Topological Similarity Measurement of Region Compositions Based on Boundary Contacts

Abstract

Topological relationship, one of the most important subjects in spatial analysis, has been studied thor oughly and applied to various fields. An important application of topological relations is the similarity measurement on spatial scenes. However, when considering the common topological relations, there exists dif ficulty in distinguishing different region compositions where complicated boundary contacts occur between regions. To address this problem, a method of describing detailed topological relations in a region compo si tion has been proposed [Lewis et al. 2013], by traversing a region's boundary and recording boundary intersections one by one. To implement similarity measurement on region compositions where complex boundary contacts happen, we propose a measuring model based on the boundary contact recording method. The model is struc tured in two steps, a preliminary matching step and an exact matching step. In the preliminary matching step, we recognize and filter the very dissimilar candidates comparing to the reference region composition; meanwhile, we obtain corresponding relations of regions between the potential candidates and the reference, and represent the correspondences with an association graph composed of nodes and edges. In the exact matching step, we encode boundary contact records to binary sequence, and adopt a sequence alignment method, an approach from bioinformatics to compare the sequences of DNA, RNA, or protein, to fulfill the topological similarity measurement for two region compositions. We illustrate the complete process of our model through a case study, and show the survey result on weight setting for criteria in the exact matching step.