The above step are repeated until infer- 
ence engine derives the correcting results for 
TM image. 
4. The discussion about the test result 
The development of simple interpreting 
system is completed in 1991. 
The system program is implement by C 
language (Qin, 1991). The author puts the sys- 
tem to the test of landuse classification using 
TM landsat image (date: Oct. 3,1984) in Bei- 
jing region. The test shows that it is of cor- 
recting in landuse type such as city land type, 
grassland type, vegetable plot type and culti- 
vated landtype. In addition,lake and river can 
be recognized by shape of water in data plane. 
Meanwhile, the author pay more attention to 
that there are namy job to do, which include 
the building geo _models and appending image 
interpreting knowledge to knowledge base ,ap- 
plying parallel processing algorithm in refer- 
366 
ence engine in future . 
Reference 
Burrough. P. A., 1986. Principles of Geo- 
graphical Information Systems for Land Re- 
Sources Assessment, Clarendon Press, Ox- 
ford ,pp. 20— 22. 
Qin Qiming, 1990. Desigh and implementation 
of Knowledge Base on LandEvaluation. In; 
14th International Congress of Soil Science, 
Japan,Vol. V. pp. 132—135. 
Qin Qiming. ,1991, A approach of recognition 
Based on Knowledge to Image ,Proceedings of 
the Chinese Joint Conference on Pattern 
Recognition and Artificial Intelligence , China, 
pp. 691—694. 
Zhao Songqiao,,1983,A new scheme for com- 
prehensive physical regionalizationin China, 
Acta Geographica Sinica , Vol. 38,No. 1,PP. 1 
— 10, 
peut 
I(M,l 
les 
nive: 
Soit 
niveez 
la fc 
Ou X 
est 
appel 
Ele,