Determining the number of bunches per plant, evaluating the average bunch weight, counting the number of berries per bunch, and estimating the average berry weight are all necessary tasks to reliably evaluate vineyard performance . These tasks have traditionally been carried out manually, which is impractical nowadays, where we seek to modernize and increase the efficiency of our viticultural systems. In this sense, the application of new artificial vision technologies for the analysis of the different components of the vineyard from images is assuming a real revolution, being able to characterize these components effectively, quickly and completely objectively.
Thus, the analysis of images taken with standard digital cameras (RGB images) is presented as an interesting option for the rapid and economic characterization of the vineyard components. As an example, it is worth highlighting a recent work (Klodt et al ., 2015) in which an efficient method is presented capable of segmenting the lateral image of a canopy into four different classes (or components): leaf, shoot, grape and bottom of image. With this information, important characteristics of the vineyard such as plant growth or canopy leaf surface can be evaluated efficiently and quickly . Likewise, this allows to quickly and objectively calculate important viticultural parameters (such as the leaf: fruit ratio, Figure 1),allowing the winegrower to act immediately . Likewise, there are studies that present the necessary methodology to automatically detect the number of berries that make up a bunch and estimate their average size, all from images taken directly in the vineyard under uncontrolled conditions (Herzog et al ., 2014 )
However, technologies based on image analysis are not limited to characterizing the components of the vineyard , but their potential in the field of viticulture is practically unlimited. Thus, the analysis of cluster images allows obtaining very valuable information for their morphological characterization , obtaining information on their size (length and width), shape and density (or compactness). These attributes are essential for obtaining quality grapes, having a clear impact on the commercial value of the grape, the health status of the bunch, and the yield of the crop. In this sense, image analysis has been proposed as an efficient method to quantify the number (and size) of free holes present in the cluster (Figure 2). According to the authors, this characteristic is closely related to the compactness of the bunch , with loose bunches presenting a greater number of free holes in their structure than compact bunches. Among others, this variable allowed the construction of a model capable of predicting cluster compactness in a non-invasive and objective way, presenting itself as an interesting alternative to the traditional evaluation method (Cubero et al ., 2015).
