The International Archives of the Photoerammetrv. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
In 2007, two new methods have been presented (Habib et al„
2007) to safe limitations and requisites of high cost of
methodologies based in Z-Buffer existent algorithms, based in
the analysis of angles along radial directions from nadiral point
to detect occlusions:
• Radial circular sweep method. Determines a map with
occlusions doing a radial sweep of the DSM increasing
the angular value of visibility a for each azimuth angle 0
(Figure 1).
Perspective center
/i \
► o, *» ao
Figure 1. Radial circular sweep method (Habib et al., 2007)
• Spiral sweep method. Introduces the variant of the spiral
sweep of DSM starting in the nadiral point checking
directly angles of visibility in radial direction.
These experimentally tested methods, present some problems
for the generation of individual orthophoto that require the
fusion of common areas to complete zones detected as
invisibles, and with no mosaiking, for which no continuous
coverts of true orthophotos are generated, having to introduce
seamlines needed to mosaic orthophotos from the
neighbourhood in a posterior stage.
3. DEVELOPMENT OF THE ALTAIS-LRTO SYSTEM
The Altais LRTO system was raised by modules, starting with
the development of a basic orthorectification method, which
could integrate progressively algorithms that resolve the
problems that actual orthophotographs have, until the most
rigorous solution was reached of True Orthophoto.
3.1 Ground Ortho Module: GO+.
Orthorectification Module to ground level based in a simple
mathematic model of collinearity equations and additional
parameters, which uses calibration data from the camera, and
inner and exterior orientation data from imagery.
The raised solution is rigorous through an orthorectification
method pixel to pixel from the most nadiral image, different
from the differential methods used in most of the existent
programs that were simplified solutions adapted to computer
technology of that moment.
3.2 Right Ground Ortho Module: RGO+.
It is an orthorectification module to ground level, which
includes an algorithm that searches imagery depending on the
incidence angle of the perspective ray. This module supposes
the introduction of a new method of Orthorectification and
Mosaiking based in an algorithm that searches optimal
photographs, not just the minimal distance with respect to the
nadiral point considered (known as MOST NADIRAL), but
also the optimal incidence angle to avoid presence of
“stretching areas” within the final digital orthoimage, method
that is called MOST RIGHT.
Figure 2. Most Right Method
Usually flights are planned to a scale or resolution that
differences of local scale of photography are absorbed. For this
height differences within the terrain are taken into account, so
that the final scale allows uniform GSD greater than the input
GSD, within an admissibility percentage.
However, the slope of the terrain is not taken into account in
these calculations, which influences directly in the scale and
therefore in the resolution of the imagery.
In figure 2, it can be seen how the solution Most Nadiral offers
less resolution for segment R 2 from the image with centre of
projection in O 1 that from the image with centre of projection in
O.
As the incidence angle i over segment R 2 is greater than angle
i 1 , we have r‘ 2 < r 2 and therefore:
E Vr =-£-> —= E Vr .
Vr 2 R' 2 R' 2 Vr 2
where E Vr i^ t ^ ie flight local scale to which the segment R,
is represented in photograph i.
If the sensor resolution is “s”, and “g” and “g 1 “ are the ground
resolutions respectively obtained from the imagery with centres
in O and O 1 , we have:
s s 1
= g
'Vr 2
This is why a priori the improvement of resolution is clear.
If g° is the nominal output resolution, it will exist a limited
incidence angle i°, which is a function of the aperture angle of
perspective ray and the slope of terrain, just as g'> g°, from
which there will be a stretch in the imagery for view angles
smaller, i.e. where the next relation is not true:
GSD input / GSD output <l.
The optimal incidence angle is obviously the one that is closest
to 90°, however it will have to be valuated the advantages and
disadvantages of using this algorithm without restrictions.