多尺度邻域特征下的机载LiDAR点云电力线分类

利用机载激光雷达技术三维测量精度高且获取快速的优点进行电力线自动分类提取已成为点云数据处理与电力应用的重要领域。针对电力线分类模型的自动化和高精度需求，本文提出了基于三维多尺度邻域特征的机载LiDAR点云电力线分类提取模型框架，主要包括4个步骤：电力线候选点滤波、多尺度邻域类型选取、形状结构特征提取和支持向量机分类。通过对2个复杂城市区域的试验数据集和8种不同邻域类型的详细结果对比分析，发现基于多尺度圆球邻域形状结构特征的分类模型结果准确率、召回率和质量分别达到97%、94%和93%，同时整体处理时间在2个试验数据中分别从366、256 s减少到274、160 s。试验结果表明，该方法在多种复杂城市场景中能够实现机载LiDAR数据的电力线较高精度分类提取。

Power line classification from airborne LiDAR data via multi-scale neighborhood features

With the rapid development of 3D accurate measurement technology of airborne LiDAR, automatic extraction of power lines from airborne laser point clouds has become an important topic in point cloud data processing and transmission line management. In this paper, we present an automated and versatile framework for power line classification, which consists of four steps:power line candidate point filtering, multi-scale neighborhood type selection, feature extraction based on geospatial structure, and SVM classification. To comprehensively evaluate the proposed algorithm, we calculate each point's feature based on eight levels of scales. Two datasets demonstrate that classification results reach up to 97%, 94%, and 93% in terms of precision rate, recall rate and overall quality. The whole processing time also decreases from 366 s, 256 s to 274 s, 160 s, respectively. Experimental results show that this method can achieve high-precision classification of power lines in complex urban environment.