On ontology in image analysis

Object reconstruction from remote sensed images is an actual research topic in computer vision, photogrammetry, and spatial information science. One of the most interesting related questions is the modeling of the uncertainty of knowledge gained by such a process. For a better understanding of the different sources and aspects of uncertainty, indeterminacy, vagueness, and the relations between them, a better understanding of the underlying ontological and epistemological foundations of imaging is necessary. An important aspect is to make the relationships between the objects in the world that are observed, for example, by means of remote sensing, and fiat objects created from the remote sensed data by means of spatial analysis explicit. Both kinds of objects are linked to each other by observation and analysis processes. The process of spatial analysis has three basic components which are modeled in this paper as mappings between fiat objects of different kind. We show that vagueness in the definition of fiat objects involved in this process results in indeterminacy of location of those objects. Definitorial vagueness and location indeterminacy of location result in uncertainty about the truth of conclusions about existence, properties, and location of objects in the world derived from knowledge about fiat objects created by spatial analysis.