抄録
Digital twins of trees have significant applications in urban landscape re-creation, growth simulation, and biomass assessment. Accurate whorl detection—identifying points where multiple branches originate from the trunk—is crucial for precise 3D tree modeling. However, existing methods face limitations in detection accuracy and predominantly rely on 2D projection approaches that result in information loss. This paper describes a new 3D whorl detection method that directly uses point cloud data through toroidal (doughnut-shaped) sectioning around the trunk. The method was evaluated using the TreeNet3D dataset on two species: small-leaved olive (25 trees) and fir (25 trees). Unlike conventional 2D projection methods, this approach preserves 3D spatial information, enabling detection even in dense foliage. The toroidal sectioning technique can be adapted to various tree species by adjusting radius parameters. This research contributes to advancing whorl detection accuracy and establishes a foundation for applications in tree age estimation, timber value assessment, and high-precision 3D tree modeling.
https://ror.org/04chrp450 
