基于U-Net双层融合模型的电气二次图纸图元匹配算法  

Element Matching Algorithm for Electrical Secondary Drawings Based on U-Net Dual-layer Fusion Model

作　　者：王磊 田雨 邹瑞睿 张礼波 WANG Lei;TIAN Yu;ZOU Ruirui;ZHANG Libo(Liupanshui Power Supply Bureau of Guizhou Power Grid Co.,Ltd.,Liupanshui 553001,China)

机构地区：[1]贵州电网有限责任公司六盘水供电局,六盘水553001

出　　处：《自动化与仪表》2025年第4期126-131,共6页Automation & Instrumentation

摘　　要：变电站二次系统包含大量电气设备和复杂逻辑控制关系,图纸种类繁多、数量庞大,导致管理困难。针对以上问题,研究提出了一种基于U-Net和传统图像处理算法相结合的图元匹配算法,以改善图纸智能识别效果。该算法采用双层结构,利用U-Net网络匹配端子排图文字标签、导线和元器件,并结合线条骨架矢量化方法提取图纸线条信息,解决线条、元器件与文字标签的匹配问题。实验表明,该算法能有效提取文本标签与导线和元器件间的对应关系,建立匹配关系,提升图纸理解效率,为变电站工作智能化提供支持,具有重要应用价值。The secondary system of substations contains a large number of electrical devices and complex logical control relationships,with a wide variety and vast quantity of drawings,leading to management difficulties.In light of the aforementioned issues,this study proposes a graphic element matching algorithm that integrates U-Net with traditional image processing techniques to enhance the effectiveness of intelligent diagram recognition.The algorithm adopts a dual-layer structure,utilizing the U-Net network to match terminal block text labels,wires,and components,while incorporating line skeleton vectorization methods to extract line information from drawings,thereby solving the matching problem between lines,components,and text labels.Experiments demonstrate that the algorithm can effectively extract the corresponding relationships between text labels,wires,and components,establish matching relationships,improve drawing comprehension efficiency,and provide strong support for the intelligentization of substation operations,holding significant practical application value.

关 键 词：深度学习 图元匹配 线条提取 图像理解 变电站智能化 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]