ULS-TLS Forest Registration Benchmark (ForestReg)

1. Dataset Description

This is the first multi-platform benchmark dataset for Unmanned Aerial Vehicle Laser Scanning (ULS) and Terrestrial Laser Scanning (TLS) point cloud registration in complex forest environments. Integrating ULS and TLS data provides a comprehensive 3D understanding of forest structures, but robustly registering these datasets remains a significant challenge due to differing perspectives, point densities, and occlusions. This benchmark aims to facilitate the development and comparative evaluation of registration algorithms tailored for such scenarios.

The dataset encompasses 17 plots from seven diverse regions, totaling approximately 1.56 billion points. It features 9 tree specics (including mixed, eucalyptus, fir, willow, poplar, birch, larch, pinus densata, and pinus sylvestris) and terrains. Each plot includes co-acquired ULS and TLS point clouds, along with manually verified ground truth transformation matrices for aligning ULS to TLS data.

The dataset is categorized into three difficulty levels (Low, Medium, High) based on calculated overlap ratios and rigid overlap (specifically focusing on tree trunks), providing a structured way to assess algorithm robustness across varying complexities.

Key Features:

• Multi-platform: First multi-platform point cloud registration dataset.
• Forest-specific: Designed for the unique challenges of forest environments.
• Large-scale & Diverse: 17 plots, 7 regions, 9 forest types, ~1.56 billion points, varied terrain.
• Difficulty Categorization: Enables nuanced algorithm performance analysis.
• Ground Truth: Manually verified 4x4 transformation matrices for accurate evaluation.

2. Dataset Structure

The dataset is organized into 17 plots. For each plot, the following data are typically provided:

• plot_id: A unique identifier for the plot (e.g., S01, S02, ..., S17).
• ULS point cloud data (e.g., ULS_norm.las)
• TLS point cloud data (e.g., TLS_norm.las)
• Ground truth transformation matrix (e.g., groundtruth.txt) aligning TLS to ULS.
• Detailed metadata for each plot is provided below.

Detailed Plot Information

The following table provides detailed information for each plot in the dataset (corresponding to Table V in the publication).

3. Dataset Creation

Curation Rationale

The dataset was created to address the lack of reliable multi-platform benchmark datasets for ULS-TLS point cloud registration specifically in complex forest environments. This hinders robust comparative studies and the development of specialized registration algorithms.

Source Data

• ULS Data: Acquired using a DJI M600 Pro UAV equipped with a RIEGL mini VUX-1 scanner system (360° scanning angle, five returns per pulse, flying altitude 50m, speed 5 m/s).
• TLS Data: Collected using a RIEGL VZ-1000 terrestrial laser scanner (Time of Flight, multi-echo).

Annotations & Categorization Workflow

The ground truth transformation matrices were generated through a meticulous process. The dataset was then categorized based on difficulty. The overall workflow is illustrated below (corresponding to Figure 1 in the publication):

Annotation Process:

1. Initial coarse registration using five different algorithms, followed by ICP (Iterative Closest Point) for fine registration.
2. Visual assessment of registration quality in CloudCompare software.
3. If visual alignment of trunks and branches was satisfactory, the optimal result was selected as ground truth.
4. Otherwise, corresponding points were manually selected for registration to establish the ground truth.

The dataset creators are the authors of the paper.

4. Citation Information

For questions about the dataset, please contact Wangjun Liu at [email protected]. If you use this dataset in your research, please cite the following paper, :

@ARTICLE{Liu2025ForestReg,
  author={Liu, Wangjun and Nie, Sheng and Xia, Shaobo and Wang, Cheng and Wang, Jinliang and Xi, Xiaohuan and Cheng, Feng},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  title={Benchmarking ULS-TLS Point Cloud Registration Algorithms in Forest Environments},
  year={2025},
  volume={},
  number={},
  pages={1-15},
  doi={10.1109/TGRS.2025.3569061}}