Minimum bounding rectangle: Difference between revisions

Revision as of 20:16, 13 February 2008

Revision as of 02:46, 26 April 2008

Line 23:

==Applications==

The purpose of an MBR is usually to provide a proxy for a more complex spatial object, that can then be easily processed for simple spatial queries based on overlapping ranges in both the x and y dimensions. This produces a coarse "spatial filter" that can either be used alone, recognising its limitations, or as a pre-filter stage to a more computationally expensive "overlapping polygon" test<ref>[http://www.oracle.com/technology/products/spatial/htdocs/data_sheet_9i/10g_spatial_locator_ds.html Oracle Corporation Data Sheet - ORACLE® Spatial Option and ORACLE® Locator: Location Features in Oracle Database 10g]</ref>. Applying the (computationally trivial) "overlapping rectangles" test as an initial step can frequently eliminate many spatial objects that cannot satisfy the designated spatial query conditions before passing to the more precise, but slow, second stage.

The degree to which an "overlapping rectangles" query based on MBRs will be satisfactory (in other words, produce a low number of "false positive" hits) will depend on the extent to which individual spatial objects occupy (fill) their associated MBR. If the MBR is full or nearly so (for example, a mapsheet aligned with axes of latitude and longitude will normally entirely fill its associated MBR in the same coordinate space), then the "overlapping rectangles" test will be entirely reliable for that and similar spatial objects. On the other hand, if the MBR describes a dataset consisting of a diagonal line, or a small number of disjunct points (patchy data), then most of the MBR will be empty and an "overlapping rectangles" test will produce a high number of false positives. One system that attempts to deal with this problem, particularly for patchy data, is [[c-squares]].