融合深度特征的输电线路金具缺陷因果分类方法

赵振兵; 张薇; 戚银城; 翟永杰; 赵文清

doi:10.13700/j.bh.1001-5965.2020.0456

融合深度特征的输电线路金具缺陷因果分类方法

doi: 10.13700/j.bh.1001-5965.2020.0456

1.
华北电力大学电气与电子工程学院, 保定 071003
2.
华北电力大学控制与计算机工程学院, 保定 071003

基金项目:

国家自然科学基金 61871182

国家自然科学基金 61773160

北京市自然科学基金 4192055

河北省自然科学基金 F2020502009

中央高校基本科研业务费专项资金 2018MS095

中央高校基本科研业务费专项资金 2020YJ006

模式识别国家重点实验室开放课题基金 201900051

详细信息

作者简介:
赵振兵   男，博士，副教授，硕士生导师。主要研究方向：电力视觉检测

张薇   女，硕士研究生。主要研究方向：目标检测与图像处理

戚银城   男，博士，教授，硕士生导师。主要研究方向：电力系统通信与信息处理

通讯作者:
赵振兵, E-mail: zhaozhenbing@ncepu.edu.cn

中图分类号: TM726;TH165⁺.3
计量
- 文章访问数: 674
- HTML全文浏览量: 174
- PDF下载量: 64
- 被引次数: 0
出版历程
- 收稿日期: 2020-08-24
- 录用日期: 2020-08-28
- 网络出版日期: 2021-03-20

Causal classification method of transmission lines fitting defect combined with deep features

1.
School of Electrical and Electronic Engineering, North China Electric Power University, Baoding 071003, China
2.
School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, China

Funds:

National Natural Science Foundation of China 61871182

National Natural Science Foundation of China 61773160

Beijing Natural Science Foundation 4192055

Natural Science Foundation of Hebei Province F2020502009

the Fundamental Research Funds for the Central Universities 2018MS095

the Fundamental Research Funds for the Central Universities 2020YJ006

Open Project Program of the National Laboratory of Pattern Recognition 201900051

More Information

Corresponding author: ZHAO Zhenbing, E-mail: zhaozhenbing@ncepu.edu.cn

摘要

摘要:
针对输电线路金具缺陷样本不足和缺陷目标形态多样化，仅仅利用深度学习模型导致金具缺陷分类准确率较低的问题，提出了一种结合深度网络和逻辑回归模型的因果分类方法。首先，通过样本扩充方法获得数量丰富化和角度多样化的数据集；然后，基于微调后的VGG16模型提取深度特征并进行特征处理，以构建符合因果关系学习的输入特征集；最后，通过全局混杂平衡进行金具缺陷特征与标签之间的因果关系学习，构建符合金具特点的因果逻辑回归模型，完成金具缺陷分类。为了证明所提方法的有效性，利用无人机实际采集的4类金具缺陷图片分别进行了实验，所使用的训练样本和测试样本数量较原始数据集提升了5倍左右。实验结果表明：所提方法可以实现对输电线路金具缺陷的精准分类，其中，防震锤相交和变形分类准确率分别达到了0.929 9和0.911 8，屏蔽环锈蚀和均压环损坏分类准确率分别达到了0.956 7和0.966 9。
- 输电线路金具缺陷 /
- 因果关系学习 /
- 深度特征 /
- 逻辑回归模型 /
- VGG
Abstract:
Aimed at the insufficient transmission lines fitting defect samples and diverse defect target shapes, a causal classification method combining deep network and logistic regression model is proposed to solve low defect classification accuracy when only using deep learning models. Firstly, rich and diverse datasets are obtained through the sample expansion method. Secondly, deep features are extracted based on the fine-tuned VGG16 model, and processed to construct an input feature set that conforms to causality learning. Finally, the causal relationship between fitting defect feature and label is learned through the global balance, and a causal logistic regression model is constructed to complete the classification of the fitting defects. Four types of fitting defect images collected by UAV are used respectively in the experiments to prove the effectiveness of the proposed method. The number of training and testing samples used is about 5 times higher than the original dataset. The experimental results show the proposed method can realize the accurate classification of the fitting defects, the classification accuracy of the shockproof hammer intersection and deformation reach 0.929 9 and 0.911 8 respectively, and the classification accuracy of the shielding ring corrosion and the grading ring damage reach 0.956 7 and 0.966 9 respectively.
- transmission line fitting defect /
- causality learning /
- deep features /
- logistic regression model /
- VGG

