Doctoral candidate : Van Tho Ngyuen
University : Institut de la Francophonie pour l’Informatique (IFI), Hanoi, Vietnam/ Université de la Rochelle, France
Contract duration : 2015-2018
Research topic — Evaluation de la qualité de bois ronds et de tronc d’arbre à partir de données 3D issues d’un LiDAR terrestre
Research team and supervising scientists —
Research team : Laboratory for Forest and Wood Resource Studies (LERFOB) – Équipe Growth, Yield and Wood Quality Team
PhD supervisors : Francis Colin (LERFOB) / Thierry Constant (INRA) / Alexandre Piboule (ONF)
_______________________
Context and state of the art — This work builds on currently developing technology for measurement tools allowing forest managers and scientists to characterize forests and individual trees. It concerns in particular using systems like terrestrial light detection and ranging (LiDAR), which can provide detailed three-dimensional images to study plots and predict the quality and value of standing trees. Previous studies have addressed how to extract information to study individual trees and their wood using 3D data describing the log (Schuut, 2004), branch characteristics (Klemmt et al.2010) and the scar stage (Stangle et al. 2013), even analyzing bark roughness for species recognition (Othmani et al. 2013).
Objectives and specific questions to be adressed — The objectives include developing algorithms to automatically analyze the relief of the trunk surface to detect areas corresponding to defects, to then define the type of defects and give their dimensions. The corresponding questions are related to: (i) the selection, and adaptation of suitable methods, within the shape recognition field, able to handle the high variability of shape existing for each type of defect, and (ii) for biological issues, a better quantification of the relationships between the outer shape of a defect and its inner characteristics.
Science and innovation issues — The development of automated methods delivering quality criteria involved in the grading of standing trees would bring a undisputable added value to monitoring a valuable forest resource, for different purposes such as forest inventory, ecosystem functioning studies or commercial valuation.
Methodological approaches and expected results — After a critical inventory of the existing methods able to (i) segment a defect zone, (ii) identify its type and (iii) deliver its characteristics, the emerging methods will be integrated into the Computree platform (http://computree.onf.fr). Efficiency of the algorithms will be assessed using an existing database built from logs scanned by terrestrial LiDAR and X-Ray computed tomography. The latter provides reference measurements of defects, especially of knots.
