This research addresses the rising critical need for automated conflation of vector/raster data as more and more high resolution data are collected for updates/improvement of graphic products employed in various mission areas of the Army and other user communities. Our conflation methodology derives from the integration of two well-established conflation approaches, and is developed with a preliminary software design in Phase I. Successful testing using Government data will be followed with a Phase II effort wherein we will complete the conflation system design and produce a prototype ready for deployment as a commercial product. Our approach starts with a deep scientific understanding of the complexities/challenges that characterize the conflation problem. We attack the misalignment issues of nonlinearity, disparate scales, uncontrolled noise, etc., on a first principles basis using rigorous mathematics rather than empirical trail and error adjustments that basically distribute the misalignment rather than attack fundamental causes. In developing our integrated methodology, we will establish the synergism achieved by combining two approaches: use of algebraic algorithms and similarity transformation of local features. We establish three fundamental research hypotheses about these complementing approaches, which will be tested and evaluated in order to provide a technical roadmap toward the optimal conflation solution.
Benefits: Our optimal conflation solution is aimed at minimizing human intervention while maximizing automation of the process. We take advantage of preprocessing activities that identify/assess available metadata and prior knowledge about data sets. We initially examine data source differences in resolution, orientation, and feature mismatching, and thereby inform/expedite the subsequent processing steps as tangible economic and efficiency benefits.
Keywords: Aligning, Vector data, conflation, image registration, invariants, parameter space,image understanding.
