US LS
nem ii
MÀ — — LÁ LL I d
Table 5
Indexing the Vector Data by Quadtree
Qing—huai Gao, Qing—yun Shi, Min—de Cheng
Information Science Center, Peking University,
100871, Beijing, P. R. China
Abstract
In the study of GIS/LIS, two kinds of data structure
are studied, raster data structure and vector data struc-
ture. It is proved by experience and theory that the vector
data has a higher compression rate and a well —organized
logical structure ( that is, an object is represented by its
boundary polygons , and a polygon is a series of vectors
V1, 02, ...,Uw , Where v; and s, have a same end
point), and some basic image operations , such as scal-
ing, rotating, calculating the area and perimeter , can be
easily done on the vector data structure. But the spatial
organization of the vector data is very loose , which leads
to low efficiency when querying the objects near or at a
given position. So, a good spatial index is needed to speed
up the response of the system.
In this paper, it is suggested that the vector data can be
organized by a quadtree — like index structure — — — —
QO (quadtree of objects). The definition of QO , the al-
gorithm for creating QO of a image, the algorithms of op-
eration on QO are given. By analysing the querying ef-
fectiveness of QO, it is shown that the time consume of
queries on the vector with QO — index is much more
smaller than that on the "pure" vector data.
Keywords ; GIS/LIS, Raster, Spatial , Data Base
$1 Introduction
The main task of the researcher on GIS is to speed up
the implementation of variant queries on the data in some
compressed form . In ordinary, the atom query of a GIS
can be categorized into two classes ; one called attribute—
based query, and the other called position—based query.
The sophisticated queries are all composed by a series of
atom queries, which makes, in the view of users, the
GIS manage the spatial data (image data) and the at-
tribute data consistently and flexibly. Obviously , the po-
sition—based query has a more important position in the
studies of GIS than the attribute— based query , and the
speed of this kind query indicates the standard of a GIS.
The class of the position—based query contains: to find
the objects near or at a given position, to open a window
83
on a image , to judge the geometric relation of two objects
( for example , whether one object contains or intersects
with the other object) , to query the objects which have
some special geometric relation with a given object, etc.
In the study of GIS/LIS , two kinds of data structure
are studied; raster data structure and vector data struc-
ture. It is proved by experience and theory that the vector
data has a higher compression rate and a well —organized
logical structure ( that is, an object is represented by its
boundary polygons , and a polygon is a series of vectors
Dis Vas +.
point), and some basic image operations , such as scal-
ing, rotating, calculating the area and perimeter , can be
easily done on the vector data structure. But the spatial
organization of the vector data is very loose , which leads
to low efficiency when querying the objects near or at a
given position. So, a good spatial index is needed to speed
up the response of the system.
In this paper, it is suggested that the vector data can be
organized by a quadtree —like index structure — — — —
— — QO (quadtree of objects ). In a node of QO , a list
of objects is stored. A object is stored in the node which
correspondences to the smallest quad block that covers this
UN , Where v; and v4; have a same end
object.
In the second section , the definition of QO and the al-
gorithm for creating QO of a image , the algorithms of
operation on QO are given. In the third section , the time
consume of queries in the class of position — based query
are analysed . In the last section a short conclusion is
given.
$2 The definition and operation on QO
Organizing objects according their 2D position is the
main idea of QO.
2.1 The definition of QO
Suppose that the image f is a function on region R with
size 2* x 2* . Let set SO = {01,02,. . . yon} be the objects
on the image , and R(o;) represents the region correspon-
dences to oi .
A QO is a tree whose every node has 4 or no sons, that