Research areas
3D Vegetation Reconstruction — Digital Twins
Together with our collaborators, we have developed algorithms for AI-based 3D tree reconstruction from photographs, point clouds, inverse procedural models of trees as L-systems, and developmental models programmed as simulation systems. The input to this problem is an existing tree represented as an image, point cloud, or geometry. The goal is to find a functional-structural plant model (digital twin) that can be used in simulating “what-if” scenarios. The figure below from inverse procedural modeling of trees shows such a scenario. The input tree on the left is analyzed, and the simulation parameters that can reproduce it are found. Having the digital twin, the tree can be regrown in a different environment (right).
See the video of the reconstruction from images: