Visualization of Uncertainties in Urban Processes
Remotely sensed scenes have a continuously increasing impact on describing, modeling and simulating landscape structures and processes, in particular for urban applications. On the other hand it has to be stated that automatic and time saving approaches for a thematic interpretation of remotely sensed scenes and series of scenes is by far not operational yet. From that we draw the conclusion also to go back to a visual approach, i.e. to improve the visual interpretation as well as the follow-up change detection process through the adaption of interactive methods and tools that are known from disciplines like Visual Analytics or Geovisualization.
This concept is even further extended through the integration of uncertainty information. Like in other change analysis applications the key problem is to differentiate between underlying acquisition and classification errors on one hand and actual changes on the other hand. The problem becomes clearly evident if one compares typical classification errors in the order of 15% and land cover changes which are usually in the same order. Hence, we develop and implement an extended conceptual design that enables the additional visualization of different uncertainty information. This includes both the further development of an user interface, and the presentation of various options for displaying uncertainty data interactively.
Furthermore, we develop methods for visualizing uncertainties in the context of fuzzy logic classifications, in particular for representing indeterminate boundaries, transition zones and varying membership values associated to single landscape elements.
Funding: HCU DigitalCity Research Group
