Near infrared hyperspectral imaging applied to mapping chemical composition in wood samples

摘要

This paper describes a method for the two-dimensional mapping of chemical composition on the transverse face of cross-sectional discs from trees. The method uses an imaging spectrograph coupled to a near infrared (NIR) camera (900-1700nm) to obtain NIR hyperspectral data sets which are processed using partial least squares regression to visualise the distribution and variation of lignin, galactose and glucose in Pinus radiata discs with R~2/standard error of performance values of 0.84/1.48 (lignin), 0.87/0.68 (galactose) and 0.87/0.95 (glucose). The hardware design and software control are described along with a method for calibration based on one dimension spatially resolved predictions of chemical composition from conventional NIR spectroscopy. The NIR imaging system was designed as a rapid and cost-effective means of mapping chemical composition over the entire disc at a spatial resolution of ～4mm~2/pixel. The resulting maps of chemical composition clearly indicate, at high spatial resolution, the extent of heterogeneity that occurs in logs.