|MORA, CHRISTIAN - Bioforest|
|SCHIMLECK, LAURENCE - University Of Georgia|
|THAI, CHI - University Of Georgia|
Submitted to: Journal of Near Infrared Spectroscopy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/3/2011
Publication Date: 11/3/2011
Citation: Mora, C.R., Schimleck, L.R., Yoon, S.C., Thai, C.N. 2011. Determination of basic density and moisture content of loblolly pine wood disks using a NIR hyperspectral imaging system. Journal of Near Infrared Spectroscopy. Volume 19 Issue 5, Pages 401-409.
Interpretive Summary: A study was conducted to investigate the potential of a hyperspectral imaging method for the estimation of two important wood properties: basic density (BD) and moisture content (MC). The objective was to determine the estimation accuracy of the hyperspectral imaging method as compared to the point-probe spectroscopic sensing method. A knowledge of both wood properties can greatly improve weight scaling of pulplogs and sawlogs and can be used to sort logs (or loads) based on quality. The study found that hyperspectral imaging can be used as a non-destructive method to measure BD and MC and has the potential for sorting wood online on a relatively large scale.
Technical Abstract: The use of near infrared (NIR) hyperspectral imaging for the estimation of basic density (BD) and moisture content (MC) of loblolly pine (Pinus taeda L.) disks is reported. A total of 125 wood disks ranging in age from 13 to 19 years were analysed. Hyperspectral images were collected using an imaging system composed of an InGaAs camera from Sensors Unlimited, Inc. (sensitive between 1,000 and 1,700 nm at 5 nm increments) and a Liquid Crystal Tunable Filter from CRI, Inc., continuously tunable between 1,000 and 1,800 nm. Owing to noise spectra were truncated to 1,005-1,645 nm and smoothed using a 5-point median filter prior to calibration development. Sixteen samples were detected as outliers, reducing the number of disks available for analysis to 109, which were subsequently split into calibration (85) and validation (24) sets. Successful models based on the hyperspectral data were obtained for disk basic density (R2C = 0.81) and moisture content (R2C = 0.77). The predictive ability of the calibrations was acceptable, with mean square errors (RMSEP) of 23.64 kg m-3 for BD and 2.08% for MC. The calibrations were compared to those obtained using NIR spectra collected from the surface of the disks using a FOSS XDS NIR System coupled with a Smart Probe Analyser using the same wavelength range and number of samples (each disk was represented by an average of 8 spectra which had been divided into four quadrants with two spectra collected per quadrant). Probe based calibrations were more successful; however, the hyperspectral based calibrations were of sufficient strength to suggest that hyperspectral imaging could be used for the estimation of BD and MC of loblolly pine disks.