The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
precise trajectories. However, in small UAs and, particularly,
in rotary wing UAs, the vibrations caused by the engines
generate “noisy” inertial observations which are known to
integrate into strong drifts when solving the INS mechanization
equations for navigation. (The same holds for manned
helicopters.) In (Wis et al., 2008), a novel method and
algorithm for real-time denoising of inertial observations and
its results will be described. The method is the natural
extension of the numerical integration methods of Ordinary
Differential Equations (ODE) where the analytical exact
integration of an interpolating polynomial is replaced by the
analytical exact integration of a fitting polynomial. Figure 5
shows preliminary results of the proposed technique for a static
acquisition time interval at the end of a PRS mission where the
blue curve corresponds to the proposed least-squares fitting
technique and the red one to a standard interpolating one.
4.2 DSO, ISO and in between
Once the tPVA task is accomplished, the SCO task must be
performed consistently with the specific cost, time and
technical requirements of the PRS mission and with data that
may be suboptimal with respect to the usual airborne standards.
SCO is usually seen as a method and procedure that can be
performed in either one of two modes, DSO and ISO, and with
absolute control functional models. In our approach, DSO and
ISO are the ends of an interval of methods where the effort of
measuring image coordinates [of tie and ground control points]
can be tuned as a function of the precision, accuracy and
reliability of project specifications (Colomina, 2007). In our
approach, as well, the SCO model can be selected from a
family of spatio-temporal absolute and relative SCO models
according, again, to project specifications (Blazquez, 2008).
This “two dimensional” approach to SCO —with the mode and
the model dimension— can be applied to
Umt <S*c)
Figure 5: Heading determination improvement with an ODE
least-squares numerical integration algorithm [preliminary
results].
any sensor and platform combination, but in the case of UAS-
based PRS is of particular relevance. We illustrate this next
with some ideas on DSO and ISO for UAS-based PRS.
On the foreseen advantages of UAS-based PRS is its flexibility,
particularly in the case of rotary wing UAs since calibration
maneuvers before and/or after the mission can be performed at
a rather low additional burden (time, cost and later
measurement effort). Thus, a DSO based mission can be
preceded or succeeded by a comprehensive and quick
calibration maneuver by acquiring “calibration” image data at
various altitudes and headings. In this way, in the bundle
adjustment with the “calibration” image data, calibration
parameters are not correlated to the image orientation elements
and in turn they are realistically determined. In other words, a
significant number of parameters of a physical-oriented
calibration model are well determined.
The above DSO related arguments hold as well for ISO, where
the standard mission can be complemented with the mentioned
calibration maneuvers in such a way, that in addition to the
physical-oriented self-calibration parameters (like the Conrady-
Brown 5 parameter set or the 8f, 8x 0 , 8y 0 one), numerical-
oriented self-calibration parameters (like the Ebner 12
parameter or Grim 44 parameter sets) can be determined. As a
result, a total calibration concept and model can developed
which consists of pre-calibration and self-calibration steps with
physical-oriented and numerical-oriented functional models
respectively. More specifically, the collinearity model can be
extended with two sets of additional parameters, the physical-
oriented one and the numerical-oriented one.
At the other end of the ISO complexity and in one of the
contexts of UAS-based PRS —that of low cost, fast mapping
and moderate accuray requirements— there are other
possibilities like expediting the bundle adjustment with the
INS/GPS derived aerial control, a small number of ground
control points and just image observations for the ground
control points and just image observations for the ground
control points. Clearly, this procedure will not deliver at the
same level of accurcacy as the usual ISO, but will be more
robust than DSO with respect to reference frame mistakes.
We conclude this section by noting that appropiate modeling
—i.e., features on the SW side— can simplify the HW
complexity, a relevant issue in UAS-based PRS. A nice
example is that of temporal calibration in ISO (Blazquez, 2008).
With this model, if the internal sensor time delays are constant
there is no need to synchronize the navigationorientation
payload to the sensor payload as the mentioned delays can be
estimated in the ISO step.
5. ON THE FEASIBILITY OF COMMERCIAL UAS-
BASED PRS
There are UAS civilian success stories, like the use of UAs in
agriculture in Japan. In this section we are interested in
discussing the feasibility of the commercial use of UAS for
PRS. Note, that we are not addressing the various forms of
remote sensing in its broad sense —embodying biological,
chemical, electromagnetic and gravity sensors. Globally, those
have already been identified as the main future application of
UAS technology. We are rather addressing the professional
mapping markets. We stand on the opinion that while some
challenges have to be faced before the use of UAS in PRS goes
universal, there are many applications that constitute both a
business opportunity today and a platform for maturing the
technology for the future.
5.1 The outstanding challenges
For an UA to be flown on a large commercial scale, three
challenges shall be faced and solved: UAS reliability, UAS
integration in the civilian airspace and UAS social acceptance
and safety reputation.
Reliability. A significant part of current UAS technology has
been developed for use in military applications, where the