Adaptive voxelization and hierarchical object extraction for multiline LiDAR point clouds in geospatial surveying and mapping

Abstract

Raw multi-line LiDAR data in geospatial surveying are characterized by complex preprocessing, substantial redundancy, and limited efficiency in hierarchical object extraction. This paper presents a mapping-oriented framework for adaptive voxelization and hierarchical object extraction directly from raw LiDAR network packets. The framework includes packet decoding, point cloud denoising, adaptive downsampling, ground segmentation, non-ground object clustering, and DEM/DSM generation. Statistical outlier removal is used to suppress noise, and a distance-aware voxelization strategy is designed to balance compression efficiency and geometric fidelity. Ground points are extracted by RANSAC-based plane fitting, and non-ground objects are organized through Euclidean clustering. Regular gridding is further applied to generate DEM and DSM products. The results demonstrate the feasibility of the proposed workflow for reducing point redundancy, retaining the main scene structure, and supporting subsequent mapping-oriented representation. The study is mainly positioned as a practical and lightweight processing route from raw multi-line LiDAR packets to DEM/DSM products.