在高压输电线路巡检中，无人机结合先进计算机视觉技术自动精确识别故障是电力巡检研究的关键课题。针对复杂环境下，传统算法检测精度低的问题，研究一种基于深度学习的输电线路多目标故障检测方法。论文的主要工作包括：

（1）针对复杂巡检环境中待检测目标存在不规则的形态以及复杂背景造成传统算法难以准确检测的问题，提出一种基于可变形卷积与瓶颈注意力模块的YOLOv7输电线路检测算法。首先，在YOLOv7检测网络的基础上，为了更加精准地提取图像中的关键信息，在骨干网络中引入可变形卷积模型，通过其可变化的卷积核，精确地获取输电线路上多种故障特征；然后，为减少复杂背景干扰，加入瓶颈注意力模块，通过其空间注意力分支，让模型更专注于输电线路故障的关键部分；其次，利用Wise-IoU损失函数进一步提高边界框回归精度并降低低质量样本的影响；最后，为了验证所提出算法的准确性，利用某电力巡检部门在近4年无人机进行的巡检所得的数据作为实验的样本。实验结果表明，所提出的算法能够在复杂的环境下，可以实现对输电线路上多故障目标的准确检测，平均精度达到了93.8%。

（2）为提升输电线路中小故障目标的检测能力，提出一种基于双分支头部解耦和注意力模型输电线路小故障目标检测算法。首先，为了解决小尺度故障目标特征信息过少问题，在YOLOv7框架下，构造浅层检测层以增强网络对小目标故障的识别能力；然后，针对故障目标易淹没在复杂背景中进而导致目标特征难以有效表达的问题，通过融合轻量化注意力模块以增强故障目标的显著度；其次，为了减少故障检测网络中分类和回归任务的差异性对检测性能的影响，构建双分支头部解耦检测器提升故障识别和定位能力；最后，为了进一步提高故障目标检测精度，在网络中引入Varifocal Loss优化网络参数，提升检测框定位精度。实验结果表明，所提出的算法能够在复杂的环境下，实现对输电线路上的小尺度故障目标的检测，平均精度达到了94.0%。

In the inspection of high-voltage transmission lines, integrating drones with cutting-edge computer vision technology for the automatic and precise identification of potential hazards and faults has become a critical and challenging subject of research in power system inspection. Addressing the identification of fault targets in complex environments, this study proposes a multi-target fault detection method for transmission lines based on deep learning. This method aims to significantly enhance the operational efficiency of transmission lines and promote the intelligent transformation of inspection processes, which is of great importance for the stability and efficient management of power systems. The main contributions of this paper include:

To address the issue of traditional algorithms' difficulty in accurately detecting targets with irregular fault features and complex backgrounds in complex inspection environments, a YOLOv7 detection algorithm for transmission lines based on deformable convolution and bottleneck attention modules is proposed. Firstly, on the basis of YOLOv7 detection network, in order to suppress the interference of complex deformation fault targets, a deformable convolution model is introduced into the backbone network. Through its variable convolution kernel, various fault features on transmission lines are accurately obtained. Then, in order to reduce the interference of complex background, the bottleneck attention module is added. Through its spatial attention branch, the model is more focused on the key part of transmission line fault. Secondly, the Wise-IoU loss function is used to further improve the accuracy of the bounding box regression and reduce the impact of low-quality samples. Finally, in order to verify the effectiveness and accuracy of the proposed algorithm, the data obtained from the inspection of the UAV carried out by a power inspection department in the past four years is used as a sample of the experiment. The experimental results show that the proposed algorithm can realize the detection of multiple fault targets on transmission lines under complex environmental conditions, and the average accuracy reaches 93.8 %.

In order to improve the detection ability of small fault targets in transmission lines, a small fault target detection algorithm for transmission lines based on double branch head decoupling and attention model is proposed. Firstly, in order to solve the problem of too little feature information of small-scale fault targets, a shallow detection layer is constructed under the YOLOv7 framework to enhance the network 's ability to identify small-scale fault targets. Then, aiming at the problem that the fault target is easy to be submerged in the complex background, which leads to the difficulty in effectively expressing the target features, the saliency of the fault target is enhanced by integrating the lightweight attention module. Secondly, in order to reduce the influence of the difference between classification and regression tasks on the detection performance in the fault detection network, a double branch head decoupling detector is constructed to improve the fault identification and location ability. Thirdly, in order to further improve the accuracy of fault target detection, Varifocal Loss is introduced into the network to optimize the network parameters and improve the positioning accuracy of the detection frame. The experimental results show that the proposed algorithm can realize the detection of small-scale fault targets on transmission lines under complex environmental conditions, and the average accuracy reaches 94.0 %.

