(a) Raw Image, RGB acquisition in leeks
(b) Green-on-Brown in red: all plants are targeted
(c) Green-on-Green: only weed is targeted (in red), crop is protected (in green)
Nowadays, artificial intelligence (AI) is at the core of farm management, from fertilization recommendations to concrete weeding solutions. Ecorobotix has revolutionized crop care by bringing very advanced technology into the field. The company offers the ultra-high precision (UHP) spot sprayer ARA, and its value is built on AI, precision, and the exact definition of the spray spot applied to the target.
ARA is not a simple sprayer. It applies plant protection products only where they are needed, and this process is driven by AI. ARA is equipped with cameras that capture live images of the field. These images are analyzed and processed by AI. Because the contrast between green plants and brown soil is high, it is relatively easy to classify each pixel as either green or brown. The resulting segmented image becomes a binary mask of green and brown pixels. This corresponds to a map where each green pixel is sprayed. This is known as green-on-brown algorithm technology. It can be used, for example, to control weeds on fallow land or before sowing. The main advantage is significant chemical savings.
To further reduce spray volumes, Ecorobotix has added a target-size feature to this all-green algorithm. This allows the operator to focus only on a specific size range of weeds that may be difficult to control with standard mixtures. A clever way to use this feature is to define a size threshold between weeds and the crop. This is especially useful in transplanted crops where no dedicated algorithm is available yet. In these cases, the size difference between transplanted crops and weeds is large enough that the threshold is easy to set. The operator can then target only green spots smaller than a defined size (and targets the weeds) within the green cover made up of larger transplanted plants and weeds.
The pinnacle of weed control, however, is green-on-green technology. Within the green pixels, the AI can distinguish weeds from crops, not only based on size but also on leaf shape, texture, position and number, as well as green shading. This technology relies on machine learning and dedicated algorithms trained for specific crops. In simple terms, field images are manually labeled plant by plant and classified as either weeds or crops. The raw image and the expected result are then provided to a computer, which identifies features and combinations of features to build patterns that allow it to recognize and classify every object in the image. This inference step corresponds to the creation of the algorithm. This algorithm is then integrated into ARA’s onboard computer and can recognize weeds within the crop, in real time in the field, and apply inputs precisely where needed.
This technology opens new possibilities in weed control and overall food security. In essence, ARA’s capabilities in these two precision spraying technologies make it possible to minimize plant protection product use, phytotoxicity induced by herbicides, maximize profitability, and support environmental stewardship, marking a new era where technology and ecology work together for the benefit of the planet. Moreover, it aligns seamlessly with European regulations and global sustainability trends, offering an environmentally friendly solution without compromising effectiveness.
