Banca de DEFESA: THIAGO VALENTE NOVAES

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : THIAGO VALENTE NOVAES
DATE: 29/06/2021
TIME: 08:30
LOCAL: Video conferência
TITLE:

THE USE OF SPECTROSCOPY IN THE NEAR INFRARED (NIR) FOR DISCRIMINATION OF AMAZON FOREST SPECIES

KEY WORDS:

NIRS. Wood identification. Wood technology

PAGES: 54
BIG AREA: Ciências Agrárias
AREA: Recursos Florestais e Engenharia Florestal
SUMMARY:

In the wood production chain, problems associated with the scientific determination of species have become one of the main obstacles in the product's valuation. Errors in the association of scientific names based on common names are rooted in forest inventories - IF and continue along the entire chain, resulting in economic and ecological losses hitherto immeasurable due to the lack of tools that can help to find and correct them. them. From this perspective, there is a clear need to improve the species determination process so that IFs are carried out in a more consistent manner. Within this context, the present study aimed to evaluate the potential of spectroscopy in the near infrared - NIR, to discriminate wood from forest species occurring in the Amazon, based on multivariate data analysis. Samples from 6 forest species were used, namely: Manilkara elata (Ducke) Chevalier; Dinizia excelsa Ducke; Goupia glabra Aubl .; Hymenaea sp .; Micropholis melinoniana Pierre and Copaifera sp. The samples were collected in the municipality of Portel / PA, within the area of the Sustainable Forest Management Plan - PMFS of the company ABC Norte-Fazenda Pacajá, and each species was represented in the study by 3 trees, with 1 disc being collected at the base of each tree. for the production of the samples. Of the 18 trees used, 350 cubic samples and 18 radial drumsticks were produced, the latter to assess the effect of the collection position in the transverse plane towards the marrow-bark. For the purposes of analysis, the type of finishing of the parts (chainsaw and circular saw), the spectral acquisition path (optical fiber and integration sphere), the type of validation of the models (cross and independent) and the application of mathematical pre-treatment to the spectral signature. With the specimens, 1,400 spectra were acquired in the cubic samples and 528 spectra in the radial drumsticks. The results indicated that the wood samples processed with a circular saw resulted in a surface with better interaction with radiation in the NIR and the models presented higher percentage values of classification. The integration sphere was the path of spectral acquisition that generated spectra that resulted in models with higher percentage values of correct classification of wood samples. Spectroscopy in the NIR associated with multivariate statistics was able to differentiate samples produced by chainsaw and circular saw with 98.4% assertiveness. For trees of the same species, the average of correct classification of the models based on NIR was above 90% and for discriminating different species the correct classification reached 99.2%. The best results for correct species classification were obtained with spectra acquired near the spinal cord, reaching 100% accuracy. The lowest percentages of species classification were obtained with the approach based on calibration with spectra collected via the integration sphere and independent validation with spectra collected via optical fiber.

BANKING MEMBERS:
Presidente - 1550131 - MARCELA GOMES DA SILVA
Externo à Instituição - ANTÔNIO JOSÉ VINHA ZANUNCIO - UFU
Externa à Instituição - AMÉLIA GUIMARÃES CARVALHO - UFU
Externa à Instituição - FERNANDA MARIA GUEDES RAMALHO - UFLA