Three-dimensional grain angle measurement of softwood (Hinoki cypress) using near infrared spatially and spectrally resolved imaging (NIR-SSRI)

摘要

As the strength of wood is greatly affected by its three-dimensional (3D) grain angles (the dive angle and the surface angle), the wood industry today requires automatic, rapid, and robust measurement techniques for measuring them simultaneously. In the present study, a near infrared spatially and spectrally resolved imaging (NIR-SSRI) system was designed in a line scan model, mainly including an NIR hyperspectral imaging camera and a halogen spotlight source (empty set 1 mm). Spatially resolved diffuse reflectance images at three target wavelengths (1002 nm, 1217 nm, and 1413 nm) were obtained from Hinoki cypress [Chamaecyparis obtusa (Siebold & Zucc.) Endl.] samples at various (0 degrees, 3 degrees, 6 degrees, ... 45 degrees) dive angles and surface angles (0 degrees, 3 degrees, 6 degrees, ... 45 degrees). The scattering patterns caused by the "tracheid effect" were almost elliptical. Subsequently, nonlinear least squares fitting was used to determine their eccentricities (e) and rotation angles (theta). The e values at each selected wavelength were highly correlated with the dive angle reference values; and the global identification model developed using Gaussian process regression (GPR) under five-fold cross-validation (CV) reached a determination coefficient (r(2)) of 0.98 with a root mean square error (RMSE) of 2.2 degrees. On the other hand, local surface angle identification models developed using linear regression analysis achieved determination coefficients higher than 0.90 on r(2) and an RMSE of CV lower than 3.8 degrees when the dive angle was lower than 30 degrees.