MOSCOW, January 30Tyumen State University scientists have developed software that can automatically recognize and measure the amount of pesticides deposited on plants. The development makes it possible to automatically measure the area of leaf coverage in real time, which allows minimizing the amount of pesticides used, reducing their costs and reducing environmental pollution, the university reported. The results were published in the journal Agriculture.
Pesticide is a substance (or mixture of substances) of chemical or biological origin intended to destroy harmful insects, rodents, weeds, pathogens of plants and animals, and also used as a defoliant, desiccant and growth regulator. Previously, pesticides were called pesticides.
Optimization of delivery methods for plant protection products is currently one of the most pressing tasks related to both food and environmental safety. According to scientists, today organosilicon agroadjuvants have become very popular in agriculture — substances that, when added to a pesticide working solution, enhance its effect on a harmful object. Their excessive use has a detrimental effect on the vital activity of pollinating insects and affects the biosafety of the cultivation region.
Scientists from Tyumen State University (TSU) used a neural network, which made it possible to increase the accuracy of measuring the area of plant treatment due to the ability to recognize dry islands (areas of the leaf surface not covered with the drug). According to them, when using standard processing algorithms and methods, these dry areas are counted in the overall coverage area, and this gives incorrect information about the effectiveness of the treatment.
“
“We have obtained several good results in the field of artificial intelligence in relation to automatic recognition of images of plant leaves,” said Fabio Grazioso, an employee of the research laboratory of photonics and microfluidics of Tyumen State University.
He added that the software is capable of automatically recognizing and measuring the amount of water or water-based pesticides and adjuvants deposited on plant leaves.
According to him, the practical significance of the development lies in the fact that it makes it possible to automatically measure the area of leaves covered by pesticides in real time; It only takes a few seconds to process each image.
“This will minimize the amount of pesticides or adjuvants used, and therefore reduce costs and reduce environmental pollution,” he noted, adding that the results are ready for use and require ordinary equipment (“a laptop and a camera are enough, and a specialized camera with special lenses can improve the quality of the results»).
The university said that the topic of artificial intelligence for automatic image recognition is very popular among many scientists, and the number of publications in this area is large.
However, there are not many examples of the use of artificial intelligence for image processing in agriculture, the university noted, emphasizing that the contribution of Tyumen State University researchers lies in the fact that they used the existing artificial intelligence model used for automatic recognition of satellite images and adapted it for use in agriculture.
“The quality of artificial intelligence recognition is highly dependent on the training process of the neural network: we manually recognized 130 leaf images and then used them to train the artificial intelligence,” explained Fabio Grazioso, adding that the results can be significantly improved, if you carry out a longer training process with a large number of images.
The research was carried out within the framework of the basic part of the state assignment of the Ministry of Education and Science.
< br />