<p>高压输电线路由于长期暴露在复杂自然环境中，受雨水腐蚀、雷击等自然条件干扰 易发生故障，严重影响电力系统的稳定运行。目前，航拍巡检是国内外输电线路巡检的 重要方式，同时也是一个发展趋势。然而，当输电线路中螺丝、垫片等小目标电气元件 处于复杂背景环境中时，由于其自身图像区域面积小和特征难以提取导致传统算法难以 准确检测。因此，面向复杂巡检环境，本文探索了基于深度注意力模型的输电线路小目 标故障检测方法，该研究对于提高输电线路巡检效率，实现智能巡检具有重要意义。 论文主要工作包括： （1）针对小目标电气元件尺寸小和部分航拍巡检图像质量不高而易造成故障区域难以被有效检测的问题，提出一种基于改进 SRCNN-YOLOv5 的输电线路小目标故障检测 算法。首先，通过超高分辨率卷积神经网络减少原始图像中的噪声和模糊，对数据集进 行优化；然后，对 YOLOv5 检测网络的尺度预测部分进行优化，以提高算法对小目标检 测的精确度；最后，为了解决原网络锚点框尺寸难以精确定位问题，通过 K-means 聚类 算法得到新的锚点框参数，进一步提高网络对目标框检测的精准度。通过将提出的方法 与原 YOLOv5 网络进行对比实验，该方法将原网络的目标检测 mAP 值提升了 1.3%。（2）针对小目标在检测过程中特征表达能力弱的问题，提出一种融合卷积块注意模 型的输电线路小目标故障检测算法。首先，通过在 YOLOv5 主干网络的每个跨阶段局部 网络层后引入卷积块注意模型来增加在复杂背景中待检测目标信息的显著度，由此来提 高网络对小目标的检测精度；然后，通过在网络颈部使用深度可分卷积改进原网络中的 普通卷积，减小原网络的参数量和运算量，进而缩短网络检测时间；最后，通过优化网 络损失函数使得算法检测效果得到进一步提升。通过与 SSD、Faster RCNN、YOLOv4 和 YOLOv5 四种经典算法比较，证明了该方法不仅使单张图片检测速度达到了 39ms， 而且该算法的检测 mAP 值可以达到 94.6%，相比原 YOLOv5 网络提高了 3.4。</p>

<p>Due to long-term exposure to complex natural environment, high-voltage transmission lines are prone to failures due to natural conditions such as rainwater corrosion and lightning strikes, which seriously affect the stable operation of the power system. At present, aerial photography inspection is an important way of inspection of transmission lines at home and abroad, and it is also a development trend. However, when the small target electrical components such as screws and gaskets in the transmission line are in a complex background environment, traditional algorithms are difficult to accurately detect due to their small image area and difficult feature extraction. Therefore, facing the complex inspection environment, this paper explores a small target fault detection method for transmission lines based on the deep attention model. This research is of great significance for improving the inspection efficiency of transmission lines and realizing intelligent inspection. The main work of the thesis includes: (1) Aiming at the small size of small target electrical components and the low quality of some aerial inspection images, it is easy to cause the fault area to be difficult to be effectively detected. An improved SRCNN-YOLOv5 based transmission line small target fault detection algorithm is proposed. First, reduce the noise and blur in the original image through the ultra-high resolution convolutional neural network, and optimize the dataset. Then, optimize the scale prediction part of the YOLOv5 detection network to improve the accuracy of the algorithm for small target detection. Finally, in order to solve the problem that the size of the original network anchor box is difficult to accurately locate, new anchor box <font color='red'>parameter</font>s are obtained through the K-means clustering algorithm, which further improves the accuracy of the target box detection by the network. By comparing the proposed method with the original YOLOv5 network, the method improves the target detection mAP value of the original network by 1.3%. (2) Aiming at the problem of weak feature expression ability of small targets in the detection process, a small target fault detection algorithm for transmission lines is proposed, which integrates the convolution block attention model. First, the convolution block attention model is introduced after each cross-stage local network layer of the YOLOv5 backbone network to increase the saliency of the target information to be detected in the complex background, thereby improving the detection accuracy of the network for small targets. Then, by using the depth separable convolution in the network neck to improve the ordinary convolution in the original network, the <font color='red'>parameter</font> amount and computation amount of the original network are reduced, and the network detection time is shortened. Finally, the algorithm detection effect is further improved by optimizing the network loss function. promote. By comparing with the four classic algorithms of SSD, Faster RCNN, YOLOv4 and YOLOv5, it is proved that this method not only makes the detection speed of a single image reach 39ms, but also the detection mAP value of this algorithm can reach 94.6%, which is improved compared with the original YOLOv5 network 3.4%.</p>

