Ismael Colomina
components of orientation are required at high accuracy levels. Any global solution, be it AT/DGPS or INS/DGPS based
direct sensor orientation, will, in general, not resolve the short wavelength components at the accuracy degree desired
for they are “global optimizers.” These are obvious facts. Nevertheless, many times, they are overlooked when trying to
stereoscopically interpret images oriented by INS/DGPS.
3 ON DIRECT VS. INDIRECT
Direct versus indirect, or INS/DGPS versus AAT/DGPS, is one of the current holy wars in photogrammetry. This section
is not one of its battles. For the reader interested in comparative analysis there are a number of papers dealing with
the topic (Colomina, 1999, Cramer, 1999, Skaloud, 1999a, Skaloud, 1999b). Inertial trajectory determination is a well
established, known procedure (Savage, 1998a, Savage, 1998b) whose integration with DGPS has already been transferred
to commercially successful systems for sensor orientation (Scherzinger, 1997). However, there is still some confusion on
the global performance of the concept, mainly because the global INS/DGPS orientation picture is not well understood.
One of the “troublesome” behaviours of INS/DGPS based orientation is the stereo model residual parallax issue, which is
addressed next.
Whatever the choice of orientation technology is and whatever the orientation convictions are, there are some generally
accepted pros and cons of the available technologies. INS/DGPS has yet to solve the calibration family of problems. This
includes the fact that image selfcalibration is lost. AAT —rather AAAT (Assisted AAT) in the author’s opinion (Colomina,
1999)— depends on image texture and slaves block geometry to the orientation task.
The criticism that INS/DGSP does not guarantee parallax-free stereo model setting is just not fair, for aerial triangulation
suffers from the same drawback. The only difference being, that not too many were curious enough or had the SW
to set-up a stereo model directly from aerial triangulation. Since stereo model set-up was —is— mainly based on the
coordinates of ground point determined in the aerial triangulation, the second orientation step is always performed. Any
residual parallax introduced by the global solution is just not seen.
After the discussion of the above section, it is apparent that the solution to the INS/DGPS orientation parallax problem is
as simple as it is for aerial triangulation (see figure 1): identify homologous points on both images, then compute a local
bundle adjustment optimization with two sets of observations, their corresponding weights and the associated models. The
first set contains the six plus six orientation parameters based on the INS/DGPS direct solution. The model is the obvious
one. The second set contains the image coordinates of the homologous points, their weights and either coplanarity
or colinearity observation equations. In other words the second step is a standard relative orientation determination
constrained by observations of the exterior orientation elements. If the exterior orientation elements’ observations are
correctly weighted, the final solution will be parallax-free and still acceptable as a global one.
4 ON SATELLITE KINEMATIC POSITIONING
It is a misunderstanding that the limiting factor in precise direct sensor orientation, specifically in metric aerial camera
orientation, is inertial technology. Contrary to that, the limiting factors are long range satellite kinematic positioning and
the stability of the mechanical assembly between the IMU (Inertial Measurement Unit) and the camera. This is known. In
this section we will comment on two ongoing developments which will have a major impact in the way trajectory deter-
mination is performed: the European Galileo system and GPS modernization; and the progress in ionospheric modeling.
In spite of the US Congress’ rejection of the $ US17 million budget request for GPS civil modernization —on both the
Block IIR and Block IIF satellites—, it seems that GPS modernization will take place whatsoever, even if the US DoD
(Department of Defense) has to become the modernization funding agency. The benefits of GPS modernization have been
summarized by K.D. Mc Donald in (Mc Donald, 1999). It includes C/A code on L2, and a third civil frequency L5 with
precision F code. With modernized GPS, stand-alone L1/L2 C/A code positioning will be at the 5 m level and precision
stand-alone L1/L2 C/A and L5 F code positioning at the 0.5 level. For geodetic applications, the third civil frequency
will allow for 5 cm RTK (Real Time Kinematic) positioning and for 5 mm after post-processing. Galileo will provide, as
well, three signals to the civil community with an even better signal-to-noise ratio. A comparative analysis of three carrier
phase positioning for Galileo and modernized GPS and the impact on integer ambiguity resolution is made in (Joosten et
al., 1990).
While we await modernized GPS and Galileo, new approaches in GPS ionospheric tomography have already been used
for cm level trajectory determination (Colombo et al., 1999) with ground reference stations located at hundreds of km.
Although less accurate (0.40 m horizontally, 1.03 m vertically, at the 1-sigma level) absolute —non-differential— C/A
code positioning (Ovstedal, 1999) is a relatively straightforward procedure likely to become of real practical interest in
192 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.