by Antonio Donnangelo
Share

Introduction
Technological innovation is increasingly present in the world of agriculture, and the sub-project AgroTwin, funded with 60k€ from the Horizon Europe ICAERUS, is a perfect example.
ICAERUS is a project focused on exploring the potential of drones as multipurpose means for agriculture, environmental monitoring and rural services.
The objective of the AgroTwin sub-project was to develop and test a Decision Support System (DSS) based on advanced algorithms of AI and computer vision for optimize phytosanitary treatments through the use of consumer-grade drones, such as the DJI Phantom 3 Professional (Fig. 1).

Fig.1: Flight preparation of the DJI Phantom 3 Professional in the test vineyard.
A Decision Support System for Sustainable Agriculture.
Image collection and algorithm testing were carried out on a Sangiovese vineyard of about 1.2 ha from the regional estates of Cesa (AR) (Fig. 2). The latter was divided into. 2 plots of similar size and vegetative vigor, one used for variable-dose dosing of agrochemicals based on the results of our DSS (test) and the other for fixed doses (control), in which standard farm doses were given.
To capture the visible (RGB) images of the vineyard, we used an easy-to-use and widely marketed drone, the DJI Phantom 3 Professional. These images were used to generate 3D point clouds that allowed us to create the digital twin (digital twin) of the vineyard at three different phenological stages.
For each flight, through proprietary algorithms, the digital model was precisely analyzed to extract key biometric parameters of individual vines of interest. Using this data, the DSS generated customized prescription maps for variable rate treatments (VRT), improving the efficiency of pesticide delivery.

Fig.2: Location of the vineyard under test and subdivision of the two plots (test, control).
Advanced Algorithms for the Analysis of Biometric Parameters.
One of the most innovative aspects of AgroTwin has been the development of advanced AI and computer vision algorithms to analyze vine biometrics. These algorithms, applied to the drone-generated 3D point clouds, allowed for accurate analysis of key parameters of interest, such as the thickness, height, and canopy volume of individual plants of interest, with an average error margin of less than 10 percent compared to manual measurements (Fig. 3, 4).

Fig.3: Evolution in different phenological stages of a test vine (thickness (m), height (m), volume (m3)).

Fig.4: Manual measurements in the field.
Through the use of these biometric parameters, algorithms have enabled the construction of a very important vigor index, the LAI (Leaf Area Index) as well as generating the LWA (Leaf Wall Area) and TRV (Tree Row Volume), which give an indication of the area and volume of the vineyard canopy.
By interpolating an appropriate number of sample plants, the algorithms automatically generated vigor of vegetation (LAI) maps and prescription maps for phytosanitary treatments (Fig. 5).

Fig.5: Vigor maps (LAI) and prescription maps (liters/hectare) for phytosanitary treatments for each phenological stage.
Sensitivity Analysis to Improve Data Collection.
Another crucial phase of the project was sensitivity analysis to find the best flight parameters of drone surveys that would provide the best trade-off between data quality versus flight time. This study identified the best flight configurations to provide the best accuracy in estimating the parameters of the canopy vines compared to manual measurements.
The best configuration, with errors of less than 10 percent compared to manual measurements, proved to be the 30-meter flight height, with 85 percent overlap (overlap) between photos and with combined camera tilts (nadiral and 30°), which significantly improved the accuracy of the collected data.
Environmental and Economic Benefits
An internationally standardized procedure (ISO 22522) and a food tracer (tartrazine) were used to evaluate the quantity and quality of pesticide deposition on the vine canopy in the two different areas. The main objective was to compare the efficiency of the two types of treatment. Measurements of the effectiveness of the treatments were carried out by the Department of Agricultural Mechanics (DAGRI) of the University of Florence.

Fig.6: The procedure involved sampling three representative vines in the areas of low, medium and high vigor, using water-sensitive maps and nylon collectors placed at three different canopy heights (H1 above the cordon, H2 in the middle of the canopy, H3 at the top).
The results showed that the average coverage was 35 percent in the VRT plot, slightly above the optimal threshold (30 percent) but the canopies of the vines were covered according to their actual biomass, while the control plot recorded an average coverage of 39 percent, without taking into account the actual biomass in the field. Finally, no diseases were detected in the tested plots and no significant differences in grape yield and quality were found at the time of evaluation.
Regarding dose, thanks to the VRT approach, an average 35% reduction in pesticide and water use has been achieved with a peak of 41% during the intermediate phenological stage. These savings can be translated into A reduction in the impact that products have on the environment, thus improving the preservation of the biodiversity present in the vineyard and limiting groundwater and air pollution.
Conclusions
The AgroTwin project represents a breakthrough in precision agriculture by exploiting imagery in agronomic decision support. Through it, it has been designed A prototype of DSS from drone images. In addition, the extent to which the integration of drones and advanced algorithms can improve the environmental and economic sustainability of agricultural operations.
The project was able to demonstrate how the use of a drone taking simple RGB photos can enable the generation of digital models of sufficient quality to analyze a crop in the field and obtain prescription maps that can be used for the most important operations, such as in the case of pesticide distribution, or in selective harvesting based on vegetative vigor, thus allowing to save valuable resources and improve production processes.
With significant savings in the amount of pesticide and water, and improved redistribution in the field, the developed DSS is positioned as an innovative and sustainable agronomic tool that allows the dictates of precision agriculture to be pursued by taking advantage of digital copies of the vineyard.
In an increasingly sustainability-driven industry, this technology is able to offer farmers a practical, cost-effective and cutting-edge solution to meet the current challenges of modern agriculture.