第17卷第5期 智能系统学报 Vol.17 No.5 2022年9月 CAAI Transactions on Intelligent Systems Sep.2022 D0:10.11992/tis.202107026 网络出版地址：https:/kns.cnki.net/kcms/detail/23.1538.tp.20220512.1332.002.html 融合外部知识的输电线路多金具解耦检测方法 翟永杰'，王乾铭，杨旭，赵振兵2，赵文清 (1.华北电力大学控制与计算机工程学院，河北保定071003：2.华北电力大学电气与电子工程学院，河北保 定071003) 摘要：为了有效解决输电线路多金具检测过程中存在的目标密集和目标间相互遮挡的问题，提出了融合外部 知识的多目标解耦检测方法(external knowledge decoupling R-CNN,EKDR-CNN)。首先通过深入分析金具数据 集的领域规则和图像信息，提取出共现和空间知识：然后使用图神经网络方法建立共现和空间知识推理模型， 将外部知识进行实例化表达：最后使用解耦模块将金具检测任务以非耦合的方式进行训练和学习。在实验阶 段，对具有14类金具的数据集，进行多种定性和定量实验。对比实验表明，EKDR-CNN的检测效果优于其他 先进目标检测模型，与原有基线模型相比，准确率提高6.6%：定性实验表明算法能够解决目标遮挡的问题，实 现密集目标的检测：消融实验表明，每种模块对模型的检测效果均有一定的提升。 关键词：输电线路；金具；深度学习；目标检测；共现知识；空间知识；知识推理模块；解耦检测 中图分类号：TP18文献标志码：A文章编号：1673-4785(2022)05-0980-10 中文引用格式：翟永杰，王乾铭，杨旭，等.融合外部知识的输电线路多金具解耦检测方法.智能系统学报，2022,17(5)： 980-989. 英文引用格式：ZHAI Yongjie,,WANG Qianming,YANG Xu,.ctal.A multi-fitting decoupling detection method for transmission lines based on external knowledgeJ.CAAI transactions on intelligent systems,2022,17(5):980-989. A multi-fitting decoupling detection method for transmission lines based on external knowledge ZHAI Yongjie',WANG Qianming',YANG Xu',ZHAO Zhenbing,ZHAO Wenqing' (1.School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China;2.School of Elec- trical and Electronic Engineering,North China Electric Power University,Baoding 071003,China) Abstract:This paper proposes a multi-object decoupling detection method based on external knowledge(EKD R-CNN) to effectively solve the problems of object density and mutual occlusion in the process of multi-fitting detection of trans- mission lines.First,the domain rules and image information of the fittings datasets are deeply analyzed to extract the co- occurrence and spatial knowledge;then,graph neural network methods are used to build co-occurrence and spatial knowledge reasoning models,to instantiate and express external knowledge;finally,the decoupling module is em- ployed to train and learn the fittings detection task in an uncoupled way.Multiple qualitative and quantitative experi- ments are conducted on datasets with 14 types of fittings in the experiment phase.The contrast experiment shows that the detection effect of EKD R-CNN is better than that of other advanced object detection models and that compared with the original baseline model,the detection accuracy of the algorithm is increased by 6.6%;the qualitative experiments suggest that the proposed algorithm can solve the problem of object occlusion,and the ablation experiment indicates that each module improves the detection effect of the model to a certain extent. Keywords:transmission line;fitting;deep learning;object detection;co-occurrence knowledge;spatial knowledge; knowledge reasoning module;decoupling detection 近年来，随着电力系统规模和城市建设的不 收稿日期：2021-07-15.网络出版日期：2022-05-13. 基金项目：国家自然科学基金项目(61773160,61871182)：北京 断发展，输电线路逐渐覆盖了苛刻的环境和复杂 市自然科学基金项目(4192055)：河北省自然科学基 地形的区域。输电线路是电力系统的重要组成 金项目(F2021502013,F2020502009,F2021502008). 通信作者：赵振兵.E-mail:zhaozhenbing(@ncepu..edu.cn. 部分。