Principle And Methods For Map Conflation Based On Least Sqaures Adjustment

As GIS data resources become more abundant, it becomes more important to integrate all these resources to get valuable information. Because different departments obtain data by various means, these data resources differ in many respects, such as data model, database structure, positional accuracy, attribute etc. It needs to get rid of these discrepancies to create a new map according to some demands. The new map will have higher positional accuracy or rich attribution. Map conflation is an effective strategy to integrate spatial information automatically.Research on map conflation started from a program conducted by US Census Bureau from 1983 to 1985. After matching corresponding features between two maps, affine transformation was employed to reconcile coordinate discrepancies and transfer attributes. The process of map conflation includes two procedures: feature matching and map merging.This paper mainly deals with the algorithms and methods of map conflation.(1) Existing feature matching methods can be classified as two categories: common methods and methods based on probability theory. Through studying on these methods, firstly matching criteria are summarized. These criteria have two categories: spatial and non-spatial. The spatial criteria include distance, shape, structure, direction and topology etc. Non-spatial criteria are unrelated with features' geometry and location. An example is features' name. Most matching methods only deal with one-to-one matching relationship and need to determine matching threshold. This paper introduces a probabilistic method for feature matching, which avoids selecting thresholds and attempts to resolve one-many and many-many matching relationship.(2) Most map-merging methods move matched points to their counterparts and utilize affine transformation to move unmatched points. Linear features are usually replaced by their counterparts, or adjusted using projection. These methods will more or less bring about shape distortion and can't handle one-...