PhD Project
Integration of plant parameters for intelligent agricultural processes
The biggest challenges in digital image analysis are the unstructured agricultural environments, different exposure, and the change of shape of the plants in different development stages. To overcome these challenges, it is necessary to collect big amounts of digital image data and to use it in an efficient way. Furthermore, it is necessary to select deep learning models that allow the integration of a priori knowledge to reduce the needed amount of data without losing accuracy.
The aim is to integrate this system to an innovative mobile field robot that will acquire and analyse digital images from the crop field and perform actions for weed control. For the actions on the field the robot will be acquitted with a robotic arm and needs precise localisation. To achieve this, it is planned to fuse data from various sensors and parameters from image analysis. This will be achieved by developing novel modelling techniques to combine different computer vision tasks.
High precision localisation of weeds can help to reduce the quantities of plant protection products due to the possibility of precise application. A mobile field robot in combination with high precision localisation can also be used for mechanical or alternative forms of weed control. Furthermore, the plant parameters stored in a database can help to optimize parts of the agricultural process chain with decision support for planning processes. This can help to promote sustainable agriculture and saving resources.