PVP-3D: 3D Point Cloud Dataset for Structural Health Monitoring of Photovoltaic Poles

The PVP-3D dataset contains 3D point clouds of support piles used in photovoltaic (PV) panel installations, acquired using aerial LiDAR sensing technology in real-world environments. The raw data was acquired with DJI Zenmusse L2 LiDAR sensor at approximately 20 meters above ground level (AGL), varying scanning modes and return settings, including dual-mesh coverage (north–south and east–west) to ensure comprehensive spatial sampling. Each point cloud is processed to remove noise, filter ground points, and standardize individual samples. The data is stored in Object File Format (OFF) and represents the 3D geometry of a single pile using Cartesian coordinates (X, Y, Z) with around 3,367 points. The dataset comprises 5,400 samples divided into three structural classes: normal, inclined, and twisted, corresponding to different geometric conditions of the piles.