Automated Surface Area Estimation of Plants based on 3D Point Clouds

Plant phenotyping is a central task in crop science and plant breeding. Since standard methods often require timeconsuming and manual observations it is indispensable to develop automatic, sensor driven methods which offer objective and fast information. Many methods rely on camera systems [2], ranging from RGB to hyper-spectral cameras. In recent years 3D sensing systems like laser scanners became increasingly popular [3, 7], since they provide structural plant parameters, which can be hardly extracted with spectral sensors. We present a pipeline for the extraction of plant surface areas, which reconstructs meshes from raw point clouds. This pipeline is completely automated with a robust set of empirically determined parameters, which we tested on different data sets. The few data set-specific parameters are determined directly from the respective data set and therefore do not need to be adjusted manually.