Research on Visual Hierarchical Early Warning of Transmission Line Channel Combined with Scene Analysis

doi:10.11985/2022.03.023

Abstract

Abstract:

In view of the problem that the transmission line spans a wide area and has many hidden targets in the transmission channel, the visual grading warning model of the transmission line channel is put forward. First of all, the depth residual network to is improved to extract the multispectral information of the input image, the noise effect is reduced by soft threshold, the accuracy of the transmission line channel scene analysis model is improved, and then YOLOv3 target detection algorithm is used to build the transmission line channel hidden danger target identification model. Aiming at the problems of hidden smoke and small targets of contruction vihicles, hard example mining strategy is adopted to reduce the impact of the picture background, and then an early warning model is build based on the transmission line channel classification early warning structure. The results show that the visual and graded early warning model of transmission line channel combined with scene analysis can scientifically and accurately reflect the hidden danger warning state of transmission line channel and provide guidance for the operation and maintenance of transmission line.

Key words: Transmission line channel visualization, scene analysis, target recognition, hierarchical warning, deep learning

CLC Number:

TM391

WU Yongquan, ZHANG Siwei, PENG Chong, JIAO Liangbao, PU Dong. Research on Visual Hierarchical Early Warning of Transmission Line Channel Combined with Scene Analysis[J]. Journal of Electrical Engineering, 2022, 17(3): 194-202.

