Authors:
Like Gobeawan
1
;
Xuan Liu
1
;
Chi Lim
1
;
Venugopalan Raghavan
1
;
Joyjit Chattoraj
1
;
Jan Schindler
2
and
Feng Yang
1
Affiliations:
1
Institute of High Performance Computing, Agency for Science, Technology and Research, 1 Fusionopolis Way #16-16 Connexis, Singapore, Singapore
;
2
Manaaki Whenua - Landcare Research, Wellington, New Zealand
Keyword(s):
Digital Twin, Tree Species Model, Point Cloud Data, Tree Branch Reconstruction, Procedural Modeling, Optimisation.
Abstract:
3D digital twin trees for a city-scale have been limited to low-resolution, static shape models due to challenges in automation/scalability, cost performance, tree growth dynamics, species complexities and compatibilities with simulations and virtual city platforms. To address those challenges for high-resolution tree models, we propose an automated workflow of generating large-scale, lightweight, dynamic digital-twin tree species models from point cloud data. Species digital twins are modelled as detailed hierarchical branch structures by solving for all species profile parameters through stages of branch reconstruction from point cloud data, species profiling by machine learning, tropism transfer, optimisation and species growth modelling based on botany and limited field survey. We show that the generated high-resolution tree models can be lightweight while representing their true species characteristics and dynamic botanical architecture (branching patterns and growth processes).