Josef Strobl

Quantifying the robustness of fuzzy rule sets in object-based image analysis

Object-based image analysis (OBIA) has become very popular since the turn of the century. For high-resolution situations, in particular, where the objects of interest are larger than pixels, methods have been developed that build on image segmentation and on the further classification of objects rather than on pixels. Many studies have shown that OBIA methods are, in principle, more transferable and reapplicable to other images. To obtain comparable results by reapplying a given rule set on (slightly) changed conditions, the rule set must either be able to adapt to the changed conditions or it must be parameterized for manual adaptation. In this context, a rule set can be seen as the more robust the less it has to be changed, and vice versa. In this article we introduce a new method to evaluate the robustness of a rule set. The main assumption is that the amount of necessary adaptations can be measured in conjunction with the quality of classification achieved. We demonstrate that the method introduced is able to (1) evaluate the robustness of a rule set and (2) identify crucial elements of a rule set that need to be reparameterized.

Optimization of scale and parametrization for terrain segmentation: An application to soil-landscape modeling

This paper presents a procedure to optimize parametrization and scale for terrain-based environmental modeling. The workflow was exemplified on crop yield data, which is assumed to represent a proxy for soil productivity. Focal mean statistics were used to generate different scale levels of terrain derivatives by increasing the neighborhood size in calculation. The degree of association between each terrain derivative and crop yield values was established iteratively for all scale levels through correlation analysis. The first peak of correlation indicated the scale level to be further retained. To select the best combination of terrain parameters that explains the variation of crop yield, we ran stepwise multiple regressions with appropriately scaled terrain parameters as independent variables. These techniques proved that the mean curvature, filtered over a neighborhood of 55m, together with slope, made up the optimal combination to account for patterns of soil productivity. To illustrate the importance of scale, we compared the regression results of unfiltered and filtered mean curvature vs. crop yield. The comparison shows an improvement of R^2 from a value of 0.01 when the curvature was not filtered, to 0.16 when the curvature was filtered within 55x55m neighborhood size. The results were further used in an object-based image analysis environment to create terrain objects containing aggregated values of both terrain derivatives and crop yield. Hence, we introduce terrain segmentation as an alternative method for generating scale levels in terrain-based environmental modeling, besides existing per-cell methods. At the level of segments, R^2 improved up to a value of 0.47.

Object-based classification of landforms based on their local geometry and geomorphometric context

Terrain as a continuum can be categorized into landform units that exhibit common physical and morphological characteristics of land surface which may serve as a boundary condition for a wide range of application domains. However, heterogeneous views, definitions, and applications on landforms yield incompatible nomenclature that lacks interoperability. Yet, there is still room for developing methods for classification of land surface into landforms that can provide different disciplines with a basis of landscape description that is also commonsense to human insight. This study proposes a method of landform classification that reveals general geomorphometry of the landscape. A set of landform classes that are commonsense to human insight and relevant to various disciplines is adopted to generate landforms at the landscape scale. The proposed classification method is based on local geometry of the surface and the geomorphometric context in a higher level framework. A set of digital terrain models (DTMs) at relevant scale is utilized where local geometry is represented with morphometric DTMs, and the geomorphometric context is incorporated through ‘relative terrain position’ and ‘terrain network.’ ‘Object-based image analysis’ tools that have the ability to segment and classify DTMs into representative terrain objects and connect those objects in a multi-level hierarchy is utilized. Ambiguities in landforms both in attribute and geographical space are represented via fuzzy classification. The proposed method is applied to two different case areas to evaluate the efficiency and stability of the outcomes. Results reveal a reasonable amount of consistency where landform classes can be utilized as general or multi-purpose regarding some ambiguity that is already inherent in landforms.

Segmentation and Object-Based Image Analysis

This chapter focuses on segmentation of remotely sensed image data and object-based image analysis. It discusses the differences between pixel-based and object-based image analysis; the potential of the object-based approach; and, the application of eCognition software for performing image segmentation and classification at different levels of detail.

Mapping Green Spaces in Bishkek—How Reliable can Spatial Analysis Be?

Within urban areas, green spaces play a critically important role in the quality of life. They have remarkable impact on the local microclimate and the regional climate of the city. Quantifying the ‗greenness of urban areas allows comparing urban areas at several levels, as well as monitoring the evolution of green spaces in urban areas, thus serving as a tool for urban and developmental planning. Different categories of vegetation have different impacts on recreation potential and microclimate, as well as on the individual perception of green spaces. However, when quantifying the ‗greenness of urban areas the reliability of the underlying information is important in order to qualify analysis results. The reliability of geo-information derived from remote sensing data is usually assessed by ground truth validation or by comparison with other reference data. When applying methods of object based image analysis (OBIA) and fuzzy classification, the degrees of fuzzy membership per object in general describe to what degree an object fits (prototypical) class descriptions. Thus, analyzing the fuzzy membership degrees can contribute to the estimation of reliability and stability of classification results, even when no reference data are available. This paper presents an object based method using fuzzy class assignments to outline and classify three different classes of vegetation from GeoEye imagery. The classification result, its reliability and stability are evaluated using the reference-free parameters Best Classification Result and Classification Stability as introduced by Benz et al. in 2004 and implemented in the software package eCognition

Information Technologies Socialise Geographies

One of the ethical tasks and practical effects of IT is bridging and spanning different locations, thereby socialising across diverse geographies of understanding. A dozen documented case studies use IT especially Geographic Information Sciences in distance learning. The underlying conceptual model of a network society combined with empirical research on long-term civilisational and economic evolution leads to a general understanding of Information Technologies as facilitators of a multi-perspectivist and multi-disciplinary construction of world views m:n type of science. Such a synopsis of education, structural evolution, social spaces and institutional change provides insight into ITs strategic role of facilitating consensus building and constructing common world views that can socially converge socialise isolated cultures of understanding. Geography is here seen as a provider of world views that emerge from communicative action. The presented cases in this paper span both geographic locations as well as constructed cultures of understanding.

GIS-basiertes Backcasting: Ein Instrument zur effektiven Raumplanung und für ein nachhaltiges Ressourcenmanagement

Der Ursprung von Backcasting liegt in den 1970er Jahren. Damals fuhrte Amory Lovins diese Methode zur Planung von Elektrizitatsangebot und -nachfrage ein und nannte es „backwards looking analysis“. Danach fand diese Methode regelmasig in Energiestudien Anwendung (Dreborg 1996, S. 814; Quist 2007, S. 18). In der Folge dauerte es einige Zeit, bevor erkannt wurde, dass Backcasting auch bei anderen Fragestellungen, wie etwa der Nachhaltigkeit und Planung eingesetzt werden kann. In seinem Buch „Backcasting for a sustainable future: the impact after 10 years“ verglich Quist (2007) vier verschiedene Backcasting-Ansatze, wobei jeder davon auf unterschiedlichen Methoden basiert.