have ODBC, DAO, OLE C++ Oracle, OLEDB, ADO,
PRO*C and OCI, etc. In view of the compatibility and
efficiency in system development, this paper adopts OCI to
develop Oracle data.
3. THE KEY TECHNOLOGY RESEARCH ABOUT
ESTABLISHING THE IMAGE DATABASE
3.1 Block, coding and index of image data
1) Block of image data
In the huge image database, scheduling and the use of the
image data is only a small part of the database. If the data files
are very large, it will directly affect the speed of data read and
implementation. Data block is the key image database
technology to organize and manage the data efficiently. In view
of practical problems of theory and application, image database
is established to adopt data block size 128x128 pixels or
256x256 pixels.
Image data are to the production of maps for the unit, each is a
map of orthophotos of the final products are stored as
documents. Virtually every map can be seen as a Segment, but
each piece of the map data is huge, maps internal data also are
partitioned once again. The partition of image block has two
ways: Strip division and massive division. Because massive
division support graphics indexing and mosaics, and the
division of Block has good aggregation properties, massive
division is used to partition the data block in the image
database.
2) Coding of image data
The basic principium of space coding is that image block is
organized by some strategies. It is a process that
two-dimensional space object mapped to a one-dimensional
space in the light of a certain coding function. The most
commonly methods of coding space data: Row Ordering,
Morton Ordering and Hibert-Peano Ordering etc. Row
Ordering and Morton Ordering are relatively simple. They
support image and block access to data directly and regional
inquiries. In order to manage and use the space data expediently,
4D products are a standard topographic map to produce and
data files are to Map No. name as the file name, in mapping
fields. For every Orthophotomap ,we can code it with Row
Ordering after partition of image according to a certain size ,
number the corresponding maps before each coding, then can
constitute the second level coding of every block.
It is impossible to partition into integral blocks with 128x128
pixels or 256x256 pixels block because of the second level
coding of maps and blocks. As shown in Figure 2.
6
60
61
62
63
64
68
66
}
5
50
51
52
53
54
55
56
4
40
41
42
43
44
45
46
3
30
31
32
33
34
35
36
2
20
21
22
23
24
25
26
1
10
11
12
13
14
15
16
0
03
01
CG
03
04
05
06
Û
1
2
3
4
$
6
Fig 2.Code way of Row Ordering
This paper the size of 1:2000 image is 1668 x 1668 pixels, it
can be divided into a total of 7x7 = 49 image blocks with 256 x
256 block to gather 2n * 2n forms(23 * 23 = 8x 8 = 64 forms),
to meet Morton Ordering or Hilbert Ordering Coding. It need to
increase the complexity of the algorithm to affect the index of
the image blocks. So this paper use Row Ordering to code. This
method is simple and support image block to access to data
directly and regional inquiries. The rows and columns code
from the lower left-hand comer of the map, from left to right
increase gradually,col= 1,2,...from the bottom up increase
gradually, row=0,1...
3) The index of image data
Window index is the most commonly data index way from the
image database. Its core is to find image data coding in the
window quickly. Then extract the corresponding image data
block from the image database according to image blocks
coding, setting the image blocks together, you can get the
image window. The process of indexing the image blocks
within the scope of window from the image database is shown
in figure 3:
Image
Scale
index
Image
Project
Index
Map
index
Image
Types
index
Image
block
index
Image
block
Record
index
Fig 3. Image Database Multi-Index
3.2 Structure and Construction of image pyramid
PCJL0CKID Jt
PCJL00K
4009-49502
D7D389D7D38...
4009-49503
B28863CCM8...
4009-49511
B2BACCB2C8...
Fig 1. Corresponding relation between block’s number and data
block
Image is classification stored and managed in the light of
resolution ratio. The resolution ratio of the bottom is highest,
and has the largest amount of data. The lower the smaller the
volume of data, so different resolution radios remote sensing
images form a tower structure. The structure is called the
image pyramid. Remote sensing database is built with image
pyramid structures which is easy to organize, store and
manage the multi-dimensional, multiple data sources of remote
sensing data, to achieve a cross-resolution of index and browse.