South Africa Wine Technical Yearbook 2025

FIGURE 2. Data acquisition of individual bunches in the field. (a) RGB image with full canopy (FC); (b) RGB image with leaf removal (LR); (c) RGB-D (Kinect mesh) with FC; and (d) RGB-D (Kinect mesh) with LR. This figure appears in one of the articles published as part of the project: https://www.mdpi.com/1424-8220/19/17/3652.

Although HSI is still costly and complex, it is evolving rapidly. This project evaluated HSI under contrasting water conditions (non-water-stressed vs. water-stressed vines), marking the first step in developing a methodology for HSI analysis and demonstrating the general feasibility of this technology. However, further experiments are needed to evaluate this technique under a broader range of real-world water stress conditions. Pruning weight assessment The concept developed for canopy characterisation was extended to assess pruning weight, an important parameter for evaluating vine balance. Weighing the

FIGURE 3. Example of the method used to assess pruning weight by digital analysis.

Conclusions The results obtained in this study indicate the potential for using non-destructive techniques (such as computer vision, HSI and spectroscopy) as vineyard monitoring tools. However, each method has its advantages and disadvantages in terms of applicability, cost and complexity. These factors must be carefully considered when evaluating and applying these techniques. Future projects in this field should focus on the operational aspects of these methods, considering elements such as automatic analysis and user interfaces.  pruning mass of a couple of vines is simple, but on a large scale, it becomes almost impossible. Several techniques and conditions were evaluated to identify the most suitable method. The results suggest that pruning weight can be accurately determined using image analysis (Figure 3).

total weight at the vine level. Colour thresholding was very effective for classifying the bunches when a white panel was used; however, monitoring multiple vines with large structures can limit the applicability of this technique. In this regard, selecting target areas within a block can be a practical solution. Other approaches, such as complex machine learning models and nighttime imagery, should be considered for future studies. Water stress detection The detection and quantification of water stress are crucial for sustainable viticulture in light of current climate change scenarios. In this study, water stress was assessed under different levels and conditions. Field spectroscopy and hyperspectral imaging (HSI), using both laboratory and field sensors, showed promising results. For field spectroscopy, the success of the method depends on a rigorous calibration and data acquisition protocol; small changes can cause significant differences in the spectral signature.

For more information, contact Carlos Poblete-Echeverría at cpe@sun.ac.za. Reference https://www.wineland.co.za/near-real-time-characterisation-of-vines/

64

TECHNICAL YEARBOOK 2025