摘要
“智能电网”的构建是现代化社会电力系统建设规划与发展的重中之重,输电走廊的数字化作为其中的重要一环,然而,现阶段输电走廊数字化仍需耗费大量人工,其自动化水平亟待提升。而输电杆塔的建模效率能否提升成为了决定整个输电线路三维建模效率的主要因素。为解决杆塔建模效率的问题,本文作为模型驱动的输电杆塔建模方法的前期工作,提出了一种基于PCA算法与YOLOv5目标检测算法结合的塔型识别方法。该方法首先使用PCA算法提取电塔的主方向,通过主方向投影实现了杆塔点云的二维转换,并将主方向投影图作为网络模型的输入,进而通过YOLOv5算法实现了电塔的塔型识别。实验表明,本文所提出的方法能够实现对输电杆塔点云数据实时精确的塔型识别:塔型识别的平均精确度map@0.5为0.867,识别速度为0.02 s。
The construction of “smart grid” is the top priority of the construction planning and development of the modern social power system, and the digitization of the transmission corridor is an important part of it. However, at this stage, the digitization of the transmission corridor still requires a lot of labor, and its automation level needs to be improved urgently. Whether the modeling efficiency of transmission towers can be improved has become the main factor that determines the efficiency of 3D modeling of the entire transmission line. In order to solve the problem of tower modeling efficiency, as the preliminary work of model-driven transmission tower modeling method, this paper proposes a tower type identification method based on the combination of PCA algorithm and YOLOv5 target detection algorithm. The method first uses the PCA algorithm to extract the main direction of the tower, realizes the two-dimensional transformation of the tower point cloud through the main direction projection, and uses the main direction projection map as the input of the network model, and then realizes the tower shape of the tower through the YOLOv5 algorithm. identify. Experiments show that the method proposed in this paper can realize real-time and accurate tower type recognition for point cloud data of transmission towers: the average accuracy of tower type recognition map@0.5 is 0.867, and the recognition speed is 0.02 s.
出处
《测绘科学技术》
2022年第3期161-172,共12页
Geomatics Science and Technology