HTML全文

图 1 本文方法总体流程

Figure 1. Overall flowchart of proposed method

下载: 全尺寸图片幻灯片

图 2 三类主要金具及其缺陷样本示例

Figure 2. Examples of three types of main fitting and their defect samples

下载: 全尺寸图片幻灯片

图 3 VGG16模型结构

Figure 3. Structure of VGG16 model

下载: 全尺寸图片幻灯片

图 4 金具缺陷数据集深度特征提取处理图

Figure 4. Extraction and processing of deep features of fitting defect dataset

下载: 全尺寸图片幻灯片

图 5 扩充后金具正常及其缺陷样本数量分布

Figure 5. Distribution of the number of fitting's normal and defect samples after expansion

下载: 全尺寸图片幻灯片

表 1 防震锤相交分类实验结果

Table 1. Classification experimental results of shockproof hammer intersection

方案	准确率	召回率	F1分数
方案1	0.803 7	0.763 2	0.805 6
方案2	0.887 8	0.807 1	0.884 6
方案3	0.929 9	0.894 7	0.891 4

下载: 导出CSV

表 2 防震锤变形分类实验结果

Table 2. Classification experimental results of shockproof hammer deformation

方案	准确率	召回率	F1分数
方案1	0.808 8	0.388 9	0.518 5
方案2	0.860 3	0.472 2	0.641 5
方案3	0.911 8	0.666 7	0.800 1

下载: 导出CSV

表 3 屏蔽环锈蚀分类实验结果

Table 3. Classification experimental results of shielding ring corrosion

方案	准确率	召回率	F1分数
方案1	0.839 5	0.861 1	0.826 7
方案2	0.882 7	0.847 2	0.865 2
方案3	0.956 7	0.902 7	0.948 9

下载: 导出CSV

表 4 均压环损坏分类实验结果

Table 4. Classification experimental results of grading ring damage

方案	准确率	召回率	F1分数
方案1	0.842 5	0.834 2	0.843 3
方案2	0.914 4	0.887 6	0.914 5
方案3	0.966 9	0.956 5	0.972 3

下载: 导出CSV

表 5 样本扩充前后本文方法实验结果

Table 5. Experimental results by proposed method before and after sample expansion

金具缺陷类别	扩充前准确率	扩充前召回率	扩充后准确率	扩充后召回率
防震锤相交	0.906 7	0.555 6	0.929 9	0.894 7
防震锤变形	0.901 5	0.333 4	0.911 8	0.666 7
屏蔽环锈蚀	0.927 5	0.833 4	0.956 7	0.902 7
均压环损坏	0.933 0	0.647 1	0.966 9	0.956 5

下载: 导出CSV

表 6 不同分类器的实验结果

Table 6. Experimental results of different classifiers

方案	准确率	召回率	F1分数
方案1	0.552 5	0.515 0	0.571 9
方案2	0.702 5	0.725 0	0.700 5
方案3	0.762 5	0.711 1	0.780 4

下载: 导出CSV

参考文献(25)

