Projects per year
Autonomous operation of robotic systems in an agricultural environment is a difficult task due to the inherent uncertainty in the environment. The robot is in a dynamic, non-deterministic and semi-structured environment with many sources of noise and a high degree of uncertainty. A novel approach dealing with uncertainty is by means of probabilistic methods. This PhD thesis studies the efficacy of probabilistic methods for autonomous robot applications in agriculture focusing on two agricultural tasks namely automatic detection of weed in a grassland and autonomous navigation of a robot in a Maize field. In automatic weed detection we look at the detection of a common weed called Rumex obtusifolius (Rumex). The suitability of image analysis for the task is examined, various existing methods are scrutinized and new probabilistic methods are proposed for robust detection of Rumex using a monocular camera in real-time. For autonomous navigation in a Maize field, probabilistic methods are developed for row following using a camera as well as a laser scanner. New sensor models are proposed to characterize the noisy measurements which are used in the navigation method for tracking the position of the robot and the plant rows. Through extensive field experiments we show that the proposed probabilistic methods are robust to varying operating conditions and conclude that probabilistic methods are essential for autonomous operation of robotic systems in an agricultural environment.
|Qualification||Doctor of Philosophy|
|Award date||3 Oct 2013|
|Place of Publication||S.l.|
|Publication status||Published - 2013|
- image analysis
- bayesian theory