[1]Wang Y, Chen L, Yao M, et al. Evaluating weather influences on transmission line failure rate based on scarce fault records via a bi-layer clustering technique[J]. IET Generation, Transmission & Distribution, 2019, 13(23): 5305-5312.

[2] Saber A, Zeineldin H, El-Fouly T, et al. Current differential relay characteristic for bipolar

HVDC transmission line fault detection[J]. IET Generation Transmission & Distribution, 2020, 14(23):5505-5513.

[3] Hazim S, Ahmad S, Ala A F, et al. Unmanned Aerial Vehicles (UAVs): A Survey on Civil

Applications and Key Research Challenges[J]. IEEE Access, 2018,7(48):572-634.

[4] Lai D, Huang Z, Qian J. Development of Hybrid VTOL Inspection UAV[J]. International

Core Journal of Engineering, 2020, 06(04):241-245.

[5] Ribeiro R G, JRC Júnior, Cota L P, et al. Unmanned Aerial Vehicle Location Routing

Problem With Charging Stations for Belt Conveyor Inspection System in the Mining

Industry[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(10):4186-4195. [6] Mfds A, Lmh A, Almm A, et al. Unmanned aerial vehicle for transmission line inspection

using an extended Kalman filter with colored electromagnetic interference - ScienceDirect[J]. ISA Transactions, 2020, 100(08):322-333. [7] Hu S, Yang S, Zhang Y. Research on UAV Technology in the Field Of Transmission Line

Inspection[J]. International Core Journal of Engineering, 2020, 06(04):173-175.

[8] Abouzahir S, Sadik M, Sabir E. Bag-of-visual-words-augmented Histogram of Oriented

Gradients for efficient weed detection[J]. Biosystems Engineering, 2021, 202(05):

179-194.

[9] Wang Y, Zhang X. Digital core image registration based on SIFT features[C]//Journal of

Physics: Conference Series. IOP Publishing, 2021, 1748(04): 042-047. [10]Abdulrahim K, Salam R A. Traffic surveillance: A review of vision based vehicle

detection, recognition and tracking[J]. Int. J. Appl. Eng. Res, 2016, 11(01): 713-726. [11]周宏宇，宋旭，刘国英．Haar 特征耦合级联分类器的车道线检测算法[J]．计算机工

程与设计，2020，41(06)：10-16．

[12]姜云土，韩军，丁建，等．基于多特征融合的玻璃绝缘子识别及自爆缺陷的诊断[J]．中

国电力，2017，50(05)：52-58．

[13]于兰英，姚波，吴文海，刘桓龙，谭万秋，等．一种基于多特征的绝缘子识别方法[J]．电

瓷避雷器，2016，271(03)：79-83．

[14]彭向阳，梁福逊，钱金菊，等．基于机载红外影像纹理特征的输电线路绝缘子自动

参考文献

61

定位[J]．高电压技术，2019，45(03)：922-928．

[15]陈文浩，姚利娜，李丰哲，等．无人机电网巡检中的绝缘子缺陷检测与定位[J].计算

机应用，2019，39(S1)：210-214．

[16]金立军，胡 娟，闫书佳，等．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(05)：1040-1045．

[17]赵江曼，孟建良．高压输电线路航拍图像目标边缘检测[J]．中国电力，2018，51(02)：

27-32．

[18]马兴誉．基于航拍图像的输电线路自动巡检相关技术研究[D]．北京：华北电力大学，

2016．

[19]翟永杰，王迪，赵振兵，程海燕，等．基于空域形态一致性特征的绝缘子串定位方

法[J]．中国电机工程学报，2017，37(05)：1568-1578．

[20]刘思言，王 博，高昆仑，等．基于 R-FCNN 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．

[21]Girshick R, Donahue J, Darrell T, et al. Richfeature hierarchies for accurate object

detection and semantic segmentation[C]//Proceedings of 2014 IEEE Conference on

Computer Vision and Pattern Recognition (CVPR). Columbus, OH, USA, 2014: 580-587. [22]Girshick R. Fast RCNN[C]//Proceedings of 2015 IEEE Internat

ional Conference on

Computer Vision (ICCV). Santiago, Chile, 2015: 1440-1448.

