Efficient Access to Spatial Data

Manipulation of spatial data frequently involves spatial operators such as intersection, containment, and adjacency. These operators are computationally much more expensive than relational operators such as selection and join [Ooi et al., 1997]. Thus it is extremely important to design efficient access methods for spatial databases. In order to efficiently access spatial data, its spatial properties need to be used in designing the access methods. Efficient access is achieved through the use of indexes that provide a linkage between spatial properties and physical storage locations of the data. Unfortunately, conventional indexing methods such as B-tree and hashing are not well suited for accessing spatial data because such methods are designed for alphanumeric data that has no spatial properties. However, many of the conventional indexing methods have been used as the starting point for designing spatial indexes. For example, kD-trees [Bentley, 1975a, Bentley and Fridman, 1979] and its variations have been derived from binary trees, R-trees [Guttman, 1984] and its descendants have their roots in the B-trees, and Grid files [Hinrichs, 1985] and their extensions use hashing functions.