Symbolic Spatial Reasoning on Object Shapes for Qualitative Matching

Using qualitative methods for spatial reasoning about object shapes is of interest in, for instance, matching. Normally matching is based on methods that require massive computational resources since most such methods work on pixel level. The method described here is qualitative and can be used to investigate objects for convexities and concavities where the position of these concavities and convexities relative each other are also of interest. Furthermore, information about local extreme points will also become available as well as various types of sub-forms. The approach that will be discussed here is based on symbolic projections. In symbolic projections strings of objects and their relations are generated from projections of the objects down to the coordinate axes. This technique is developed further in this paper so that relational strings made up by sequences of angles are used in the matching process together with projections of slopes from parts of the object contours.