激光点云输电线精细提取的残差聚类法

Extraction of power lines from laser point cloud based on residual clustering method

针对输电线点云数据中存在缺失、噪声等复杂环境,提出了一种基于模型残差聚类的激光点云电力线精细提取方法。首先根据归一化高程阈值分割去除近地面点,在此基础上,采用自适应维度特征和方向特征粗提取电力线点;然后以抛物线模型为约束条件,采用改进的建模方法,确定模型残差并对其进行密度聚类,根据聚类结果实现单根电力线精细提取;最后讨论了关键参数的选择对提取结果的影响。两景实测数据试验表明:该方法能快速实现点云部分缺失、噪声干扰等复杂环境下的电力线精细提取,无须电力线数目、点云密度等先验知识,对不同类型分裂导线提取均具有很好的适用性。单根电力线提取准确率达99.17%以上,模型误差最大值为0.167 m,中误差最大值为0.079 m。

Aiming at the complex environment such as missing and noise in power line cloud data,a precise power line extraction method based on model residual clustering from LiDAR point is proposed.Firstly,the near-ground points are removed according to the normalized elevation threshold segmentation.The power line points are roughly extracted using adaptive dimension features and directional features.Secondly,the improved modeling method is adopted to determine the model residual error with the constraint condition of the parabolic model.The result obtained by density clustering on the model residual error is used to extract the single power line point.Finally,the influence of the selection of key parameters on the extraction results is discussed.Two experimental results show that the method can quickly extract power line from point cloud with partial missing and noise interference,without prior knowledge such as the number of power lines and density of point cloud,etc.Which has good applicability for different types of bundle conductor extraction.the accuracy of single power line extraction is more than 99.17%,the maximum error of model fitting is 0.167 m,and the maximum mean square error of model fitting is 0.079 m.