Adaptive Local Maxima Windows for Tree Segmentation: A Point Process Perspective
Résumé
The growing accessibility of Light Detection And Ranging (LiDAR) data brings out novel perspectives that are crucial for tracking forest growth and enhancing resource management amid climate change. Utilizing these data to propose decision-support tools involves a vital step of segmenting individual trees. Although most of the segmentation algorithms in the literature are designed to be unsupervised, some hyperparameters, which are likely to be site-or species-specific, need to be set by the users, thus increasing the risk of bias. In this direction, based on point process theory, we introduce formal implementation guidelines to refine the window size selection for the class of Local Maxima algorithms, a widely adopted class of methods for tree segmentation. This methodology can also be applied to incomplete plot measurements. We apply this methodology to an open dataset to ensure the reproducibility of the results. The method achieves an F 1 -score in the 55 -90% range, depending on the positioning of the tree within the canopy relative to its neighboring trees, i.e., its dominance status.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |