Dataset containing spectral data from hyperspectral imaging and sugar content measurements of grapes berries in various maturity stage

In the dataset presented in this article, two hundred and seventy four trays containing one hundred berries were measured by a hyperspectral camera in the visible/near-infrared spectral domain. This dataset was formed to study the use of hyperspectral imaging for maturity monitoring of grape berries [2]. This dataset contains reflectance spectra from hyperspectral camera of grape berries of three different varieties and chemical composition (sugar content).


a b s t r a c t
In the dataset presented in this article, two hundred and seventy four trays containing one hundred berries were measured by a hyperspectral camera in the visible/near-infrared spectral domain. This dataset was formed to study the use of hyperspectral imaging for maturity monitoring of grape berries [2] . This dataset contains reflectance spectra from hyperspectral camera of grape berries of three different varieties and chemical composition (sugar content Value of the Data • These spectra were acquired under controlled conditions over a range of sugar values representative of all stages of maturity. • This dataset is useful for testing and comparing prediction methods.
• This dataset can be used to build models to predict the sugar content of grapes.
• This dataset is intended for scientists to test new methods (variable selection, data exploration, regression methods) or to have reference data to guide their future experiments.

Data Description
Chemical and Nir spectra measurements were made on 274 samples (one sample is composed by one hundred berries) from three different grape varieties ( Table 1 ). The table containing the dataset ( DATASET.csv ) is represented so that the rows correspond to the samples and the columns correspond to the variables. The first column corresponds to the tray key, the second column to the variety, the third column to the sugar content and the following columns correspond to reflectance spectra values obtained on the indicated spectral bands. For the three varieties, sugar content values are similar and comprised between 100 and 300 (g/L), representative of all stages of maturity.
Grape berries were sorted using NaCl densimetric baths and were placed on a tray for hyperspectral acquisition.
From these one hundred berries, an average spectrum was calculated from each image. At this end, 274 reflectance spectra were obtained on three different grape varieties with two red grape varieties (Syrah and Fer Servadou) and one white grape variety (Mauzac).  To illustrate, PLSR results are shown for total sugar content for each grape variety (cf. Table 2 ).

Samples and analyses
The sampling of grape berries started one or two weeks after veraison and before harvest in summer 2020 on plots of the experimental vineyard Domaine Exp érimental Viticole Tarnais located in Gaillac (France). The berries belonged to three grape varieties, two red grape varieties (Syrah and Fer Servadou) and one white grape variety (Mauzac). For each variety, thirty bunches were collected approximately once a week.
Grape berries were prepared in the laboratory. They were cut at the pedicel to preserve the whole fruit. Sorting was carried out by batches of the same degree of ripeness using sodium chloride (NaCl) baths. For this purpose, twelve NaCl baths of increasing concentrations from 70 to 190 g/L were prepared to classify the berries according to their berry density corresponding to sugar concentrations from 110 to 279 g/L. [1] . Berry musts were obtained with one hundred berries of the same degree of ripeness and sugar content measurements were made with a refractometer (HI-96816, Hanna Instruments).

Hyperspectral images and NIR spectra acquisition
Reflectance spectra were acquired before preparing a hundred berry must. These berries were placed on a tray to make the berries visible. These measurements were made with a hyperspectral camera (Specim IQ, Specim, Finland) with a spectral range of 400 nm to 10 0 0 nm and a spectral resolution equal to 7 nm.
The camera was positioned at 1.5m from the stage. A halogen lamp was used for lighting (Arrilite 750 Plus ARRI, Munich, Germany). Constant angles of -50 • and 50 • were maintained between the axes of the halogen lamp and the axis of the hyperspectral camera. A certified reflectance standard (Labsphere, SRS-40-010) was placed next to the tray containing the hundred bays in order to know the reference reflected intensity ( I o ( λ)). This procedure allows to standardise the images coming from the non-uniformities of the instrumentation (light source, lens, detector). For each image, spectra of the berries were selected by using the Spectral Angle Mapper (SAM) and were averaged.

Data analysis
All the calculations were run under Matlab (The Mathworks, Natick, MA, USA). PLS-R algorithm was used to perform model sugar content [3] . Model results for each varieties were evaluated on the basis of the coefficient of determination (R 2 cv ) and the root mean square standard error of cross-validation (RMSE cv ).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Spectral dataset of grape berries from hyperspectral imaging for maturity monitoring