[1]刘颖, 王玉, 于凤洋, 等. 基于GM(1, 1)法的吉林省“十四五”电力需求量预测分析[J]. 绿色科技, 2022, 24(18): 232-236.

[2]韩先才, 孙昕, 陈海波, 等. 中国特高压交流输电工程技术发展综述[J]. 中国电机工程学报, 2020, 40(14): 4371-4386+4719.

[3]单葆国, 刘青, 张莉莉等. 新形势下“十四五”后三年中国电力需求形势研判[J]. 中国电力, 2023, 56(03): 1-11.

[4]姚尚衡, 傅晨玲, 曾金灿, 等. 智能电网技术创新对电力行业绿色转型的非线性影响[J]. 智慧电力, 2024, 52(02): 1-7+16.

[5]成建宏. 机器视觉在输电线路巡检机器人中的应用综述[J]. 自动化技术与应用, 2019, 38(04): 83-87.

[6]韦舒天, 李龙, 岳灵平, 等. 输电通道人机协同巡检方式的探索[J]. 浙江电力, 2016, 35(03): 10-13.

[7]Zhang Y, Yuan X, Li W, et al. Automatic power line inspection using UAV images[J]. Remote Sensing, 2017, 9(8): 824.

[8]Yang L, Fan J, Liu Y, et al. A review on state-of-the-art power line inspection techniques[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(12): 9350-9365.

[9]张庆静, 耿辉. 红外线成像仪在高压输电线路中的应用[J]. 科技创新与生产力, 2011(02): 100-101.

[10]王卓远, 张迪.紫外线检测技术在高灵敏度绝缘子远距离放电监测中的应用[J]. 通信电源技术, 2017, 34(06): 75-77.

[11]Sikorski W. Development of acoustic emission sensor optimized for partial discharge monitoring in power transformers[J]. Sensors, 2019, 19(8): 1865.

[12]Li J, Zhang M. Physics and applications of Raman distributed optical fiber sensing[J]. Light: Science & Applications, 2022, 11(1): 128.

[13]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(01): 266-275.

[14]方挺, 韩家明. 航拍图像中绝缘子缺陷的检测与定位[J]. 计算机科学, 2016, 43(S1): 222-225.

[15]Zhao Z, Liu N, Wang L. Localization of multiple insulators by orientation angle detection and binary shape prior knowledge[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2015, 22(6): 3421-3428.

[16]姚春羽, 金立军, 闫书佳. 电网巡检图像中绝缘子的识别[J]. 系统仿真学报, 2012, 24(09): 1818-1822.

[17]林聚财, 韩军, 陈舫明, 等. 基于彩色图像的玻璃绝缘子缺陷诊断[J]. 电网技术, 2011, 35(01): 127-133.

[18]马帅营, 安居白, 陈舫明. 基于区域定位的绝缘子图像分割[J]. 电力建设, 2010, 31(07): 14-17.

[19]张少平, 杨忠, 黄宵宁, 等. 航拍图像中玻璃绝缘子自爆缺陷的检测及定位[J]. 太赫兹科学与电子信息学报, 2013, 11(04): 609-613.

[20]Zhai Y, Chen R, Yang Q, et al.Insulator ault detection based on spatial morphological features of aerial images[J].IEEE Access, 2018, 6: 35316-35326.

[21]Du L, Zhang R, Wang X. Overview of two-stage object detection algorithms[C]//Journal of Physics: Conference Series. IOP Publishing, 2020, 1544(1): 012033.

[22]Zhang Y, Li X, Wang F, et al. A comprehensive review of one-stage networks for object detection[C]//2021 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC). IEEE, 2021: 1-6.

[23]Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 580-587.

[24]Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[25]Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28.

[26]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.

[27]赵文清, 程幸福, 赵振兵, 等. 注意力机制和Faster RCNN相结合的绝缘子识别[J]. 智能系统学报, 2020, 15(01): 92-98.

[28]杨莉, 王金宇. 基于改进R-FCN的多尺度输电线路鸟巢检测方法[J]. 数据通信, 2023(06): 29-33.

[29]李鑫, 刘帅男, 杨桢, 等. 基于改进Cascade R-CNN的输电线路多目标检测[J]. 电子测量与仪器学报, 2021, 35(10): 24-32.

[30]郝艳捧, 张伟勋, 梁苇, 等. 输电线路覆冰导线Mask R-CNN图像识别与分割[J]. 广东电力, 2022, 35(02):1-10.

[31]Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition(CVPR). Las Vegas, USA: IEEE, 2016: 779-788.

[32]Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition(CVPR). Hawaii, USA: IEEE, 2017: 7263-7271.

[33]郝帅, 张旭, 马旭, 等. 基于CBAM-YOLOv5的煤矿输送带异物检测[J]. 煤炭学报, 2022, 47(11): 4147-4156.

[34]Liu W, Anguelov D, Erhan D, et al. Ssd: Single shot multibox detector[C]//Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14. Springer International Publishing, 2016: 21-37.

[35]阎光伟, 王婷. 基于平衡特征融合SSD的绝缘子检测方法[J]. 中国科技信息, 2021(10): 99-100.

[36]熊鸿军, 熊旺.基于改进的YOLOv3输电线路异物检测[J].上海电机学院学报, 2023, 26(06):350-355+366.

[37]陈杰, 安之焕, 唐占元, 等. 基于改进YOLOv4模型的无人机巡检图像杆塔缺陷检测方法研究[J]. 电测与仪表, 2023, 60(10): 155-160.

[38]季世超, 曲星合, 宋庆彬, 等. 基于改进YOLOv5的轻量化绝缘子缺陷检测算法[J].浙江电力, 2023, 42(12):64-72.

[39]杨校李, 高林, 赵晓雨, 等. 基于改进YOLOv7-tiny算法的输电线路螺栓缺销检测[J]. 湖北民族大学学报(自然科学版), 2023, 41(03): 314-321.

[40]彭紫扬, 陈诺天, 易俊飞, 等. 基于改进RetinaNet算法的输电线路电力器件及异常目标检测[J]. 湖南电力, 2023, 43(05): 79-84.

[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]Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. pmlr, 2015: 448-456.

[43]Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25.

[44]Nair V, Hinton G E. Rectified linear units improve restricted boltzmann machines[C]//Proceedings of the 27th international conference on machine learning (ICML-10). 2010: 807-814.

[45]Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. nature, 1986, 323(6088): 533-536.

[46]LeCun Y, Bottou L, Orr G B, et al. Efficient backprop[M]//Neural networks: Tricks of the trade. Berlin, Heidelberg: Springer Berlin He idelberg, 2002: 9-50.

[47]Maas A L, Hannun A Y, Ng A Y. Rectifier nonlinearities improve neural network acoustic models[C]//Proc. icml. 2013, 30(1): 3.

[48]Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. nature, 1986, 323(6088): 533-536.

[49]Cortes C, Vapnik V. Support-vector networks[J]. Machine learning, 1995, 20: 273-297.

[50]Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.

[51]黄凤荣, 李杨, 郭兰申等.基于Faster R-CNN的零件表面缺陷检测算法[J].计算机辅助设计与图形学学报, 2020, 32(06):883-893.

[52]He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969.

[53]Cai Z, Vasconcelos N. Cascade r-cnn: Delving into high quality object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 6154-6162.

[54]Kirillov A, Girshick R, He K, et al. Panoptic feature pyramid networks[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 6399-6408.

[55]Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988.

[56]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.

[57]He K, Zhang X, Ren S, et al. Identity mappings in deep residual networks[C]//Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part IV 14. Springer International Publishing, 2016: 630-645.

[58]Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2117-2125.

[59]Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988.

[60]Tan M, Pang R, Le Q V. Efficientdet: Scalable and efficient object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 10781-10790.

[61]高敏, 李元. 基于YOLOv7-CA-BiFPN的路面缺陷检测[J/OL]. 计算机测量与控制: 1-12[2024-03-01].

[62]Loshchilov I Hutter F. Sgdr: Stochastic gradient descent with warm restarts[J]. arXiv preprint arXiv:1608.03983, 2016.

[63]张文铠, 刘佳. 基于改进YOLOv7的口罩佩戴检测算法[J]. 北京信息科技大学学报(自然科学版), 2024, 39(01): 55-60.

[64]Dai J, Qi H, Xiong Y, et al. Deformable convolutional networks[C]//Proceedings of the IEEE international conference on computer vision. 2017: 764-773.

[65]Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.

[66]王安静, 袁巨龙, 朱勇建, 等. 基于改进YOLOv8s的鼓形滚子表面缺陷检测算法[J]. 浙江大学学报(工学版), 2024, 58(02): 370-380+387.

[67]Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.