Three-dimensional Reconstruction of Fruit Trees by a Shape from Silhouette Method

Abstract

In order to robotically prune a dormant fruit tree, the branches must be identified in three-dimensional space. Furthermore, the branches need to be measured in order to determine which branches should be pruned. Both the identification and measurement of branches can be accomplished by generating a three-dimensional reconstruction of the tree, and by making measurements on the reconstruction. In this work, we use a computer vision-based method to generate the three-dimensional reconstruction. Images of the tree are acquired in a laboratory setting, and then a shape from silhouette (SFS) method of our own design is used to reconstruct the shape of the object. Although we use the SFS method for leafless apple trees, this method is appropriate for many complex objects. We present an overview of our method of SFS and experimental results in a laboratory setting.