Artificial intelligence/machine learning agriculture, food and animal sciences

Machine learning based modelling and artificial intelligence applications for agriculture, food and animal sciences.

For more information or to work with us, please contact:

Our group has made several contributions in the way that data from different technologies (UAV, remote sensing, sensor networks, among others) are analysed to obtain meaningful information for decision making.

These have been in terms of water and fertiliser management using sensor networks and UAV through machine learning and artificial intelligence. Water savings was our first objective since Australia is a water deficit country that requires irrigation technology for commercial agriculture.

Another climate change-related problem to solve was the effect of bushfires on viticulture, mainly related to smoke contamination and smoke taint in wines. We developed digital tools based on non-destructive proximal remote sensing, such as infrared thermal imagery and near-infrared spectroscopy (NIR) and machine learning to detect contamination in canopies, berries, and wines with high accuracy and using portable devices. Hence, winegrowers could detect levels of contamination in the field and apply amelioration management techniques in almost real-time, instead of sending samples to specialised laboratories and wait for results that could take days or weeks. These new technologies can be applied to predict smoke taint in berries and wines before winemaking, so grape growers and winemakers can make critical decisions without incurring in further operational costs and reducing losses.

DAFW is now publishing results from digital approaches to manage big data from vineyards and wineries accumulated for years to characterise wine quality traits while the grapevines are growing. Models that we are developing can predict quality traits and full aroma profiles of wines early in the season and at harvest. The implementation of these technologies gives information to winegrowers when to harvest to obtain the maximum quality potential or wine style required objectively.

Some DAFW recent research is based on the relationship between the level and patterns of berry cell death in the mesocarp with wine quality, which can be assessed using handheld NIR technology, as smoke taint and plant water status, offering multiple uses. Berry cell death is linked to quality traits and the development of flavour and aroma in berries and final wines, which makes the wine style. We found that different cultivars have different patterns of cell death that can be characterised using artificial neural network algorithms for classification.

This research could potentially take out the guesswork from viticulture and winemaking, creating “digital art” when producing wines, which is more objective. Similar digital tools have been applied to animal science to assess stress levels and welfare and their effects on food products, such as milk volume and quality traits and meat quality. New digital sensor developments from DAFW have been currently applied to the detection of pests and diseases in crops.