Triangulation based topology approach for 2D point sets registration

Abstract

Registration is designed to overcome global spatial correspondence ambiguities of different spatial databases and is required as a first stage in integration. It aims at ensuring an accurate and qualitative match while avoiding erroneous or coinciding data-features in the final product. The algorithm presented in this research paper describes an automatic constraint-free robust homological feature based registration of two 2D point sets. The points represent features of interest that exist in spatial databases, e.g. maps and topographic datasets. The algorithm entails an iterative aggregative voting process that consists of a set of qualitative statistical quantifications, based on the correspondences of triangulation structure, aimed at evaluating the geometric similarity of the data. The algorithm is structured to overcome data ambiguities, including data outliers and noise. The automatic aggregative voting algorithm replaces the need for a-priori spatial knowledge or the use of manual or semi-automatic registration that is prone to error. A comparison of the proposed automatic registration with alternative commonly used processes is presented, showing better and robust results. As such, the presented algorithm proves as an important stage towards the qualitative integration and enhancement of spatial databases.