249 
other individuals assigned to the category, the correct 
identification rate is 94.726%. This shows that the selection 
and screening of the training samples have high rate of correct 
identification. They used for classification will be a good 
separation effects. The correct identification rate which is 
calculated by confusion matrix is an indirect measure of 
officially supervised and supervised classification mapping. 
Adopt of this approach, based on our land surface 
characteristics of the mining area classification system, the 
district stratification of 1981 respectively. 1995, 2001, remote 
sensing images for supervised classification, and reached a 
basic classification accuracy, ensure the correct classification 
rate of 94%. 
3 ANALYSIS OF LAND SURFACE 
EVOLUTION THROUGH MULTI-PHASE REMOTE 
SENSING IMAGE 
There are two ways that Satellite remote sensing images will be 
used in regional land surface dynamics of the evolution: One 
is directly based on remote sensing imagery, through an 
integrated multi-temporal image analysis and evaluation; 
another is the use of information extraction, classification 
results in GIS support, combining spatial analysis model 
analysis, which is Classification comparison. In this study we 
mainly uses the above two methods of the study area which 
were monitored for the analysis. 
visnsdmqmos mo rì ■„ 
3.1Remote sensing Monitoring and analysis based on image 
•jrif oi ¿fumndt 'jit 
^Tmies 
Land type 
1976-1981 
1981-1995 
1995-2001 
2001-2005 
Water 
-3.4 
-8.5 
-13.5 
3 
Residential 
land 
4.4 
11.8 
18.4 
2.7 
Dry land 
5.3 
14.4 
7.5 
2.2 
Water land 
-5.7 
-14.6 
-9.9 
-4.3 
Road 
-0.6 
-3.1 
-2.5 
-3.6 
Table .3-1 change table of different time type size 
■ ■■ !-. > 
Area static of every land type of different years 
‘jdt 'io bna srii ri-. 
□ water land 
■ residential 
land 
□ road 
□ dry land 
■ water 
Fig.3-1 Area static of every land type of different years 
Area static of every years of different land type 
water land residential land road 
land type 
dry land 
Fig.3-2 Area static of every year of different land type 
We used principal component analysis model for the 1995 and 
2001 two-phase TM Image Analysis of Evolution, we select 
the most informative bandsTM3, TM4, TM5 for each phase, 
and make principal component analysis. Make analysis of the 
PCA images, find that PCI, PC2 and PC3 covering all image 
information. Analysis of the principal components, we can see 
that: first and second principal component mainly reflects the 
image of the phase of relative stability, which is primarily 
used to identify unchanged, the third principal component was 
mainly reflects some changes.