Deep quantification of a refined adulterant blended into pure avocado oil

Pérez-Calabuig, Ana M.; Pradana-López, Sandra; Ramayo-Muñoz, Andrea; Cancilla, John C.; Torrecilla, José S.

doi:10.1016/j.foodchem.2022.134474

Cited by 5 publications

(1 citation statement)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite its potential, not many reports can be found in the scientific literature on the assessment of the quality parameters of avocado oil using DIC. 2023) used optical images and artificial neural networks to achieve a 95% accuracy classification of avocado oil blended with a range of 1 to 15% refined olive oil [34]. A higher number of reports exist regarding the successful quantification of quality parameters in avocado oil using smartphone images, such as peroxide values [21], chlorophyll and carotenoids [22], and total sterols [20].…”

Section: Resultsmentioning

confidence: 99%

Exploring Strategies to Mitigate the Lightness Effect on the Prediction of Soybean Oil Content in Blends of Olive and Avocado Oil Using Smartphone Digital Image Colorimetry

de Carvalho,

da Silva Mutz,

Machado

et al. 2023

Foods

View full text Add to dashboard Cite

Extra virgin olive oil (EVOO) and avocado oil (AVO) are recognized for their unique sensory characteristics and bioactive compounds. Declared blends with other vegetable oils are legal, but undeclared mixing is a common type of fraud that can affect product quality and commercialization. In this sense, this study explored strategies to mitigate the influence of lighting in order to make digital image colorimetry (DIC) using a smartphone more robust and reliable for predicting the soybean oil content in EVOO and AVO blends. Calibration models were obtained by multiple linear regression using the images’ RGB values. Corrections based on illuminance and white reference were evaluated to mitigate the lightness effect and improve the method’s robustness and generalization capability. Lastly, the prediction of the built model from data obtained using a distinct smartphone was assessed. The results showed models with good predictive capacities, R2 > 0.9. Generally, models solely based on GB values showed better predictive performances. The illuminance corrections and blank subtraction improved the predictions of EVOO and AVO samples, respectively, for image acquisition from distinct smartphones and lighting conditions as evaluated by external validation. It was concluded that adequate data preprocessing enables DIC using a smartphone to be a reliable method for analyzing oil blends, minimizing the effects of variability in lighting and imaging conditions and making it a potential technique for oil quality assurance.

show abstract