Schöler, Florian Erwin: 3D Reconstruction of Plant Architecture by Grammar-based Modeling and Markov Chain Sampling. - Bonn, 2014. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37307
@phdthesis{handle:20.500.11811/6164,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37307,
author = {{Florian Erwin Schöler}},
title = {3D Reconstruction of Plant Architecture by Grammar-based Modeling and Markov Chain Sampling},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2014,
month = sep,

note = {The world population is growing rapidly. As a consequence increasingly more products of crop plants are needed. Here, plant breeding gains a special significance. The goal of plant breeding is to develop new cultivars that need less nutrients and at the same time are more resistent to diseases and pests. A deciding factor is the mapping from gene analysis to the phenotype of the plant, i.e., all traits that can be observed from the outside. A major impairment is the so-called phenotyping bottleneck, i.e., the fact that phenotyping is too time-consuming and that current, manual methods are subjective and imprecise.
This thesis presents a procedure that is a step toward eliminating the phenotyping bottleneck. A method is presented for the reconstruction of plant architecture from sensor data with the goal of applying it in an automated phenotyping. Several contributions are presented: The reconstruction procedure has three sub-steps that allow phenotyping in increasingly fine detail. A model of the investigated plant, which comprises components, topology and geometry, serves as a compact description of all allowed structures and makes the intrinsic complexity and variability manageable. Above that, the parameters of the model are assigned valid value ranges by an automated skeletonization method. The introduction of the model into the reconstruction allows to regard not only the sensor data as source of evidence, making possible the handling of massive occlusions. In phenotyping coarse categorizations can be replaced by quantified measures and new traits can be investigated.
The grapevine plant, especially the grape cluster, serves as application. The grape cluster is a highly complex construct of branches and berries, whose different development stages offer different challenges. In addition, the grape cluster enables a yield estimate and the detection of diseases. The presented procedure is applied on grape clusters in three different development stages and on the venation of vine leaves. A detailed presentation and evaluation is given for the grape cluster development stage of full ripeness. In principle, the presented procedure is extensible to other plants that, like the grape cluster, can be defined by a tree-like branching structure.},

url = {https://hdl.handle.net/20.500.11811/6164}
}

The following license files are associated with this item:

InCopyright