Characterisation of cereal morphology through proximal stereo vision under contrasting nitrogen inputs

This study aimed at determining the capacity of a proximal stereo vision system to characterise the architecture of winter wheat and spring barley canopies. Images were acquired by a nadir pair of cameras, in two fields dedicated to fertilisation trials (Gembloux, Belgium). Soil and plants were separated through Support Vector Machines. Dense 3D point clouds were computed, describing canopy height. The study of mean leaf angles and leaf angle distributions revealed different dynamics for wheat and barley but little effect of fertilisation. Due to its low cost and its compactness, such a 3D vision system is suitable for farm applications.