Improved Yield Prediction for the Australian Wine Industry
The project aims to improve upon current yield estimation methods, with a focus on techniques drawn from
the field of computer vision and automation.
Initially the project will assess, design and benchmark computer vision techniques in yield estimation
from various stages of the wine growth cycle. The project will then continue into the development
of a range of mobile low cost devices that can be used to help farmers and wineries gain a more accurate
estimate of yield in the field.
- Develop the ability to rapidly sense the amount of fruit on vines after fruit set in situ for two cultivars and two commonly-used
trellising systems at different times during the season.
- Describe the relationship between sampling frequency and the error of prediction for fruit sensing in vineyards with differing
trellising systems and cultivars at a vineyard block level and across multiple blocks after fruit set and close to harvest,
- Determine the ability of sensing approaches in the field and in the laboratory to accurately assess berry size, a key
parameter needed to predict final bunch weight.
- Explore the potential for image analysis to determine potential inflorescence size and yield after budburst, based on the
number of inflorescences and their branching patterns, to improve "early season" forecasts.
- Develop a field-mobile system, that integrates into existing vineyard operations and is based on image analysis, to predict
yield after budburst, after fruit set, at veraison and immediately prior to harvest.
> Australian Grape and Wine Authority Project Summary