Satellites, AI Unlock Cane Disease Detection 

A ground-breaking tool developed by James Cook University researchers that harnesses the power of artificial intelligence and satellite imagery could help farmers win the fight against a major sugar cane disease. 

The tool has already shown promise in the Herbert River District, north of Townsville. 

JCU Research Officer Ethan Waters, Associate Professor of Electrical and Electronic Engineering Mostafa Rahimi Azghadi, and Master of Data Science Senior Lecturer Dr Carla Ewels will use a grant from Australia’s Economic Accelerator to continue to fine-tune their prototype technology over the next year. 

“The original idea was to utilise satellite imagery data and apply machine learning techniques to process that data in order to detect diseases and abnormalities in sugarcane,” Associate Prof Azghadi said. 

“This project is a collective effort to move this promising technology beyond its initial phase that Ethan has developed and bring it to industry.” 

The project evolved out of Mr. Waters’ Honours thesis in 2022, which saw him and his supervisors, Associate Professor Azghadi and Dr Ewels, partner with Ingham-based Herbert Cane Productivity Services Limited (HCPSL) to develop a tool which could detect Ratoon Stunting Disease (RSD) across 72 cane paddocks in the Herbert River District. 

“Ratoon Stunting Disease is a really big issue for not just HCPSL but the global sugar community because it has no visual symptoms but it affects the way water propagates through the stalk. It means the cane yield can be reduced by up to 60 per cent,” Mr. Waters said. 

“We discovered that we could use multi-spectral satellite imagery in near-infrared and short-wave infrared regions of the electromagnetic spectrum to get an indication of how much water is in the vegetation which would identify if the disease was in cane. We then trained the machine learning algorithms to identify the disease from that satellite iagery.” 

Associate Prof Azghadi said early detection of cane diseases or abnormalities would save farmers financial pain as they could take steps to minimise crop loss. 

“In the worst-case scenario with RSD, losing 60 per cent of your yield means roughly 60 per cent of your annual income if you look at the commercial cane sugar content of your crop,” he said. 

“So, if you can detect those diseases as early as possible, you can prevent the spread of the disease and improve your yield.” 

The team will also be working with growers in the Burdekin and Tully regions and other industry partners interested in harnessing the innovative tool, which can currently analyse five different cane varieties. 

“The grant focus is on detecting RSD first but we have added scope to expand this technology to other diseases in order to be an effective, real world solution,” Mr. Waters said. 

“We know it’s not just one disease that impacts the sugar cane industry and hoped to eventually expand the abilities of the tool to account for more sugar cane varieties and how other factors, such as weather conditions and soil type, could affect the crop.” 

“Over the past two years we’ve been able to work with the industry, gain their trust and gather more data. And that will only improve more as we head into the future,” he said. 

“That will allow us to build much larger, more generalised AI models and go to all different areas of Queensland, ideally expanding to anywhere there is sugarcane.”