An optimisation model for linear feature matching in geographical data conflation

Issues of heterogeneity and incompatibility in geospatial data become increasingly important as data sources become more abundant. Scientific research and decision-making usually require geospatial data from a variety of sources, since it is not realistic to collect all data directly; therefore, it is important to effectively utilise data created by various agencies using different methodologies under different circumstances. The term conflation here refers to the problem of combining incompatible geospatial data. One crucial component in conflation is feature matching, which is a prerequisite for the subsequent steps such as feature transformation. Although previous research has provided different methods of feature matching for specific applications, most of them have relied on a greedy strategy to execute the matching process. This article develops a new optimisation model to improve linear feature matching in situations with one-to-one, one-to-many and one-to-none correspondences by extending the optimised feature matching method proposed by Li and Goodchild (Automatically and accurately matching objects in geospatial datasets. In: Proceedings of theory, data handling and modelling in geospatial information science. Hong Kong, 26-28 May, 2010). Considering all possible matched pairs simultaneously, this new model achieves a high percentage of correctly matched features by maximising the total similarity between all matched pairs. When autocorrelated distortions exist in the datasets, an affine transformation can be integrated into the feature matching to improve the matching results. In addition, this study takes advantage of the asymmetry of a dissimilarity metric - directed Hausdorff distance - to address one-to-many correspondences.