Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards synthetic generation of realistic wooden logs

Fedor Zolotarev, Borek Reich, Tuomas Eerola, Tomi Kauppi, Pavel Zemcik

TL;DR

The paper addresses the scarcity of realistic training data for surface-based log measurement and sawmilling optimization by proposing a novel, fully synthetic 3D log generator. It introduces a three-component, log-centric model that separately parameterizes knot growth, centerline shape, and surface texture, enabling joint generation of internal knot distributions and external surface geometry. A nine-parameter knot model derived from CT data, together with a surface model that captures base shape, surface knots, grain, and high-frequency texture via Gabor noise, constitutes the core technical contribution. The framework is validated against CT reconstructions of eight Scots pine logs, demonstrating accurate fitting of knots and the ability to generate plausible new logs, with code released for public use to support pretraining of surface-based log measurement and virtual sawing workflows. Overall, this approach enables large-scale synthetic data generation to drive improvements in knot detection, internal knot distribution modeling, virtual sawing, and sawing angle optimization, bridging the gap between surface measurements and internal wood structure.

Abstract

In this work, we propose a novel method to synthetically generate realistic 3D representations of wooden logs. Efficient sawmilling heavily relies on accurate measurement of logs and the distribution of knots inside them. Computed Tomography (CT) can be used to obtain accurate information about the knots but is often not feasible in a sawmill environment. A promising alternative is to utilize surface measurements and machine learning techniques to predict the inner structure of the logs. However, obtaining enough training data remains a challenge. We focus mainly on two aspects of log generation: the modeling of knot growth inside the tree, and the realistic synthesis of the surface including the regions, where the knots reach the surface. This results in the first log synthesis approach capable of generating both the internal knot and external surface structures of wood. We demonstrate that the proposed mathematical log model accurately fits to real data obtained from CT scans and enables the generation of realistic logs.

Towards synthetic generation of realistic wooden logs

TL;DR

The paper addresses the scarcity of realistic training data for surface-based log measurement and sawmilling optimization by proposing a novel, fully synthetic 3D log generator. It introduces a three-component, log-centric model that separately parameterizes knot growth, centerline shape, and surface texture, enabling joint generation of internal knot distributions and external surface geometry. A nine-parameter knot model derived from CT data, together with a surface model that captures base shape, surface knots, grain, and high-frequency texture via Gabor noise, constitutes the core technical contribution. The framework is validated against CT reconstructions of eight Scots pine logs, demonstrating accurate fitting of knots and the ability to generate plausible new logs, with code released for public use to support pretraining of surface-based log measurement and virtual sawing workflows. Overall, this approach enables large-scale synthetic data generation to drive improvements in knot detection, internal knot distribution modeling, virtual sawing, and sawing angle optimization, bridging the gap between surface measurements and internal wood structure.

Abstract

In this work, we propose a novel method to synthetically generate realistic 3D representations of wooden logs. Efficient sawmilling heavily relies on accurate measurement of logs and the distribution of knots inside them. Computed Tomography (CT) can be used to obtain accurate information about the knots but is often not feasible in a sawmill environment. A promising alternative is to utilize surface measurements and machine learning techniques to predict the inner structure of the logs. However, obtaining enough training data remains a challenge. We focus mainly on two aspects of log generation: the modeling of knot growth inside the tree, and the realistic synthesis of the surface including the regions, where the knots reach the surface. This results in the first log synthesis approach capable of generating both the internal knot and external surface structures of wood. We demonstrate that the proposed mathematical log model accurately fits to real data obtained from CT scans and enables the generation of realistic logs.

Paper Structure

This paper contains 23 sections, 10 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related work
  3. Synthetic log generation
  4. Knot growth modeling
  5. Log surface modeling
  6. Synthetic log model
  7. Model components
  8. Knot model
  9. Centerline model
  10. Log surface model
  11. Log thickness.
  12. Surface curvature.
  13. Surface knots.
  14. High frequency features.
  15. Experiments
  16. ...and 8 more sections

Figures (7)

  • Figure 1: An example of a generated log using the proposed model.
  • Figure 2: Comparison of knot points in (\ref{['subfig:cartesian']}) Cartesian and (\ref{['subfig:logcentric']}) Modified log-centric coordinates. Axis $s$ is in degrees.
  • Figure 3: Plot of log thickness along the length of a log. Vertical lines correspond to the centers of knot clusters. Fitted function is plotted with dashed line.
  • Figure 4: Surface knots (\ref{['subfig:surfaceKnots']}) as seen on the heightmap, (\ref{['subfig:surfaceDoGs']}) fitted models of surface knots and (\ref{['subfig:surfaceRecs']}) fitted knots on the reconstruction.
  • Figure 5: Knot model: (a) fitted models for a cluster of knots; (b) generated knot cluster.
  • ...and 2 more figures