基于关键帧提取与图像识别的电网设施数字化重建  

3D Reconstruction of Power Grid Facilities Based on Keyframe Extraction and Image Recognition

作　　者：陈志永 韩璟琳 刘洋 李光毅 蒋雨 蔡瑞天 CHEN Zhiyong;HAN Jinglin;LIU Yang;LI Guangyi;JIANG Yu;CAI Ruitian(State Grid Hebei Electric Power Co.Ltd.,Economic and Technical Research Institute,Shijiazhuang 050021,China;State Grid Hebei Electric Power Co.Ltd.,Shijiazhuang 050000,China;State Grid Hebei Electric Power Co.,Ltd.,Xiong'an New Area Power Supply Company,Baoding 071000,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]国网河北省电力有限公司经济技术研究院,河北石家庄050021 [2]国网河北省电力有限公司,河北石家庄050000 [3]国网河北省电力有限公司雄安新区供电公司,河北保定071000 [4]华北电力大学电气与电子工程学院,北京102206

出　　处：《中国电力》2024年第9期217-223,共7页Electric Power

基　　金：国家电网有限公司科技项目(5100-202113564A-0-5-SF)。

摘　　要：提出一种基于视频关键帧提取与图像识别的电网设施数字化重建方案。首先,通过移动作业终端拍摄待重建设施的视频序列,提取视频的关键帧并利用YOLOv5网络识别并提取电力设施锚框去除环境背景,构造图像集;然后,以运动恢复结构为原理进行三维重建,提出改进定向快速旋转算法提高相机位姿计算精度,进而实现电网设施三维模型重建;最后,以变压器为例,验证所提方法可以有效完成电力设施三维模型的重建。A digital reconstruction scheme for power grid facilities based on video keyframe extraction and image recognition is proposed.Shoot a video sequence of the facility to be reconstructed through a mobile homework terminal,extract key frames from the video,and use YOLOv5 network to recognize and extract anchor frames of power facilities to remove environmental background,constructing an image set;Then,based on the principle of motion recovery structure,3D reconstruction is carried out,and an improved directional fast rotation algorithm is proposed to improve the accuracy of camera pose calculation.Thereby achieving the reconstruction of 3D models of power grid facilities;Finally,taking common transformers in the power grid as an example,it is verified that the proposed method can effectively complete the 3D model reconstruction of power facilities.

关 键 词：数字孪生 关键帧提取 图像识别 

分 类 号：TM73[电气工程—电力系统及自动化] TP391.41[自动化与计算机技术—计算机应用技术]