Counting Trees from Satellite Imagery with Noisy Supervision

2026-06-23

AI Summary

This work addresses the challenges of large-scale individual tree counting, including ambiguous crown boundaries, high costs of manual annotation, and substantial label noise in LiDAR data. The authors propose a density-matching framework based on unbalanced optimal transport, formulating tree counting as a spatial density estimation problem under noisy supervision. A reliability-aware self-correction mechanism is introduced to iteratively refine the supervisory signal using transport residuals, enabling a unified approach to both isolated tree localization and dense forest density estimation. By integrating multi-source satellite imagery and LiDAR data, the method achieves state-of-the-art performance on the TinyTrees benchmark—a dataset spanning three continents and three satellite sensors—significantly outperforming existing detection-, regression-, and distribution-matching-based baselines, thereby demonstrating its effectiveness and robustness.

Abstract

Counting individual trees is a fundamental task for environmental monitoring, yet remains largely unexplored with satellite imagery. At these resolutions, isolated trees may still be identifiable, but crown boundaries become ambiguous in dense forests, making the notion of an individual tree inherently ill-defined. Moreover, large-scale manual annotations of individual trees are prohibitively expensive. While scalable supervision can be derived from airborne LiDAR, the resulting annotations are noisy and difficult to exploit effectively. We address these challenges by formulating tree counting as a spatial density matching problem supervised through Unbalanced Optimal Transport. This formulation naturally accommodates both precise localization of isolate trees and robust density estimation in dense forests. We further introduce a self-correction mechanism that leverages transport residuals to progressively refine noisy supervision during training. We evaluate our approach on TinyTrees, a new benchmark spanning three continents and three satellite sensors, comprising over 215 million tree annotations (including 773K manually verified instances) across 23,000 sq.km. Our method consistently outperforms detection-based, regression-based, and transport-based distribution-matching baselines, demonstrating the effectiveness of unbalanced transport and reliability-aware supervision for large-scale tree counting from satellite imagery. Code, data and models are available at https://github.com/dgominski/treematch.

Problem

• tree counting
• satellite imagery
• noisy supervision
• dense forests
• large-scale annotation

Innovation

• Unbalanced Optimal Transport
• Spatial Density Matching
• Noisy Supervision
• Self-correction Mechanism
• Tree Counting