Non-destructive evaluation of dry matter in ‘Edward’ mango by reflectance spectroscopy

Abstract

Mango is a very popular climacteric fruit in America and Europe. Within the internal properties of mango, dry matter is a suitable indicator to estimate the final quality of mango, however, the measurement of this indicator requires destructive testing and high time consumption. Therefore, this research creates a new spectral database of Edward mango to build models based on Partial Least Squared Regression (PLSR) and Principal Component Regression (PCR). Our research analyzes a total of 18 PCR models and 18 PLSR models, where 4 types of transformations on the dependent variable (logarithmic, square root, square and none transformation), 3 types of reflectance-based feature extractors (logarithmic, first derivative and none transformation), and 3 preprocessing techniques (Standard Normal Variate (SNV), Multiplicative Signal Correction (MSC) and none preprocessing) have been studied. The research proposes a double cross-validation both to determine the optimal number of components and to obtain the final metrics. The best model has an RMSE of 1.6142 %MS and an RMSE of 0.6102 in the scaled dimension. The model used 3 components, did not use transformation, used R reflectance as the independent variable and MSC as the preprocessing technique