194 
1 FINE PROCESSING OF TM IMANGE DATA 
At present, most products of the TM-CCT image data and image 
from are in the form of rough product at the ground receiving 
station. The residuals of these images are about 5-7 pixels and 
their planimetric accuracy is about 120-150m on ground.In order to 
improve their geometric accuracy and transform them into a cer 
tain map projection system as well as to match TM-image data with 
non-remote sensed data, the fine processing of the image data is 
necessery.The procedure of fine processing for our investigation 
is shown as follows. 
1-1 RECTIFICATION BY GROUND CONTROL POINTS 
The original TM image is rectified into UTM projection system 
by using the polynomial method during the fine processing.For this 
reason we have choosen 40 nature ground objects from large scale 
topographic map as control points. 
Due to the number of control points more than the number of 
polynomial coefficients,a least squares adjustment should be done. 
During the adjustment, gross errors have been detected in 7 
control points. After rejecting these points the adjustment is 
performed by the remaining 33 points and the standard errors for 
control points are listed in Tabel 1. The 3rd-order polynomial is 
recommendable. 
1 order 
polynomial 
2 order 
po1yno mai1 
3 order 
polynomial 
x-direction 
32.34206390 
29.88285828 
25.93333054 
y-direction 
26.39077377 
26.10504331 
24.68536568 
Table 1. Standard error after adjustment 
1-2 RESAMPLING 
Three methods for the intensity resampling are tested 1 
Considering the spectral intensity of feature and geometric accu 
racy of image. Bi-Linear interpolation method is used for resampl 
ing with the pixel size of 20mx20m. 
Bi-Linear interpolation I(p) = £ £ I(i,J)*W(i,j) (1) 
i=iJ=i 
2 MATCHING AND REVERSE MATCHING 
2-1 MATCHING BETWEEN FINELY PROCESSED IMAGE AND NON-REMOTE DATA 
Non-remote sensing data refers to these data which are acquii— 
ed by non-remote sensing ways (such as topography, soi1,landuse, 
climate, population and economy).Matching among multi-type of data 
is very important for classification and recognition of TM image 
in analysis and evaluation of resource and environment. Befor 
matching it with the fine processing image,the mon-remote sensed 
data have to be digitized,and they must have the same projection 
system of the fine processed image. Then the lst-order polynomial 
can be used to match them. Diagram of matching see FIGURE1.