Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data

Detalhes bibliográficos
Autor(a) principal: Hein, Paulo Ricardo Gherardi
Data de Publicação: 2015
Tipo de documento: Artigo
Idioma: por
Título da fonte: Repositório Institucional da UFLA
Texto Completo: http://www.cerne.ufla.br/site/index.php/CERNE/article/view/106
http://repositorio.ufla.br/jspui/handle/1/14652
Resumo: Near infrared (NIR) spectroscopy is a fast and efficient technique to predict a range of wood traits; however, methods for extracting useful information from the NIR spectra could be improved. Thus, the aim of this study was to evaluate the statistic performance of two regression methods for estimating the basic density in Eucalyptus urophylla x grandis wood from near infrared spectroscopic data. The predictive models calibrated by principal component regression (PCR) or partial least square regression (PLSR) method provided fine correlations. The coefficients of determination (R²cv) of the PCR models ranged from 0.78 to 0.85 with standard error of cross-validation (SECV) and the ratio of performance to deviation (RPD) varying from 32.8 to 41.2 kg/m³ and from 1.6 to 1.9, respectively. The PLSR models presented R²cv with relatively lower magnitude (from 0.65 to 0.78); but also lower SECV (from 29.8 to 38.9 kg/m³) and higher RPD values (from 1.6 to 2.1). In short, PCR method provides higher R² between Lab-measured and NIR-predicted values while PLSR produces lower standard errors of cross-validations. For both regression methods, the pre-treatments on NIR spectra, and the wavelength selection improved the calibration statistics, reducing the SECV and increasing the R²cv and the RPD values. Thus, PCR and PLS regression can be applied successfully for predicting basic density in Eucalyptus urophylla x grandis wood from the near infrared spectroscopic data.
id UFLA_31011c2b906cf755366c07f9f423577a
oai_identifier_str oai:localhost:1/14652
network_acronym_str UFLA
network_name_str Repositório Institucional da UFLA
repository_id_str
spelling Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic dataMétodos de regressão multivariada para estimativa da densidade básica da madeira de Eucalyptus por espectroscopia no infravemelho próximoChemometricsEucalyptusWood - Basic densityNear infrared spectroscopyQuimiometriaEspectroscopia no infravermelho próximoMadeira - Densidade básicaNear infrared (NIR) spectroscopy is a fast and efficient technique to predict a range of wood traits; however, methods for extracting useful information from the NIR spectra could be improved. Thus, the aim of this study was to evaluate the statistic performance of two regression methods for estimating the basic density in Eucalyptus urophylla x grandis wood from near infrared spectroscopic data. The predictive models calibrated by principal component regression (PCR) or partial least square regression (PLSR) method provided fine correlations. The coefficients of determination (R²cv) of the PCR models ranged from 0.78 to 0.85 with standard error of cross-validation (SECV) and the ratio of performance to deviation (RPD) varying from 32.8 to 41.2 kg/m³ and from 1.6 to 1.9, respectively. The PLSR models presented R²cv with relatively lower magnitude (from 0.65 to 0.78); but also lower SECV (from 29.8 to 38.9 kg/m³) and higher RPD values (from 1.6 to 2.1). In short, PCR method provides higher R² between Lab-measured and NIR-predicted values while PLSR produces lower standard errors of cross-validations. For both regression methods, the pre-treatments on NIR spectra, and the wavelength selection improved the calibration statistics, reducing the SECV and increasing the R²cv and the RPD values. Thus, PCR and PLS regression can be applied successfully for predicting basic density in Eucalyptus urophylla x grandis wood from the near infrared spectroscopic data.A espectroscopia no infravermelho próximo (NIR) é uma técnica rápida e eficiente para estimar características da madeira, no entanto, métodos para extrair as informações úteis a partir do NIRS podem ser melhorados. Assim, este estudo foi realizado com o objetivo de avaliar o desempenho estatístico de dois métodos de regressão para estimar a densidade básica da madeira de Eucalyptus urophylla x grandis a partir da espectroscopia no NIR. Os modelos preditivos calibrados pelos métodos da regressão de componentes principais (PCR) ou regressão de mínimos quadrados parciais (PLSR) apresentaram boas correlações. Os coeficientes de determinação (R²cv) dos modelos PCR variaram de 0,78 a 0,85, com erro padrão na validação cruzada (SECV) e a relação do desempenho de desvio (RPD) variando de 32,8-41,2 kg/m³ e 1,6-1,9 kg/m³, respectivamente. Os modelos PLSR apresentaram R²cv com magnitude relativamente menor (de 0,65 a 0,78), mas também menores valores de SECV (29,8-38,9 kg/m³) e maiores de RPD (de 1,6 a 2,1 kg/m³). De maneira geral, o método de PCR oferece maior R² entre os valores medidos e estimados, enquanto a PLSR produz menor erro padrão de validação cruzada. Para ambos os métodos de regressão, o pré-tratamento dos espectros NIR e a seleção de comprimento de onda melhoraram as estatísticas de calibração, reduzindo o SECV e aumentando os valores de R²cv e de RPD. Assim, as regressões PCR e PLS-R podem ser aplicadas com sucesso para a estimativa da densidade básica da madeira de Eucalyptus por espectroscopia no infravermelho próximo.CERNE2015-05-122017-08-01T20:15:29Z2017-08-01T20:15:29Z2017-08-01info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionapplication/pdfapplication/pdfhttp://www.cerne.ufla.br/site/index.php/CERNE/article/view/106HEIN, P. R. G. Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data. Cerne, Lavras, v. 16, p. 90-96, jul. 2010. Suplemento.http://repositorio.ufla.br/jspui/handle/1/14652CERNE; VOL 16, No 5 (2010) - SUPLEMENTO EBRAMEM; 090–096CERNE; VOL 16, No 5 (2010) - SUPLEMENTO EBRAMEM; 090–0962317-63420104-7760reponame:Repositório Institucional da UFLAinstname:Universidade Federal de Lavras (UFLA)instacron:UFLAporhttp://www.cerne.ufla.br/site/index.php/CERNE/article/view/106/80Copyright (c) 2015 Paulo Ricardo Gherardi HeinAttribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/info:eu-repo/semantics/openAccessHein, Paulo Ricardo Gherardi2021-04-06T18:01:28Zoai:localhost:1/14652Repositório InstitucionalPUBhttp://repositorio.ufla.br/oai/requestnivaldo@ufla.br || repositorio.biblioteca@ufla.bropendoar:2021-04-06T18:01:28Repositório Institucional da UFLA - Universidade Federal de Lavras (UFLA)false
