激光雷达数据下架空输电线路点云场景分割及净空入侵检测

Segmentation and clearance inspections on overhead transmission powerline corridor based on LiDAR point clouds

无人机搭载激光雷达(LiDAR)所获取的输电线路点云数据,有效提高了线路净空巡检的效率。然而,数据处理中存在大场景点云分类精度不高、净空隐患检测效率低的问题。依托典型架空输电线路激光雷达点云数据,本文提出了一种面向净空检测的轻量点云分割网络(LP网络),在多股电力线并行情况下建立二次函数,模拟电力线中线,结合规范快速检测净空隐患。通过两条输电线路测试,LP网络的平均交并比(IoU)达到0.895,所需时间为9.53 s;单座电塔净空入侵检测时间在1 min内;在交叉跨越区域,净空入侵检测也取得了准确结果。研究成果可用于自动准确识别线路隐患,为线路巡检提供决策支持。

The point clouds of transmission powerlines collected by UAV equipped light detection and ranging(LiDAR)effectively prompts the efficiency of clearance inspection in powerline patrol.However,there are problems in data processing including low classification performance of large-scene point clouds and low efficiency of clearance detection.A Light-weight Powerline(LP)network for classifying point clouds is investigated based on the practical LiDAR data from typical overhead transmission lines.A quadratic curve function of a simulated center powerline is fitted to represent the multiple powerlines,and the clearance is quickly measured between the center powerline and ground objects according to specifications.Two transmission power lines were applied to check the performance of the proposed method.The mean intersection over union(IoU)of LP network is 0.895 and the testing time is 9.53 s.The running time of measuring the clearance between two transmission towers is less than 1 min.On important areas such as crossing other powerlines and railways,the clearance detection also achieved satisfied results.The research results could identify clearance obstacles automatically and accurately,and could provide decision support for powerline inspection.