The design of the cell tree: an object-oriented index structure for geometric databases

Abstract

The design of the cell tree, an object-oriented dynamic index structure for geometric databases, is described. The data objects in the database are represented as unions of convex point sets (cells). The cell tree is a balanced tree structure whose leaves contain the cells and whose interior nodes correspond to a hierarchy of nested convex polyhedra. This index structure allows quick access to the cells (and thereby to the data objects) that occupy a given location in space. The cell tree is designed for paged secondary memory to minimize the number of disk accesses occurring during a tree search. Point locations and range searches can be carried out very efficiently using the cell tree.<>