Computer vision for global analysis of crop canopy structure

The structure of a vegetation canopy plays a central role for its interaction with the environment. The canopy structure is often represented with two global parameters: the leaf area index and the leaf angle distribution. This project works towards a theoretical basis for the development of operational methods for automated estimation of these parameters using ground-based imaging sensors. The project especially focuses on a method based on so-called gap fraction analysis where the canopy structure is inferred by measuring at different zenith viewing angles the probability of seeing through the canopy. For this purpose, colour images (mainly RGB) as well as laser range measurements are considered. Problems that are dealt with include robust and accurate segmentation of images into vegetation and soil, handling of "mixed pixels", and viewpoint planning and sampling strategy. Experiments are carried out in cooperation with the Danish Institute of Agricultural Sciences, including greenhouse experiments as well as field trials with integration of various sensors on the API field robot platform. Funded through the ACROSS project. (Kristian Kirk, Erik Granum, Hans Jørgen Andersen)