The aim of the project is to develop new AI-based methods for agricultural issues such as the optimization of pesticide use or fertilizer quantities applied through early detection mechanisms. The data and insights gained are then to be made available to end users or developers in cloud applications via open interfaces.
To get closer to these goals, the IPZ team - in close cooperation with the Heinrich Hertz Institute of the Fraunhofer Gesellschaft - is also supporting the investigations in the final year of the project. This is done with the help of pot and field trials, which serve as the basis for multispectral recordings. The recording period for the early recognition of yellow rust extended from mid-May to mid-June. During this period, several dozen individual images were taken daily from a specially established pot trial using a tripod camera. In parallel, further multispectral images were generated from a 2,000 m² field trial using a drone.
Training of AI algorithms
The background to these recordings is the creation of a data basis for training AI algorithms, including additional insights into disease dynamics. In order to achieve this, numerous bonitings from the ground were additionally carried out in the field and pot trials and markings were set so that the valuable bonituring data can be linked to the sensor recordings.
As the season progresses, the focus of the trial will switch from winter wheat to sugarbeet from July onwards. Here, it will then be tested whether AI algorithms can also help with the early detection of Cercospora leaf spot.
DLG Competence Center Agriculture, Digital Agriculture