Search for multi-component objects in large spatial databases

Abstract

Many queries to spatial database systems concern spatial objects that may be conceptually modeled in terms of multiple components satisfying constraints on both attributes and spatial relationships. We investigate the performance of a new algorithm called Forward Constraint Propagation (FCP) that dynamically assembles the spatial locations of a multi-component object from a spatial database given a description of its components and their spatial relationships. The database stores the spatial locations of individual components in either object-based or location-based form. FCP embeds spatial constraint propagation within a heuristic tree search algorithm developed for the general constraint satisfaction problem. An experimental investigation based on a quadtree-structured database of “real-world” objects indicates that FCP outperforms eight alternative procedures in terms of time complexity.