留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

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

赵振兵, 张薇, 戚银城, 等 . 融合深度特征的输电线路金具缺陷因果分类方法[J]. 北京航空航天大学学报, 2021, 47(3): 461-468. doi: 10.13700/j.bh.1001-5965.2020.0456
引用本文: 赵振兵, 张薇, 戚银城, 等 . 融合深度特征的输电线路金具缺陷因果分类方法[J]. 北京航空航天大学学报, 2021, 47(3): 461-468. doi: 10.13700/j.bh.1001-5965.2020.0456
ZHAO Zhenbing, ZHANG Wei, QI Yincheng, et al. Causal classification method of transmission lines fitting defect combined with deep features[J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(3): 461-468. doi: 10.13700/j.bh.1001-5965.2020.0456(in Chinese)
Citation: ZHAO Zhenbing, ZHANG Wei, QI Yincheng, et al. Causal classification method of transmission lines fitting defect combined with deep features[J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(3): 461-468. doi: 10.13700/j.bh.1001-5965.2020.0456(in Chinese)

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
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

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

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
  • 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。

     

  • 图 1  本文方法总体流程

    Figure 1.  Overall flowchart of proposed method

    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

    图 3  VGG16模型结构

    Figure 3.  Structure of VGG16 model

    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

    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

    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
  • [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.
  • 加载中
图(5) / 表(6)
计量
  • 文章访问数:  278
  • HTML全文浏览量:  20
  • PDF下载量:  53
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-08-24
  • 录用日期:  2020-08-28
  • 刊出日期:  2021-03-20

目录

    /

    返回文章
    返回
    常见问答