Using hyperspectral images to map moisture content and basic density of boards and frozen and thawed logs
| dc.contributor.advisor | Leblon, Brigitte | |
| dc.contributor.author | Goyaghaj, Ataollah Haddadi | |
| dc.date.accessioned | 2023-03-01T16:17:10Z | |
| dc.date.available | 2023-03-01T16:17:10Z | |
| dc.date.issued | 2016 | |
| dc.date.updated | 2020-11-24T00:00:00Z | |
| dc.description.abstract | The purpose of this research was to investigate the use of near infrared hyperspectral imaging (NIR-HSI) for in-line moisture content (MC) and basic specific gravity (BSG) estimation of thawed and frozen logs as well as of boards. We also developed a method to classify the logs according to their MC, BSG, species, and log state (frozen and thawed). Samples from three different species (black spruce, quaking aspen, balsam poplar) for logs and one species (subalpine fir) for board were collected and dried in different steps. We also considered frozen samples for logs. For each step hyperspectral NIR images and weight measurements were acquired. The images were subjected to the following processing. They were first calibrated into reflectance. Then, bad pixels were found and replaced by a corrected value using a median filter. A new method was developed to find and remove abnormal spectra. It consisted of a combination of the boxplot method and principle component analysis (PCA). The remaining spectra were converted into absorbance spectra. The raw absorbance spectra were subjected to several spectral transformations, such as the multiplicative scatter correction (MSC), as well as the first, and second derivatives. For the board, the best PLS model was found in using raw spectra for both MC and BSG estimation and had an RMSEV of 10.8% and 0.007, respectively. For the log samples, PLS models were calibrated by considering two factors: log state (thawed and frozen conditions) and species, and their combination. Then the models were applied to the whole board images in order to produce 2D images of MC and BSG. Models were better with thawed logs than with frozen logs. The models estimated MC with an RMSE that varies between 2.94% in the case of black spruce to 15.49% in the case of balsam poplar. The modelโs accuracy for BSG estimation was the best when all the three species were used together (RMSEV=0.036). PLS discriminant analysis (PLS-DA) was also applied to sort log samples into three MC or BSG classes, species, or the log state (frozen and thawed). The overall accuracy of PLS-DA models were above 72% for both MC and BSG sorting and above 85% for the species and log state sorting. Finally, the Kubelka-Munk theory equations were employed to calculate several wood optical properties from visible-near-infrared reflectance spectra acquired over thin samples of quaking aspen and black spruce. The properties included absorption and scattering coefficients, transport absorption, reduced scattering, and penetration depth. The sample MC was then estimated using PLS regression method from the absorption and scattering coefficient spectra. Absorption coefficient spectra between 800 and 1800 nm can provide PLS models having an acceptable accuracy for MC estimation (๐ ๐ถ๐2=0.83 and RMSECV=2.32%), regardless of the species. | |
| dc.description.copyright | ยฉ Ataollah Haddadi (2016) | |
| dc.format | text/xml | |
| dc.format.extent | xviii, 233 pages | |
| dc.format.medium | electronic | |
| dc.identifier.other | Thesis 9751 | |
| dc.identifier.uri | https://unbscholar.lib.unb.ca/handle/1882/13311 | |
| dc.language.iso | en_CA | |
| dc.publisher | University of New Brunswick | |
| dc.rights | http://purl.org/coar/access_right/c_abf2 | |
| dc.subject.discipline | Forestry and Environmental Management | |
| dc.title | Using hyperspectral images to map moisture content and basic density of boards and frozen and thawed logs | |
| dc.type | doctoral thesis | |
| thesis.degree.discipline | Forestry and Environmental Management | |
| thesis.degree.fullname | Doctor of Philosophy | |
| thesis.degree.grantor | University of New Brunswick | |
| thesis.degree.level | doctoral | |
| thesis.degree.name | Ph.D. |
Files
Original bundle
1 - 1 of 1