Remote sensing technology, such as hyperspectral imaging, in combination with machine learning algorithms, has emerged as a viable tool for rapid and nondestructive assessment of wine grape ripeness. However, the differences in terroir, together with the climatic variations and the variability exhibited by different grape varieties, have a considerable impact on the grape ripening stages within a vintage and between vintages and, consequently, on the robustness of the predictive models. To address this challenge, we present a novel one-dimensional convolutional neural network architecture-based model for the prediction of sugar content and pH, using reflectance hyperspectral data from different vintages. We aimed to evaluate the model’s generalization capacity for different varieties and for a different vintage not employed in the training process, using independent test sets. A transfer learning mechanism, based on the proposed convolutional neural network, was also used to evaluate improvements in the model’s generalization. Overall, the results for generalization ability showed a very good performance with RMSEP values of 1.118 °Brix and 1.085 °Brix for sugar content and 0.199 and 0.183 for pH, for test sets using different varieties and a different vintage, respectively, improving and updating the current state of the art.

The performance of a support vector regression (SVR) model with a Gaussian radial basis kernel to predict anthocyanin concentration, pH index and sugar content in whole grape berries, using spectroscopic measurements obtained in reflectance mode, was evaluated. Each sample contained a small number of whole berries and the spectrum of each sample was collected during ripening using hyperspectral imaging in the range of 380–1028 nm. Touriga Franca (TF) variety samples were collected for the 2012–2015 vintages, and Touriga Nacional (TN) and Tinta Barroca (TB) variety samples were collected for the 2013 vintage. These TF vintages were independently used to train, validate and test the SVR methodology; different combinations of TF vintages were used to train and test each model to assess the performance differences under wider and more variable datasets; the varieties that were not employed in the model training and validation (TB and TN) were used to test the generalization ability of the SVR approach. Each case was tested using an external independent set (with data not included in the model training or validation steps). The best R² results obtained with varieties and vintages not employed in the model’s training step were 0.89, 0.81 and 0.90, with RMSE values of 35.6 mg·L⁻¹, 0.25 and 3.19 °Brix, for anthocyanin concentration, pH index and sugar content, respectively. The present results indicate a good overall performance for all cases, improving the state-of-the-art results for external test sets, and suggesting that a robust model, with a generalization capacity over different varieties and harvest years may be obtainable without further training, which makes this a very competitive approach when compared to the models from other authors, since it makes the problem significantly simpler and more cost-effective.

The high quality of Port wine is the result of a sequence of winemaking operations, such as harvesting, maceration, fermentation, extraction and aging. These stages require proper monitoring and control, in order to consistently achieve the desired wine properties. The present work focuses on the harvesting stage, where the sugar content of grapes plays a key role as one of the critical maturity parameters. Our approach makes use of hyperspectral imaging technology to rapidly extract information from wine grape berries; the collected spectra are fed to machine learning algorithms that produce estimates of the sugar level. A consistent predictive capability is important for establishing the harvest date, as well as to select the best grapes to produce specific high-quality wines. We compared four different machine learning methods (including deep learning), assessing their generalization capacity for different vintages and varieties not included in the training process. Ridge regression, partial least squares, neural networks and convolutional neural networks were the methods considered to conduct this comparison. The results show that the estimated models can successfully predict the sugar content from hyperspectral data, with the convolutional neural network outperforming the other methods.

The residues generated in the wine industry (pomace, stems, seeds, wine lees, and grapevine shoots) are a potential source of bioactive compounds that can be used in other industries despite being sometimes underestimated. Different extraction methods using various solvents and extraction conditions are currently being investigated. Due to its natural occurrence in wines, safe behavior, and low toxicity when compared to other organic solvents, ethanol is used as an extracting agent. The aim of this study was to identify the winery by-product from the Região Demarcada do Douro and its corresponding extraction solvents that yields the most favorable results in (poly)phenols content and antioxidant capacity. To achieve this, five different ratios of ethanol: water, namely 0:100, 25:75, 50:50, 75:25, and 100:0 (v/v), for extracting the phenolic compounds were employed. Afterwards, the determination of total phenolic content (TPC), ortho-diphenols content (ODC), and flavonoid content (FC) as well as the antioxidant capacity of the obtained extracts using three different methods was performed. Since the best results of the spectrophotometric assays were obtained mostly with hydroethanolic extracts of stems (50:50, v/v), identification by HPLC-DAD has carried out. It was possible to conclude that the Tinta Roriz variety displayed the highest number of identified (poly)phenols.