377
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
average flight line 1
->o
average flight line 2
intercept of both flight lines
Figure 5: The displacement of object points above the reference
plane referring to its position in terms of distance to the related
flight line
The dislocation of a point depends on the height of the flight,
the point height (distance from reference plane) and the
distance of the point from the related flight line. The object
height, the only unknown, is estimated roughly and is used to
estimate the maximum displacements in order to optimize the
range to be searched in. Thus, the ambiguities and the
processing time get significantly smaller.
Figure 6: Fusion of the two input images
Figure 7: Median-filtered result of the correlation with template
sizes from 3x3 to 9x9 pixels
In Figure 6 one can clearly recognize the displacement of
building edges in vertical direction. In Figure 7 it can be seen
how the double displayed building edges are corrected to one
line. To close single missing pixels the result image has been
filtered with a median filter of size 3x3. The ambiguities, which
can't be excluded totally and the simplifications made in the
beginning are the reasons for some loss of accuracy resp.
sharpness.
6. CONCLUSIONS
The new approach for the generation of true orthoimages has
been tested with real data. The simple image and recording
geometry, which was earlier assumed in simulated studies, gets
lost if real image data from an airborne pushbroom scanner are
concerned. Several factors lead to variations from the ideal
flight conditions. For this reason some pre-processing of the
raw data is necessary. This prepares the basis for the
determination of corresponding points and the extraction of true
ground information.
A characteristic property of the data is the permanent change of
orientation parameters for any scanned line. Related
assumptions have been defined and corrections have been
introduced. From the study it turned out that the image data was
suitable, because only minor deviations from the ideal
configuration occurred. Thus, the investigations of the
geometry and the implementation of automatic correlation
confirmed generally the practicability of the new approach.
However, more detailed studies and the development of
effective matching procedures are necessary for future deve
lopment. Then it can be expected, that the method offers a
reliable and very productive alternative option to generate true
orthoimages. Particularly, the independence from elevation
models and the possible improvement of the visibility of occlu
ded areas in certain urban regions are significant advantages.