The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part Bl. Beijing 2008
1304
where
GM
~2 .(X 2 0 +Y 0 2 +Z 2 o f' 2
and ti and t 2 is the acquisition time of the corresponding
framelets in image 1 and 2 respectively.
Combining the equations (4) based on (2), B x is defined as
follows:
B x =x Lr x Li
— u x • (/| — t 2 — dt) + A • X 0 • {t 2 -F dt + 2 • ?2 ■ dt — )
The above equation clearly shows the important role of the
UCL sensor model approach because it relates directly the B x
only with the state vector of the center point of the first
pushbroom image, which are the unknown parameters in this
equation.
Going one step further the acquisition time t 1 and t 2 can be
described as:
t x = x, • interval
t 2 = x 2 ■ interval
where interval is the acquisition time of a framelet which is
assumed to be constant (Michalis and Dowman, 2004).
Finally B x is defined as follows based on the previous
equations (with exactly the same procedure the B y and B z are
also calculated):
B x -u x ■ [interval • (x, - x 2 ) - dt\ +
+ A ■ X o ■ [interval 2 ■ (x 2 - x 2 ) + 2 • x 2 • interval ■ dt + dt 2 ]
B y = u y • [interval ■ (x, - x 2 ) - dt\ + ^
+ A ■ Y o ■ [interval 2 ■ (x 2 - x, 2 ) + 2 • x 2 • interval ■ dt + dt 2 \
B z = u z ■ [interval • (x, - x 2 ) - dt] +
+ A • Z 0 ■ [interval 2 ■ (x 2 - x 2 ) + 2 • x 2 • interval • dt + dt 2 ]
The coplanarity equation f c is developed as a combination of
equations (3) and (5) where the unknown parameters are the
elements of the exterior orientation of both stereo images
which are:
• State vector of the center point of the first image ( 6
unknowns)
• The coefficients of the first order rotation
polynomials of both images( 12 unknowns)
Thus, the coplanarity equation f c is defined as follows:
fc=UX 0 Jo,Z 0 ,u x ,u y ,u 2 ,
i i ■> IP\2 » $\ l ? &\2 •> ^*111 ^"l2 ’ ®21 ’ ^22 ’ fill ’ ^22 ’ ^*21 ’ ^*22 ) — ®
5.3.3. Coplanarity role in exterior orientation
determination and in DEM extraction process. The
coplanarity equation can be used in combination with the
collinearity equations during the computation of the exterior
orientation providing additional equations and stability. In
more detail in the resection process three equations (two
collinearity equations plus one coplanarity equation) for each
GCP and one coplanarity equation for each tie point can be
applied, improving the accuracy and the stability of the
solution.
Traditionally in perspective center geometry the epipolarity is
used for the production of normalized images. In this paper a
different approach is followed as a first step. The coplanarity
condition is a robust and rigorous equation which can be used
easily and straightforwardly as a geometrical constraint in the
matching procedure of the DEM generation process. In more
detail: just after the correlation process the coplanarity
equation can be applied to all the conjugate points that are
extracted by the correlation in order to see if this equation is
fulfilled, using a threshold related to RMSE of the resection
solution. The points that do not pass this test are blunders or
correlation errors in general.
6. EVALUATION
6.1. Introduction.
6.1.1. RPCs model. The Cartosat data are distributed with the
rational polynomials coefficients. In EPS there is a module
where Cartosat RPCs could be imported and used for the
orientation of the images.
6.1.2. UCL model. The UCL sensor model could be solved
directly using navigation data, without GCPs (Michalis and
Dowman, 2004, 2005). Unfortunately, because in the case of
Cartosat no navigation data is provided the exterior orientation
parameters should be calculated using GCPs. The total number
of exterior orientation parameters of the two Cartosat along
track stereo images is eighteen. The state vector of the center
framelet of the first image represents six of these unknown
parameters, while the corresponding state vector of the second
image is calculated from the previous one by the Kepler
equation. The other twelve unknown parameters are the
rotation angles of the two images; six rotations for each image
which are the coefficients of the polynomials. Thus, at least
five GCPs are needed for the solution, when the collinearity
equations are used. It will be shown that with the simultaneous
use of the coplanarity equation it is possible to reach accurate
solution with four GCPs. Moreover even in the case where
enough GCPs are available for a solution only with the
collinearity equation the coplanarity ensures a more precise
solution.
6.2. Evaluation strategy
As it shown in figure 1 the GCPs are well distributed on the
images. Nine of the total 36 available points are used as GCPs
in the evaluation process of both models, while the rest 25 (two
are not well identified on the images) are used as ICPs
(Independent Check Points). The GCPs location is not
important for the exterior orientation solution (Michalis and