Agriculture Technology and Adoption Centre (AgTAC) Projects Drones & AI team up for healthy land management practices

Drones & AI team up for healthy land management practices

Lucy Gardner in the field with a drone.

Researchers are using high resolution drone imagery and AI technology to develop new tools for land condition assessment in the rangelands of north-west Queensland.

Drones allow people to capture high resolution images that can be highly useful for agricultural and rangeland management. For example, drone imagery can provide information about the health and condition of pasture in paddocks, the amount and distribution of weeds, as well as the locations of roads, fences and water sources.

To harness the full potential of such imagery, we need to be able to easily detect and classify key features present in each photograph.

Recent developments in artificial intelligence (AI) are opening the door to the more accurate and efficient classification of these features, helping to improve planning and management processes.

Lucy Gardner in field. Lucy Gardner with a soil core in a paddock in north-west Queensland.

We hope this approach will enable rapid and accurate assessment of Mitchell grass tussock density within a paddock. If new images are collected, these can also be input into the model, allowing one to track changes in tussock density through time. This approach could be utilised by graziers, natural resource management groups and other stakeholders to aid in land management decisions. - Dr Jack Koci

JCU Honours Student and AgTAC member, Lucy Gardner, supervised by Dr Jack Koci, is part of the team including Southern Gulf NRM, Maxus AI and the Tropical North Queensland Drought Hub, trialling the potential of integrating drones and AI technology.

The trial is being undertaken across the Southern Gulf of Queensland and is focussing initially on the identification and classification of Mitchell grass tussocks (Astrebla spp.).

Lucy Gardner in field with drone.

At each site, Lucy flies a drone collecting images of the paddock. The images are the fed into an AI platform. Individual tussocks of Mitchell grass appearing in a small subset of the images are then identified and manually labelled by Lucy. The labelled images are then used to train the AI model to automatically identify Mitchell grass tussocks in all other collected images.

Drone image of grasslands.

Example of Mitchell Grass tussocks identified by the AI Platform.

The project is supported by the Tropical North Queensland Drought Hub with funding provided by the Australian Government’s Future Drought Fund and is operating from 2022 to 2023.