The targeted application of pesticides is one of the areas of interest for so-called ‘precision agriculture’. With the help of specialised drones, farmers can monitor their fields, and use the data to determine which places are most affected by weeds and need to be treated. Artificial intelligence could make it easier for growers to evaluate images, and scientists from the Faculty of Agronomy at Mendel University (AF MENDELU) are currently working on optimising the entire process, while also focusing on methods to find difficult-to-detect weeds.
Optimising the use of pesticides in the cultivation of agricultural crops is one of the long-term goals of European policy. “The results of a number of studies show significant savings on herbicides when they are applied in a targeted manner, either in the form of point application or area variable dosing. The volume of spraying used can be reduced by 30 to 80%,” explained Vojtěch Slezák from the Institute of Agrosystems and Bioclimatology of AF MENDELU.
The result of targeted application is not only less burden on the environment, but also lower costs for growers. Farmers obtain the necessary data from images taken by drones. “We observe a boom in the use of unmanned aircraft among Czech agronomists from 2022, when an affordable and user-friendly drone entered the market,” explained the researcher.
Agricultural drones work with sensors to record RGB images and spectral images. “Their uses vary as needed with plant growth stages in mind,” said Slezák.. “For example, with spectral images, we can take advantage of the fact that weeds and crops have different spectral reflectances in their growth stages. This means that we are able to distinguish the two plants in the image.”
However, the quality, evaluation, and subsequent practical use of the data depends on a number of factors during the entire process. The goal of MENDELU scientists is therefore to create a set of recommendations that farmers can follow when monitoring weeds.
“The targeted application process involves three phases,” said Slezák. “The first is choosing a suitable drone and setting the correct flight parameters, for example flight height. The second area is the evaluation of data with the help of advanced algorithms based on machine learning – that is, data analysis with the use of artificial intelligence. The third category is the actual application of the spray to the land so that everything works as it should. Simply put, we want to choose the most economical and efficient solution for the agronomist.”
As part of the experiment, the researchers will test three types of commercially available agricultural drones. The data will be collected during this spring and autumn on the plots of agricultural enterprises in Kroměříž and Znojmo. Over the rest of the year, the researchers will use the acquired data sets to “teach” artificial intelligence to recognize weeds. At the same time, experts want to focus on cereals and the weeds growing in them. Most of the research currently published is on broad-row crops such as sugar beet. We want to focus on cereals where weed detection is more difficult, ” said the researcher.
In particular, Slezák would like to pay attention to the ragwort, a dangerous weed which infests winter cereals. “Essentially, it is the detection of a sedge plant in a stand of sedge plants. The difference is not very noticeable to the naked eye. It will be so interesting to see how artificial intelligence will deal with recognizing this weed,” he explained.
Slezák also sees great potential for the future in the use of drones, which is reflected in the growing interest in these technologies among agronomists. “Drones help farmers not only in detecting weeds, but also in monitoring active burrows of voles. With their help, we can detect biotic and abiotic damage to the vegetation, dose herbicides in a targeted manner, or skip places when fertilising,” he said.
The resulting recommendations regarding weed monitoring should be available to farmers at the end of this year.
