A method to extract morphological traits of plant organs from 3D point clouds as a database for an architectural plant model

Abstract

Combining process-based and three-dimensional (3D) structural models for specific crops to functional–structural plant models (FSPMs) enable ecophysiologists to investigate the interaction of single plants or plant stands with their biotic and abiotic environment in a unique way. The present study was part of a collaborative research program on the development of a FSPM for the sample plant ( Hordeum vulgare L.). The emphasis of this paper is put on two main aspects. First, improved generic and flexible functions are formulated for modeling the shape of leaves and stems of graminaceous plants as organ-related triangulated surfaces, where the parameters may be directly interpreted in terms of morphological traits. The proposed functions constitute the structural model, which is amplified by topological information to a so-called architectural model. Second, we suggest a new approach to parameterize these functions based on 3D point cloud data obtained by digitization of entire plants. Since no automated technique is available to process 3D point clouds in a way appropriate for parameterization of the architectural model, the required algorithms are developed and implemented in Matlab ®. Our approach comprises the following steps. First, the measured set of points is partitioned into subsets representing each organ. Each subset is then divided further to represent organ segments. Next, the centroid of each partial point cloud representing an organ segment is computed. The sequence of these centroid points describes the organ axis. By means of the architectural model for leaves and stems, triangulated surfaces are assembled from the computed organ axis points and from user-defined initial values for the various parameters in the model (e.g. maximum leaf width). Finally, the parameters in the functions describing leaf and stem surfaces are estimated by fitting computed triangulated surfaces into the related point cloud using least squares optimization. Hence, the proposed method allows the use of 3D point clouds obtained with modern 3D digitizing techniques for the parameterization of an organ-based architectural model.