Automatic detection of the heartwood/sapwood boundary within Norway spruce (Picea abies (L.) Karst.) logs by means of CT images

An algorithm that automatically detects the boundary between heartwood and sapwood within fresh Norway spruce (Picea abies (L.) Karst.) logs was developed. Stacks of digital images of the internal structure of the logs were obtained through computed tomography. Boards, oriented in the longitudinal direction of the logs and passing about through the pith line, were reconstructed from stacks of images. The reconstructed boards were inspected by three operators who independently mapped the heartwood/sapwood boundaries using an image manipulation program. Their digitised mappings were collated and then sorted into two sets. One set signified close agreement between readings of the three operators, the other signified greater variation. The data sets were used as a basis for validation of the algorithm. The algorithm initially detected the pith location. Then, for the heartwood containing no knots, the boundary was detected by drawing (rasterising) 360 radii (with an angle of 1° between adjacent radii) from the pith to the first pixel whose intensity exceeded a given threshold. A post-processing was developed to separate knots in the heartwood from sapwood by longitudinal interpolations of the boundary. The automatically detected boundary was found to be in good agreement with the mappings of the operators (mean and median differences of 1.8 and 0.9 mm, respectively, where mappings of operators were concordant), especially at the bottom of the stems and for the oldest trees.