Because chemical herbicides are taboo in organic vegetable farming, weeding is still done by hand. But manually chopping between the delicate plants is laborious, time-consuming and expensive. Researchers now want to change that: with a laser module that uses artificial intelligence to detect weeds and eliminate them with pinpoint accuracy.
In organic farming, weed control is one of the biggest practical challenges. Unlike conventional agriculture, chemical herbicides are not allowed to be used, so farmers often have to rely on manual labor. Although this is environmentally friendly, it is labor-intensive and quickly reaches its limits, especially on larger areas.
Accurate to the millimeter
To make farmers’ work easier, researchers from the Leibniz Institute for Agricultural Engineering and Bioeconomy (ATB) in Potsdam have now developed a new, innovative system. “Imagine an attachment module for tractors or a small field robot that is equipped with a camera,” explains Karuna Koch from ATB. “While driving over the vegetable dam, the camera records images that an integrated computer analyzes in real time using AI. It decides which plants should be removed and controls a high-precision tool. In the JaetRobi project, we have developed a module that automatically detects weeds and removes them with a laser with millimeter precision.”
Specifically, the team uses blue diode lasers for laser-assisted weed removal, whose light is particularly strongly absorbed by plant tissue. In contrast to infrared lasers, which lose part of their energy to the humidity in the air and to water on the leaves, blue laser light penetrates directly into the tissue. This means that a lower power is sufficient and the laser beam acts like a precise cut. The targeted weeds can be destroyed so efficiently and precisely, as the researchers report. At the same time, large-scale burns are avoided and the risk of heat damage to the surrounding crops is reduced.
Accurate thanks to AI
However, a precise laser alone is not enough to ensure that weeds can be eliminated with pinpoint accuracy. The system for identifying the “culprits” must also work reliably. “Our application scenario focuses on weeds from the cotyledon stage until the first true leaves appear,” explains Koch. “Our tests are currently running with plants up to two centimeters in size. So that we can precisely distinguish them from carrots, for example, during the crossing, we first created an image database and annotated weeds and crops. This enabled us to train an AI that now distinguishes weeds from carrots with an accuracy of 94 percent, starting at the cotyledon stage.”
However, the laser weeder is still in development and is not yet used regularly in organic farming. Until then, the researchers want to give the system a few upgrades. Among other things, in the future it will only remove weeds that actually affect crops. The rest should remain standing, which would further accelerate the work and at the same time promote biodiversity in the field.
Source: Leibniz Institute for Agricultural Engineering and Bioeconomy eV (ATB)