296 
Ac = the area of the minimal excircle 
P = the perimeter of the ground feature 
L = the length of the long axis 
Moreover, it can be expressed by the rectangle, the roundness, 
the ratio of length and width etc. 
The rectangle. E.g. 
R = A 0 /A r . (5) 
where A 0 = the area of the ground object 
A r = the area of MER 
The ratio of length and width. E.g. 
A = L/W. (6) 
where L = the length of the ground object 
W = the width of the ground object 
The roundness. E.g. 
C = P 2 ! A. (7) 
where C = the roundness 
P = the perimeter of the ground object 
A = the area of the ground object 
2.3.2 Remote Sensing Image Information Extraction 
Based on Edge Knowledge: The different ground features 
have the different edges, and the parameters which describe the 
ground feature’s edge are linear shape, toothed shape, undee 
shape etc. The edge knowledge can be used for the ground 
feature’s identification of the location and the attribute: First, it 
extracts the edge information with the edge enhancement, and 
then ascertains the ground feature’s attribute with it’s edge 
knowledge when it is used for the identification of the location 
and the attribute; It ascertains the attribute of the extracted 
information ulteriorly when it is used for ascertaining the 
attribute. 
2.4 Remote Sensing Image Information Extraction Based 
on Other Assistant Knowledge 
2.4.1 Remote Sensing Image Information Extraction 
Based on Process Knowledge: It can carry out remote sensing 
information extraction based on the ground feature’s process 
knowledge when we have the multi-period remote sensing 
image. The different ground features have the different 
variation cycles. The ground features have the unique change 
characteristic within their variation cycles. It can establish the 
extraction models based on the process knowledge according to 
these characteristic. 
2.4.2 Remote Sensing Image Information Extraction 
Based on Assistant Knowledge of GIS: In information 
extraction, besides using remote sensing data, it needs to use a 
mass of correlative data, and these data are the graphics data 
and the non-graphics data from GIS. The graphics data are all 
kinds of existing pictures, for example, the present land-use 
map, the topographic map, the slope map, the aspect map; the 
non-graphics data are the statistical data such as population, 
society and economy etc[5]. There are two steps when uses the 
graphics data, the first step: mine knowledge; the second step: 
contact the graphics data with the remote sensing data in order 
to support information extraction with the mined knowledge. 
Figure 1. The flow chart of remote sensing information 
extraction based on knowledge 
3. THE REALIZATION OF OBTAINING KNOWLEDGE 
AUTOMATICALLY SPATIAL DATA MINING AND 
KNOWLEDGE DISCOVERY 
Spatial Data Mining or Knowledge Discovery from Spatial 
Database, is to extract the spatial mode and characteristic, the 
universal relation of the spatial and non-spatial data and the 
other universal data characteristic of keeping in the database of 
being interested by users from the spatial database. We give a 
more concise definition: Spatial Data Mining and Knowledge 
Discovery is the process of extracting the spatial and non- 
spatial mode and the universal characteristic which are 
connotative and interested by users from the spatial database. 
Data Mining and Knowledge Discovery is a new subject 
developed on the basis of intercrossing many subjects and 
technologies mutually such as Machine Learning, Computer 
Visualization, Pattern Recognition, Statistic Analysis, Data 
Base and Artificial Intelligence. SDMKD has the close