If a piece of valuable wood has some defects, it cannot be used to make certain objects, but some others can be made avoiding the defect zone.
Aim of the project is to utilize at best the wood blocks by reducing discards. This is performed by introducing a new software module in the mechanical process that from walnut blanks conducts to the production of specific pieces. The software receives in input the slices acquired by a tomograph through which the wood transits before being passed to the next machine of the production line. A 3D shape of the wood block is reconstructed and a deep-learning neural net classifies the voxels into good or a defect class. A priori knowledge of the 3D designs of each single product and its characteristics permits the algorithm to choose the most suitable one to be built from that block.
Duration: From November 2018 to January 2020 (14m).
Project partially funded by Provincia Autonoma di Trento under the work programme FESR 2014-2020 LP 6/1999
D. Boscaini (FBK), F. Poiesi (FBK), S. Messelodi (FBK), A. Younes (Meccanica del Sarca S.p.A.), D. A. Grande (Meccanica del Sarca S.p.A.), Localisation of defects in volumetric Computed Tomography scans of valuable wood logs - 1st Int. Workshop on Industrial Machine Learning, 25th International Conference on Pattern Recognition - ICPR, Milano, Italy, January 2021 [pdf]
Contact: Stefano Messelodi (firstname.lastname@example.org)
Team: Davide Boscaini, Stefano Messelodi, Fabio Poiesi, Carla Maria Modena, Dimitri Giordani