1ality 
ment" 
y and 
M.Sc. 
p" JJ. 
rated 
(XVII, 
jatial 
Dept. 
KNOWLEDGE NETWORK AND INFERENCE MODEL 
ON INTELLIGENT INTERPRETATION OF TM IMAGE 
Qiming Qin 
(Postd octor ,Peking University Mathematical Department , Beijing ,China , 1t 1261) 
Abstract 
The TM image understanding based on knowledge is the key to completeintelligent interpretation 
of image. In this paper ,the author discussedthe model on knowledge network which is the tree of re- 
gional knowledge ,and given the algorithm on operation of knowledge node, furthermore, put forword 
the model of correlation analysis ,according tothe model, the intelligent interpretation of a simple lan- 
duse type is implemented by inference. 
Keywords: Interpretation of image; Knowledge network ; Correlation analysis ;Inference; 
1. Building the model of knowledge network 
Geographical region _ oriented knowledge 
network is developedbased on the features of 
knowledge and knowledge representation 
structure. The knowledge network can be de- 
scribed as follows; 
The knowledge network has many nodes, 
each node represents a region. the region is a 
hierarchic system according to the basic idea 
of comprehensive natural regionalization 
(Zhao, 1983), a large region includes middle 
regionsand a middle region includes smaller 
regions. A region tree is designed to represent 
the relationship among region and knowledge 
of background and interpretation. The region 
tree is defined as follows: 
(1) root node is the largest region ,it cov- 
ers a set of all regions which discussed. Root 
node has the unique geo — code, stores the 
common knowledge of background and inter- 
pretation which is fit to all regions. 
(2) Not —leaf node is the region which in- 
cludes middle or smaller regions, it is indicat- 
ed by geocode and has unique knowledge of 
background and interpretation different from 
other brather node’s knowledge. 
(3) Leaf node is the smallest region in 
which TM image interpretation is completed. 
363 
Leaf node is appended with the knowledge 
which can be used to image recognition of this 
region and has geocode too. 
Geocode indicates the hierarchy of re- 
gions. Each node has the unique geocode, 
amount of number in node shows the lever of 
region. Root node has the one Geocode, first 
lever has two geocode and records second 
geocode in storage structure,and second lever 
has three geocode and records third geocode in 
storage structure,and so on. Leaf node is inn 
lever of region hierarchic system and records 
last geo— code in storage structure. 
2. The algorithm on operation of knowledge 
node 
Here given three major algorithms, in 
which the storage structure of knowledge node 
and variables are printed as follows : 
geocode: this integer indicates code of re- 
gion node; 
leftlink :left pointer link son node; 
uplink :up pointer link father node; 
rightlink; right pointer link brother 
node ; 
vatiables in algorithm include : 
count is the integer counter of depth in 
region tree.