2024-02-12 10:30:44
To reduce the quantities of herbicides used in agriculture, a set of technologies is now available to specifically treat weeds. This strategy requires mapping a plot, then the use of image analysis techniques, and finally the use of specific spraying equipment.
In the absence of agreement between EU member states, the European Commission has decided to extend the use of glyphosate for ten years, until 2033. While the usual use of this herbicide in agriculture is to spray it evenly over an entire plot, another, more economical approach involves targeted treatment of only weeds. This strategy can also be applied to other substitute products, even if they are few in number. It was born in the United States in the 1990s with precision agriculture and the development of GPS. Since then, it has arrived in Europe and the technologies to implement it have evolved.
The first step requires mapping a field with the goal of identifying weeds. Different means are available and differ in particular depending on the degree of resolution of the image obtained. With the SPOT 7 satellite, it is possible to obtain images where each pixel represents a square of a ground surface of 8 meters on each side. Since weeds are often distributed by spots in a plot, this technology is suitable when their size is greater than 8 meters.
For more precision, it is necessary to use proxidetection, a more precise image collection technique working with sensors positioned very close to their targets. Service providers offer the use of drones whose altitude can be varied, for example two meters above the crops. “For even greater resolutions, on-board sensors and cameras can be used and are positioned at the front of the sprayer, explains Jean-Paul Douzals, spraying R&D manager at Inrae. This solution is particularly suitable for detecting seedlings which have just emerged from the ground. The idea is to start as early as possible in order to save the quantities of herbicides sprayed. »
Technologies in development for detection within a row
For image analysis, there are two main types of technologies. The first consists of looking for “green on brown”, we use the term “Green on Brown”, with the aim of defining vegetation indices. The most used wavelength range is visible and near infrared. The sensors measure reflectance, that is to say the light emitted by objects: the brown of the ground being reflected in the red, while the plants reflect in the green. This type of approach is suitable for detection between rows of crops, except in certain specific situations where a plant cover is maintained on the ground, as may be the case in viticulture.
To detect weeds within the rows of a crop, it is then necessary to carry out a search for “Green on Green”, also called “Green on Green”. Multispectral techniques must then be used, particularly in the infrared. The objective is to distinguish between two major classes of plants. Monocots which correspond in particular to cereals and which are characterized by more or less thin and elongated leaves. And dicotyledons, which include other types of plants, including a significant number of weeds. It is possible to differentiate these two classes thanks to their spectral coverage which is often not the same as well as the shape of the plants which is not identical.
“This technique is still very much at the R&D stage, because there are many problems to resolve in terms of image processing, adds the head of INRAE. We are not in the industrial sector and there can be difficulties, particularly with sunlight and shadows on the plants. In this area, we are seeing strong development in the use of AI, particularly neural networks and deep learning. The aim is to achieve a minimum detection threshold by limiting the percentage of false negatives, that is to say cases where weeds are considered as crop plants, or vice versa false positives.
Very high resolution sprayers on the market
All sprayer manufacturers today market their own system for applying a localized herbicide treatment, with the particularity that they can achieve greater or lesser precision. On certain sprayers, management is carried out in sections, this means that certain parts of the treatment boom are activated or not according to needs. Weed identification is done either in real time with sensors and cameras on board the front of the sprayer. Or in two stages with first mapping the plot, for example using a drone, then importing the data into the tractor interface.
“Some sprayers operate with precision down to the nozzle and allow each of them to be activated or not individually, completes Jean-Paul Douzals. They are generally spaced every 50 cm on a treatment ramp. We are also seeing the emergence of very high resolution systems, as is the case with the one developed by the company Ecorobotix and called ARA. The processing ramp of this equipment makes it possible to achieve very fine spatial resolution since the nozzles are spaced every 4 cm. »
Even if they might avoid the movement of large agricultural machinery in the fields, the use of drones to apply the treatment is not authorized by the regulations. They indeed fall into the category of aerial treatments which are prohibited in Europe, unless exempted.
As for the savings in herbicides achieved thanks to this localized treatment strategy, it depends on several parameters, in particular the overall level of infestation of the plot, the type of weeds responsible for the infestation and finally the degree of resolution of the means of treatment used.
1707767030
#Reduce #quantities #herbicides #applying #spot #treatment
