Measuring Quality of Geographical Information using Imperfect Reference Data

N. Charlier, G. De Tr, S. Gautama, and J. Verstraete (Belgium)


Soft Computing, Fuzzy Sets, Geographical Information Systems, Image Processing, Qualitative Spatial Reasoning, Imperfect Data Modelling


In the past, theories of quality measuring of imperfect data have always assumed the use of perfect reference data. In practice, this isn’t always the case. In our new approach we consider the possibility of measuring quality using imper fect reference data. Especially the use of classified satel lite images is considered. The question we want to answer is what can be said about the quality of a questioned ob ject taking the quality of the reference data as input: is it good, is it bad or is it unknown? In first instance, an im perfection model for the reference data is built based on possibility theory. From this we extract a fuzzy egg-yolk model, consisting of two parts: a fuzzy set of what must be in the object and a fuzzy set of what might be in the object. A special buffering operation is defined to build in tolerance in relation to the mistakes in the classified satel lite images. Operators are defined for measuring quality using this model. Finally the method is tested and evalu ated. It is shown that bad quality can often be stated, but that it is difficult to confirm good quality when the quality of the reference data is insufficient.

Important Links:

Go Back