[1]	赵强. 输电线路金具理论与应用[M]. 北京: 中国电力出版社, 2013: 2-12. ZHAO Q. Theory and application of transmission line fittings[M]. Beijing: China Electric Power Press, 2013: 2-12(in Chinese).
[2]	DENG C, WANG S, HUANG Z, et al. Unmanned aerial vehicles for power line inspection: A cooperative way in platforms and communications[J]. Journal of Communications, 2014, 9(9): 687-692. doi: 10.12720/jcm.9.9.687-692
[3]	TONG W, YUAN J, LI B. Application of image processing in patrol inspection of overhead transmission line by helicopter[J]. Power System Technology, 2010, 34(12): 204-208. http://www.researchgate.net/publication/283525036_Application_of_image_processing_in_patrol_inspection_of_overhead_transmission_line_by_helicopter
[4]	金哲, 尹洪, 吴启进. 典型500 kV输电线路地线金具腐蚀及磨损事件机理分析[J]. 电工技术, 2017(8): 66-69. doi: 10.3969/j.issn.1002-1388.2017.08.026 JIN Z, YIN H, WU Q J. Analysis on the corrosion and wear event mechanism of the ground wire of typical 500 kV transmission line[J]. Electric Engineering, 2017(8): 66-69(in Chinese). doi: 10.3969/j.issn.1002-1388.2017.08.026
[5]	陆旭, 罗汉武, 李文震, 等. 电力金具图像故障状态评估[J]. 红外技术, 2020, 42(7): 632-636. https://www.cnki.com.cn/Article/CJFDTOTAL-HWJS202007006.htm LU X, LUO H W, LI W Z, et al. Evaluation of image failure state of power fittings[J]. Infrared Technology, 2020, 42(7): 632-636(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-HWJS202007006.htm
[6]	吴坤祥, 朱迪锋, 许杨勇, 等. ±800 kV特高压输电线路耐张线夹未压区鼓胀缺陷分析[J]. 浙江电力, 2017, 36(7): 11-13. https://www.cnki.com.cn/Article/CJFDTOTAL-ZJDL201707003.htm WU K X, ZHU D F, XU Y Y, et al. Analysis of bulging defects in uncompressed zone of tension clamp in ±800 kV UHV transmission line[J]. Zhejiang Electric Power, 2017, 36(7): 11-13(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZJDL201707003.htm
[7]	金立军, 胡娟, 闫书佳. 基于图像的高压输电线间隔棒故障诊断方法[J]. 高电压技术, 2013, 39(5): 1040-1045. doi: 10.3969/j.issn.1003-6520.2013.05.003 JIN L J, HU J, YAN S J. Image-based fault diagnosis method for spacers of high-voltage transmission lines[J]. High Voltage Engineering, 2013, 39(5): 1040-1045(in Chinese). doi: 10.3969/j.issn.1003-6520.2013.05.003
[8]	胡彩石, 吴功平, 曹珩, 等. 高压输电线路巡线机器人障碍物视觉检测识别研究[J]. 传感技术学报, 2008, 21(12): 2092-2096. doi: 10.3969/j.issn.1004-1699.2008.12.028 HU C S, WU G P, CAO H, et al. Research on visual inspection and recognition of obstacles for high-voltage transmission line patrol robot[J]. Chinese Journal of Sensors and Actuators, 2008, 21(12): 2092-2096(in Chinese). doi: 10.3969/j.issn.1004-1699.2008.12.028
[9]	宋伟, 左丹, 邓邦飞, 等. 高压输电线防震锤锈蚀缺陷检测[J]. 仪器仪表学报, 2016, 37(S1): 113-117. https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB2016S1019.htm SONG W, ZUO D, DENG B F, et al. Corrosion defect detection of earthquake hammer for high voltage transmission line[J]. Chinese Journal of Scientific Instrument, 2016, 37(S1): 113-117(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB2016S1019.htm
[10]	付晶, 邵瑰玮, 吴亮, 等. 利用层次模型进行训练学习的线路设备缺陷检测方法[J]. 高电压技术, 2017, 43(1): 266-275. https://www.cnki.com.cn/Article/CJFDTOTAL-GDYJ201701035.htm FU J, SHAO G W, WU L, et al. Line equipment defect detection method using hierarchical model for training and learning[J]. High Voltage Engineering, 2017, 43(1): 266-275(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-GDYJ201701035.htm
