A Rule-based Approach for the Conflation of Attributed Vector Data

Abstract

In this paper we present a complete approach for the conflation of attributed vector digital mapping data such as the Vector Product Format (VPF) datasets produced and disseminated by the National Imagery and Mapping Agency (NIMA). While other work in the field of conflation has traditionally used statistical techniques based on proximity of features, the approach presented here utilizes all information associated with data, including attribute information such as feature codes from a standardized set, associated data quality information of varying levels, and topology, as well as more traditional measures of geometry and proximity. In particular, we address the issues associated with the problem of matching features and maintaining accuracy requirements. A hierarchical rule-based approach augmented with capabilities for reasoning under uncertainty is presented for feature matching as well as for the determination of attribute sets and values for the resulting merged features. Additionally, an in-depth analysis of horizontal accuracy considerations with respect to point features is given. An implementation of the attribute and geometrical matching phases within the scope of an expert system has proven the efficacy of the approach and is discussed within the context of the VPF data.