Field robots employing a new artificial intelligence algorithm can help with the early identification of issues affecting the health and productivity of corn.
Using a machine learning algorithm to identify and classify images that can be used on board field robots, database engineer Justyna Stypułkowska from Institute of Aviation in Warsaw tested the system’s ability to identify different corn growth stages, hydration levels in the crop and the presence of pathogens and pests.
She found that it was able to detect the chosen parameters to a high degree of accuracy, and in different corn fields.
Rigorous training required
In the initial stages of the research, she ‘trained’ the machine learning system on the different developmental stages of corn and the crop’s hydration levels using a test data set of several thousand images, collected from an experimental plot in a Polish village. For training on the effects of different pathogens and pests, she used an external data set, which featured common rust, grey spot and damage by fall armyworm caterpillars, among other plant health issues.
She then went on to test the performance of the algorithms using a series of images of corn, from the experimental site, different fields, as well as when the camera was placed on a robot prototype in larger fields. Its performance was cross-checked by the human eye, confirming the accuracy of its analysis.
The most useful analyses can be achieved by linking output data from the AI technology to data from a GPS transmitter, she noted in a paper presented at 18th Conference on Computer Science and Intelligence Systems — something she said “is not complicated.” The system can be used as part of a computing unit attached to a field robot, or with a standard office-based computer.
Multiple benefits, broad application
Application of the technology as part of precision agriculture will have benefits for manufacturers of agricultural machinery, farmers, consumers and the environment, Stypułkowska continued.
Pathogens, plant development stages and hydration can all be accurately assessed without additional soil sensors, making incorporation of the model into new technologies advantageous, she noted.
The benefits to farmers include early detection of potential issues affecting their crops, which would allow for reduced application of synthetic chemicals and prevent over-drying of fields, increasing yields and making savings, as well as reduced costs by the automation itself. Reduced use of crop protection products and fertilisers would provide the environmental and consumer benefits.
“The system under discussion, which will be well received by precision farming machinery producers and agricultural producers, will have tangible benefits…This solution will be able to find interest not only in Poland, but also in the rest of Europe and around the world,” she concluded.
“[The system] is easily transferable to other crop species than just corn and can be an invaluable aid to the implementation of modern precision agriculture in many key plant food crops in the world.”
