Comparative study on the use of three different near infrared spectroscopy recording methodologies for varietal discrimination of walnuts

• Autor(a) principal: Nogales-Bueno, Julio
• Data de Publicação: 2020
• Outros Autores: Feliz, Luis, Baca-Bocanegra, Berta, Hernandéz-Hierro, José Miguel, Heredia, Francisco José, Mota Barroso, João Manuel, Rato, Ana Elisa
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10174/33552; https://doi.org/Julio Nogales-Bueno, Luis Feliz, Berta Baca-Bocanegra, José Miguel Hernández-Hierro, Francisco José Heredia, João Manuel Barroso, Ana Elisa Rato. Comparative study on the use of three different near infrared spectroscopy recording methodologies for varietal discrimination of walnuts. Talanta, Volume 206,2020.https://doi.org/10.1016/j.talanta.2019.120189; https://doi.org/10.1016/j.talanta.2019.120189
• topic: Near infrared hyperspectral imaging; Walnuts; Chemometrics; Discriminant analysis; FT-NIRPortable MicroNIR

Walnut fruit (Juglans regia L.) is an internationally well-known product with an important tradition of consumption. Its health benefits and economic importance in the food industry make this nut an interesting research topic. In this feasibility study, 200 walnut samples of 5 different varieties were collected and their near infrared (NIR) spectra were recorded with 3 different devices: a benchtop Fourier transform near infrared (FT-NIR) spectrograph, a dispersive hyperspectral imaging camera and a portable NIR dispersive spectrograph. Discriminant analyses were applied and different methods for the varietal discrimination of walnuts were obtained and compared. Up to 96 and 84% of correct identification were respectively obtained in internal (training set) and external validations. Better results were obtained covering the entire shell surface than collecting a unique random spectrum per sample. Moreover, FT-NIR and hyperspectral tools produced classification models with a lower classification error in internal and external validations than the portable NIR one.