Conception and Implementation of a Case Specific Uncertainty Chain in the processing of Remote Sensing Scenes.

This contribution deals with alternative and innovative approaches for modeling and visualizing uncertainties that arise in the context of a remote sensing analysis. Concerning the modeling stage, measures will be generated that yield more spatially and thematically differentiated information in contrast to the commonly used overall accuracy and Kappa-Index. For this purpose, we propose an expanded categorization of uncertainties related to the carriers of uncertainty (polygonal object, object boundary) and the origin of uncertainty information (comparison with ground truth, quality of class assignment, evaluation by means of a priori expert knowledge). Based on these general uncertainty types we present adapted visualization forms (Noise Annotation Lines and Ghost Lines, respectively). Finally, the implementation of these concepts up to the utilization in a web portal, i.e. towards the closing of the "uncertainty chain", as part of the project "KLI-WAS" is described.