Hans-Gerd Maas
capable of indicating singularity situations, but cannot determine the direction of the singularity with sufficient preci-
sion.
e Reduction of singularities: The problem of the singularities in gable direction as addressed in chapter 2.3 and 4.2 can
be reduced by implementing the knowledge that points on a roof cannot be occluded and thus should not be ex-
cluded, even if they fall into an occlusion zone. Provided that a patch contains a complete roof, this option may be
used to solve the singularity. Nevertheless, the information for the shift in gable direction will be derived from only a
few points at the roof edges, so that the determinability of the parameter remains weak. Moreover, this option will
conflict with the restriction of matching to planar patches as described in chapter 4.1.
e Use of features: As an alternative to the matching based on local point clouds as shown in the previous chapters,
strip discrepancies can also be detected via a comparison of parameters of objects or object parts (such as buildings
or roof planes) modelled independently in both strips. While this option may offer more flexibility concerning short-
comings of the local contrast situation and effects caused by occlusions, it can realistically only be applied to high
density laserscanner data with a density of at least one point per square meter.
e Use of reflectance data: Several airborne laserscanner systems deliver a reflectance signal derived from the intensity
of the backscattered pulse-echo in addition to the actual height measurement. This reflectance value is sometimes
being used in segmentation tasks. As the reflectance value is perfectly referenced with the height data points, it may
also be used for the determination of planimetric shifts between neighbouring strips.
This option may especially be relevant in rather flat areas, where the patch contrast situation in height data only al-
lows the determination of the vertical shift parameter. In such situations, the use of reflectance values may
complement well for the determination :
of the two planimetric shift parameters.
Figure 6 shows an example for the
potential of simultaneous matching in
height and reflectance data, where the
height data allows only for the deter-
mination of a vertical shift parameter,
while the reflectance data allows for
the determination of the two pla-
nimetric shift parameters.
Figure 6: Road crossing in high-resolution
laserscanner data: Height and reflectance
image.
7. Conclusion
Least squares matching applied to laserscanner data in a TIN structure can be a very valuable tool for the analysis and
improvement of the quality of laserscanner data, both in vertical and horizontal direction. Special attention has to be paid
to the avoidance of systematic errors caused especially by occlusions and to a realistic estimation of the precision and
determinability of parameters.
Taking these necessities into consideration, shift parameters between homologous patches in neighbouring or crossing
data strips can be determined with a precision in the order of one centimeter in vertical direction and about one decimeter
in horizontal direction, related to a dataset with an average point spacing of 1.8 meter. This is sufficient to determine strip
discrepancies significantly, so that least squares matching may be used as a tool for the improvement of the geometric
quality of laserscanner data.
Acknowledgement
The author would like to thank Fotonor (Sandefjord, Norway) and Geodelta (Delft, The Netherlands) as well as the
Rijkswaterstaat Survey Department (Delft, The Netherlands) for providing data for the work presented here.
References:
* Behan, A., 2000: On the Matching Accuracy of Rasterised Scanning Laser Altimeter Data. IAPRS, Vol. 33, Part 2
Cramer, M., 1999: Direct geocoding - is aerial triangulation obsolete? Proceedings 47" Photogrammetric Week (Eds.
Fritsch/Spiller), Wichmann Verlag, pp. 59-70
554 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
Ebne
imag
Fôrs
Grüt
Phot
Huis
syste
261
Kili:
Maa
Proc
Maa
airb
Maa
Jour
de N
Stro
van
Dep
Oud
scan
Pos
grar
Las:
Vos
Aut
van
Scie