Figures/Tables 13

References 23

[1]	段家振. 试析输电线路通道可视化系统关键技术[J]. 科技创新与应用, 2019(30):151-152.
	DUAN Jiazhen. Analysis of key technologies of transmission line channel visualization system[J]. Science & Technology Innovation and Application, 2019(30):151-152.
[2]	王炜, 王树军, 徐硕, 等. 输电线路通道可视化运维管理体系的应用[J]. 电气时代, 2018(11):77-79.
	WANG Wei, WANG Shujun, XU Shuo, et al. Application of visual operation and maintenance management system for transmission line channels[J]. Electric Times, 2018(11):77-79.
[3]	吕志来, 刘浩, 李海, 等. 输电线路通道可视化系统关键技术研究及实践[J]. 电力信息与通信技术, 2016, 14(9):52-57.
	LÜ Zhilai, LIU Hao, LI Hai, et al. Research and practice on key technologies of transmission line channel visualization system[J]. Electric Power Information and Communication Technology, 2016, 14(9):52-57.
[4]	叶俊健, 邓伟锋, 徐常志, 等. 基于深度强化学习与图像智能识别的输电线路在线监测系统[J]. 工业技术创新, 2020, 7(3):72-75.
	YE Junjian, DENG Weifeng, XU Changzhi, et al. Transmission line online monitoring system based on deep reinforcement learning and image intelligent recognition[J]. Industrial Technology Innovation, 2020, 7(3):72-75.
[5]	孟秀军, 谷连军. 基于Skyline的输电线路巡检数据可视化系统研究[J]. 信息与电脑, 2020, 32(6):141-144.
	MENG Xiujun, GU Lianjun. Research on transmission line inspection data visualization system based on Skyline[J]. Information and Computer, 2020, 32(6):141-144.
[6]	徐常志, 邓伟锋, 李耀均, 等. 输电线路图像可视化监测终端无线通信解决方案[J]. 工业技术创新, 2020, 7(6):99-102,111.
	XU Changzhi, DENG Weifeng, LI Yaojun, et al. Wireless communication solution for visualized monitoring terminal of transmission line image[J]. Industrial Technology Innovation, 2020, 7(6):99-102,111.
[7]	曾昌健, 李丽, 郑国华, 等. 基于VR技术的输电线路运检一体化研究[J]. 山东农业大学学报, 2021, 52(2):304-307.
	ZENG Changjian, LI Li, ZHENG Guohua, et al. Study on the integration of transmission line transport and inspection based on VR technology[J]. Journal of Shandong Agricultural University, 2021, 52(2):304-307.
[8]	江泽涛, 秦嘉奇, 胡硕. 基于多路卷积神经网络的多光谱场景识别方法[J]. 计算机科学, 2019, 46(9):265-270.
	JIANG Zetao, QIN Jiaqi, HU Shuo. Multi-spectral scene recognition method based on multi-channel convolutional neural network[J]. Computer Science, 2019, 46(9):265-270.
[9]	戴玉超, 张静, 何明一. 深度残差网络的多光谱遥感图像显著目标检测[J]. 测绘学报, 2018, 47(6):873-881.
	DAI Yuchao, ZHANG Jing, HE Mingyi. The salient target detection of multispectral remote sensing image based on deep residual network[J]. Acta Geodaetica et Cartographica Sinica, 2018, 47(6):873-881.
[10]	李一松. 基于卷积神经网络的多光谱图像多标签场景分类[J]. 电子设计工程, 2018, 26(23):25-29.
	LI Yisong. Multi-spectral image multi-label scene classification based on convolutional neural network[J]. Electronic Design Engineering, 2018, 26(23):25-29.
[11]	江泽涛, 刘小艳, 胡硕. 基于 CNN 的红外与可见光融合图像的场景识别[J]. 计算机工程与设计, 2019:2289-2294.
	JIANG Zetao, LIU Xiaoyan, HU Shuo. Scene recognition of infrared and visible light fusion image based on CNN[J]. Computer Engineering and Design, 2019:2289-2294.
[12]	张忠星, 李鸿龙, 张广乾, 等. CCNet:面向多光谱图像的高速船只检测级联卷积神经网络[J]. 红外与毫米波学报, 2019(3):290-294.
	ZHANG Zhongxing, LI Honglong, ZHANG Guangqian, et al. CCNet:High-speed ship detection cascaded convolutional neural network for multispectral image[J]. Journal of Infrared and Millimeter Waves, 2019(3):290-294. doi: 10.11972/j.issn.1001-9014.2019.03.006
[13]	姚建华, 吴加敏, 杨勇, 等. 全卷积神经网络下的多光谱遥感影像分割[J]. 中国图象图形学报, 2020, 25(1):180-192.
	YAO Jianhua, WU Jiamin, YANG Yong, et al. Multispectral remote sensing image segmentation under a fully convolutional neural network[J]. China Journal of Graphic Graphics, 2020, 25(1):180-192.
[14]	刘佶鑫, 魏嫚. 可见光-近红外HSV融合的场景类字典稀疏识别[J]. 计算机应用, 2018, 38(12):3359-3366.
	LIU Jixin, WEI Man. Sparse recognition of scene dictionary based on visible light-near infrared HSV fusion[J]. Journal of Computer Applications, 2018, 38(12):3359-3366.
[15]	REDMON J, FARHADI A. YOLO9000:Better,faster,stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017:7263-7271.
[16]	赵飞扬, 罗兵, 林国军, 等. 基于改进YOLOv3的火焰检测[J]. 中国科技论文, 2020, 15(7):820-826.
	ZHAO Feiyang, LUO Bing, LIN Guojun, et al. Based on improved flame detection of YOLOv3[J]. Chinese Science and Technology Papers, 2020, 15(7):820-826.
[17]	钟映春, 孙思语, 吕帅, 等. 铁塔航拍图像中鸟巢的YOLOv3识别研究[J]. 广东工业大学学报, 2020, 37(3):42-48.
	ZHONG Yingchun, SUN Siyu, LÜ Shuai, et al. Research on YOLOv3 recognition of bird’s nest in aerial images of iron tower[J]. Journal of Guangdong University of Technology, 2020, 37(3):42-48.
[18]	于晓英, 苏宏升, 姜泽, 等. 基于 YOLO 的铁路侵限异物检测方法[J]. 兰州交通大学学报, 2020, 39(2):37-42.
	YU Xiaoying, SU Hongsheng, JIANG Ze, et al. YOLO-based railway intrusion detection method for foreign bodies[J]. Journal of Lanzhou Jiaotong University, 2020, 39(2):37-42.
[19]	刘肯, 何姣姣, 张永平, 等. 改进 YOLO 的车辆检测算法[J]. 现代电子技术, 2019, 42(13):47-50.
	LIU Ken, HE Jiaojiao, ZHANG Yongping, et al. Improved YOLO vehicle detection algorithm[J]. Modern Electronics Technique, 2019, 42(13):47-50.
[20]	江枭宇, 李忠兵, 张军豪, 等. 基于NCS2神经计算棒的车辆检测方法[J]. 计算机工程, 2021, 47(3):298-303.
	JIANG Xiaoyu, LI Zhongbing, ZHANG Junhao, et al. Vehicle detection method based on NCS2 neural computing stick[J]. Computer Engineering, 2021, 47(3):298-303.
[21]	李清, 杨晓辉, 刘振声, 等. 基于灰色聚类分析的输电线路舞动分级预警方案[J]. 电测与仪表, 2020, 57(17):45-51.
	LI Qing, YANG Xiaohui, LIU Zhensheng, et al. Graded early warning scheme for transmission line galloping based on gray cluster analysis[J]. Electrical Measurement and Instrumentation, 2020, 57(17):45-51.
[22]	徐铭铭, 牛荣泽, 谢芮芮, 等. 多源信息融合的配电网重复多发性停电在线监测与预警技术[J]. 山东科学, 2020, 33(4):117-123. doi: 10.3976/j.issn.1002-4026.2020.04.015
	XU Mingming, NIU Rongze, XIE Ruirui, et al. Multi-source information fusion online monitoring and early warning technology for repetitive power outages in distribution networks[J]. Shandong Science, 2020, 33(4):117-123. doi: 10.3976/j.issn.1002-4026.2020.04.015
[23]	陈尚, 蒋毅, 宋珍. 电力电缆火灾风险评价模型与预警信号分级[J]. 南方电网技术, 2020, 14(4):3-7,16.
	CHEN Shang, JIANG Yi, SONG Zhen. Power cable fire risk evaluation model and early warning signal classification[J]. Southern Grid Technology, 2020, 14(4):3-7,16.

