A field-proven technical guide to real-time AI Green-on-Green spot spraying for corn, lentils, sunflower and soybean — with herbicide savings, working speeds and payback timelines based on WeedEye deployments across Central Europe.
Green-on-Green (GoG) means detecting and spraying weeds that grow inside a green, actively-growing crop — as opposed to Green-on-Brown, where weeds are sprayed on bare soil. GoG is the harder computer-vision problem: the AI must distinguish weed species from the crop leaf-by-leaf, in real time, from a moving boom.
WeedEye's FALCON cameras stream frames to an on-boom AI unit that classifies each square of the field and fires the corresponding nozzle within ~5 ms. The result: only weeds get sprayed, even in a dense post-emergence canopy.
Recommended application windows, target weed spectrum, and observed savings for the four most common Green-on-Green crops in the WeedEye field programme.
Zea mays
Agronomy note: Row-crop geometry lets the FALCON camera learn a strong crop-vs-weed model. Pair Green-on-Green with a reduced-rate residual for late-season escapes.
Lens culinaris
Agronomy note: Lentils have a narrow herbicide toolbox and low tolerance to over-application. Green-on-Green protects yield while cutting the most expensive chemistry.
Helianthus annuus
Agronomy note: High per-hectare chemistry cost makes sunflower one of the fastest payback crops for Green-on-Green. Ambrosia hotspots are treated selectively.
Glycine max
Agronomy note: Ideal for resistance management: spot-treat escapes with an alternative mode of action instead of blanket re-spraying.
Three engineering choices make in-crop detection reliable in real field conditions.
High-frame-rate cameras mounted every ~1.5 m across the boom, with active exposure control for sunlight, shadow, and dust.
Crop-specific neural networks run at the edge — no cloud round-trip — for classification and nozzle firing in ~5 ms.
Each season we retrain models against local weed pressure — from Ambrosia in sunflower to resistant Amaranthus in soy.
