基于BIM的变电站仪表检测研究  

Research on BIM for Automatic Panel Detection in Transformer Substation

作　　者：蔡航 邵家玉[1,2] 张宁 Cai Hang

机构地区：[1]东南大学自动化学院,江苏南京210096 [2]复杂工程系统测量与控制教育部重点实验室,江苏南京210096 [3]东南大学教育部ITS工程研究中心,江苏南京211189

出　　处：《工业控制计算机》2025年第1期19-21,共3页Industrial Control Computer

摘　　要：变电站内部环境危险复杂,仪表图像数据采集难度大,图像检测算法模型泛化能力弱。为解决此问题,提出了基于建筑信息模型的仪表检测数据增强方法。首先将建筑信息模型数据渲染为三维模型,采集图像,生成无标注数据集;再根据仪表参数计算仪表的三维几何体在二维图像上的投影,为数据集添加标注;接着为图像添加真实纹理,提高数据集的泛化能力;最后使用基于现实图像数据训练的仪表检测模型对生成的数据集进行验证。结果表明,提出的方法可基于建筑信息模型信息生成海量仿真数据集,降低数据采集和数据标注的难度与工作量。The internal environment of the substation is dangerous and complex,and it is difficult to collect instrument image data.The generalization ability of the automatic reading algorithm model is weak.To address this issue,a BIM based instrument detection optimization algorithm has been proposed.Firstly,render the BIM data into a 3D model,collect images,and generate an unlabeled dataset.Then,calculate the projection of the three-dimensional geometry of the instrument on the two-dimensional image based on the instrument parameters,and add annotations to the dataset.Finally,the instrument detection model trained on real-world image data is used to validate the generated dataset.The results show that the proposed method can generate massive simulation datasets based on BIM information,reducing the difficulty and workload of data collection and annotation.

关 键 词：BIM 变电站 目标检测 图像增强 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]