4.2 Muti-point matching method 
the area 
exist but 
In order to extract a GCP in 
where many pointable objects are 
no single object satisfies necessary 
accuracy. we tried to develop a method to 
Fitts cup the ‘accuracy of a GCP “by "using 
multiple points ‘andigetting an average of 
them. The process and algorithm of the 
multi-point matching method are as 
follows. 
(1) Measurement of non-accurate GCPs 
Some ten points in a’ local window are 
measured both on an image and on the 
ground, whichever the accuracy of each 
point is good or mot. 
(2) Average matching of points 
Correspondence between image coordinates 
and geodetic coordinates is given by least 
square adjustment using those points in 
the window. 
(8) Screening of points 
Among the “points in the window, those 
points which residuals exceed certain 
Then average 
excluding 
tolerance are to be omitted. 
matching is repeatedly executed 
the omitted points. 
(4)*Adoption as :a-GCP 
The center of the points included in the 
final matching calculation is adopted as a 
GCP. 
B. "PLOTTING LIMIT ESTIMATION 
The resolving power of panchromatic SPOT 
image is stated as 10m on the ground, 
however, it is not exactly mean that every 
objects larger than 10m can be plotted and 
those smaller than 10m can not be plotted. 
Practically, narrow street may sometimes 
clearly be seen, while large buildings may 
sometimes be difficult to be identified. 
  
Therefore, it was inspected whether ground 
objects expressed on a topographic maps 
were able to be identified on SPOT images. 
5.1 Study flow 
Two SPOT images of Japan and one image of 
Peru were chosen for the study. 30 sites 
of 5km by 5km wide were selected as test 
sites of Japan and four sites of 
approximately 10km by 10km wide of Peru. 
In each test sitos, ground objects 
expressed on corresponding topographic 
maps of 1:25,000 of Japan and 1:50,000 of 
Peru were examined if they were able to be 
identified or not. The results were 
displayed on the corresponding maps. 
Figure 5-1 and 5-2 show examples of the 
results. 
5.2 Plottable objects 
(1) Water 
Since water absorbs infrared ray, water 
339 
  
    
  
@-—® : Buildings 
: Railroad @—@ : Petrorium tanks 
Interpreted objects (Japan) 
© 
Figure 5-1 
  
  
  
  
  
  
: Y Ed 
RSS 
Interpreted objects (Peru) 
  
Figure 5-2 
can clearly be discriminated from others. 
[t is Possible to plot oven a thin stream 
or a reservoir except the case that it 
flows in deep forest. 
(2) Road 
Wide roads and highways are possible to be 
plotted except the case it run through 
densely urbanized area. Plottability 
depends on gray level contrast against its 
surroundings. If there exist very little 
contrast between a road and its 
neighborhood, it is hard to be plotted 
even it runs through country side. 
(3) Railway 
Railway is more difficult to be identified 
than road, in general. 
(4) Buildings 
Plottability of buildings depends rather 
on its contrast than its size. Buildings 
larger than SPOT image resolution are not 
always identified, while smaller 
independent building may sometimes be 
identified.