Using AI algorithm to improve strawberry disease detection

University of Florida Institute of Food and Agricultural Sciences (UF/IFAS) scientists are endeavouring to find new ways to help strawberry growers control diseases that can damage the crop.

In newly published research, Won Suk “Daniel” Lee, a professor of agricultural and biological engineering and Natalia Peres, a professor of plant pathology, show how artificial intelligence (AI) can improve leaf wetness detection (LWD).

Florida farmers have traditionally used the UF/IFAS-designed Strawberry Advisory System (SAS) to tell them when to spray fungicides to prevent plant diseases. SAS works with data generated by Florida Automated Weather Network stations near farms — in this case, near strawberry fields. It uses leaf wetness duration to help growers estimate the risk of their fruit getting infected with a fungal disease such as botrytis fruit rot and anthracnose.

These traditional methods for measuring LWD have been reliant on sensors with known limitations in accuracy and reliability, and difficulties with calibrating. To overcome these limitations, this study introduced an algorithm for leaf wetness detection systems using high-resolution imaging and deep learning technologies, including convolutional neural networks (CNNs).

The scientists trained the algorithm to use the images and detect wetness, and found that AI technology improved the accuracy of wetness detection. Nearly 96% of the time, the algorithm found moisture on the reference plate in comparison with manual observations, and a nearly 84% accuracy rate was observed when comparing with the current sensors and models in SAS.

“Ultimately, we want to replace the current wetness sensors with an imaging system, because the current sensors are difficult to calibrate and not always reliable,” Lee said.

“Using the AI system, we can detect wetness and consequently forecast the disease better, so we can help growers. The implementation of this advanced detection system within SAS may improve decisions about fungicide applications and may facilitate the implementation of leaf wetness detection for disease forecasting to other crop systems.”

Read the full findings here.

Image credit: iStock.com/firemanYU