Image analysis to assess wood variability in longleaf pine cross-sectional disks

Image analysis is an important method for rapidly measuring wood property variation, but it is infrequently applied to disks collected from forestry studies. The objective of this study was to compare image estimated wood and bark volumes and diameters to reference measurements, and to extract more information from the images including the shape (out of round index, eccentric pith) and the amount and location of severe compression wood. A total of 1,120 disks were cut from multiple height levels of 48 defect-free and 56 defect-containing longleaf pine (Pinus palustris) trees from 16 stands across Georgia (U.S.). Disks were machined on one transverse surface using a computer numeric controlled router to prepare a clean surface for imaging. Three images; one under white light, second under blue light, and third under blue light with a green longpass filter, were taken for each disk. Volumes and diameters estimated from images were in close agreement with reference methods. Linear models fitted as measured versus image volumes for wood and bark had coefficient of determination (R2) values of >0.99 and 0.96. Linear models fitted as measured versus image diameters had R2 values of >0.99. Out of round index and pith eccentricity values calculated from images showed a moderate positive correlation (R=0.43). Algorithms developed were able to correctly identify severe compression wood, but not mild to moderate compression wood. Severe compression wood was moderately correlated to out of round index (R=0.54) and pith eccentricity (R=0.48). More than 98% of the disks having severe compression wood came from defect-containing trees