le,
let
nd
3S
th
ne
se
It
se
ce
es
ng
he
ter
Ismael Colomina
take off the airplane will head to M; (3), fly stripwise (4) and go to Ma (5). It will then fly M» (6) and then fly back to A,
possibly after some waiting circles before landing (7).
So far, nothing new. Let us ask ourselves this question: when to switch on and off the trajectory and attitude related
instruments like GPS receivers and inertial units. The T.O.P. approach answer to the question is that GPS and inertial
observations should be collected from take off to landing, as opposed to the current practice of switching the position
and attitude instruments on/off a few minutes before/after the image acquisition process starts/ends. Apparently, this is
inconsistent with the results presented in (Colombo et al., 1999) where GPS integer cycle ambiguities are solved on-the-fly
epoch-by-epoch. This is actually not the case as it is described next when re-reading the mission process.
During phase 1, sensor positioning can be performed at a few cm accuracy level because of a close GPS reference station.
Also, during phase 1, the IMU will be able to align with the help of GPS. After phase 3, the IMU will be calibrated and
some GPS tropospheric modeling process will have been started and stabilized. During phase 4, the imaging phase, the
airplane will fly over M, and, while in there, a few times close to where the second ground reference station is. This
will allow, again, for an independent check/determination of the GPS integer cycle ambiguities. (In phase 4, the actual
imaged area on the ground is restricted to the mapping area of interest since sensor orientation is performed primarily
by INS/DGPS.) Phase 6 is, again, an imaging phase whose "short" trajectory is better dtermined as a subtrajectory of a
longer trajectory, which is accomplished by phases 5 and 7. Note that with the current GPS constellation, the integer cycle
ambiguities, which we assumed to be reliably estimable in phases 1 or 4, are likely to be maintained for some period of
phases 3 and 5. The situation will improve in the future if the GPS modernization initiative prospers (Mc Donald, 1999).
And it will be further improved by the simultaneous use of a modernized GPS system and the future European GALILEO
system.
So far we did not mention the permanent GPS stations. They are likely to be located not exactly in the mapping areas but
they can provided either additional raw GPS observations or derived products if they are related to networks like those
of the IGS (http://igscb.jpl.nasa.gov/) or EUREF (http://homepage.oma.be/euref/). The derived products include precise
ephemerides, satellite clock corrections, coordinates for the stations themselves referred to well defined geodetic reference
frames, and periodic ionospheric models. The above information seems to be enough not to use the ad-hoc GPS reference
stations (Colombo et al., 1999). This is a major step forward. Even for some medium to small scale applications the
straightforward absolute positioning techniques described in (Ovstedal, 1999) are likely to be sufficient.
The INS/DGPS post-processed trajectory transferred to the sensor instrumental reference frame is optimal in the trajectory
sense but not in the sensor orientation sense since correct interpretation requires a well calibrated system. Calibration is a
rather general term and, in the case of traditional photogrammetric cameras, includes the calibration of the fundamental
camera constants —a must— and selfcalibration —an option—. With INS/DGPS the fundamental camera constants are
calibrated and checked periodically. Selfcalibration is largely mission/time dependent and if there is a need for it, then the
INS/DGPS process has to be followed by homologous point matching and a bundle adjustment. This does not necessarily
mean that we are back to digital AAT since, given the image orientation parameters, the number of homologous points
that have to be matched is relatively small. For the same reason, the robustness of the estimation of 10 to 40 additional
selfcalibrating parameters is high. A lightweight AAT package or even manual, traditional measurement of a few points
would compare favorably to the quality-checks and editing of digital AAT.
The above methodology corresponds to the ideas given in (Colomina, 1999): concentrate the efforts in developing methods
to fully exploit the potential of INS/DGPS and then, if needed, complement INS/DGPS technology for orientation with
image processing technology and bundle adjustment for fine calibration; in any case, keep the existing bundle programmes
or a lightweight version for sensor calibration.
6 THE T.O.P. PROCESS
In section 5, a concept for an optimal combination of INS/DGPS and lightweight AAT has been described. Nevertheless,
for a concept to be brought into actual practice a decision making process is needed. Contributions to this are discussed
in the next subsections.
6.1 Define the project quality parameters
It is said that business is about making money out of satisfying customers. The project quality parameters are the weights
to balance the relative importance of technical specifications, time to market and cost of the service. Those three main
parameters are then modulated by market, external factors and by corporate, internal factors. Once this becomes clear,
what the customer wants and how it is to be delivered (progressively, at once, etc.), then the technology options to do the
orientation are likely to show up naturally.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 195