As the result of comparison, the minimum absolute sum method 
were adopted in this study. A center of n x n window or the 
subdivided area to minimize the absolute sum of difference 
between stereo images is the conjugate point. In this study, 
the window size were varied from 3x3, 5x5, 9x9, 11x11 
and 33 x 33. 
Table 3 shows the result of auto-correlation with respect to 
mismatch and computing time for sampled points. Table 4 
shows the accuracy of three dimensional measurement as 
compared with the corresponding observed coordinates which 
are given by DTM. Table 5 shows mean error of height 
measurement with respect to number of G.C.P. and stereo 
angle, by independent solution. Table 6 shows mean error of 
height measurement with respect to number of G.C.P. and 
stereo angle by integrated solution. Fig.4 shows the 
comparison between observed contour map produced by DTM and 
calculated contour map produced by auto-corretation. 
CONCLUSION 
a. Simulated stereo imagery which would be generated by 
linear array sensor is useful for evaluating the methodo 
logies of orientation and automated three dimensional 
measurement. 
b. Exterior orientation parameters as the function of time 
or line can be determined by using ground control points 
and/or satellite data. Accuracy of exterior orientation 
was not more than +1 pixel for the test area with fourteen 
control points. 
c. Minimum absolute sum method was adopted for automated 
three dimensional measurement. Accuracy of height measure 
ment was 1.00 pixel for stereo angle of 45°, 0.963 pixel 
for stereo angle of 60°.