Original papers Comparison of different approaches for the prediction of sugar content in new vintages of whole Port wine grape berries using hyperspectral imaging Véronique M. Gomes a,⇑ , Armando M. Fernandes b , Arlete Faia d,e , Pedro Melo-Pinto a,c a CITAB-Centre for the Research and Technology of Agro-Environmental and Biological Sciences, Universidade de Trás-os-Montes e Alto Douro, 5000-801 Vila Real, Portugal b INOV – INESC Inovação, Rua Alves Redol, 9, 1000-029 Lisboa, Portugal c Departamento de Engenharias, Escola de Ciências e Tecnologia, Universidade de Trás-os-Montes e Alto Douro, 5000-801 Vila Real, Portugal d BioISI – Biosystems & Integrative Sciences Institute, University of Lisboa, Faculty of Sciences, Campo Grande, Lisboa, Portugal e Departamento de Genética e Biotecnologia, Escola das Ciências da Vida e Ambiente, Universidade de Trás-os-Montes e Alto Douro, 5000-801 Vila Real, Portugal article info Article history: Received 1 August 2016 Received in revised form 13 April 2017 Accepted 10 June 2017 Keywords: Prediction Hyperspectral imaging PLSR Neural networks Grapes berries abstract Two different approaches, PLS regression and neural networks, were compared for monitoring the quality of grapes using sugar content predictions based on hyperspectral imaging. The present work expands the result analysis and updates the state-of-the-art published in a conference article of the authors which concern the prediction of sugar content for vintages not used in model creation when the measured sam- ples are composed of a small number of whole berries. This is highly innovative. The prediction models were established upon training under each approach and the generalization ability of both methodologies was determined through using n-fold-Cross-Validation and test sets. Sugar content was estimated using a model trained with spectra from samples of 2012. The test sets were composed of samples with six whole berries of 2012 or 2013 vintages. The results for PLS regression and Neural Networks for a test set with 2012 samples, were 0.94 °Brix and 0.96 °Brix for the root mean square error (RMSE), and 0.93 and 0.92 for squared correlation coeffi- cients (R 2 ), respectively, for each approach. When using test data containing 2013 samples, the RMSE val- ues were 1.34 °Brix and 1.35 °Brix, and the R 2 values were 0.95 and 0.92. These errors are competitive with those of works from other authors executed under less demanding conditions. The results obtained suggest that when combining hyperspectral imaging with appropriate chemometric techniques or machine learning algorithms, it is possible to have a satisfactory generalization for vintages not employed in model creation. Ó 2017 Elsevier B.V. All rights reserved. 1. Introduction Over the last years, Portugal has become very competitive in producing wines. One of the most famous Portuguese wines with renowned quality is Port wine which is a unique fortified wine pro- duced from grapes that grow in Douro appellation. In this work, we will focus on the development and comparison of two chemomet- ric and machine learning algorithms models, namely, PLSR and neural networks, for sugar content prediction in samples com- posed by only six whole Port wine grape berries, using hyperspec- tral imaging data collected in reflectance mode. The major novelty is that the grapes from one of the vintages used for test were not employed in the model creation. This evaluation, rare in scientific literature, is extremely relevant since testing a model with the same vintage employed in the training model does not ensure that the model will be adequate when tested with future vintages. In addition, the determination of sugar content using a small number of berries per sample is also uncommon and is more difficult than using a large number of berries. The relevance of the method developed for sugar content mea- surement comes from the need to maintain prominence in current markets by ensuring the high quality of the wines produced through the continuous improvement of the winemaking process. By measuring sugar content of grapes, which allows to evaluate the degree of ripening and is directly related with the alcoholic strength of the wine produced, it is possible to harvest wine grapes at the optimal point of maturity and to select them according to their quality making it possible to improve the quality of the grapes and resulting wines. Due to large differences in terroir, http://dx.doi.org/10.1016/j.compag.2017.06.009 0168-1699/Ó 2017 Elsevier B.V. All rights reserved. ⇑ Corresponding author. E-mail address: veroniquegomes@gmail.com (V.M. Gomes). Computers and Electronics in Agriculture 140 (2017) 244–254 Contents lists available at ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag