Virtual plant architectures give opportunities to understand plant growth and interaction with their environment (temperature, light, pest and disease propagation, etc.). Specific virtual models called Functional-Structural Plant Models (FSPM) combine detailed representation of the structure and the geometry of plants with functional processes of their development. In this talk, I will present my work on the definition of such models, starting from their acquisition to their simulation. I will first present a method to reconstruct plant architecture from laser scans. Recently, we developed a software pipeline for processing point clouds acquired using terrestrial laser scanner to reconstruct the structure of trees. We currently use it, through a graphical editor we built, to reconstruct different cultivars of apple trees. To validate our approach, we also developed an evaluation pipeline that makes it possible to compare the structure and the geometry of two reconstructions. Using this method, we compare our reconstruction with different methods of the literature. From these structures, analysis can also be derived to formulate hypotheses on the development of the plants. To assess such hypotheses, we also developed a simulation framework, namely L-Py, which encapsulates the L-systems formalism in the Python modelling language and makes it possible to build comprehensive model of plant growth. I will illustrate the use of this framework with an on-going work on the definition of a developmental model of mango tree architecture. The appearance of the different organs (growth units, inflorescences, fruits) of the mango tree are know to be affected by structural and temporal architectural factors. We defined developmental rules accordingly and used generalized linear models to estimate the corresponding probabilities. Using such model, we were able to assess the influence of the structural factors on the development of the tree. All the presented methods are available in the OpenAlea platform and can be combined and adapted for other modelling scenario.

Contact :

Frédéric BOUDON