基于YOLO-2MCS的输电线路走廊隐患目标检测方法

doi:10.19595/j.cnki.1000-6753.tces.230666

摘要
图/表
参考文献
相关文章 (3)

全文: PDF (1969 KB)
输出: BibTeX | EndNote (RIS)

摘要

输电线路在跨越高速铁路、高速公路和重要输电通道场景下易受到外力破坏,可能严重影响输电线路安全可靠运行。针对此问题,该文通过构建输电线路走廊隐患目标数据集,提出新模型YOLO-2MCS用于输电线路走廊隐患目标检测。使用混合数据增强策略对数据集进行有效扩充,以提高模型在复杂场景下的泛化性和鲁棒性;在EfficientRep骨干网络引入卷积注意力机制模块,有效提升模型对多尺度目标的检测能力;构建使用softplus激活函数的双向特征金字塔结构加强模型特征学习能力;在检测头使用SIoU损失函数进一步提升模型检测精度。实验结果表明,相较于原YOLOv6网络,该模型在0.5:0.95的严苛阈值下平均精度均值提升4.4%;将该模型与主流的检测模型Faster R-CNN、YOLOX、YOLOv5和YOLOv7模型分别进行对比评估,该模型的检测精度、检测速度、模型复杂度均获得最优化性能,其平均检测速度高达约300帧/秒,且内存大小仅为40.7 MB,同时满足在边缘计算设备上部署的要求。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	郑含博
	胡思佳
	梁炎燊
	黄俊杰
	汪涛

关键词 ：输电线路走廊, 防外破, 目标检测, 注意力机制

Abstract：

Transmission lines are vulnerable to external breakage when crossing high-speed railways, highways and important transmission channels (referred to as “Three-Span”). The occurrence of such accidents is highly random, and may causes damage to transmission equipment, power grid tripping, power outages, and even accidental electric shock and other safety accidents, which seriously affect the safe, reliable and stable operation of transmission line. The existing super-high/UHV transmission line inspection methods include manual inspection, robot inspection, helicopter inspection and UAV inspection, etc., and gradually form the maintenance and operation mode with UAV inspection as the main and manual inspection as the supplement. However, UAV inspection also has limitations such as difficult operation, limited flight distance, and many uncertainties in the field, which is not suitable for large-scale promotion.
Aiming at the limitations of the existing transmission line inspection methods and the lack of the external force damage object dataset, this paper collected and constructed a hidden danger object dataset of transmission line corridor, which contains three typical hidden danger object categories and 8 654 targets in total. The hybrid data augmentation strategy is used to effectively enrich the dataset to improve the generalization and robustness of the model in complex scenarios and avoid the model overfitting problem during training caused by a single scenario. The convolutional block attention module (CBAM) is introduced in the EfficientRep backbone network to reduce the feature loss in the extraction process of the original backbone network and improve the model's ability to identify and locate occluded objects. The bidirectional feature pyramid network using the softplus activation function enhance the feature learning ability of the model and the convergence speed during training. The SIoU loss function that defines the angle cost function is introduced in the detector, which can make the loss function in the training process converge as soon as possible and complete the parameter adjustment of backpropagation, so that the model can obtain better and faster object positioning performance.
The results of ablation experiments showed that the strategies proposed in this paper can significantly improve the detection accuracy and detection speed of the model without unduly affecting the complexity of the model in the hidden danger object dataset of transmission line corridor constructed in this paper. Compared with the original YOLOv6 model, the average accuracy of the new model is improved by 4.4% under the strict threshold of 0.5:0.95, the average detection speed is as high as about 300 frames per second, and the memory size is only 40.7MB. The results of comparative experiments showed that the detection accuracy and detection speed of the proposed method are significantly better than the Faster R-CNN, YOLOX, YOLOv5 and YOLOv7 under the same hyperparameters and the same dataset. The visualization results showed that when there is a hidden danger object that are occluded, the proposed model can also better identify and locate it.
Based on the above experimental results, the YOLO-2MCS model proposed in this paper can not only accurately identify the categories of external force damage in the transmission line corridor monitoring scene, but also quickly and accurately locate the hidden danger object, all while meeting the needs of devices installed on the mobile edge. The hidden danger object dataset of transmission line corridor constructed in this paper provides effective support for subsequent model training and iteration, and effectively develops the intelligent development of the prevention of external force damage in transmission line corridor.

