A relational learning method for pattern and object recognition

Abstract In this paper we consider how shape (including patterns and objects) can be encoded in terms of a relational learning method which simultaneously derives features, their attribute ranges and the dependencies which best describe their specific shapes. To illustrate this approach we consider two problems in the context of pattern and object recognition. First, the problem of determining what constitutes `features' or `parts' of patterns? Second, the problem of what constitutes acceptable variations of shape in a recognition process? In the former case we examine polyhedral approximations to 3D objects while in the latter case we explore range-based objects defined by surfaces of arbitrary shape and form. The results demonstrate the robustness and explanatory power of the approach.