场景类别	隐患目标类别	分级预警库
任意场景	烟雾&山火	一级+红色信号①
任意场景	异物(风筝、损坏塑料大棚、损坏布艺广告牌等)	一级+红色信号②
三跨(高速铁路、高速公路、河流)	任意隐患	一级+红色信号③
固定施工-隐患高发段	高危工程机械(吊车、混凝土浇灌车)	一级+红色信号④
高茎植物监控场景	高危&常规工程机械	二级+黄色信号⑤
固定施工-隐患低发段	高危工程机械	二级+黄色信号⑥
固定施工-隐患低发段	常规工程机械(混凝土搅拌车、挖掘机、铲车)	三级+蓝色信号⑦

序号	隐患目标的定义
①	烟雾&山火
②	异物(风筝、损坏塑料大棚、损坏布艺广告牌等)
③	高危工程机械(吊车、混凝土浇灌车)
④	常规工程机械(混凝土搅拌车、挖掘机、铲车)

序号	场景类别的定义
①	三跨(高速铁路、高速公路、河流)
②	固定施工-隐患高发段
③	高茎植物监控场景
④	固定施工-隐患低发段

隐患目标类别	标记类别名称	数据量
烟雾	smoke	750
山火	wildfire	750
异物	foreign-matter	1 500
吊车	crane	3 600
混凝土浇灌车	concrete-pour-truck	500
混凝土搅拌车	concrete-mixer-truck	500
挖掘机	excavator	4 600
铲车	forklift	1 000

原YOLOv3目标检测网络			加入难负样本挖掘策略
隐患目标类别	查全率	查准率	隐患目标类别	查全率	查准率
烟雾	0.831	0.852	烟雾	0.853	0.922
山火	0.923	0.922	山火	0.942	0.965
异物	0.803	0.723	异物	0.806	0.731
吊车	0.872	0.885	吊车	0.914	0.903
混凝土浇灌车	0.891	0.858	混凝土浇灌车	0.926	0.873
混凝土搅拌车	0.908	0.839	混凝土搅拌车	0.956	0.862
挖掘机	0.912	0.901	挖掘机	0.934	0.905
铲车	0.781	0.752	铲车	0.783	0.737