Abstract Technologies utilising laser dot scanning to measure surface fibre orientation have evolved into advanced, system-ready solutions for machine strength grading. Although these methods significantly enhance predictive models of timber strength, they generally rely on surface-based interpolations that may not fully represent internal fibre architecture. In this study, a gradient structure tensor (GST) approach is investigated to estimate the normal direction of growth layers and infer internal fibre orientation. The method is compared with existing internal fibre orientation determination (IFOD) techniques combining laser dot measurements and destructive serial sectioning (DSS), and is assessed through local bending stiffness profiles derived from digital image correlation (DIC) tests and finite element (FE) simulations. Results indicate that accurate characterisation of internal fibre orientation, particularly when supported by laser dots-based measurements, enables highly reliable predictions of timber mechanical performance. DSS-based implementations yield determination coefficients of R² 0. 8-0. 9, while GST applied directly to DSS images—despite the images stack modest quality and without filtering optimisation—remains computationally efficient and shows promising correlation (R² 0. 5) for tomographic applications. The proposed methodology provides a basis for generating robust fibre orientation datasets and for developing data-driven models capable of inferring internal architecture from surface or CT information. These outcomes open perspectives for improving mechanical grading procedures, integrating knot modelling and elastic property refinement, and ultimately reconstructing fibre orientation at the scale of entire logs for forestry and industrial use.
Penvern et al. (Thu,) studied this question.