金具是输电线路中广泛使用的铁制或铝制DOI: 10.11992/tis.202107026 网络出版地址: https://kns.cnki.net/kcms/detail/23.1538.tp.20220512.1332.002.html 融合外部知识的输电线路多金具解耦检测方法 翟永杰1 ，王乾铭1 ，杨旭1 ，赵振兵2 ，赵文清1 （1. 华北电力大学 控制与计算机工程学院, 河北 保定 071003; 2. 华北电力大学 电气与电子工程学院, 河北 保 定 071003） 摘 要：为了有效解决输电线路多金具检测过程中存在的目标密集和目标间相互遮挡的问题，提出了融合外部 知识的多目标解耦检测方法 (external knowledge decoupling R-CNN, EKD R-CNN)。首先通过深入分析金具数据 集的领域规则和图像信息，提取出共现和空间知识；然后使用图神经网络方法建立共现和空间知识推理模型， 将外部知识进行实例化表达；最后使用解耦模块将金具检测任务以非耦合的方式进行训练和学习。在实验阶 段，对具有 14 类金具的数据集，进行多种定性和定量实验。对比实验表明，EKD R-CNN 的检测效果优于其他 先进目标检测模型，与原有基线模型相比，准确率提高 6.6%；定性实验表明算法能够解决目标遮挡的问题，实 现密集目标的检测；消融实验表明，每种模块对模型的检测效果均有一定的提升。 关键词：输电线路；金具；深度学习；目标检测；共现知识；空间知识；知识推理模块；解耦检测 中图分类号：TP18 文献标志码：A 文章编号：1673−4785(2022)05−0980−10 中文引用格式：翟永杰, 王乾铭, 杨旭, 等. 融合外部知识的输电线路多金具解耦检测方法 [J]. 智能系统学报, 2022, 17(5): 980–989. 英文引用格式：ZHAI Yongjie, WANG Qianming, YANG Xu, et al. A multi-fitting decoupling detection method for transmission lines based on external knowledge[J]. CAAI transactions on intelligent systems, 2022, 17(5): 980–989. A multi-fitting decoupling detection method for transmission lines based on external knowledge ZHAI Yongjie1 ，WANG Qianming1 ，YANG Xu1 ，ZHAO Zhenbing2 ，ZHAO Wenqing1 (1. School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, China; 2. School of Elec￾trical and Electronic Engineering, North China Electric Power University, Baoding 071003, China) Abstract: This paper proposes a multi-object decoupling detection method based on external knowledge (EKD R-CNN) to effectively solve the problems of object density and mutual occlusion in the process of multi-fitting detection of trans￾mission lines. First, the domain rules and image information of the fittings datasets are deeply analyzed to extract the co￾occurrence and spatial knowledge; then, graph neural network methods are used to build co-occurrence and spatial knowledge reasoning models, to instantiate and express external knowledge; finally, the decoupling module is em￾ployed to train and learn the fittings detection task in an uncoupled way. Multiple qualitative and quantitative experi￾ments are conducted on datasets with 14 types of fittings in the experiment phase. The contrast experiment shows that the detection effect of EKD R-CNN is better than that of other advanced object detection models and that compared with the original baseline model, the detection accuracy of the algorithm is increased by 6.6%; the qualitative experiments suggest that the proposed algorithm can solve the problem of object occlusion, and the ablation experiment indicates that each module improves the detection effect of the model to a certain extent. Keywords: transmission line; fitting; deep learning; object detection; co-occurrence knowledge; spatial knowledge; knowledge reasoning module; decoupling detection 近年来，随着电力系统规模和城市建设的不 断发展，输电线路逐渐覆盖了苛刻的环境和复杂 地形的区域[1]。输电线路是电力系统的重要组成 部分。金具是输电线路中广泛使用的铁制或铝制 收稿日期：2021−07−15. 网络出版日期：2022−05−13. 基金项目：国家自然科学基金项目 (61773160, 61871182)；北京 市自然科学基金项目 (4192055)；河北省自然科学基 金项目 (F2021502013, F2020502009, F2021502008). 通信作者：赵振兵. E-mail：zhaozhenbing@ncepu.edu.cn. 第 17 卷第 5 期 智 能 系 统 学 报 Vol.17 No.5 2022 年 9 月 CAAI Transactions on Intelligent Systems Sep. 2022