Using near-infrared hyperspectral images on subalpine fir board. Part 1: Moisture content estimation |
| |
| Authors: | Ataollah Haddadi James Burger Brigitte Leblon Zarin Pirouz Kevin Groves Joseph Nader |
| |
| Institution: | 1. Faculty of Forestry and Environmental Management, University of New Brunswick, Fredericton, NB, Canadaata.haddadi@unb.ca bleblon@unb.ca;3. Burgermetrics, Riga, Latvia;4. Faculty of Forestry and Environmental Management, University of New Brunswick, Fredericton, NB, Canada;5. FP Innovations – Wood Products, Vancouver, BC, Canada;6. FP Innovations – Forest Operations, Pointe-Claire, QC, Canada |
| |
| Abstract: | AbstractIn this study, moisture content (MC) images of subalpine fir (abies lasiocarpa Hook) boards were derived from near-infrared hyperspectral images in the 947–1637 nm range. One hundred and seven cubic samples with the size of 4 cm were prepared from 14 boards. All samples were dried to various MCs during several steps until being completely dried. Hyperspectral images and weight measurements were acquired over each sample at each drying step. The samples have MC ranging from 1% to 137% (dry basis). The images were first calibrated into reflectance. Then, bad pixels were found and replaced by a corrected value using a median filter. A modified version of the boxplot method was used to find abnormal spectra that were then removed. The remaining spectra were converted into absorbance spectra. They were then split into a calibration and a validation data-set according to the boards they were extracted from to build and validate a partial least squares (PLS) regression model between the near-infrared absorbance spectra and the measured MCs. The PLS model was applied first to the sample images, then to the whole board images in order to produce 2D images of MC. |
| |
| Keywords: | Hyperspectral imaging board distribution of moisture content near-infrared PLS subalpine fir |
|
|