[11]	李辉, 钟平, 戴玉静, 等. 基于深度学习的输电线路锈蚀检测方法的研究[J]. 电子测量技术, 2018, 41(22): 54-59. https://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201822014.htm LI H, ZHONG P, DAI Y J, et al. Research on transmission line corrosion detection method based on deep learning[J]. Electronic Measurement Technology, 2018, 41(22): 54-59(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201822014.htm
[12]	汤踊, 韩军, 魏文力, 等. 深度学习在输电线路中部件识别与缺陷检测的研究[J]. 电子测量技术, 2018, 41(6): 60-65. https://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201806011.htm TANG Y, HAN J, WEI W L, et al. Research on deep learning in component identification and defect detection in transmission lines[J]. Electronic Measurement Technology, 2018, 41(6): 60-65(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-DZCL201806011.htm
[13]	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2014: 1-14.
[14]	MARCUS G. Deep leaening: A critical appraisal[EB/OL]. (2018-01-02)[2020-08-01]. http://export.arxiv.org/abs/1801.00631.
[15]	GHOSH J, JUDEA P. Causality: Models, reasoning and inference[J]. International Statistical Review, 2011, 79(2): 289-290. doi: 10.1111/j.1751-5823.2011.00149_16.x
[16]	LOPEZPAZ D, NISHIHARA R, CHINTALA S, et al. Discovering causal signals in images[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2017: 58-66.
[17]	ZHONG J, SUN Y, YU Y, et al. Attribute-guided network for cross-modal zero-shot hashing[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 31(1): 321-330. http://ieeexplore.ieee.org/document/8686336/
[18]	郭琳, 秦世引. 遥感图像飞机目标高效搜检深度学习优化算法[J]. 北京航空航天大学学报, 2019, 45(1): 159-173. doi: 10.13700/j.bh.1001-5965.2018.0239 GUO L, QIN S Y. Deep learning and optimization algorithm for high efficient searching and detection of aircraft targets in remote sensing images[J]. Journal of Beijing University of Aeronautics and Astronautics, 2019, 45(1): 159-173(in Chinese). doi: 10.13700/j.bh.1001-5965.2018.0239
[19]	杨博雄, 杨雨绮. 利用PCA进行深度学习图像特征提取后的降维研究[J]. 计算机系统应用, 2019, 28(1): 279-283. https://www.cnki.com.cn/Article/CJFDTOTAL-XTYY201901044.htm YANG B X, YANG Y Q. Research on dimensionality reduction after deep learning image feature extraction using PCA[J]. Computer Systems & Applications, 2019, 28(1): 279-283(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-XTYY201901044.htm
[20]	JOSE C. A fast on-line algorithm for PCA and its convergence characteristics[J]. IEEE Transactions on Neural Network, 2000, 4(2): 299-305. http://ieeexplore.ieee.org/document/889421/citations
[21]	李航. 统计学习方法[M]. 北京: 清华大学出版社, 2012: 77-87. LI H. Statistical learning method[M]. Beijing: Tsinghua University Press, 2012: 77-87(in Chinese).
[22]	SHEN Z, CUI P, KUANG K, et al. Causally regularized learning with agnostic data selection bias[C]//ACM Multimedia. New York: ACM Press, 2018: 411-419.
[23]	梁杰, 陈嘉豪, 张雪芹, 等. 基于独热编码和卷积神经网络的异常检测[J]. 清华大学学报(自然科学版), 2019, 59(7): 523-529. https://www.cnki.com.cn/Article/CJFDTOTAL-QHXB201907004.htm LIANG J, CHEN J H, ZHANG X Q, et al. One-hot encoding and convolutional neural network based anomaly detection[J]. Journal of Tsinghua University(Science and Technology), 2019, 59(7): 523-529(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-QHXB201907004.htm
[24]	KUANG K, CUI P, LI B, et al. Estimating treatment effect in the wild via differentiated confounder balancing[C]//ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM Press, 2017: 265-274.
[25]	NEAL P, STEPHEN B. Proximal algorithms[J]. Foundations and Trends in Optimization, 2014, 1(3): 127-239.