A method for building a three-dimensional (3-D) model of a vegetation canopy is presented. It is based on stereovision, the purpose of which is to reconstruct scenes in 3-D space from a pair of images taken from different viewpoints. We applied this method to maize photographs recorded from above by two photographic cameras and obtained a 3-D model of the vegetation. This model enables us to analyse the canopy geometrical structure and to carry out different simulation procedures. Estimations of leaf position and orientation, and of the leaf area distribution, are shown. Possible ways to automate and accelerate the processing, and some applications, are discussed.