This Plant Patch Uses AI to Detect Crop Disease and Drought 

This Plant Patch Uses AI to Detect Crop Disease and Drought 

Diseases in crops are devastating. They cause economic losses, and some can negatively affect human and animal health. Some which have grown successfully were met with rejection due to the risks.  

Farmers and scientists have been tireless when it comes to providing solutions that can increase resistance as well as reduce risks.  

Now, we have AI for many things. Scientists have figured out if we can improve agriculture with it and have begun experimenting. What they’ve come up with are a high-tech plant patch and an algorithm to protect plants and crops from disease and other threats. 

Researchers at the North Carolina State University were the ones who made the patch. It’s an electronic device which we can stick on leaves to monitor the plant for different pathogens (like viral and fungal infections) and stresses such as drought. 

In testing, the patch was able to detect a viral infection in tomato plants before growers would be able to detect any visible symptoms of disease. 

“This is important because the earlier growers can identify plant diseases or fungal infections, the better able they will be to limit the spread of the disease and preserve their crop,” said the corresponding author Dr Qingshan Wei.  

Dr Wei added that when growers can identify stresses such as irrigation water contaminated by saltwater intrusion, they will be able to come up with solutions better, and in turn, improve crop yield. 



The need for AI assistance 

So far, scientists have tried to improve agriculture by phenotyping. It’s a process when they grow genotypes of the crop, infecting them with the disease, and looking for symptoms. Phenotyping is successful when it identifies resistant genotypes that don’t develop symptoms, or less severe symptoms. 

When symptoms develop, researchers try to identify the genes related to disease resistance and then put those genes in high-performing hybrids of the crop. 

Despite the technology, a lot of phenotyping is done manually, so it’s a long repetitive process. That’s why researchers utilize artificial intelligence to accelerate the process.  

“Normally, we look at a petri dish of kernels and then give it a subjective rating. It’s very mind-numbing work. You have to have people specifically trained and it’s slow, difficult, and subjective,” said Jessica Rutkoski, co-author of a study from the University of Illinois College of Agricultural, Consumer and Environmental Sciences which developed the algorithm. 

In the University of Illinois study, the researchers started with algorithms similar to those used by tech giants for object detection and classification. However, they needed to develop the capability to discern minute differences in diseased and healthy plants (in this study, wheat kernels). 

Co-author Girish Chowdhary said that their algorithm’s uniqueness is that they’ve trained the network to detect minutely damaged kernels with good enough accuracy using just a few images. 

“We made this possible through meticulous pre-processing of data, transfer learning, and bootstrapping of labeling activities. This is another nice win for machine learning and AI for agriculture and society,” Girish said. 

Combining AI with plants 

We haven’t reached a point where plants have a synthetic AI stalk just like in science fiction. Maybe we’ll get there someday, but for now, AI is just a means to an end. 

For the NC State University study, Dr Wei and team developed the patch based on a previous prototype, which detected plant disease by monitoring volatile organic compounds (VOCs) emitted by plants. 

According to the NC State University researchers, the plants emit different combinations of VOCs under different circumstances. And by targeting VOCs that are relevant to specific diseases or plant stress, the sensors can alert gardeners and farmers to specific problems. 

“The new patches incorporate additional sensors, allowing them to monitor temperature, environmental humidity, and the amount of moisture being ‘exhaled’ by the plants via their leaves,” said co-corresponding author Professor Yong Zhu. 

The AI patches are only 30 millimeters (about 1.18 in) long, consisting of a flexible material containing sensors and silver nanowire-based electrodes. 

As I mentioned, the technology is not inside the plants, but rather placed on the underside of the leaves because those areas have a higher density of stomata. Stomata are pores that allow the plant to “breathe” by exchanging gases with the environment. 



Testing patches on tomato plants 

The NC State study tested the patches on tomato plants grown in greenhouses. The patches had different combinations of sensors. 

They found that the tomato plants were infected with three different pathogens: tomato spotted wilt virus (TSWV); a fungal infection called early blight, and late blight, which is a type of pathogen called an oomycete. 

As a part of the experiment, the plants were also exposed to a variety of stresses such as overwatering, drought conditions, lack of light, and high salt concentrations in the water. 

After getting the data, the researchers input them into an artificial intelligence program to determine which combinations of sensors worked most effectively to identify both disease and stress. 

According to Dr Wei, the results produced by the AI detection of all challenges were promising. 

“For example, we found that using a combination of three sensors on a patch, we were able to detect TSWV four days after the plants were first infected. This is a significant advantage since tomatoes don’t normally begin to show any physical symptoms of TSWV for 10 to 14 days,” Dr Wei said. 

Not without challenges 

AI on plants is a new technology, so there are challenges that researchers need to overcome. For the NC State study, the team need to make the patches wireless, which to them is a relatively simple challenge.  

Second, they need to test the patches in the field—outside greenhouses—to ensure that this technology works well under real-world conditions. 

The team has been looking for industry and agriculture partners to help them move forward with developing and testing the patch in actual conditions. 

As for the University of Illinois study, the team has found that when compared to humans rating disease damage on kernels in the lab, the algorithm was only 60% as accurate.  

However, the researchers are not discouraged, as their initial tests didn’t use a large number of samples to train the model. They’ve been adding more data and expect to achieve greater accuracy with additional tweaking. 

According to Rutkoski, what’s important is to create an online portal where growers could upload cell phone photos of wheat kernels for automatic scoring of fusarium damage. 

Despite the challenges, the researchers of both studies are one step forward to ensure plant resistance to diseases and make agriculture industry thrive more. 

“This could be a significant advance to help growers prevent small problems from becoming big ones and help us address food security challenges in a meaningful way,” Prof Zhu said. 



Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.