Ripeness evaluation of white grape (Vitis Vinifera L.) by using of a prototype of LED based optical device

Establishing ripeness at harvest is a crucial issue since fruit quality is closely related to it. The search for non-destructive methods which could explore a large number of samples and give a rapid and comprehensive overview of ripening would be helpful. The aim of the research was to design, build and test a prototype of a low-cost and user-friendly device to support small-scale growers in determining the optimal harvest date according to grape ripening degree. The acquisition of the spectral reflectance at four specific wavelengths (630, 690, 750 and 850 nm) was proposed. Nevertheless, attention was given to the modularity and versatility of the system. Light emitting diode (LED) technology was chosen as the light source, to achieve the capability to individually adjust the light emission intensity for each measurement channel. A customized optical fiber was specifically designed for the application in order to optimize the radiation transmission. The prototype was tested for a rapid estimation of ripening parameters of grape for Franciacorta wine directly in field. At the same time, spectral measurements on grape berries were performed (for data comparison) with a commercial portable vis/NIR spectrophotometer operating in the wavelength range of 400-1000 nm, directly in the field too. A total of 95 bunches of white grapes were collected and non-destructive analyses were carried out on each sample using both the LED based system and the commercial vis/NIR spectrophotometer. The correlation between the spectral data matrix from the commercial device and the reference parameters (SSC and TA) were carried out. Finally, a classification analysis was performed. The percentages of correctly and misclassified grape samples were determined according to ripeness thresholds for the analyzed parameters (ripe > 17 °Brix for SSC and < 7 g dm-3 for TA) suggested by the Franciacorta Consortium. Overall PPV results deriving from the simplified system showed a slightly low classification capabilities (SSC on berry = 79.0%, SSC on bunch = 77.5%, TA = 84.3%) compared to the commercial device (SSC on berry = 84.6%, SSC on bunch = 89.5%, TA = 90.5%). This means that only a small loss of information was noticeable between the PLS models calculated using 2048 wavelengths and the MLR models calculated employing only the four channels of the simplified device, especially for the TA estimation. Classification performances related to the two optical instrumentations reflect the trend of the quantitative PLS and MLR models confirming the small loss of information using only four wavelengths and, consequently, the applicability of the simplified LED system.