Wood Species Identification by Near-Infrared Spectroscopy

An optimum artificial neural network and a partial least square with discriminant analysis regression were developed and tested for accuracy in distinguishing two wood species by using near-infrared (NIR) spectrum. A mixed population of kiln-dried wood boards of western hemlock (Tsuga heterophylla (Raf.) Sarg.) and amabilis fir (Abies amabilis (Dougl.) Forbes) were scanned by NIR and then a random sub-set was water saturated under vacuum conditions and scanned again. This design aimed to capture the effect of moisture content above the fibre saturation point on the separation algorithms. Our results revealed that both modelling techniques can be effective tools for species recognition achieving correct identification of over 86% for fir and 94% for hemlock on either kiln-dried or fully saturated boards.