Reconstruction for Sparse View Tomography of Long Objects Applied to Imaging in the Wood Industry

2024-03-05

arXiv.org

AI Summary

To address the need for online quality inspection of long cylindrical logs in the timber industry, conventional sparse-view X-ray computed tomography (CT)—using only five source positions—fails to faithfully reconstruct critical 3D biological structures (e.g., knots, heartwood, sapwood) due to severe information deficiency in single-layer 2D projections. This paper proposes a cross-slice feature-coupled sparse-view CT reconstruction method that breaks the traditional paradigm of independent single-slice reconstruction. For the first time under sequential scanning geometry, it introduces a structure-guided joint reconstruction mechanism leveraging anatomical consistency across adjacent slices. Built upon the Learned Primal-Dual framework, the method tightly integrates the physical forward imaging model with data-driven priors. Experiments demonstrate substantial improvements in quantitative metrics (e.g., +3.2 dB PSNR, +0.08 SSIM), high-fidelity 3D structural recovery, robust identification of key biological features, and compliance with industrial real-time inspection requirements.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Computer Vision: Multi-modal Vision
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

In the wood industry, logs are commonly quality screened by discrete X-ray scans on a moving conveyor belt from a few source positions. Typically, two-dimensional (2D) slice-wise measurements are obtained by a sequential scanning geometry. Each 2D slice alone does not carry sufficient information for a three-dimensional tomographic reconstruction in which biological features of interest in the log are well preserved. In the present work, we propose a learned iterative reconstruction method based on the Learned Primal-Dual neural network, suited for sequential scanning geometries. Our method accumulates information between neighbouring slices, instead of only accounting for single slices during reconstruction. Our quantitative and qualitative evaluations with as few as five source positions show that our method yields reconstructions of logs that are sufficiently accurate to identify biological features like knots (branches), heartwood and sapwood.

Problem

• Reconstructing 3D logs from sparse 2D X-ray slices
• Preserving biological features with limited source positions
• Accumulating information across slices for accurate reconstruction

Innovation

• Learned Primal-Dual neural network reconstruction
• Accumulates information between neighboring slices
• Works effectively with only five source positions