Seed World

Data Used for Automated Weed Identification

small green plant covered under bell jar isolated on white

Machine-based weed control solutions are progressing with spectroscopy and data analysis according to a release. A solution was needed, as the sensors have difficulty identifying the weeds from the crops.

The Brazil research team successfully used the technology to identify three species of morning glory, ivy leaf, Japanese and hairy wood, that are commonly found in sugarcane fields in Brazil according to an article in Weed Science journal. Infrared spectroscopy was used to collect spectral data from lab-grown weed specimens. The accuracy raters were “99.3% for ivy leaf morning glory, 98.5% for Japanese morning glory, and 98.7% for hairy woodrose morning glory.”

“Developing reliable sensor-based techniques for the identification of weed species is an important first step towards highly targeted weed management,” says Andreisa Flores Braga, Ph.D., of Sao Paulo State University. “With smart sensors that can reliably distinguish weeds from crops in the field, we will have the information needed to guide mechanized sprayers and apply post-emergent herbicides to specific weeds.”

More information: Discrimination of morningglory species using near-infrared spectroscopy and multivariate analysis“.