Agricultural Robotics

Even in agriculture the use of high tech is increasing rapidly. In the not too distant future, swarms of robots which drive over agricultural fields will be a familiar picture. Agricultural engineering is becoming an area of global significance for food and energy. For proper self-localization and object detection / recognition the use of several sensors like GPS, laser scanner and cameras is needed. The sensors deliver data which can be used to determine the optimum sowing density, the optimum use of fertilizers, more accurate predictions on crop yields and better assessment of disease infestation of plants.

In contrast to applications in industry the challenge in the field of agriculture is the changing and not exactly specified environment. This makes a safe navigation and object detection a difficult task. We focus on the detection of plants and other vegetation. For that purpose we use a 3D LIDAR sensor and a PMD camera along with other cameras. The fusion of the data from the different sensors is needed to get the best results.

Within a MINT scholarship of the state of Baden-Württemberg we work in co-operation with the Robert Bosch GmbH to enhance the navigation algorithms of agricultural robots through image processing, e.g. terrain classification.


An image and the corresponding 3D scan of a plant

Central research questions are:

  • Detecting and removing shadows from images
  • Obtaining illumination-invariant color images
  • Real-time 3D scan matching and registration
  • Fusion of 3D and camera data for object detection
  • Navigation of outdoor robots


Contact

Stefan Laible
Tel: (+49/0) 7071 / 29 78983
Email: stefan.laible at uni-tuebingen.de