Key words： transmission line corridor prevention of external force damage object detection attention mechanism

收稿日期: 2023-05-14

PACS:

TM85

基金资助:

国家自然科学基金（52367014, 52277139）和广西科技基地和人才专项（2020AC19010）资助项目

通讯作者: 郑含博男,1984年生,副教授,博士生导师,研究方向为电力设备的状态评估和智能诊断。E-mail：hanbozheng@163.com

作者简介: 胡思佳女,1999年生,硕士研究生,研究方向为电力设备的图像识别与智能诊断。E-mail：husj78@126.com

引用本文:

郑含博, 胡思佳, 梁炎燊, 黄俊杰, 汪涛. 基于YOLO-2MCS的输电线路走廊隐患目标检测方法[J]. 电工技术学报, 0, (): 230666-230666. Zheng Hanbo, Hu Sijia, Liang Yanshen, Huang Junjie, Wang Tao. A hidden danger object detection method for transmission line corridor based on YOLO-2MCS. Transactions of China Electrotechnical Society, 0, (): 230666-230666.

链接本文:

https://dgjsxb.ces-transaction.com/CN/10.19595/j.cnki.1000-6753.tces.230666 https://dgjsxb.ces-transaction.com/CN/Y0/V/I/230666

[1] 和敬涵, 罗国敏, 程梦晓, 等. 新一代人工智能在电力系统故障分析及定位中的研究综述[J]. 中国电机工程学报, 2020, 40(17): 5506-5516.
He Jinghan, Luo Guomin, Cheng Mengxiao, et al.A research review on application of artificial intelligence in power system fault analysis and location[J]. Proceedings of the CSEE, 2020, 40(17): 5506-5516.
[2] 国家电网公司. 架空输电线路“三跨”重大反事故措施(试行)[Z]. 2016.
[3] 沙洁韵. “三跨”输电线路智慧巡检中的车辆检测技术研究[D]. 西安: 西安电子科技大学, 2021.
[4] 国家电网有限公司安全监察部. 国家电网有限公司2021年生产安全事故事件分析报告[M]. 北京: 中国电力出版社, 2022.
[5] 陆佳政, 周特军, 吴传平, 等. 某省级电网220kV及以上输电线路故障统计与分析[J]. 高电压技术, 2016, 42(1): 200-207.
Lu Jiazheng, Zhou Tejun, Wu Chuanping, et al.Fault statistics and analysis of 220 kV and above power transmission line in province-level power grid[J]. High Voltage Engineering, 2016, 42(1): 200-207.
[6] 刘传洋, 吴一全. 基于深度学习的输电线路视觉检测方法研究进展[J/OL]. 中国电机工程学报, 2022: 1-24[2023-05-12]. http://kns.cnki.net/kcms/detail/11.
2107.TM.20220831.1053.002.html.Liu Chuanyang, Wu Yiquan. Research progress of vision detection methods based on deep learning for transmission lines[J/OL]. Proceedings of the CSEE, 2022: 1-24[2023-05-12]. http://kns.cnki.net/kcms/detail/11.2107.TM.20220831.1053.002.html.
[7] 彭向阳, 易琳, 钱金菊, 等. 大型无人直升机电力线路巡检系统实用化[J]. 高电压技术, 2020, 46(2): 384-396.
Peng Xiangyang, Yi Lin, Qian Jinju, et al.Practical research on patrol inspection system of electric power line based on large scale unmanned helicopter[J]. High Voltage Engineering, 2020, 46(2): 384-396.
[8] 仲林林, 胡霞, 刘柯妤. 基于改进生成对抗网络的无人机电力杆塔巡检图像异常检测[J]. 电工技术学报, 2022, 37(9): 2230-2240, 2262.
Zhong Linlin, Hu Xia, Liu Keyu.Power tower anomaly detection from unmanned aerial vehicles inspection images based on improved generative adversarial network[J]. Transactions of China Electrotechnical Society, 2022, 37(9): 2230-2240, 2262.
[9] Li Jinheng, Shuang Feng, Huang Junjie, et al.Safe distance monitoring of live equipment based upon instance segmentation and pseudo-LiDAR[J]. IEEE Transactions on Power Delivery, 2023, 38(4): 2953-2964.
