Non-destructive evaluation of soluble solid content in fruits with various skin thicknesses using visible–shortwave near-infrared spectroscopy

Abstract

Visible–shortwave near-infrared spectroscopy has been used for internal quality measurement, but the optical penetration to the thickness of fruit skin becomes a challenge. This research aimed to develop partial least square regression model for the soluble solid content (SSC) measurement of fruits having various skin thicknesses, namely dragon fruit, tomato, guava, sapodilla, and banana. The spectra of each fruit were taken in a reflectance mode over a wavelength range of 400–1,000 nm. The best models obtained from banana and sapodilla yielded determination coefficient of prediction (R 2 p) of 0.88 and 0.90 and root mean square error of prediction (RMSEP) 0.39 and 0.38°Brix, respectively. The banana and sapodilla SSC prediction models should be able to be used carefully in a variety of applications. Tomato and guava had moderately thinner skin but had the lower R 2 p of 0.64 and 0.76 and the RMSEP of 0.17 and 0.26°Brix, respectively. The poorest model was yielded by dragon fruit, which had the thickest skin with the R 2 p of 0.59 and the RMSEP of 0.40°Brix. The model for guava, although having low R 2 p, can still be utilized as a screening criterion and in some other ‘approximate’ applications. However, the SSC prediction model for tomatoes and dragon fruit is not recommended to use and requires additional research. In addition to the effect of skin thickness, other fruit morphological influences the result of this study. Internal structure and seed number influence the reflection optical geometry, which also affects the SSC prediction model.

Keywords

• fruit
• soluble solid content
• partial least square regression
• vis–swnir