Automated knot detection for high speed computed tomography on Pinus sylvestris L. and Picea abies (L.) Karst. using ellipse fitting in concentric surfaces

• 88–94% of all reference knots detected.
• 1% falsely detected knots.
• Robust with regard to varieties in log features and in between species.

High speed industrial computed tomography (CT) scanning of sawlogs is new to the sawmill industry and therefore there are no properly evaluated algorithms for detecting knots in such images. This article presents an algorithm that detects knots in CT images of logs by segmenting the knots with variable thresholds on cylindrical shells of the CT images. The knots are fitted to ellipses and matched between several cylindrical shells. Parameterized knots are constructed using regression models from the matched knot ellipses. The algorithm was tested on a variety of Scandinavian Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies (L.) Karst.) with a knot detection rate of 88–94% and generating about 1% falsely detected knots.