1187
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
is completed with field data to obtain cartography of urban
areas, the digitization of more objects during the stereoplotting
could reduce field data capture. Unfortunately no analysis has
yet been done to know in detail the impact of this reduction,
although that the estimated percentage will not achieve in any
case more than 4 or 5%. In small scale flights it has not been
possible to extend the flight period because the requirements in
radiometry for orthophoto products require flights taken from
May to September.
The second advantage is that usually the sharpness of the DMC
images is higher than the obtained in the analogical ones at any
scale, so the vector digitization is more comfortable for the
operators.
The main disadvantage is a slight loss of the relief feeling. In
large scale digital flights, with 7.5cm GSD, it is quite difficult
to measure the height in objects lower than 20cm, like sidewalk
curbstones or the rails of railways. These short measurements in
height were possible using images taken by analogical cameras.
This slight imprecision in the component H introduces some
insecurity and uncomfortable feeling during the stereoplotting
process.
Another disadvantage is the big differences in contrast between
dark and bright areas. If the photogrammetric system doesn’t
include some tools to modify more or less automatically the
histogram during the stereoplotting process, the digitization
process is harder than in the case of using analogical images
where the contrast was not so high.
The last disadvantage is related to the reduction of the area
covered by the digital stereopairs. It implies to manage twice as
much files than in the case of analogical stereopairs and
represents an extra cost during the stereoplotting process if the
system hasn’t tools to change automatically to the adjacent
stereopair.
6. IMAGE RESOLUTION
Traditionally, the resolving power of an optical system was
measured by the visual identification of targets, amongst which
the USAF 1951 test stood out due to the fact that it has been
used more than others. With the same purpose electronic
methods have been developed to assess the performance of a
system, based on profiles of sinusoidal intensity and supported
by the Modulation Transfer Function (MTF) and the Edge
Spread Function (ESF). Our methodology is based on the latter.
A software tool is implemented with the aim of providing a
resolving measurement of an image in pixel magnitude. The
program processes a region of interest that contains a single
contour and makes a minimum square adjustment over the bi-
dimensional function of the edge that we model as a sigmoid
function:
f(j’i) P4 J -P 3 (/-cos P 5 +ysin P 5 -P 2 ) W
P k e%k = 1,...,5.
The five parameters are estimated in a least squares adjustment
and the computed function is derived for obtaining the LSF
(Line Spread Function). We consider the FWHM (Full Width at
Half Maximum) over the LSF to be the measurement value.
6.1 Results obtained with aerial images (DMC) and
Siemens star targets
The algorithm has been applied on a Siemens star target painted
on a canvas surface of 100m 2 that has been captured in four
DMC aerial images. The pattern used for this study can be seen
on the top-right part of Figure 5. On each of the four images
used for the test, the pattern was located at a different distance
from the centre. The average resolution obtained for each image
is shown in table 9.
Figure 5: Siemens star target used for the resolution study
Image
Distance
to the
centre
(pixel)
Resolution
(pixel)
Image 1
1772
0.83
Image 2
1519
0.79
Image 3
4241
1.07
Image 4
4543
1.32
Table 9: Resolution measurements as function of the distance
from the image centre
These results show that the distance to the centre of the image
greatly affects the resolution. The main reason that explains this
behaviour is the formation of the large format image from
oblique component images. The 12pm size pixel gives us a
nominal resolution of 84 1/mm. But, in fact, the resolution is not
constant in the whole area of the large format virtual image.
The sensor is tilted along the two axis x and y. This factor has
been studied by Honkavaara et al. 2006 who obtained a
resolution of 53 and 84 1/mm perpendicular to the direction of
flight and around 60 and 84 1/mm in the direction of flight.
6.2 Results obtained with aerial images (DMC) on urban
settings.
Finally, the aim is to apply concepts and techniques developed
to images captured with DMC that do not have, necessarily,
artificial models. Due to their radiometric and morphological
characteristics (close to the USAF model) we chose to apply the
software on images of urban settings. Adequate radiometric
contours are considered to be road markings and, more
precisely, pedestrian crossings, see Figure 6.