Article Preview
Top1. Introduction
At 3D GeoInfo 2012, we presented an innovative and efficient way to generate “Virtual Forests” from remote sensing data (Bücken & Roßmann, 2013). Individual trees are delineated from normalized digital surface models and annotated with height and species. This approach is the first step towards various forestall simulation applications based on real-world data like the simulation of forest machines (Figure 1), a flight simulator, or a tree growth simulation. To provide a basis for an efficient and modern data management of such vast datasets, a database-driven method for 3D simulation systems previously presented at 3D GeoInfo 2010 is used (Hoppen, Roßmann, Schluse, & Waspe, 2010). It provides a persistence layer and a common data schema for simulation systems. Now, it is enhanced by techniques for database-driven, distributed data management and simulation, and for data versioning.
Figure 1. A driver training with the forest machine simulator
In this new paper, we focus on the integration, enhancement, and on future trends regarding these two core technologies of a large-scale 4D forest simulation and information system. In particular, algorithms for the attribution of the individual tree, details on the GML-based (OGC, 2014) object-oriented schema family ForestGML for forestry data, and the concept of database-driven communication are presented. Overall, a shared world model is efficiently managed in a geo database and filled using modern techniques of semantic world modelling. The latter transform remote sensing data into a semantic object representation that can be used for the various simulation scenarios as mentioned above. Furthermore, data versioning can be used to analyse past scenarios like a windthrow, where the corresponding storm loss must be calculated. Furthermore, even simulated or predicted future values can be managed in a database for conservation, analysis, and comparison. These two concepts – simulation and versioning – add a fourth dimension yielding a 4D forest simulation and information system. Furthermore, given the performance of today’s database systems, it even becomes feasible to use the presented system for a multi-client simulation. Here, different clients are simultaneously working with the shared world model, while their actions’ effects are distributed over the very same active geo database system.
The paper is organized as follows. In the next section we give an overview of related work. In Section 3, the tree extraction approach is introduced and current results are presented. Subsequently in Section 4, the database interface is introduced, including database versioning and data streaming. Here, we give an insight into the systems 4D capabilities and current developments and show how the database interface and the tree extraction interact. Subsequently, we give details on the new ForestGML schema family used for data modelling in the Virtual Forest in Section 5, some applications benefitting from realistic tree data are presented in Section 6. We conclude in Section 7 and end with references and acknowledgements.
TopThere are several approaches for the delineation and attribution of individual trees from remote sensing data documented in literature. (Garcia, Suárez, & Patenaude, 2007) compare four common algorithms for single tree delineation from nDSM (normalized digital surface model) datasets and point out their advantages and disadvantages. (Reitberger, 2010) provides algorithms that work on full waveform data. The volumetric algorithm used in this paper focuses on nDSM-data. It can detect up to 90 percent of the trees in a spruce forest that is ready to harvest. (Hyyppä & Inkinen, 1999) estimate the diameter at breast height of a tree depending on its height and crown diameter but do not specify the parameters of their heuristic as it might vary for different areas.