3D digitizing bsed on tree topology: application to study the variability of apple quality within canopy | [Numérisation en 3D de la topologie des arbres: application à la variabilité de la qualité des pommes à l'intérieur de la canopée]

For several years we have been addressing the problem of designing dedicated databases to quantitatively explore plant architectures. In a first step, plants were decomposed into constituents, at different levels of details, and their topology represented by specific models. In a second step, a geometric description of plant constituents was integrated into the databases. This paper illustrates how these databases can be used to study the variability of apple quality at harvest time, accounting for the relative location of both fruit and vegetative growth within the canopy. Two training systems, i.e. bent and not bent, were compared for "Fuji" variety. Two trees per modality were described. The topology was organized into four scales of including both fruits and different kinds of vegetative shoots. The geometry of each constituent was described with a few parameters that were measured in the field, and its spatial location was recorded using a 3D digitizer (3SPACE~ FASTRAK, Polhemus Inc.). Geometric parameters and spatial coordinates were considered as constituent properties and were attached to the corresponding constituents in the database. At harvest, the main usual quality parameters were analyzed for each individual fruit (height and diameter, refractometric and starch indexes, color, firmness). These parameters were integrated into the database, as variables attached to fruit entities. Thrce-dimensional representations of the digitized apple trees were reconstructed using AMAPmod software. The visual comparison between 3D reconstructions and photographs gave us a natural feedback of the measurements, at whole plant scale as well as at branch scale. The fruit quality variability within the canopy was explored a posteriori building fruit samples, according to different criteria (spatial location, topological location, number of leaves of the associated bourse-shoot) and taking advantage of the multiscale description of the tree. This approach highlighted that fruit characteristics at harvest are influenced by the vegetative growth considered at different scales within the canopy: (1) the total vegetative growth of the whole tree, (2) the relative arrangement of the shoots within the canopy and their location with respect to the fruits,(3) the local vegetative context of the fruits