CLAIM - Classification Assessment using an Integrated Method
With the increasing importance of information derived from remotely sensed data for planning and decision processes also the necessity increases to generate binding and sound statements about the geometric and thematic uncertainties that go along with this information. In contrast to classical approaches problems arise in the course of an a priori evaluation of classification results, in particular for those that are based on spatial high resolution data. Consequently, new developments are necessary which are treated in the project Classification Assessment using an Integrated Method (CLAIM) which is performed by the authors and which is funded by the German Research Foundation (DFG).
A key problem during the evaluation process is caused by the imperfections in the reference (ground truth) data itself. Because of the increased demands concerning better geometrical and thematic accuracies of modern sensor systems these uncertainties becoming even more dominant and should not be neglected anymore. Hence, we propose an integrated approach that considers not only uncertainties in the classified but also in the reference data. Furthermore the phenomenon of indeterminate boundaries gains in importance which can be treated for example with methods known from fuzzy logic theory. But on the other hand so far no comprehensive solutions are known for the evaluation of uncertainties along indeterminate boundaries or within the resulting fuzzy transition zones.