248 
2.3 Bands combination 
Multi-spectral remote sensing data has many bands, how to 
make bands combination is the key. Band chosen should adhere 
to the following three princilies[3-7].l) selected the three bands 
to large volume of information; 2) selected the three bands 
band combinations with other features has a good separability. 
• information of band combination (within the coverage of all 
the features of the total amount of radiation) the largest of the 
Satellite 
Year 
Frame 
Time 
Bands 
Landsat 185*185 (MSS 80) 
1976 
128/31 
09.16 
4 bands 
Landsat 185*185 (MSS 80) 
1981 
128/31 
10.14 
4 bands 
Landsat 185*185 (TM 30) 
1995 
119/32 
05.15 
7 bands 
Landsat 185*185 (ETM) 
1996 
120/31 
05.24 
7 bands 
Landsat 185*185 (ETM) 
1996 
120/32 
05.24 
7 bands 
Landsat 185*185 (TM 30) 
2001 
119/31 
08.11 
13 bands 
Landsat 185*185 (TM 30) 
2001 
119/32 
08.11 
9 bands 
Landsat 185*185 (ETM) 
2001 
119/32 
09.03 
9 bands 
QUICKBIRD16.5*16.5(PAN 0.61) 
2005 
10.20 
3 bands 
CEBERS 113*113 (CCD 19.5) 
2005 
368/53 
05.18 
5 bands 
Table. 1-1 remote sensing data source 
Land type 
No. 
Meaning 
Water 
1 
Include water, regional water, rivers, reservoirs, ponds. 
Residential 
land 
2 
Residential sites and mining sites, land transport. 
Dry land 
3 
Main planting com, soybeans, and other crops 
Water land 
4 
Main rice planting 
Sink land 
5 
Mining Subsidence 
Table.2-1 system of mine land change classification 
Before 
classification 
After classification 
Water 
Construction 
Dry land 
Water land 
Sink land 
Water 
238 
1 
0 
0 
1 
Residential land 
2 
512 
14 
4 
7 
Dry land 
0 
3 
603 
8 
1 
Water land 
6 
11 
5 
420 
6 
Sink land 
8 
6 
7 
5 
170 
Sum 
254 
533 
629 
437 
185 
Percentage 
93.7% 
96.06% 
95.87% 
96.11% 
91.89% 
Table .2-2 classification confusion matrix of study area 
three-band (single wave of the standard deviation, correlation 
coefficient between the band and identify common index OIF), 
Together with the features of different spectral characteristics, 
determine the best band combinations. 
2.4 Supervised classification of remote sensing images 
Based on the district tiered 
The existing image classification methods are generally the 
same standard of the whole image classification, However, 
remote sensing images spectrum feature are more complicated, 
even after treatment, the images are often arise from different 
spectrum, foreign body with the spectrum, Hill shadow effects 
and significantly different regions or different seasonal 
characteristics of the images are quite different phenomena 
Therefore, the district is closer to the idea of remote sensing 
image characteristics, inline with the interpretation of ideas. 
1 n Mine resources and environmental monitoring 
classification system 
2-, Supervised classification of mine district remote sensing 
image 
First, chose classification samples according with the 
classification system of mine district, in the process of 
selection as far as possible to ensure accuracy of the sample. 
After obtaining samples of repeated screening and finally 
obtain the classification training template. We choose sample 
of the 1995 TM image of mine district, obtain the region’s 
classification confusion matrix, the table below. 
From Table 2-2, we can see that the classification of the 
various types of samples before the individual classification 
after most still belong to their original category. Only a few