[10] 宋立业, 刘帅, 王凯, 等. 基于改进EfficientDet的电网元件及缺陷识别方法[J]. 电工技术学报, 2022, 37(9): 2241-2251.
Song Liye, Liu Shuai, Wang Kai, et al.Identification method of power grid components and defects based on improved EfficientDet[J]. Transactions of China Electrotechnical Society, 2022, 37(9): 2241-2251.
[11] Wu Xiongwei, Sahoo D, Hoi S C H. Recent advances in deep learning for object detection[J]. Neurocomputing, 2020, 396: 39-64.
[12] Zheng Hanbo, Cui Yaohui, Yang Wenqiang, et al.An infrared image detection method of substation equipment combining iresgroup structure and CenterNet[J]. IEEE Transactions on Power Delivery, 2022, 37(6): 4757-4765.
[13] Ren Shaoqing, He Kaiming, Girshick R, et al.Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[14] He Kaiming, Gkioxari G, Dollár P, et al.Mask R-CNN[C]//2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2980-2988.
[15] Cai Zhaowei, Vasconcelos N.Cascade R-CNN: delving into high quality object detection[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 6154-6162.
[16] 苟军年, 杜愫愫, 刘力. 基于改进掩膜区域卷积神经网络的输电线路绝缘子自爆检测[J]. 电工技术学报, 2023, 38(1): 47-59.
Gou Junnian, Du Susu, Liu Li.Transmission line insulator self-explosion detection based on improved mask region-convolutional neural network[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 47-59.
[17] Redmon J, Farhadi A.YOLO9000: better, faster, stronger[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017: 6517-6525.
[18] Redmon J, Farhadi A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08)[2023-09-07]. https://arxiv.org/abs/1804.02767.
[19] Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23)[2023-09-01]. https://arxiv.org/abs/2004.10934
[20] Ge Zheng, Liu Songtao, Wang Feng, et al. YOLOX: exceeding YOLO series in2021[EB/OL]. (2021-08-06) [2023-09-01]. https://arxiv.org/abs/2107.08430.
[21] Li Chuyi, Li Lulu, Jiang Hongliang, et al. YOLOv6: a single-stage object detection framework for industrial applications[EB/OL]. (2022-09-07)[2023-09-01]. https://arxiv.org/abs/2209.02976.
[22] Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[EB/OL]. (2022-07-06) [2023-09-01]. https://arxiv.org/abs/2207.02696.
[23] Liu Wei, Anguelov D, Erhan D, et al.SSD: single shot multibox detector[C]//Computer Vision-ECCV 2016, Amsterdam, the Netherlands, 2016: 21-37.
[24] 郑含博, 李金恒, 刘洋, 等. 基于改进YOLOv3的电力设备红外目标检测模型[J]. 电工技术学报, 2021, 36(7): 1389-1398.
Zheng Hanbo, Li Jinheng, Liu Yang, et al.Infrared object detection model for power equipment based on improved YOLOv3[J]. Transactions of China Electrotechnical Society, 2021, 36(7): 1389-1398.
[25] 龙乐云, 周腊吾, 刘淑琴, 等. 改进YOLOv5算法下的输电线路外破隐患目标检测研究[J]. 电子测量与仪器学报, 2022, 36(11): 245-253.
Long Leyun, Zhou Lawu, Liu Shuqin, et al.Identification of hidden damage targets by external forces based on domain adaptation and attention mechanism[J]. Journal of Electronic Measurement and Instrumentation, 2022, 36(11): 245-253.
