The combination of artificial intelligence and high-resolution photography may help U.S. cotton farmers distinguish between two common insect pests, informing their control strategies.
Using machine learning techniques, a team of researchers produced a system which can differentiate between eggs laid by tobacco budworm, Chloridea virescens, and cotton bollworm, Helicoverpa zea, two species responsible for the majority of damage to cotton in parts of the U.S.
Know your enemy
Recognising which of the species are on your crops is essential for farmers growing cotton which is genetically-modified to express the insect-killing toxins of the bacteria Bacillus thuringiensis (Bt). Nearly 90% of cotton grown in the U.S. has the Bt modification, and its prevalence has resulted in H. zea building resistance, requiring additional Bt applications.
The eggs of both species are around 0.5mm in diameter, and traditionally require a biologist using a microscope to tell them apart. However, the system developed in the study was shown to be able to identify eggs to over 99% accuracy, after a period of ‘training’ using around 5500 images.
Not yet smartphone-ready
The system could lead to the development of smartphone applications with which farmers could photograph eggs and get an instant identification, while the methodology could replicable in targeting other problem pests, the researchers said.
However, before such practical applications will be possible, some potential issues do need ironing out, they explained. The model may have more difficulties identifying species with a complex background such as leaves, while it will also need to be able to identify different developmental stages of the eggs to the same level of accuracy. As such, more practical training using more variable conditions is needed.
You can read the full paper in Agriculture.
