Colour in thermally modified wood of beech, Norway spruce and Scots pine. Part 2: Property predictions from colour changes

Abstract In the woodworking industry, image analysis is routinely used for quality control and for matching and classification during various processes. An extension of these automated systems for the prediction of physical properties of thermally modified wood (TMW) is enticing, because to date there is no generalised procedure for the quality assurance of TMW. In this work, the feasibility of predicting 13 physical parameters from the analysis of colour changes is demonstrated using small thermally modified specimens of three wood species. Simple linear regression models for anti-swelling efficiency, nominal density, heat-induced weight loss and 10 strength parameters in six forms of stress were all very or highly significant, with R2 statistics for the best predictor from 0.24 to 0.94. ΔE* was found to be a better predictor than ΔL* for most properties. Multiple linear regression with 11 colour variables increased the prediction ability of most models in terms of R2, although these improvements varied with the property and species concerned. The best models altogether were obtained by partial least squares regression, with relative prediction error values >0 in all cases. Our results demonstrate that physical properties of small specimens of TMW can be efficiently predicted with only one after treatment measurement of colour in the CIEL*a*b* colour space by means of image analysis of TMW surfaces. We anticipate that our approach would be a starting point for more refined modelling of physical properties of larger wood members and other properties of interest in TMW (e.g., decay resistance).