Professor Mostafa Rahimi Azghadi. Credit: JCU
Machine learning methods and satellite data are being used to spot sugarcane disease early.
Researchers at James Cook University are combining machine learning and satellite data to spot sugarcane disease before it would normally begin to reveal itself.
Led by Professor Mostafa Rahimi Azghadi, the team has developed a software tool and tested its ability to accurately tell the difference between healthy and diseased sugarcane.
This is the first use satellite data to target asymptomatic Ratoon Stunting Disease (RSD).
“RSD can affect the yield of sugar by up to 60% and it’s highly contagious. But being asymptomatic, you can’t see it with the naked eye until the latter stages of the growing season,” said Prof Azghadi.
RSD is usually identified by cutting and sampling sugarcane by hand, and then sending the samples to laboratories for DNA analysis.
“It’s time consuming and expensive, especially if you want to do it at larger scale as every test costs about 10-15 dollars,” said Prof Azghadi.
“Depending on the sugarcane variety, our method was between 86 and 97% accurate… which is on par or better than other crop disease detection tools.”
Like a check-up with your GP
The study made use of various machine learning techniques to detect the presence of RSD in different varieties of sugarcane, using vegetation indices derived from the freely available Sentinel-2 data.
The scientists arranged for ground truth samples to be taken across 76 sugarcane blocks in the Herbert region of Queensland, Australia. This dataset was obtained by trained field agronomists from Herbert Cane Productivity Services.
The ground-truth data “was then used to extract the 76 sampled blocks from the Sentinel-2 imagery, ensuring that each pixel within the block geometries was labelled with both disease status and variety,” the researchers write in their paper.
The result was a finding that machine learning algorithms “can effectively classify RSD across several varieties with freely available satellite-based multispectral data”.
“RSD in sugarcane is just our first successful case study … our approach can be extended to other crops and other crop health challenges,” said Professor Azghadi.
“The long-term objective is to develop an early-warning tool that identifies disease risk and tracks overall crop health, making it easier to manage the health and vitality of farmer’s crops.”
“It’ll be a bit like a regular check-up with your GP, but for sugarcane and other crops.”
A promising approach
Writing in the journal Information Processing in Agriculture, the researchers say that “Our study highlights the potential of satellite-based remote sensing as a cost-effective and efficient method for large-scale sugarcane disease detection alternative to traditional manual laboratory testing methods”.
Highlighting the benefits the method brings when compared with traditional means of detecting RSD, they go on to say that “Freely available satellite-based remote sensing offers a cost-effective and efficient alternative to the traditional, resource-intensive methods of identifying and managing disease in sugarcane”.
“In particular, free publicly accessible multispectral satellite data can alleviate the financial burden of purchasing expensive spectrometer images, and promote the widespread adoption of this advanced technology.”
“These promising initial results, coupled with the efficiency of classifying 76 blocks within minutes as opposed to months, indicates the potential benefits of implementing a large-scale health monitoring system using satellites and machine learning.”