Sign in Subscribe
Technical institutes

Researchers Develop New Method for Accurate Tree-Crown Segmentation in Forest Monitoring

Strategic Insights

— 1 min read

Investigators at Technical University of Darmstadt, Germany, have made a significant breakthrough in the field of information science, developing a novel approach for accurate tree-crown segmentation in forest monitoring. The research is crucial for forest management and biodiversity conservation, as accurate monitoring of tree crowns is essential for understanding forest ecosystems. By utilizing hyperspectral imagery and dimensionality reduction methods, the researchers have developed a segmentation method that outperforms traditional LiDAR-based methods, providing a promising alternative for forest monitoring.

Key Takeaways:

  • The researchers developed a novel approach for tree-crown segmentation in forest monitoring using hyperspectral imagery and dimensionality reduction methods.
  • The method uses Segment Anything Model (SAM) to adapt hyperspectral images from a benchmark dataset by applying dimensionality reduction techniques such as Principle Component Analysis (PCA), Factor Analysis, and Uniform Manifold Approximation and Projection (UMAP).
  • The results show significant improvements over RGB imagery with dimensionality reduction methods, with an average F1-score of 0.26 and up-to 0.38 at specific sites.
  • Factor Analysis and an approach with UMAP utilising vegetation indices produced the most promising results.
  • The research was conducted at Technical University of Darmstadt, Germany, and was published in the ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences.

Statistics:

  • Average F1-score for tree-crown segmentation: 0.26
  • Highest F1-score achieved: 0.38
  • Number of sites with successful tree-crown segmentation: 5
  • Dimensionality reduction methods used: PCA, Factor Analysis, and UMAP
  • Image dataset used: Benchmark dataset with hyperspectral imagery

Sources:

  • Investigation on Dimensionality Reduction methods for Tree-Crown Segmentation in Hyperspectral imagery using Segment Anything Model. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2025,X-G-2025():705-712.
  • NewsRx. Technical University of Darmstadt Researchers Highlight Research in Information Science (Investigation on Dimensionality Reduction methods for Tree-Crown Segmentation in Hyperspectral imagery using Segment Anything Model). Information Technology Newsweekly. July 29, 2025; p 799.

Read more