Charles Sturt University
Graham Centre for Agricultural Innovation

Weed Manager’s Guide to Remote Detection

Researchers

Project news

Harnessing remote detection technologies to fight weeds in complex landscapes

Summary

Testing  limitations and developing guidelines for the use of remote sensing technology across airborne platforms for weed detection in mixed landscapes.

Aim

This project will:

  • test the limitations of remote sensing technologies across various airborne platforms (UAV and satellite), to detect weeds in mixed landscapes
  • use machine learning to model relationships between low-resolution satellite imagery and corresponding high-resolution images, thereby enhancing the ability for landholders to use cheaper satellite imagery for detection of weeds.
  • develop guidelines in the use of remote sensing technology for weed detection in mixed landscapes, an online portal and community of practice for information and data-sharing amongst the weeds community.

What's involved

Study sites have been identified within three different landscapes across NSW housing the three model species:

1) Kosciusko National Park – Hawkweeds
2) Snowy Monaro region – African lovegrass
3) North and South Coast – Bitou bush


Site ground surveys will be undertaken four times at various plant phenological stages to facilitate recording of site vegetation and landscape features. These will correspond with all imagery taken simultaneously at the same sites.

The relationships between accuracy and spatial resolution for sites will reveal limits of detection in relation to operational parameters. Machine learning and image processing will determine detection limitations and optimal spatial scales for each weed system.

A set of guidelines will be developed from project results to provide broad instructions on the use of remote sensing technology for weed detection, and for application to other weed species and landscapes.

The online portal and community of practice will be developed to facilitate data and resource sharing, networking and support for end users within the remote weed detection community..

Hillary Cherry (DPIE), Assoc Prof Lihong Zheng (Charles Sturt), Dr Remy Dehaan, Dr Jane Kelly, Wendy Menz (DPIE).It brings together drone, machine learning and remote sensing researchers and weed experts from Charles Sturt, the Queensland University of Technology, and weed management experts from NSW National Parks and Wildlife Service.

This project is supported through funding from the Australian Government’s Established Pest Animals and Weeds Management Pipeline Program - Advancing Pest Animal and Weed Control Solutions

Other partners include South East, North Coast and Murray Local Land Services, Mid Coast, Eurobodalla and Bega Valley Shire Councils, the Illawarra District Weeds Authority, and XAg Australia.