| 激光雷达的多源数据融合点云分类算法研究 |
 点此下载全文 |
| 引用本文：杨治，曾寰，涂起龙.激光雷达的多源数据融合点云分类算法研究[J].计算技术与自动化,2023,(2):100-107 |
| 摘要点击次数: 189 |
| 全文下载次数: 0 |
|
|
| 中文摘要:根据单档输电线空间分布特性，提出了改进随机采样一致的输电线点云分割方法。首先优化初始样本点选择原则、引入最小二乘原理参数求解等改进策略，提高了随机采样一致性算法输电线模型重建精度；然后以直线-抛物线方程为单根输电线识别的约束条件，利用逐根提取方式实现输电线激光点云分割。选择两组典型代表性的机载激光点云数据进行实验分析，该方法有效解决了数据缺失、点云噪声等复杂背景环境的输电线激光点云分割，准确率、召回率和整体精度最小值分别为99.19％、99.25％、99.10％。较之已有方法，本文方法具有点云分割精度高、算法普适性强的优势；随机采样一致性（RANSAC）算法是常见的激光点云分割方法，但该算法推广至输电线场景时存在点云分割效率低、抗噪性差等不足，不利于高精度的输电线模型重建及后续线路风险检测。 |
| 中文关键词:机载激光雷达 改进随机一致性采样 输电线 点云分割 |
| |
| Research on Point Cloud Classification Algorithm of Multi-Source Data Fusion for Lidar |
|
|
| Abstract:According to the spatial distribution characteristics of single-file transmission lines, this paper proposes a method to improve the cloud segmentation of transmission lines with consistent random sampling. Firstly, the initial sample point selection principle and the introduction of the least squares principle parameter solution and other improvement strategies were introduced to improve the reconstruction accuracy of the transmission line model of the random sampling consistency algorithm. The linear-parabolic equation is then used as the constraint for the identification of a single transmission line, and the laser point cloud segmentation of the transmission line is realized by means of root-by-root extraction. Two sets of typical representative airborne laser point cloud data were selected for experimental analysis.This method effectively solves the laser point cloud segmentation of transmission line in complex background environments such as data deletion and point cloud noise, and the minimum values of accuracy, recall rate and overall accuracy are 99.19%, 99.25% and 99.10%, respectively. Compared with the existing methods, it has the advantages of high precision of point cloud segmentation and strong universality of the algorithm. Random consistent sampling (RANSAC) is a common laser point cloud segmentation method, but when the algorithm is generalized to the transmission line scene, there are shortcomings such as low point cloud segmentation efficiency and poor noise resistance, which is not conducive to high-precision transmission line model reconstruction and subsequent line risk detection. |
| keywords:airborne lidar improved random consistent sampling power line point cloud segmentation |
| 查看全文 查看/发表评论 下载pdf阅读器 |
