Project Number
DVO 07
Project title
Near real-time characterisation of grapevines for more efficient vineyard management
Project leader
Carlos Poblete-Echeverría
Institution
Department of Viticulture and Oenology, Stellenbosch University
EXECUTIVE SUMMARY
Objectives and Rationale
This project was implemented to investigate the quantification of canopy architecture using new platforms, with the goal of defining canopy characteristics at the plant level in a vineyard. The concepts developed during the project were also expanded to include water stress determination, pruning evaluation, and yield estimation.
Methods
Laboratory and field experiments were conducted to evaluate different image-based approaches (machine vision) to determine canopy characteristics, yield (bunch weight) and pruning weight. Spectrometry and Hyperspectral image (HSI) analysis were assessed as alternative methods to detect water stress at leaf level.
Key results
The remote techniques used for canopy characterisation (RGB images, LiDAR, and multispectral images) and pruning mass estimation (RGB and RGB-D) showed good results, with agreements around 80% under optimal conditions. Regarding yield estimation, our results confirm that RGB and RGB-D methods are effective in estimating bunch weight in both laboratory and field conditions when the bunches were fully exposed (leaf removal treatment) as well as in standard conditions (no canopy interventions). Field spectroscopy and hyperspectral imaging (HSI) (acquired at the laboratory and field) showed promising results; however, further experiments are needed to evaluate these techniques in a broader range of water stress conditions since, in this project, the detection of water stress with these techniques was evaluated under particular contrasting water conditions (non-water-stressed vs water-stressed vines).
A key conclusion of the discussion
Several fast and noninvasive computer vision and remote sensing techniques were implemented to assess grapevine features (Canopy characterisation, Yield, Water stress, and Pruning mass). The results obtained in this project confirm the potential of the studied techniques; however, each method has its pros and cons, which must be considered when evaluating and using these techniques in practice.
Take home message for the industry
From a research point of view, the project was very successful in terms of evaluating and implementing the studied techniques and obtaining promising results. However, from a practical point of view, future projects should focus on operational aspects, considering elements such as automatic analysis and user interface.