Conflation

The word conflation refers to the process (compilation, reconciliation) or the resulting product of combining two or more versions of a complex information entity, such as a manuscript, map, or database, to produce a new improved version of the same entity. A map conflation is the process or the output map resulting from merging information from two or more maps of the same region. Spatial data conflation is the process of, or the output data set or database resulting from, merging two or more spatial data sets or databases for the same area or region. Although the terms conflation and data fusion are sometimes used interchangeably, data fusion is technically the more general concept of merging any two data sets, whereas conflation merges data sets that are presumed to each contain large subsets of common features. Map conflation is considered successful only after many or most features on one map have been matched to counterpart features on the other.

The first prototype semiautomated, computerized map conflation system was built in 1985 for a study of the feasibility of merging urban map data (Digital Line Graph [DLG] files) from the U.S. Geological Survey (USGS) with urban map data (Geographic Base File/Dual Independent Map Encoding [GBF/DIME] files) from the U.S. Bureau of the Census. At that time, long before the current widespread proliferation of spatial data sets, only government agencies had access to multiple map data sets of large coverages; hence, only government agencies had the opportunity, the resources, and the incentive to try to develop computer programs to conflate such data sets.

The USGS/Census semiautomated map conflation system employed a match-and-merge conflation strategy: Find duplicate features in the two [Page 569]map versions, then display all the unique features (duplicated or otherwise) exactly once to create a conflated map of the two versions. More specifically, the steps are as follows:

• 1.
  Using a variety of map-feature similarity measures, find candidate matching pairs of features (every pair will have one element from each of the two map versions).
• 2.
  Apply a topology-preserving geometric transformation (commonly referred to as rubber sheeting) to all the points of one of the map versions to bring candidate-matched features into exact alignment with each other.
• 3.
  Simultaneously display the two realigned and newly registered map versions for visual assessment of the effect of collocating the candidate-match pairs and displaying each pair as a single feature.

Since 1985, many private companies, academic institutions, and government agencies have continued developing spatial data conflation software. Although the matching criteria and alignment transformations vary slightly from system to system, all systems use the two-part feature-matching/feature-alignment model to generate graphic displays of the results of the conflation processing.

Map-to-map conflation has been expanded to include map-to-image conflation and image-to-image conflation. Working with images requires an extra step of image feature recognition, but once image features have been identified, those features can be matched and merged (i.e., rubber sheeted) to align with map features or other image features using previously developed matching and rubber-sheeting tools.