[26] 唐志国, 曹智, 何宁辉. 卷积神经网络迁移学习在局部放电类型诊断中的应用[J]. 高压电器, 2022, 58(4): 158-164.
Tang Zhiguo, Cao Zhi, He Ninghui.Application of convolutional neural network transfer learning in partial discharge type diagnosis[J]. High Voltage Apparatus, 2022, 58(4): 158-164.
[27] Yetgin Ö E, Benligiray B, Gerek Ö N.Power line recognition from aerial images with deep learning[J]. IEEE Transactions on Aerospace and Electronic Systems, 2019, 55(5): 2241-2252.
[28] 李斌, 屈璐瑶, 朱新山, 等. 基于多尺度特征融合的绝缘子缺陷检测[J]. 电工技术学报, 2023, 38(1): 60-70.
Li Bin, Qu Luyao, Zhu Xinshan, et al.Insulator defect detection based on multi-scale feature fusion[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 60-70.
[29] Pacal I, Karaboga D.A robust real-time deep learning based automatic polyp detection system[J]. Computers in Biology and Medicine, 2021, 134: 104519.
[30] Woo S, Park J, Lee J Y, et al.CBAM: convolutional block attention module[C]//Computer Vision - ECCV 2018, Munich, Germany, 2018: 3-19.
[31] Ding Xin, Wang Z J, Welch W J.Subsampling generative adversarial networks: density ratio estimation in feature space with softplus loss[J]. IEEE Transactions on Signal Processing, 2020, 68: 1910-1922.
[32] Wang Jingyao, Yu Naigong. UTD-Yolov5: a real-time underwater targets detection method based on attention improved YOLOv5[EB/OL]. (2022-07-02) [2023-09-01]. https://arxiv.org/abs/2207.00837.
[33] Gevorgyan Z. SIoU loss: more powerful learning for bounding box regression[EB/OL].2022: arXiv: 2205.12740. https://arxiv.org/abs/2205.12740
[34] Xu Shangliang, Wang Xinxin, Lv Wenyu, et al. PP-YOLOE: an evolved version of YOLO[EB/OL]. (2022-05-25)[2023-09-01]. https://arxiv.org/abs/2203.16250.
[35] Li Chuyi, Li Lulu, Geng Yifei, et al. YOLOv6 v3.0: a full-scale reloading[EB/OL]. (2023-01-13)[2023-09-01]. https://arxiv.org/abs/2301.05586.
[36] Wang C Y, Mark Liao H Y, Wu Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Seattle, WA, USA, 2020: 1571-1580.
[37] Ding Xiaohan, Zhang Xiangyu, Ma Ningning, et al.RepVGG: making VGG-style ConvNets great again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, 2021: 13728-13737.
[38] Yin Lifeng, Hong Pujiang, Zheng Guanghai, et al.A novel image recognition method based on DenseNet and DPRN[J]. Applied Sciences, 2022, 12(9): 4232.
[39] Zheng Zhaohui, Wang Ping, Liu Wei, et al. Distance-IoU loss: faster and better learning for bounding box regression[EB/OL]. (2019-11-19)[2023-09-01]. https://arxiv.org/abs/1911.08287.
[40] Yang Zuomin, Wang Xianlun, Li Jianguang.EIoU: an improved vehicle detection algorithm based on VehicleNet neural network[J]. Journal of Physics: Conference Series, 2021, 1924(1): 012001.