Effect of surface roughness on near infrared models for wood density analysis

The near-infrared spectroscopy (NIR) coupled with multivariate analysis to predict density of wood samples with two surface roughness levels was analyzed in order to investigate the effect of surface roughness on NIR predicting wood density in this paper. The main results were described as follows: (1) The predicting results of density model based on samples with smooth surface performed better than that based on samples with coarse surface. The correlation coefficients of calibration model and validation model of samples with coarse surface based on the NIR which were obtained from cross section of samples were 0.81 and 0.74, and that of the samples with smooth surface were 0.94 and 0.88 respectively. (2) The NIR with wavelength ranges of 350∼500 nm and 2 350 ∼ 2 500 nm showed significant spectral response to wood surface roughness. (3) It was suggested that the surface roughness of prediction samples should be same to that of the calibration samples.