GAP-Tree Extensions Based on Skeletons

Abstract

The Generalized Area Partitioning tree (GAP-tree) is a model that supports on-the-fly generalisation of planar partitions of polygon objects. This paper focuses on solving some limitations of the existing GAP-tree and explores two extensions. The first extension enables the handling of a disjoint polygonal cluster (within the partition) and allows aggregation operations to be performed. The skeleton partitioning model, which is based on the constrained Delaunay triangulation for the polygonal cluster, is used to define the bridge areas between aggregated polygons. The second extension involves the improvement of removing insignificant objects by separating them into parts around the adjusted skeleton and assigning these parts to different neighbours. The adjusted skeleton is defined by the compatibility between the removed object and its neighbours, which considers not only topological relationships but also importance and semantic similarity. This process again uses the Delaunay triangulation. The algorithm is given to construct the extended GAP-tree. NR - 13 PU - SPRINGER-VERLAG BERLIN PI - BERLIN PA - HEIDELBERGER PLATZ 3, D-14197 BERLIN, GERMANY