One new artificial intelligence (AI) system is being tested in the UK to identify weeds in plantations through cameras installed in agricultural sprayers. The technology promises to revolutionize the application of herbicides, reducing costs for farmers and reducing the amount of chemicals applied to the soil, explains the “BBC”.
The solution is being developed by the company Rothamsted Researchbased in Harpenden (Hertfordshire), which trained cameras capable of recognizing blackgrass (a type of weed) at different stages of growth. AI analyzes captured images in real time and tells the sprayer where to apply product, ensuring chemicals are only applied when and where they are needed.
David Comont, an ecologist at Rothamsted Research, explains that the system combines decades of scientific knowledge about this cutting-edge weed. “By directing herbicides only to places where there is an infestation, we are able to reduce the amount applied, save the producer money and maintain control over blackgrass”, he states.
Rothamsted Research worked with Bosch to train the AI by capturing thousands of images of agricultural fields. The algorithm was programmed to analyze photographs collected by dozens of cameras mounted on the sprayer boom. The technique works as a dynamic map: whenever the system detects black grass in an area of the land, the sprayer acts just at that point, rather than treating the entire area.
For three years, the project collected around 5000 images and mapped more than 100 hectares. According to data scientist Muhammad Kassemfrom Bosch, the model reached a accuracy of about 85%even in different crops such as wheat, barley or beans. “We captured so many images that we ended up needing several days just to download and clean them before training the AI,” he explained.
Funding for the initiative came from Agricultural Innovation Program of the United Kingdom Department for Food, Agriculture and Rural Affairs and UK Research and Innovation, a government institution that funds research in the country.