Abstract

In this study, a method suitable for on-line rapid species classification of western hemlock and amabilis fir (hem-fir) green mix of timber was developed and tested with the use of near infrared spectroscopy (NIRS) and chemometrics. The spectra of 600 wood specimens obtained from each species were collected over the wavelength range of 350–2500 nm. They were thereafter pretreated by smoothing; first derivative, second derivative and standard normal variate and calibration models were developed using the wavelength range of 800–1800 nm by partial least squares-linear discriminant analysis (PLS-LDA) and least squares-support vector machines (LS-SVM). The effects of wood surface (transverse, tangential and radial), wood zone (heartwood and sapwood) and sample moving speed (0, 0.5 and 1 m/s) were also explored. LS-SVM is superior to PLS-LDA in terms of classification performance at moving conditions. The best results were obtained using the LS-SVM method when spectra were collected on the transverse surface at a speed of 1 m/s and pretreated by smoothing. The sensitivity, specificity, and accuracy for both calibration and validation sets were all 100%. Collectively, NIR spectroscopy combined with LS-SVM is capable of on-line species separation of green hem-fir mix prior to wood drying.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call