Resinwood formation in Scots pine (Pinus sylvestris) resulting from Cronartium pini infection represents a critical quality defect that substantially reduces timber value and complicates industrial processing decisions in the Nordic Forest industry. Current timber grading protocols rely exclusively on visual surface inspection, which cannot detect internal resin accumulation patterns essential for optimal sawmill operations and volume yields. This knowledge gap generates cascading economic inefficiencies throughout timber processing, as sawmills cannot optimise cutting patterns without comprehensive internal defect information.

This doctoral research addresses these fundamental limitations by developing and validating non-destructive methods for detecting and segmentation of internal resinwood distribution using X-ray computed tomography (CT) imaging. The pathophysiological basis for this approach stems from the host defence mechanism whereby C. pini infection triggers systematic hyperproduction of oleoresin compounds that infiltrate wood tissues, generating distinguishable radiographic density variation, so the research work bridges this gap by integrating two approaches:

Algorithm Development: An automated segmentation pipeline combining Gaussian Mixture Models (GMM), multi-directional ray-casting, and morphological refinement was developed to distinguish resinwood from healthy tissues in CT imagery. The method processes ~7,000 slices per log, achieving high sensitivity for sapwood (0.98) and heartwood (0.80) segmentation. Resinwood detection attained moderate precision (0.25), limited by density overlaps with heartwood and CT image artefacts.Qualitative Feature Identification: CT analysis of paired green- and dry- state specimens revealed three diagnostic resinwood signatures: (a) growth-ring distortions, (b) non-concentric cambial development, and (c) ground-glass opacity patterns. Resin extraction validated these features, confirming increase resin content in CT-identified zones (41% vs. control: 2%, p < 0.001).Key novel scientific contribution:

Evaluated probabilistic framework for resinwood segmentation through industrial CT imaging.Dual-state (green/dry) CT scans enabling the utilisation of comparison imaging data for resinwood mapping in logs.Open-source implementation supporting model development in pilot testing for sawmill-related research and development.Challenges persist in distinguishing pathological resin accumulation from natural impregnation heartwood density gradients and moisture driven influenced false detections. Future work will integrate multi-threshold GMMs, more noise reduction, and spatial feature descriptors (e.g., resin morphology) to enhance specificity. This research establishes a foundation for data-driven defect detection and have potential to improve volume yield in production and processing optimisation in digitally assisted sawmilling.