[23]Ren S, He K, Girshick R, et al. Faster RCNN: towards real.time object detection with

regionproposal networks[J]. IEEE transactions on pattern analysis and machine

intelligence, 2017, 39(06):1137-1149.

[24]白洁音，赵瑞，谷丰强，等．多目标检测和故障识别图像处理方法[J]．高电压技术，

2019，45(11)：3504-3511．

[25]林刚，王波，彭辉，等．基于改进 Faster-RCNN 的输电线巡检图像多目标检测及定

位[J]．电力自动化设备，2019，39(05)：213-218．

[26]Redmon J, Divvala S, Girshick R, et al. You onlylook once: unified, real.time object

detection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern

Recognition (CVPR). Las Vegas, NV, USA, 2016:779-788. [27]Liu W, Aaguelov D, Erhan D, et al. SSD: singleshot MultiBox detector[C]//Proceedings of

the 14thEuropean Conference on Computer Vision. Amsterdam, the Netherlands, 2016:

21-37. [28]Miao X, Liu X, Chen J, et al. Insulator detection in aerial images for transmission line

inspection using single shot multibox detector[J]. IEEE Access, 2019, 07(02):9945-9956. [29]田庆，胡蓉，李佐勇，等．基于 SE-YOLOv5s 的绝缘子检测[J]．智能科学与技术学

报， 2021，03(03)：312-321．

[30]Chollet F. Xception: Deep learning with depthwise separable

西安科技大学全日制工程硕士学位论文

62

convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern

recognition. 2017: 1251-1258.

[31]Everingham M, Van Gool L, Williams C K, et al. The pascal visual object classes (voc)

challenge[J]. International journal of computer vision, 2010, 88(02): 303-338.

[32]Lin T-Y, Maire M, Belongie S, et al. Microsoft coco: Common objects in context[C]. in:

European conference on computer vision, 2014: 740-755.

[33]Dai Y, Wu Y, Zhou F, et al. Attentional local contrast networks for infrared small target

detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(11):

9813-9824.

[34]Han C, Gao G, Zhang Y. Real-time small traffic sign detection with revised

Faster-RCNN[J]. Multimedia Tools and Applications, 2019, 78(10):13263- 13278.

[35]Lin T Y, Dollar P, Girshick R, et al. Feature pyramid networks for object

detection[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern

Recognition (CVPR). Honolulu, HI, USA, 2017:936-944.

[36]Araki R, Onishi T, Hirakawa T, et al. Mt-dssd: Deconvolutional single shot detector using

multi task learning for object detection, segmentation, and grasping detection[C]//2020

IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020:

10487-10493.

[37]Zhang S, Zhu X, Lei Z, et al. Faceboxes: A CPU real-time face detector with high

accuracy[C]//2017 IEEE International Joint Conference on Biometrics (IJCB). IEEE, 2017: 1-9.

[38]Kisantal M, Wojna Z, Murawski J, et al. Augmentation for small object

detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern

Recognition, 2019: 2-10.

[39]Mahmoud M A B, Guo P, Fathy A, et al. SRCNN-PIL: Side Road Convolution Neural

Network Based on Pseudoinverse Learning Algorithm[J]. Neural Processing Letters, 2021, 53(6): 4225-4237.

[40]Ecke G A, Papp H M, Mallot H A. Exploitation of image statistics with sparse coding in

the case of stereo vision[J]. Neural Networks, 2021, 135(08): 158-176. [41]Lecun Y, Bottou L,Bengio Y, et al. Gradient-based learning applied to document

recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324. [42]Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional

neural networks[C].in: Advances in neural information processing systems, 2012:1097-1105.

