Tree extraction in open-pit mine restoration areas using unmanned aerial vehicle LiDAR and you only look once-based deep learning
- 1Nanjing Center of Geological Survey, China Geological Survey
Abstract
Accurate monitoring of tree recovery in restored open-pit mines is essential for evaluating rehabilitation effectiveness and guiding post-closure management. Using the Huafeng open-pit limestone quarry (Qingyang County, Anhui Province) as a case study, we integrated 2025 unmanned aerial vehicle (UAV)-borne LiDAR with 2024 UAV red, green, and blue (RGB) orthophotos to develop a cross-temporal tree extraction and forest-change assessment workflow. A 0.5 m canopy height model (CHM) was derived from high-density LiDAR point clouds, and individual crown tops and canopy boundaries were delineated by combining local maximum detection with watershed segmentation, yielding 5,863 individual trees under predefined height and crown-size constraints. These LiDAR-derived tree objects were projected onto RGB imagery to generate training/validation labels for a You Only Look Once (YOLO)v9 detector, which was applied to 2024 orthophotos to map tree distribution and forest/non-forest patterns. The net forest increase of approximately 1.34 ha represented a 15.8% relative increase compared with 2024. Non‑forest areas declined by 24.1%, from 5.56 ha to 4.22 ha. Newly established forest was concentrated on the pit‑bottom platform, terraced slopes with topsoil cover, and the former industrial site, whereas persistent non‑forest patches were predominantly associated with steep bare‑rock slopes. These contrasting spatial patterns highlight priority areas for targeted soil conservation and replanting, and demonstrate the value of multi‑source UAV remote sensing for reproducible restoration monitoring.
Related
- Trajectory tracking control of unmanned aerial vehicle with disturbance observer and robust adaptive neural dynamic surface
- High-resolution 3D turbulence measurements with different quadcopter uncrewed aerial systems: verification and validation for single- and multi-vehicle measurements
- Measuring atmospheric microplastic composition and occurrence from an uncrewed aerial system
- Phototrophy improves the aerial fitness in a photoheterotrophic Methylobacterium isolated from clouds
- Seatizen Atlas: a collaborative dataset of underwater and aerial marine imagery
BibTeX
@article{wang2026tree,
title = {{Tree extraction in open-pit mine restoration areas using unmanned aerial vehicle LiDAR and you only look once-based deep learning}},
author = {Shangxiao Wang and Shengjun Xiao and Xiaonan Niu and Yanwei Sun and Leli Zong and Yi Liu and Ming Zhang},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-61759-1},
url = {https://doi.org/10.1038/s41598-026-61759-1},
note = {CC BY},
}
RIS
TY - JOUR
TI - Tree extraction in open-pit mine restoration areas using unmanned aerial vehicle LiDAR and you only look once-based deep learning
AU - Shangxiao Wang
AU - Shengjun Xiao
AU - Xiaonan Niu
AU - Yanwei Sun
AU - Leli Zong
AU - Yi Liu
AU - Ming Zhang
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-61759-1
UR - https://doi.org/10.1038/s41598-026-61759-1
AB - Accurate monitoring of tree recovery in restored open-pit mines is essential for evaluating rehabilitation effectiveness and guiding post-closure management. Using the Huafeng open-pit limestone quarry (Qingyang County, Anhui Province) as a case study, we integrated 2025 unmanned aerial vehicle (UAV)-borne LiDAR with 2024 UAV red, green, and blue (RGB) orthophotos to develop a cross-temporal tree extraction and forest-change assessment workflow. A 0.5 m canopy height model (CHM) was derived from high-density LiDAR point clouds, and individual crown tops and canopy boundaries were delineated by combining local maximum detection with watershed segmentation, yielding 5,863 individual trees under predefined height and crown-size constraints. These LiDAR-derived tree objects were projected onto RGB imagery to generate training/validation labels for a You Only Look Once (YOLO)v9 detector, which was applied to 2024 orthophotos to map tree distribution and forest/non-forest patterns. The net forest increase of approximately 1.34 ha represented a 15.8% relative increase compared with 2024. Non‑forest areas declined by 24.1%, from 5.56 ha to 4.22 ha. Newly established forest was concentrated on the pit‑bottom platform, terraced slopes with topsoil cover, and the former industrial site, whereas persistent non‑forest patches were predominantly associated with steep bare‑rock slopes. These contrasting spatial patterns highlight priority areas for targeted soil conservation and replanting, and demonstrate the value of multi‑source UAV remote sensing for reproducible restoration monitoring.
SN - 2045-2322
ER -