dc.title.none.fl_str_mv Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
Métodos de regressão multivariada para estimativa da densidade básica da madeira de Eucalyptus por espectroscopia no infravemelho próximo
title Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
spellingShingle Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
Hein, Paulo Ricardo Gherardi
Chemometrics
Eucalyptus
Wood - Basic density
Near infrared spectroscopy
Quimiometria
Espectroscopia no infravermelho próximo
Madeira - Densidade básica
title_short Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
title_full Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
title_fullStr Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
title_full_unstemmed Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
title_sort Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data
author Hein, Paulo Ricardo Gherardi
author_facet Hein, Paulo Ricardo Gherardi
author_role author
dc.contributor.author.fl_str_mv Hein, Paulo Ricardo Gherardi
dc.subject.por.fl_str_mv Chemometrics
Eucalyptus
Wood - Basic density
Near infrared spectroscopy
Quimiometria
Espectroscopia no infravermelho próximo
Madeira - Densidade básica
topic Chemometrics
Eucalyptus
Wood - Basic density
Near infrared spectroscopy
Quimiometria
Espectroscopia no infravermelho próximo
Madeira - Densidade básica
description Near infrared (NIR) spectroscopy is a fast and efficient technique to predict a range of wood traits; however, methods for extracting useful information from the NIR spectra could be improved. Thus, the aim of this study was to evaluate the statistic performance of two regression methods for estimating the basic density in Eucalyptus urophylla x grandis wood from near infrared spectroscopic data. The predictive models calibrated by principal component regression (PCR) or partial least square regression (PLSR) method provided fine correlations. The coefficients of determination (R²cv) of the PCR models ranged from 0.78 to 0.85 with standard error of cross-validation (SECV) and the ratio of performance to deviation (RPD) varying from 32.8 to 41.2 kg/m³ and from 1.6 to 1.9, respectively. The PLSR models presented R²cv with relatively lower magnitude (from 0.65 to 0.78); but also lower SECV (from 29.8 to 38.9 kg/m³) and higher RPD values (from 1.6 to 2.1). In short, PCR method provides higher R² between Lab-measured and NIR-predicted values while PLSR produces lower standard errors of cross-validations. For both regression methods, the pre-treatments on NIR spectra, and the wavelength selection improved the calibration statistics, reducing the SECV and increasing the R²cv and the RPD values. Thus, PCR and PLS regression can be applied successfully for predicting basic density in Eucalyptus urophylla x grandis wood from the near infrared spectroscopic data.
publishDate 2015
dc.date.none.fl_str_mv 2015-05-12
2017-08-01T20:15:29Z
2017-08-01T20:15:29Z
2017-08-01
dc.type.driver.fl_str_mv info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
format article
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://www.cerne.ufla.br/site/index.php/CERNE/article/view/106
HEIN, P. R. G. Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data. Cerne, Lavras, v. 16, p. 90-96, jul. 2010. Suplemento.
http://repositorio.ufla.br/jspui/handle/1/14652
url http://www.cerne.ufla.br/site/index.php/CERNE/article/view/106
http://repositorio.ufla.br/jspui/handle/1/14652
identifier_str_mv HEIN, P. R. G. Multivariate regression methods for estimating basic density in Eucalyptus wood from near infrared spectroscopic data. Cerne, Lavras, v. 16, p. 90-96, jul. 2010. Suplemento.
dc.language.iso.fl_str_mv por
language por
dc.relation.none.fl_str_mv http://www.cerne.ufla.br/site/index.php/CERNE/article/view/106/80
dc.rights.driver.fl_str_mv Copyright (c) 2015 Paulo Ricardo Gherardi Hein
Attribution 4.0 International
http://creativecommons.org/licenses/by/4.0/
info:eu-repo/semantics/openAccess
rights_invalid_str_mv Copyright (c) 2015 Paulo Ricardo Gherardi Hein
Attribution 4.0 International
http://creativecommons.org/licenses/by/4.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
application/pdf
dc.publisher.none.fl_str_mv CERNE
publisher.none.fl_str_mv CERNE
dc.source.none.fl_str_mv CERNE; VOL 16, No 5 (2010) - SUPLEMENTO EBRAMEM; 090–096
CERNE; VOL 16, No 5 (2010) - SUPLEMENTO EBRAMEM; 090–096
2317-6342
0104-7760
reponame:Repositório Institucional da UFLA
instname:Universidade Federal de Lavras (UFLA)
instacron:UFLA
instname_str Universidade Federal de Lavras (UFLA)
instacron_str UFLA
institution UFLA
reponame_str Repositório Institucional da UFLA
collection Repositório Institucional da UFLA
repository.name.fl_str_mv Repositório Institucional da UFLA - Universidade Federal de Lavras (UFLA)
repository.mail.fl_str_mv nivaldo@ufla.br || repositorio.biblioteca@ufla.br
_version_ 1815439001825312768