[43]Lu S, Lu Z, Zhang Y D. Pathological brain detection based on AlexNet and transferlearning[J]. Journal of computational science, 2019, 30(01): 41-47. [44]Huang G, Liu S, Van der Maaten L, et al. Condensenet: An efficient densenet using

learned group convolutions[C]//Proceedings of the IEEE conference on computer vision

and pattern recognition, 2018: 2752-2761.

[45]Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]. in:Proceedings of the

IEEE conference on computer vision and pattern recognition, 2015:1-9. [46]Gao S H, Han Q, Li D, et al. Representative batch normalization with feature

calibration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern

Recognition, 2021: 8669-8679.

[47]Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer

vision[C]. in: Proceedings of the IEEE conference on computer vision and pattern

recognition, 2016: 2818-2826.

[48]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]. in:

Proceedings of the IEEE conference on computer vision and pattern recognition, 2016:

770-778.

[49]Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural

networks[C]. in: Proceedings of the IEEE Conference on Computer Vision and Pattern

Recognition, 2017: 1492-1500.

[50]Chollet F. Xception: Deep learning with depthwise separable convolutions[C]. in:

Proceedings of the IEEE conference on computer vision and pattern recognition, 2017:

1251-1258.

[51]Yu F, Koltun V, Funkhouser T. Dilated residual networks[C]. in: Proceedings of the

IEEE conference on computer vision and pattern recognition, 2017: 472-480. [52]Zhang X, Zhou X, Lin M, et al. Shufflenet: An extremely efficient convolutional neural

network for mobile devices[C]. in: Proceedings of the IEEE Conference on Computer

Vision and Pattern Recognition, 2018: 6848-6856.

[53]Shankar K, Lakshmanaprabu S K, D Gupta, et al. Optimal feature-based multi-kernel

SVM approach for thyroid disease classification[J]. The Journal of Supercomputing, 2020, 76(28):1-16.

[54]Ssr A, Skv B, Yk C. Review on recent development in infrared small target detection

algorithms[J]. Procedia Computer Science, 2020, 167(05):2496-2505. [55]Redmon J, Farhadi A. YOLO9000: Better, Faster, Stronger[J]. IEEE, 2017:6517-6525. [56]Redmon J, Farhadi A. YOLOv3:An incremental improvement [EB/OL]. [2018-09-30].

https://arxiv.org/abs/1804.02767.

[57]Bochkovshiy A, Wang C Y, Liao H Y M. YOLOv4:Optimal speed and accuracy of object

detection[J/OL].[2020-04-23].https://arxiv.org/abs/2004.10934.

[58]Gonon L, Schwab C. Deep ReLU network expression rates for option prices in

high-dimensional, exponential Lévy models[J]. Finance and Stochastics, 2021, 25(4):

615-657.

[59]He K, Zhang X, Ren S, et al. Delving deep into rectifiers: Surpassing human-level

performance on imagenet classification[C]//Proceedings of the IEEE international

conference on computer vision, 2015: 1026-1034.

[60]Govender P, Sivakumar V. Application of k-means and hierarchical clustering techniques

for analysis of air pollution: A review (1980–2019)[J]. Atmospheric Pollution Research, 2020, 11(01): 40-56.

[61]Zhang S, Zhang C, You Z, et al. Asynchronous stochastic gradient descent for DNN

training[C]//IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2013: 6660-6663.

[62]Sifre L, Mallat S. Rigid-Motion Scattering for Texture Classification[J]. Computer

Science, 2014, 3559(01):501-515.

[63]Hu J, Shen L, Sun G. Squeeze.and.excitation networks[C]//Proceedings of the IEEEConference on Computer Vision and Pattern Recognition(CVPR), 2018: 7132-7141.

[64]Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention

module[C]//Proceedings of the European Conference on Computer Vision(ECCV), 2018:

3-19.

[65]Lin T Y, Goyal P, Girshick R, et al. Focal Loss for Dense Object

Detection[C]//Proceedings of IEEE Transactions on Pattern Analysis and MachineIntelligence, 2017:2999-3007