A multilayered urban tree dataset of point clouds, quantitative structure and graph models

Area of Science:

• Urban ecology
• Remote sensing
• Computer vision

Background:

• Urban trees are crucial for human health, especially with increasing urbanization and climate change.
• Remote sensing technologies offer advanced methods for urban tree research.
• The TreeML-Data dataset aims to support these research efforts.

Purpose of the Study:

• To introduce and describe the TreeML-Data dataset.
• To provide a comprehensive resource for studying urban trees.
• To facilitate the development and evaluation of models in related scientific fields.

Main Methods:

• Compilation of labelled point clouds from street scans and individual trees (leaf-off).
• Inclusion of quantitative structure models (QSM), tree structure measurements, and graph models.
• Cross-checking of tree structure measurements and QSM against manual measurements.

Main Results:

• The dataset contains 3,755 leaf-off point clouds of individual trees and 40 street scanning projects.
• Quantitative structure models and tree structure measurements are provided.
• Tree diameter at breast height (DBH) measurements show a 4.3% deviation from manual measurements.

Conclusions:

• The TreeML-Data dataset is a reliable resource for urban tree research.
• The dataset supports diverse scientific disciplines including remote sensing, computer vision, and urban forestry.
• Quality control ensures the dataset's suitability for generating and evaluating scientific models.