace: Deep Convolutional Generative Adversarial Network (DCGAN) Based Ball-bearing Failure Detection Method[J]. Information Sciences (S1674-733X), 2020, 542: 195-211.
[16]	Wang S, Yang Y, Wu Z, et al. Data Augmentation Using Deep Generative Models for Embedding Based Speaker Recognition[J]. IEEE-ACM Transactions on Audio, Speech, and Language Processing (S2329-9290), 2020, 28: 2598-2609.
[17]	Zhu J Y, Park T, Isola P, et al. Unpaired Image-toimage Translation Using Cycle-consistent Adversarial Networks[C]//IEEE International Conference on Computer Vision. Piscataway: IEEE Press, 2017: 2242-2251.
[18]	刘哲良, 朱玮, 袁梓洋. 结合全卷积网络与CycleGAN的图像实例风格迁移[J]. 中国图象图形学报, 2019, 24(8): 1283-1291.
	Liu Zheliang, Zhu Wei, Yuan Ziyang. Image Instance Style Transfer Combined with Fully Convolutional Network and CycleGAN[J]. Journal of Image and Graphics, 2019, 24(8): 1283-1291.
[19]	Ma Z, Li J, Wang N, et al. Semantic-related Image Style Transfer with Dual-consistency Loss[J]. Neurocomputing (S0925-2312), 2020, 406: 135-149.
[20]	Xu Z, Wilber M, Fang C, et al. Adversarial Training for Fast Arbitrary Style Transfer[J]. Computers & Graphics- UK (S0097-8493), 2020, 87: 1-11.
[21]	马岽奡, 唐娉, 赵理君, 等. 深度学习图像数据增广方法研究综述[J]. 中国图象图形学报, 2021, 26(3): 487-502.
	Ma Dongao, Tang Ping, Zhao Lijun, et al. Review of Data Augmentation for Image in Deep Learning[J]. Journal of Image and Graphics, 2021, 26(3): 487-502.
[22]	Zhang K, Cao Z, Wu J. Circular Shift: An Effective Data Augmentation Method for Convolutional Neural Network on Image Classification[C]//IEEE International Conference on Image Processing, Electr Network: IEEE Press, 2020: 1676-1680.
[23]	Gupta A, Vedaldi A, Zisserman A. Synthetic Data for Text Localisation in Natural Images[C]//IEEE Conference on Computer Vision and Pattern Recognition, Piscataway: IEEE Press, 2016: 2315-2324.
[24]	柴伟佳, 王连明. 卷积神经网络的多字体汉字识别[J]. 中国图象图形学报, 2018, 23(3): 410-417.
	Chai Weijia, Wang Lianming. Recognition of Chinese Characters Using Deep Convolutional Neural Network[J]. Journal of Image and Graphics, 2018, 23(3): 410-417.
[25]	Ekin D, Cubuk B Z, Dandelion M, et al. AutoAugment: Learning Augmentation Policies from Data[EB/OL]. (2018-05-24) [2019-04-11]. .
[26]	翟永杰, 杨旭, 王金娜, 等. 平行视觉框架下深度卷积神经网络可视化分析[J]. 计算机工程与应用, 2020, 56(19): 139-145.
	Zhai Yongjie, Yang Xu, Wang Jinna, et al. Visual Analysis of Deep Convolutional Neural Networks in Parallel Vision[J]. Computer Engineering and Applications, 2020, 56(19): 139-145.
[27]	王坤峰, 鲁越, 王雨桐, 等. 平行图像：图像生成的一个新型理论框架[J]. 模式识别与人工智能, 2017, 30(7): 577-587.
	Wang Kunfeng, Lu Yue, Wang Yutong, et al. Parallel Imaging: A New Theoretical Framework for Image Generation[J]. Pattern Recognition and Artificial Intelligence, 2017, 30(7): 577-587.
[28]	Springenberg J T, Dosovitskiy A, Brox T, et al. Striving for Simplicity: the All Convolutional Net[EB/OL]. (2014-12-21) [2015-4-13]. .
[29]	Krizhevsky A, Sutskever I, Hinton G E. ImageNet Classification with Deep Convolutional Neural Networks[J]. Advances in Neural Information Processing Systems (S1049-5258), 2012, 25: 1097-1105.

评价指标	方法	卷积神经网络
评价指标	方法	AlexNet	VGG11	VGG13	ResNet18	Inception_v3
准确率/%	不进行数据扩充	93.10	89.34	89.32	80.98	62.60
	未补偿的人工图像进行数据扩充	95.25	90.56	90.18	79.41	63.35
	基于CycleGAN的图像风格迁移数据扩充	95.03	90.96	90.08	77.70	63.31
	本文方法	96.40	91.02	91.26	82.44	64.29
AUC	不进行数据扩充	0.98	0.94	0.95	0.86	0.61
	未补偿的人工图像进行数据扩充	0.98	0.97	0.95	0.86	0.63
	基于CycleGAN的图像风格迁移数据扩充	0.99	0.98	0.96	0.83	0.62
	本文方法	0.99	0.99	0.98	0.90	0.62
迭代次数	不进行数据扩充	80	100	100	776	531
	未补偿的人工图像进行数据扩充	50	80	110	456	382
	基于CycleGAN的图像风格迁移数据扩充	75	130	130	918	687
	本文方法	70	90	115	526	443

方法	卷积神经网络
方法	AlexNet	VGG11	VGG13	ResNet18	Inception_v3
基于CycleGAN的图像风格迁移数据扩充	27.06 G
本文方法	1.34 G	14.99 G	22.40 G	3.63 G	11.43 G