Development of a tree shape reconstruction technique using Blender and terrestrial laser data

Abstract

This study proposes a method for reconstructing realistic tree shapes using Blender and terrestrial laser scanning data. With the growing importance of digital twins in fields such as urban planning and disaster prevention, accurate three-dimensional tree models are essential. While terrestrial laser scanning can capture tree crowns, occlusion by branches and leaves often results in missing internal structure. To address this limitation, we employed the Blender add-on Mtree to generate tree models whose parameters were optimized through numerical methods. Validation was conducted through two-dimensional contour similarity analysis with OpenCV’s algorithm named matchShapes, vertical distribution of maximum distances at different crown heights, and voxel-based comparisons of internal structures. Results showed the specific species such as Katsura Tree (Cercidiphyllum japonicum) and Japanese black pine (Pinus thunbergii Parl) achieved high reproducibility through the both two- and three-dimensional evaluations, while closed upper crown trees such as poplar (Populus) were more difficult to reconstruct due to Mtree’s parameter constraints. Furthermore, voxel analysis revealed that Blender models contained more voxels than the laser data, suggesting internal structures were successfully successful reconstructed. This study demonstrates that optimization with Blender-based modeling can complement laser measurements and reduce field survey labor and costs to measure tree structure.