|. Istanbul 2004 
nanual. As the 
> radiation from 
, the seam lines 
the seam lines 
adjacent stereo 
or every stereo 
polygons are 
e relationships 
or between the 
es: The left and 
ly according to 
ir model in the 
one of the two 
to one of the 
f the area. 
Iterior and 
M 
  
1ge 
L 
ir models 
  
hoto-pair 
  
Yoasicked | 
mosaic 
1age 
troduction | 
nate | 
  
asure 
del | 
urable seamless 
and a mosaic 
form a seamless 
he stereo model 
some operations 
oscopic viewing 
ice. However, it 
e linage of each 
sasurable stereo 
  
  
   
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
42. 3D measurement in measurable seamless stereo model 
The process for measurement on such a measurable seamless 
stereo model is illustrated in Figure 3. 
  
The corresponding points in the measurable seamless stereo 
model (Manually measured or by image matching) 
  
  
  
  
: t i 
E pixel coordinate (1, J) in The pixel coordinate (Ig, Js) in 
  
  
the orthoimage the stereomate 
  
  
: 
The planemteric coordinates 
(Xs, Ys) in the stereomate 
ge] DTM | — 
Y 
The planemteric coordinates 
(X^, Y?) in the orthoimage 
  
  
The planemteric coordinates 
(X, Y) in the orthoimage 
  
  
  
  
  
  
  
  
  
  
  
  
  
—p 
Y 
The 3D coordinates The 3D coordinates 
(X, Y, 2) (CY, Z7) 
  
  
  
  
  
  
i 
  
Searching the original photo pair, which the corresponding point| 
comes from 
  
  
: | 
  
  
The left (right) photo The right (left) photo 
coordinates (x , y) coordinates (x' , y?) 
  
  
  
  
| Space intersection | 
: 
The ground point 
coordinates (Xp, Yp, Zp) 
  
  
  
  
  
  
Figure 3. The 3D measurement in measurable seamless stereo 
The 3D measurement steps in detail are as follows. 
(i). For a given point on the stereo model, the corresponding 
point on the stereomate is found by image matching. The 
image coordinates are recorded separately as (I, J) in the 
mosaic orthoimage and (I, J,) in the stereomate. 
(ii). The pixel coordinates are separately transformed to ground 
planimetric coordinates, to become (X,Y) and (X,Y). 
(iii). For the planimetric coordinates (X, ,Y,) in the stereomate, 
the planimetric coordinates (X, Y" ) in the orthoimage before 
introducing parallaxes can be computed according to the 
parallax function and DTM. 
(iv). For the planimetric coordinates of a corresponding point 
pair, (X, Y) and (X^, Y^), the heights Z, Z^ are bilinearly 
interpolated from DTM. The 3D coordinates of the 
corresponding point (X, Y, Z) and (X', Y', Z^) are then 
obtained. 
(v). The valid mosaic polygon to which the corresponding point 
belongs is searched according to the planimetric ground 
coordinates (X, Y) in the orthoimage. Then the original 
photo pair, from which the corresponding point comes, is 
also searched according to the valid mosaic polygon. 
(vi) The coordinates (X, Y, Z) and (X', Y, Z^) can be 
transformed into photo coordinates (x, y) and (x', y’) using 
the known interior and exterior orientation elements of the 
photos according to collinear equation. 
(vii). The ground point (Xp, Yp, Zp) can then be computed 
according to the forward intersection. 
5. EXPERIMENTAL TESTING 
According to theory and procedures described in the previous 
two sections, a prototype the software SOD (Stereo Orthoimage 
Database) is developed. Using in Visual C++. 
Two sets of data with different photo scales and terrain types are 
used to test the presented method and the software. The 
parameters of the data sets are listed in Table 2. The pixel sizes 
of the orthoimages for data sets I and II are corresponding 
approximately to the footprint of the raw image pixels. 
  
  
  
  
  
  
  
  
  
  
  
  
model 
Table 2. The experimental data sets parameters 
Item Data Set I Data Set II 
Principle distance 153.710mm 304.034mm 
Photo scale 1:25,000 1:8,000 
Format 23cm X 23cm 23cm X 23cm 
Overlap 60% 60% 
Photo type Panchromatic False Color 
Pixel size 25um 25um 
Average flight height 4,225m 2,090m 
Landform Hill mountain City area 
GSD of DEM 12.50m 5m 
GSD of orthoimage 0.65m 0.2m | 
Data Range Five strips with five stereo models per strip Three strips with nine stereo models per strip 
  
  
  
  
  
Figure 4 shows the result of the measurable seamless stereo 
model generated from Data Set II. Figure 4a illustrates the valid 
Mosaic polygons of the stereo photo-pair in the block area. 
Every stereo pair has a valid mosaic polygon. Figure 4b is the 
measurable seamless stereo model in red and green 
complementary color mode. The stereo model can be directly 
observed with red/green glasses. The 3D measurement of terrain 
surface points and objects are performed on both two data sets. 
For Data set I, twenty ground points were selected to the 
accuracy test. These points are separately measured from 
measurable seamless stereo model and from the original 
photo-pair model at the SOCET SET v4.4.0 of Leica 
Geosystems, Inc. The latter is used as the benchmark for 
accuracy assessment. The maximum error and RMS error are 
listed in table 3. Figure 5 illustrates the error distribution of 
these twenty points.