A new SmartForest article is out on how AI can help draw forest stand boundaries, based on the work of Håkon Næss Sandum, Hans Ole Ørka, Oliver Tomic, Erik Næsset, and Terje Gobakken.
Semantic segmentation of forest stands using deep learning
In forest management, dividing forests into stands is an important part of planning, inventories, and decision-making. Today, this work is often done manually by experts interpreting aerial images — a process that can be time-consuming and sometimes subjective.
In this study, the authors explored whether deep learning can support this task by automatically identifying forest stand boundaries using aerial imagery and laser-scanning data.
The results are promising: the AI-based approach was able to produce stand boundaries that followed the same general patterns as expert interpretation, especially in well-managed spruce forests with clear stand structures. This suggests that AI could help make forest inventory work faster, more consistent, and more efficient. At the same time, forests are complex, and the model struggled more in areas with irregular stand shapes and more variation.
This is an exciting step towards more automated and data-driven forest management.
the link to the article: https://www.silvafennica.fi/article/25010