Artificial neural network as alternative method for prediction of sugar and acidity using near-infrared spectroscopy in table grapes
In table grape production it is critical to have an accurate method to assess the external and internal quality of the fruits. Many fruit quality attributes that affect consumer acceptance and price are still tested using traditional approaches, which are either subjective or time-consuming. In this sense, Near-Infrared spectroscopy has been successful to determine maturity quality parameters non-destructively and rapidly. Furthermore, this technique enables the detection of several internal and external fruit attributes simultaneously. However, for determining quality parameters using Near infrared (NIR) spectral data are necessary to implement several chemometrics procedures (outlier detection, spectral pre-processing, variable selection, calibration and validation) to build accurate models. This study analyses the use of a multivariate predictive modeling technique-artificial neural network (ANN) as an alternative to conventional Partial least squares (PLS) modeling using NIR spectral data of whole table grape bunches to determine the total soluble solids (TSS), titratable acidity (TA) and TSS/TA ratio. The results of this study show that the ANN-based models developed using NIR data fits well with the reference data and can be used as an alternative method for predicting purpose with high accuracy.
Poblete-Echeverría, C., Daniels, A.J., Nieuwoudt, H.H. and Opara, U.L. (2020). Artificial neural network as alternative method for prediction of sugar and acidity using near-infrared spectroscopy in table grapes. Acta Hortic. 1292, 321-328
whole table grape bunches, total soluble solid, titratable acidity, sugar-acid ratio, near-infrared spectroscopy, non-destructive measurements, machine learning, artificial neural networks