Determination of quality attributes and ripening stage using vis-nir spectroscopy in intact seriguela and umbu fruits

Autor: Patrícia De Araujo Souza, Iara Jeanice Souza Ferreira, Daniel dos Santos Costa
Rok vydání: 2022
Předmět:
Zdroj: Engineering in Agriculture; Vol. 30 (2022): Publicação Contínua; 127-141
Revista Engenharia na Agricultura-REVENG; v. 30 (2022): Publicação Contínua; 127-141
Engenharia na Agricultura
Universidade Federal de Viçosa (UFV)
instacron:UFV
ISSN: 2175-6813
1414-3984
Popis: To produce seriguela and umbu on a large scale, it is important to detect the ripening stages and quality attributes of the fruits, to define the ideal harvest point. Thus, this study aimed to determine, in a non-destructive way, the quality attributes and ripening stages of intact seriguela and umbu fruits using Vis-NIR spectroscopy. A total of 150 seriguela fruits and 150 umbu fruits were used, at different ripening stages, and subjected to spectral analysis and reference laboratory testing to determine total soluble solids (TSS) and firmness. Spectral data were subjected to different pre-processing techniques. Regression and classification models were created through the statistical learning and machine learning methods. The models with the best performance for TSS were RF (R2P = 0.94) and PLSR (R2P = 0.68), and for firmness were PLSR (R2P = 0.92) and RF (R2P = 0.58), for seriguela and umbu, respectively. The model with the best performance in the classification was LDA, with a precision greater than 95% to discriminate the ripening stages of both fruits. Therefore, the Vis-NIR spectroscopy is a potential tool to determine the quality attributes and ripening stages, in a non-destructive way, of intact seriguela and umbu fruits. To produce seriguela and umbu on a large scale, it is important to detect the ripening stages and quality attributes of the fruits, to define the ideal harvest point. Thus, this study aimed to determine, in a non-destructive way, the quality attributes and ripening stages of intact seriguela and umbu fruits using Vis-NIR spectroscopy. A total of 150 seriguela fruits and 150 umbu fruits were used, at different ripening stages, and subjected to spectral analysis and reference laboratory testing to determine total soluble solids (TSS) and firmness. Spectral data were subjected to different pre-processing techniques. Regression and classification models were created through the statistical learning and machine learning methods. The models with the best performance for TSS were RF (R2P = 0.94) and PLSR (R2P = 0.68), and for firmness were PLSR (R2P = 0.92) and RF (R2P = 0.58), for seriguela and umbu, respectively. The model with the best performance in the classification was LDA, with a precision greater than 95% to discriminate the ripening stages of both fruits. Therefore, the Vis-NIR spectroscopy is a potential tool to determine the quality attributes and ripening stages, in a non-destructive way, of intact seriguela and umbu fruits.
Databáze: OpenAIRE
Externí odkaz: