Original research articles

Using of a combined approach by biochemical and image analysis to develop a new method to estimate seed maturity stage for Bordeaux area grapevine


Aim: The importance of phenolic maturity (depending on tannins and anthocyans) of the grape is crucial at harvest and determines the final quality of wine. The work presented here aims to characterize the evolution of phenolic maturity of seeds for 3 varieties combining macroscopic analysis and biochemical analyzes of these tannins at phenological stages of interest.

Methods and results: Using R software for macroscopic analyzes have been shown that the color varies dramatically (from green to dark brown) in the two months between the bunch closure and maturity. Biochemical analysis (HPLC measurement) shows that tannins in seeds are increasing from bunch closure to early veraison and decrease after this step until maturity.

Conclusion: All together these results have shown that the color variation is correlated to the tannins content in the seeds.

Significance of the study: Nowadays, no easy ways of prediction of phenolic maturity are known. The aim of this work is to use these results (usually considered independently) to have an knowledge of seed level of phenolic maturity necessary without biochemical analysis to establish a date of great harvest for the most favorable conditions for the extraction of tannins required for the organoleptic quality of a wine .The originality of this work is to use the combined visual seeds and its biochemical composition in tannins (correlation established by CPA). Forward, these results will help to develop a decision support tool based on simply system to acquiring seed image easily usable by winemakers.


Grape maturity at harvest is crucial and determines the final quality of wine. The maturity of grapevine berries is the combination of optimal technological and phenolic maturity. If technological maturity is well documented and described as sugar and acidity balance (Jones and Davies, 2000; Ollat et al., 2002; Doshi et al., 2006; Castellarin et al., 2016), phenolic maturity depends on various secondary metabolites including tannins (Geny et al., 2003; Ali et al., 2010; Pantelić et al., 2016; De Santis et al., 2017). These phenolic compounds play an important role in red wine organoleptic characteristics, especially the mouthfeel properties like bitterness and astringency. They are also involved in the colour stabilization and evolution of red wine due to their interaction with anthocyanins (Singleton, 1992; Boulton, 2001; Hanlin et al., 2010). Tannins have been partially characterized (Kennedy et al., 2000; Downey et al., 2003; Bogs et al., 2005) and we know that they come from the skin (Cerpa-Calderón and Kennedy, 2008) and seeds of grapes (Kennedy et al., 2000; Downey et al., 2003; Cadot et al., 2006). To determine the phenolic maturity, methods such as chemical analysis as well as pH, anthocyanin or flavan-3ol measurement exist (Glories’ method; Ribéreau-Gayon et al., 2006). These methods are effective but require laboratory analyses. During the past 10 years, other subjective methods have been developed to determine phenolic maturity. These experiments are based on the knowledge of oenologists and their ability to taste berries and seeds (Letaief et al., 2013). In 2005, Ristic and Iland (2005) initiated research addressing seed development and tannin accumulation and showed correlations with colour evolution. Fredes et al. (2010) used information from this study to develop a comparison method of seed colour against a colour scale. Recently, new approaches such as sensory and instrumental texture measurement or FT-NIR analysis for grape seed maturity characterization have been initiated (Torchio et al., 2012; Letaief et al., 2013; Rodríguez-Pulido et al., 2014; Brillante et al., 2015).

In this paper, both the colour and the tannin composition of grape seeds were analyzed. The originality of our work is to use simple acquisition of numerical data and combine these virtual seed colours with tannin composition (correlation established by principal component analysis (PCA)). This combination allows avoidance of colour scales. These results will help develop a decision support tool based on seed image acquisition.

Materials and Methods

Plant material

Experiments were performed during the growing season of 2013 on three grape cultivars (Vitis vinifera L. Cabernet-Sauvignon – clone no. 337, Sauvignon blanc – clone no. 242 and Tannat – clone no. 474) from an experimental vineyard in the Bordeaux area (44.791730 N, 0.579145 E), which are known to present different tannin concentrations in wine (Ribéreau-Gayon et al., 2006).

Ten grape clusters were removed at four points, from “bunch closure” to “ripening” (stages 33 to 38 according Eichhorn and Lorenz, 1977), which correspond to key steps in the synthesis of seed tannins (Darné, 1991).

Numerical data acquisition

Twelve grape seeds, randomly selected from each cultivar, were scanned with an Epson scanner (Epson Perfection 3200 Photo) on the ventral side. The statistical software R (R Development Core Team, 2008) was used to analyze the variations in seed colour according to the maturation status and the variations in colour inside the seed. Images were read using the R package EBImage (Pau et al., 2010). The image background was excluded to analyze only the seeds using the R packages EBImage and rtiff (Kort, 2015; Pau et al., 2010). The green frame of the red/green/blue (RGB) images gave the best contrast between seeds of different maturation status. To detect several colours in the image, the histogram of the green frame was seen as a mixture of Gaussian distributions where each Gaussian distribution corresponded to one colour. The R package MCLUST (Fraley and Raftery, 2006) was used to detect from 1 to 3 Gaussian distributions that composed the histogram and estimated the mean, variance and weight in the mixture distribution of each Gaussian. For each seed, the number of pixels contained in each Gaussian was calculated. The images were recoloured with blue, green or red according to the probability of membership of each pixel to each cluster.

Chemical analysis

Extraction of phenolics from seed samples was performed according to Pekić et al. (1998). After fine grinding of seed samples (Retsch Ultra Centrifugal Mill ZM200, Retsch GmbH & Co, Haan, Germany), two successive macerations were carried out for 3 h each, at room temperature by stirring 0.5 g of seed powder in 40 mL of MeOH/12 N HCl (99.9:0.1; v/v). After incubation, total phenols were directly measured at 280 nm after diluting to 1:100. Data were expressed in mg total phenols (tannins) per g fresh weight, in mg per seed or in mg per berry and represent the mean (± standard deviation) of three replicates.

Statistical analysis

All results reported here are means ± standard error of the mean of at least three independent experiments. An ANOVA was performed to test the cultivar effect on mean total tannin concentration. The ANOVA test was run using R software. PCA correlation was established between tannin concentration and colour clusters using R software.

Results and discussion

This work aimed to combine macroscopic analysis of the evolution of seeds for three contrasting tannin grape varieties (Cabernet-Sauvignon, Tannat, Sauvignon blanc) and chemical analyses identifying tannin concentrations at phenological stages of interest (bunch closure, beginning of veraison, end of veraison and ripening).

The macroscopic analyses showed a similar pattern of change over time between the seeds, whatever the cultivar: seeds went through a growth phase (during the first two stages studied) characterized by an elongation, a broadening and an increase in seed surface and, towards the end of ripening, a withering that became maximum at maturity (data not shown).

Besides the physical pattern of evolution, the colour varied dramatically in the two months between bunch closure and fruit maturity regardless of the cultivar (Figure 1A). It went from a green to a dark brown pattern, which was quantified using an R program. This program helped to identify three clusters of pixels according to their colour (represented in false blue, green or red colours) in each seed analyzed and to determine the proportion of pixels displaying the same colour (Figures 1B and 2).

Figure 1. Pictures of seeds from three grape cultivars at four phenological stages

(A) Real seeds and (B) clusters of colours after image treatment.

In the 2013 vintage, cluster evolution was different for the three grape varieties (Figure 2). Red and green were the two major clusters found (Figures 1B and 2). Cabernet-Sauvignon and Sauvignon blanc had the same progression pattern, with a drastic increase in the proportion of pixels belonging to the red cluster (Figure 1B) and a notable decline in that belonging to the green cluster between the beginning and the end of veraison. This result was more important for Sauvignon blanc (decrease from 92% to 36% of the pixels belonging to the green cluster) than for Cabernet-Sauvignon (75% to 40%). Tannat showed a delay in red cluster changeand at the end of veraison, only 15% of the pixels belonged to the red cluster and < 50% (41%) were measured at ripening (Figure 2).

Figure 2. Colour cluster separation of three grape varieties and four phenological stages

Analysis of grape seed tannin evolution has normally been performed according to Ristic and Iland (2005) by comparison of colours against a colour scale and has been confirmed for different grape varieties (Rodríguez-Pulido et al., 2012a,b; Zuñiga et al., 2014); however, nothing has been done on Bordeaux grape cultivars. Here, extracted colours have been compared to each other. By using digital imagery, the use of a colour scale was unnecessary. Numerical data have already been used in precision viticulture studies (Rodríguez-Pulido et al., 2012a,b; Whalley and Shanmuganathan, 2013; Zuñiga et al., 2014), all of which have investigated ripeness estimation and phenolic maturity of grape seeds using DigiEye® imaging or NIR hyperspectral imaging. However, if these methods are scientifically approved, their adoption by wine growers is still difficult. All these results led us to use a common scanner and software. Literature and our results allow us to consider evidence that artificial colour variation over time could be representative of seed maturity status. But colours alone are not sufficient to estimate phenolic maturity of seeds. Therefore, we determined total tannins to get a biochemical profile of our seeds. Total tannins in seeds increased from bunch closure to veraison and decreased from veraison to fruit maturity in all grape cultivars. Nevertheless, there were some differences between grapevine varieties. Total tannins were 1.5-fold and 1.2-fold higher for Cabernet Sauvignon than for Sauvignon blanc and Tannat, respectively, in seeds and also in berries at the end of veraison (Figure 3).

Figure 3. Total tannin determination

Evolution of mean total tannins in (A) mg / g fresh material, (B) mg / seed and (C) mg / berry. For all graphs: Cabernet-Sauvignon is represented by a solid line, Sauvignon blanc by a dashed line and Tannat by a dotted line. BC = bunch closure, BV = beginning of veraison, EV = end of veraison and R = ripening.

Veraison is the key point for tannin accumulation in grapes. In fact, subsequent to that point, tannins are not accumulated and display a decrease that is cultivar-dependent. For Cabernet Sauvignon, total tannins decreased 1.5 fold between the end of veraison and harvest (Figure 3A). The slope is less important (1.2 fold) but still significant for Sauvignon blanc and Tannat, in seeds and even more in berries (Figure 3B and C). For Sauvignon blanc, a large proportion of total tannins declined from the start of veraison (12.4 mg/berry) to reach 7.7 mg/berry at the end of veraison (Figure 3C). Overall, the tannin profile in seeds and, as a consequence, in berries, is consistent with literature. In grape seeds, tannin biosynthesis starts with seed development. This synthesis pathway, with the increase in seed tannin synthesis before veraison followed by an important decrease from the end of veraison until harvest, has previously been reported (Romeyer et al., 1985; Katalinić and Maleš, 1997; Kennedy et al., 2000). Our observations are also consistent with profiles obtained for Shiraz (Downey et al., 2003), Sangiovese (Filippetti et al., 2015)and Cabernet franc (Cadot et al., 2006). Taken independently, these results do not predict the level of phenolic maturity necessary to establish an optimal harvest date for the extraction of tanninsand hence for the aromatic quality of wine. So, a combination of macroscopic determination and biochemical measurement by PCA allows linkage of those two major metrics – colour clusters and tannin composition – of grape seeds. First, PCA allowed demonstration of the correlation between colour clusters and chemical analysis. It showed that the cluster represented in green appeared linked to total tannins (Figure 4A): total tannin measurement was linked to the green cluster (mostly observed at veraison) at nearly 58%. Second, another PCA allowed us to test the correspondence between grape cultivar and phenological stage. It showed that three groups were apparent (Figure 4B). One of them grouped bunch closure and the beginning of veraison (5-6) of Sauvignon blanc and (9) Tannat. The second one grouped bunch closure and the beginning of veraison (1-2) of Cabernet Sauvignon and the end of veraison (10) of Tannat. Finally, a third group emerged with the end of veraison and ripening (3-4) of Cabernet-Sauvignon and (11) Tannat.

This result is in accordance with the macroscopic analysis, which showed a delay of maturity for Tannat until the end of veraison that disappeared by ripening (Figures 1 and 2).

This first study gives us many interesting results to establish solid perspectives. Although we need to confirm these initial conclusions with seeds from other vintages and cultivars, these preliminary observations show that grape seed colour seems to be linked to phenolic maturity and convince us to persevere in developing a new way to establish the phenolic maturity of grape berry by image analysis using common informatics tools.

Figure 4. Principal component analysis (PCA) between colours, total tannins and phenological stages

(A) Visualization of correlation between colour cluster and total tannins. (B) Cluster of grape cultivar and phenological stage. 1 to 4 (Cabernet-Sauvignon), 5 to 7 (Sauvignon blanc) and 8 to 11 (Tannat) represent bunch closure (BC), beginning of veraison (BV), end of veraison (EV) and ripening (R), respectively.


  • Ali K., Maltese, F., Choi, Y. H. and Verpoorte, R., 2010. Metabolic constituents of grapevine and grape-derived products. Phytochemistry Reviews, 9, 357–378. doi:/10.1007/s11101-009-9158-0
  • Bland R. D., Clarke T. L. and Harden L.B., 1976. Rapid infusion of sodium bicarbonate and albumin into high-risk premature infants soon after birth: a controlled, prospective trial. American Journal of Obstetrics and Gynecology, 124, 263–267. doi:/10.1016/0002-9378(76)90154-X
  • Bogs J., Downey, M. O., Harvey, J. S., Ashton, A. R., Tanner, G. J. and Robinson S. P., 2005. Proanthocyanidin synthesis and expression of genes encoding leucoanthocyanidin reductase and anthocyanidin reductase in developing grape berries and grapevine leaves. Plant Physiology, 139, 652–663. doi:/10.1104/pp.105.064238
  • Boulton R., 2001. The copigmentation of anthocyanins and its role in the color of red wine: a critical review. American Journal of Enology and Viticulture, 52, 67–87.
  • Brillante L., Gaiotti, F., Lovat, L., Vincenzi, S., Giacosa, S., Torchio, F. and Tomasi D., 2015. Investigating the use of gradient boosting machine, random forest and their ensemble to predict skin flavonoid content from berry physical–mechanical characteristics in wine grapes. Computers and Electronics in Agriculture, 117, 186–193. doi:/10.1016/j.compag.2015.07.017
  • Cadot Y., Mi-ana-Castelló, M. T. and Chevalier M., 2006. Anatomical, histologicaland histochemical changes in grape seeds from Vitis vinifera L. cv Cabernet franc during fruit development. Journal of Agricultural and Food Chemistry, 54, 9206–9215. doi:/10.1021/jf061326f
  • Castellarin S.D., Gambetta G. A., Wada H., Krasnow M. N., Cramer G. R., Peterlunger E. Shackel K.A. and Matthews M.A., 2016. Characterization of major ripening events during softening in grape: turgor, sugar accumulation, abscisic acid metabolism, colour developmentand their relationship with growth. Journal of Experimental Botany, 67, 709–722. doi:/10.1093/jxb/erv483
  • Cerpa-Calderón F.K. and Kennedy J.A., 2008. Berry integrity and extraction of skin and seed proanthocyanidins during red wine fermentation. Journal of Agricultural and Food Chemistry, 56, 9006–9014. doi:/10.1021/jf801384v
  • Darné G., 1991. Recherches sur la composition en anthocyanes des grappes et des feuilles de vigne (Doctoral dissertation, Bordeaux 1).
  • De Santis D., Frangipane, M. T., Brunori, E., Cirigliano, P. and Biasi, R., 2017. Biochemical markers for enological potentiality in a grapevine aromatic variety under different soil types. American Journal of Enology and Viticulture, 68, 100–111. doi:/10.5344/ajev.2016.15123
  • Doshi P., Adsule, P. and Banerjee K., 2006. Phenolic composition and antioxidant activity in grapevine parts and berries (Vitis vinifera L.) cv. Kishmish Chornyi (Sharad Seedless) during maturation. International Journal of Food Science and Technology, 41, 1–9. doi:/10.1111/j.1365-2621.2006.01214.x
  • Downey M.O., Harvey J. S. and Robinson S.P., 2003. Analysis of tannins in seeds and skins of Shiraz grapes throughout berry development. Australian Journal of Grape and Wine Research, 9, 15–27. doi:/10.1111/j.1755-0238.2003.tb00228.x
  • Eichhorn K.W. and Lorenz D.H., 1977. Phenological development stages of the grape vine. Nachrichtenblatt des Deutschen Pflanzenschutzdienstes, 29, 119-120.
  • Filippetti I., Movahed, N., Allegro, G., Valentini, G., Pastore, C., Colucci, E. and Intrieri, C., 2015. Effect of post-veraison source limitation on the accumulation of sugar, anthocyanins and seed tannins in Vitis vinifera cv. Sangiovese berries: late trimming delays sugar accumulation in grapes. Australian Journal of Grape and Wine Research, 21, 90–100. doi:/10.1111/ajgw.12115
  • Fraley C. and Raftery A.E., 2006. MCLUST Version 3: An R Package for Normal Mixture Modeling and Model-Based Clustering. Technical report, DTIC.
  • Fredes C., Von Bennewitz, E., Holzapfel, E. and Saavedra, F., 2010. Relation between seed appearance and phenolic maturity: a case study using grapes cv. Carménère. Chilean Journal of Agricultural Research, 70, 381–389. doi:/10.4067/s0718-58392010000300005
  • Geny L., Saucier, C., Bracco, S., Daviaud, F. and Glories, Y., 2003. Composition and cellular localization of tannins in grape seeds during maturation. Journal of Agricultural and Food Chemistry, 51, 8051–8054. doi:/10.1021/jf030418r
  • Hanlin R.L., Hrmova M., Harbertson J.F. and Downey M.O., 2010. Review: condensed tannin and grape cell wall interactions and their impact on tannin extractability into wine. Australian Journal of Grape and Wine Research, 16, 173–188. doi:/10.1111/j.1755-0238.2009.00068.x
  • Jones G.V. and Davis R.E., 2000. Climate influences on grapevine phenology, grape compositionand wine production and quality for Bordeaux, France. American Journal of Enology and Viticulture, 51, 249–261.
  • Katalinić V. and Maleš P., 1997. Compositional changes in grape polyphenols throughout maturation. Journal of Wine Research, 8, 169–177. doi:/10.1080/09571269708718117
  • Kennedy J.A., Matthews M.A. and Waterhouse A.L., 2000. Changes in grape seed polyphenols during fruit ripening. Phytochemistry, 55, 77–85. doi:/10.1016/S0031-9422(00)00196-5
  • Kort E., 2015. Read and Write TIFF Files.
  • Letaief H., Maury C., Symoneaux R. and Siret R., 2013. Sensory and instrumental texture measurements for assessing grape seed parameters during fruit development. Journal of the Science of Food and Agriculture, 93, 2531–2540.
  • Ollat N., Diakou-Verdin, P., Carde, J.-P., Barrieu, F., Gaudillère, J.-P.and Moing, A., 2002. Grape berry development: a review. Journal International des Sciences de la Vigne et du Vin, 36, 109-131. doi.org/10.20870/oeno-one.2002.36.3.970
  • Pantelić M.M., Dabić Zagorac D.Č., Davidović S M., Todić S.R., Bešlić Z.S., Gašić U.M., Tešić Z. and Natić M.M., 2016. Identification and quantification of phenolic compounds in berry skin, pulpand seeds in 13 grapevine varieties grown in Serbia. Food Chemistry, 211, 243–252.doi:/10.1016/j.foodchem.2016.05.051
  • Pau G., Fuchs, F., Sklyar, O., Boutros, M. and Huber, W., 2010. EBImage--an R package for image processing with applications to cellular phenotypes. Bioinformatics, 26, 979–981.doi:/10.1093/bioinformatics/btq046
  • Pekić B., Kovač, V., Alonso, E. and Revilla, E., 1998. Study of the extraction of proanthocyanidins from grape seeds. Food Chemistry, 61, 201–206. doi:/10.1016/S0308-8146(97)00128-3
  • R Development Core Team: R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2007.
  • Ribéreau-Gayon P., Dubourdieu, D. and Donèche B., 2006. Handbook of Enology (2nd ed). Chichester, West Sussex, England; Hoboken, NJ: John Wiley. doi:/10.1002/0470010398
  • Ristic R. and Iland P.G., 2005. Relationships between seed and berry development of Vitis vinifera L. cv Shiraz: developmental changes in seed morphology and phenolic composition. Australian Journal of Grape and Wine Research, 11, 43–58. doi:/10.1111/j.1755-0238.2005.tb00278.x
  • Rodríguez-Pulido F.J., Ferrer-Gallego R., Lourdes González-Miret M., Rivas-Gonzalo J.C., Escribano-Bailón M.T. and Heredia F.J., 2012a. Preliminary study to determine the phenolic maturity stage of grape seeds by computer vision. Analytica Chimica Acta, 732, 78–82. doi:/10.1016/j.aca.2012.01.005
  • Rodríguez-Pulido F.J., Gómez-Robledo L., Melgosa M., Gordillo B., González-Miret M.L. and Heredia F.J., 2012b. Ripeness estimation of grape berries and seeds by image analysis. Computers and Electronics in Agriculture, 82, 128–133. doi:/10.1016/j.compag.2012.01.004
  • Rodríguez-Pulido F.J., Hernández-Hierro J.M., Nogales-Bueno J., Gordillo B., González-Miret M.L. and Heredia F.J., 2014. A novel method for evaluating flavanols in grape seeds by near infrared hyperspectral imaging. Talanta, 122, 145–150. doi:/10.1016/j.talanta.2014.01.044
  • Romeyer F.M., Macheix J.-J. and Sapis J.-C., 1985. Changes and importance of oligomeric procyanidins during maturation of grape seeds. Phytochemistry, 25, 219–221. doi:/10.1016/S0031-9422(00)94532-1
  • Singleton V.L., 1992. Tannins and the qualities of wines. In R. W. Hemingway and P. E. Laks (Eds.), Plant Polyphenols: Synthesis, Properties, Significance, pp. 859–880. Springer US. doi:/10.1007/978-1-4615-3476-1_51
  • Torchio F., Giacosa S., Segade, S. R., Mattivi F., Gerbi, V. and Rolle L., 2012. Optimization of a method based on the simultaneous measurement of acoustic and mechanical properties of winegrape seeds for the determination of the ripening stage. Journal of Agricultural and Food Chemistry, 60, 9006–9016. doi:/10.1021/jf302548t
  • Whalley J.L. and Shanmuganathan S., 2013. Applications of image processing in viticulture: a review. 20th International Congress on Modelling and Simulation, Adelaide, Australia, 1–6 December 2013.
  • Zuñiga A., Mora M., Oyarce M. and Fredes C., 2014. Grape maturity estimation based on seed images and neural networks. Engineering Applications of Artificial Intelligence, 35, 95–104. doi:/10.1016/j.engappai.2014.06.007


Amélie RABOT

Affiliation : Université de Bordeaux Institut des Sciences de la Vigne et du Vin UR Oenologie EA 4577 - USC 1366 INRA
Country : France
Biography :

210 Chemin de Leysotte - CS 50008
33882 Villenave d'Ornon cedex



No supporting information for this article

Article statistics

Views: 646


XML: 23

PDF: 121