To determine if image sensing technology can be used to improve the accuracy of block-level crop yield forecasts and to design a suitable system for large scale image collection with image processing methods that are robust.
Specifically, the project will develop three mobile, spatially aware systems based on a high precision sensor suite, a low cost sensor suite and smart devices to acquire data and associated software.
At various stages during each season, wineries ask grapegrowers to provide forecasts of yield. Differences between expected intake and actual delivery can lead to a series of problems spreading throughout the process of wine production and distribution, affecting harvest organisation, regional pricing negotiations, intake scheduling, tank space allocation, investment in winery capital equipment and the development of marketing strategies for domestic and export markets. Apart from the substantial economic benefits of improved crop forecasting, it is an essential first step to successful yield regulation. Currently, best practice systems used after fruit set, and based on manual measurements, are on average ± 15% out, which is substantially higher than the ± 5% winemakers and wine companies would like. The level of accuracy, the cost of labour and training requirements of current systems are major impediments to adoption. This project will substantially improve block level forecasts after fruit set and early in the season and also remove the other barriers to adoption through the application of image sensing technology. Using this approach, it is considered that the goal of ± 5% accuracy after fruit set is achievable for block level forecasts after fruit set.
Through the use of proximal sensing image technology, the project seeks to directly estimate fruit load and berry size at any time during the growing season. In contrast with remote sensing approaches that just assess the canopy, this project proposes to use image technologies to assess the fruit load and berry size directly. Early season estimates of yield will also be made by assessing inflorescence number and branching patterns. In tandem with existing yield estimation practices, the technology is claimed to enable accurate block-level forecasts to be made.
Chardonnay and Shiraz blocks in Orange and the Clare Valley will be used to trial the technology, on two trellis types – single wire VSP and single wire sprawl. The sensing technique will be compared against manual measurements of yield components. It is further intended to develop mobile systems to collect and process field data.
This project will provide the capability to produce more accurate, timely and less costly block level yield forecasts after fruit set and early in the season, through the application of image sensing technologies and mobile systems.