Sparse Tensor Networks for Individual Tree Diameter Estimation in Mixed Forests
Authors: Clara LΓΆΓl, Fabian Ewald Fassnacht, Viacheslav Barkov
Subject: Computer Vision and Pattern Recognition (cs.CV)
arXiv: 2610.37784 [cs.CV] β Submitted 27 October 2026
Abstract
Measurements of tree diameter at breast height (DBH) are fundamental to forest inventories, but manual surveys across entire stands or regions are impractical. UAV-borne laser scanning (ULS) enables scalable, automated forest surveys, yet canopy occlusion makes accurate DBH estimation from ULS point clouds non-trivial. The limited stem-level point density of ULS data constrains approaches that rely on detailed stem reconstruction, such as Quantitative Structure Models (QSMs) and circle fitting. Empirical machine learning methods based on predefined point cloud metrics offer an alternative. Newer deep learning approaches provide an even more principled solution, allowing models to learn directly from point clouds. Sparse tensor networks have shown particular promise, transforming forestry tasks such as tree species classification.
In this work, we investigate whether these strengths can be extended to estimate individual tree DBH from ULS data across four mixed-species temperate forest plots in Germany. We further advance existing sparse tensor networks by introducing TreeNext, a lightweight deep learning architecture for direct individual tree parameter estimation from voxelized LiDAR point clouds. We present a fully automated forest inventory pipeline spanning individual tree instance segmentation through DBH regression, and we quantify the impact of segmentation errors by comparing performance on automatically segmented versus ground-truth instances.
TreeNext achieves an $R^2$ of 0.8355 on automatically segmented trees and 0.9380 on ground-truth segments, outperforming strong tabular baselines including Random Forest and XGBoost. Segmentation quality emerges as the dominant error source in the automated pipeline. TreeNext also proves highly data-efficient, performing well even in small-data scenarios. Our findings establish sparse tensor networks as a practical foundation for automated individual tree DBH estimation from ULS.
Key Contributions
- TreeNext architecture: A lightweight sparse tensor network for direct DBH regression from voxelized LiDAR point clouds.
- End-to-end pipeline: Fully automated processing from individual tree instance segmentation to DBH estimation.
- Error analysis: Quantification of segmentation quality as the dominant error source in automated forest inventories.
- Strong empirical results: Superior performance over Random Forest and XGBoost baselines, with notable data efficiency.
Metadata
- DOI: https://doi.org/10.48550/arXiv.2610.37784
- Version: arXiv:2610.37784v1 [cs.CV]
via ArXiv CV