Automatic knot segmentation in CT images of wet softwood logs using a tangential approach

• A robust algorithm was proposed for segmenting knots in CT images of softwood logs.
• The method was applied to 125 knots of five wood species.
• The RMSE is comparable to RMSD between two repetitions of manual measurements.
• The method works as well in wet sapwood as in heartwood.
• The software will be published under the GPL license and available online soon.

Computed Tomography (CT) is more and more used in forestry science and wood industry to explore internal tree stem structure in a non-destructive way. Automatic knot detection and segmentation in the presence of wet areas like sapwood for softwood species is a recurrent problem in the literature. This article describes an algorithm named TEKA able to segment knots even into sapwood and other wet areas by using parallel tangential slices into the log that enable to follow the knot from the stem pith to the bark. On each tangential slice, knot pith is detected, then knot diameter is estimated by analyzing gray level variations around the knot pith. A validation was performed on 125 knots from five softwood species. The CT slice resolution ranged from 0.4 to 0.8 mm/pixel with an interval between slices of 1.25 mm. Compared to manual diameter measurements performed on the same CT slices, the TEKA algorithm led to a RMSE of 3.37 mm and a bias of 0.81 mm, which is rather good compared to other algorithms working only in heartwood.