Large-scale building instance segmentation in diverse urban scenes based on LiDAR data
5 affiliations
- 1Department of Geography, The University of Hong Kong, Hong Kong, 999077, China; The University of Hong Kong Shenzhen Institute of Research and Innovation, Shenzhen, China
- 2Department of Geography, The University of Hong Kong, Hong Kong, 999077, China; Institute for Climate and Carbon Neutrality, The University of Hong Kong, Hong Kong, China
- 3Department of Geography, The University of Hong Kong, Hong Kong, 999077, China
- 4Department of Geography, The University of Hong Kong, Hong Kong, 999077, China; The University of Hong Kong Shenzhen Institute of Research and Innovation, Shenzhen, China; Institute for Climate and Carbon Neutrality, The University of Hong Kong, Hong Kong, China; Corresponding authors.
- 5School of Geospatial Engineering and Science, Sun Yat-sen University, Zhuhai, 519082, China; Corresponding authors.
Abstract
Building 3D profiles constitute critical factors in urban environmental analyses, including assessments of urban heat island effects, carbon emissions, population distribution, and related challenges. LiDAR data provide comprehensive high-resolution 3D information, enabling the acquisition of detailed 3D building profiles. However, in dense urban environments, buildings are in diverse shapes and heights, and are close to one another, making it challenging to extract building instances. In this paper, we propose BuildGroup, a novel building instance segmentation model for both Airborne Laser Scanning (ALS)-based and Unmanned Aerial Vehicles (UAV) datasets. It employs the progressive learning strategy in the feature aggregation, which uses prior 2D knowledge from the Segment Anything Model (SAM) to infer spatial relationships within point clouds and mitigate cross-source discrepancies. An offset enhancement module is introduced to mitigate the effects of diverse point densities, improving instance clustering by compacting point cloud distributions. Additionally, an edge-enhanced loss function prioritizes boundary point accuracy during training to enhance morphological precision. Furthermore, we present a building LiDAR dataset with instance-level building information, which fills a key gap in real-world, large-scale urban point cloud segmentation, particularly for ALS-based sparse façade data. Experimental results validate the framework’s capability to produce accurate building instance information across varying urban densities. All code and data are available at https://github.com/littlebeen/BuildGroup .
Related
- Research on the air supply pattern of stratified air conditioning device in large-scale building space
- Large-scale bioacoustic detection using semantic segmentation: a deep learning framework applied to fin whale calls in ocean-bottom seismometer recordings
- Fixed-ground multisensor monitoring resolves longitudinal vegetation condition and fuel-state structure in Southern California chaparral and oak scenes
- Same computation, different scenes: a single-site comparison of Sentinel-2 L2A composites from Google Earth Engine and Microsoft Planetary Computer
- Looking up, looking down: zenith-to-nadir lidar comparisons for cloud scenes (IPRAL and ATLID/EarthCARE)
BibTeX
@article{sui2026largescale,
title = {{Large-scale building instance segmentation in diverse urban scenes based on LiDAR data}},
author = {Jialu Sui and Luoma Wan and Rui Liu and Longfei Ren and Hongsheng Zhang and Yiping Chen},
journal = {International Journal of Applied Earth Observations and Geoinformation},
year = {2026},
month = oct,
doi = {10.1016/j.jag.2026.105642},
url = {https://doi.org/10.1016/j.jag.2026.105642},
note = {CC BY},
}
RIS
TY - JOUR
TI - Large-scale building instance segmentation in diverse urban scenes based on LiDAR data
AU - Jialu Sui
AU - Luoma Wan
AU - Rui Liu
AU - Longfei Ren
AU - Hongsheng Zhang
AU - Yiping Chen
PY - 2026
DA - 2026/10/30
JO - International Journal of Applied Earth Observations and Geoinformation
DO - 10.1016/j.jag.2026.105642
UR - https://doi.org/10.1016/j.jag.2026.105642
AB - Building 3D profiles constitute critical factors in urban environmental analyses, including assessments of urban heat island effects, carbon emissions, population distribution, and related challenges. LiDAR data provide comprehensive high-resolution 3D information, enabling the acquisition of detailed 3D building profiles. However, in dense urban environments, buildings are in diverse shapes and heights, and are close to one another, making it challenging to extract building instances. In this paper, we propose BuildGroup, a novel building instance segmentation model for both Airborne Laser Scanning (ALS)-based and Unmanned Aerial Vehicles (UAV) datasets. It employs the progressive learning strategy in the feature aggregation, which uses prior 2D knowledge from the Segment Anything Model (SAM) to infer spatial relationships within point clouds and mitigate cross-source discrepancies. An offset enhancement module is introduced to mitigate the effects of diverse point densities, improving instance clustering by compacting point cloud distributions. Additionally, an edge-enhanced loss function prioritizes boundary point accuracy during training to enhance morphological precision. Furthermore, we present a building LiDAR dataset with instance-level building information, which fills a key gap in real-world, large-scale urban point cloud segmentation, particularly for ALS-based sparse façade data. Experimental results validate the framework’s capability to produce accurate building instance information across varying urban densities. All code and data are available at https://github.com/littlebeen/BuildGroup .
SN - 1569-8432
SN - 1872-826X
ER -