Remote sensing of priority weeds in sugarcane

We are looking for students with experience in one or more of the following: deep learning, computer vision, remote sensing, drone image capture and analysis and drone pilot skills.

About the project

Distribution of troublesome weeds at a district level is largely unknown. The lack of information on distribution of troublesome weeds means that:

• threats can be prioritised incorrectly. Some emerging weeds can be overlooked, and no control strategies developed.
• spread of emerging weeds over time and space is unknown, leading to a reactive approach instead of proactive district strategies that could prevent quick dispersal.

Aerial imagery of cane farms from drone for the purpose of weed scouting and spot spraying currently need manual interpretation by an agronomist to be converted into weed maps and spray maps. This task is time consuming and limit the adoption of the spot spray technology using drones.

Aim: This project aims to facilitate the development and early deployment of a platform that maps weeds from drone imagery and generates spray maps; and a spot-and-spray system with a clear commercialisation pathway for sugarcane. It also aims to explore diagnostic technology using satellite imagery to identify the target weeds and map their distribution at a paddock, farm and district levels.

Approach: Images of six target weeds in sugarcane paddocks in North and Far North Queensland will be captured by InFarm using their drones and camera sensors from January 2025 to April 2027. The project will also work with AI models and techniques for accurate detection of the target weed species using InFarm drone and potentially using satellite imagery.   

The project will implement the AI models and techniques into the InFarm processing pipeline and capture and process pre-trial imagery using small drones to guide project data collection decisions. 

Methods may include:

• Satellite imagery
• Drone sensors
• AI for weed identification: Train AI detection models (e.g., object detection, instance segmentation) using state-of-the-art techniques such as Convolutional Neural Networks or multi-modal vision-language models
• Weed map generation

Location

You will be enrolled at the James Cook University and will spend time at the SRA Meringa Research Station in Gordonvale, QLD and work remotely with InFarm.

As a student in the Training Centre, we’ll also cover your travel to Centre Forums, Training Retreats, and conferences. This will include networking with like-minded people by the beach in tropical North Queensland, in the bush outside Canberra, and in capital cities around Australia.

How to apply

Submit an expression of interest form by 22 January 2025 on our website: https://plantbiosecuritycentre.edu.au/remote-sensing-of-priority-weeds-in-sugarcane/