{
|
|
|-
ll:
!
1A
i
i
^
C. Vincent Tao
PCI software supports several sensor models including the rational function model. The results computed using the PCI
RFM module are also given in Table 2. Both the root mean square (RMS) of the residuals and the maximum errors at
control points and checkpoints in image are listed. Based on the analysis of these results, the following observations are
obtained:
= Both the rational function models and the polynomial model (i.e., p2=p4=1 case) can reach the accuracy around 1
pixel.
= In general, the cases with rational components are slightly better than the polynomial models.
= The fitting accuracy of the cases with four-degree is very close to those with six-degree. It is understandable since
the aerial photogrammetry data set may not involve high order distortions.
= The accuracy of the direct solution method is comparable with that of the iterative solution method when the
number of control points (i.e., 50 or more) are enough. The iterative solution is often slightly better than the direct
solution when only a few more control points than the minimum number of control points required are available.
5 CONCLUSIONS
An overview of sensor models are given in this paper. These models are the collinearity equations based differential
rectification model, the polynomial model, the projective transform model, the extended direct linear transform model
and the rational function model. The remarks on their properties, and advantages and disadvantages are also discussed.
It can be observed that the RFM is essentially a generic form of the all described models.
Unlike the physical sensor models, the rational function model needs no knowledge specific to each type of imaging
sensor physics, such as orbit parameters and platform orientation parameters. It has been noticed that rigorous models
are not always available, especially for images from commercial stateliness (e.g., IKONOS), where the rigorous sensor
model is hidden to the end user. Therefore, the RFM becomes an alternative sensor model to image rectification.
For aerial optical images (i.e., frame sensor type), the test results show that both the rational function model and the
polynomial model can reach reasonably good accuracy. Because the fitting accuracy of the cases with four-degree REM
is almost the same with those with six-degree RFM, high order forms are often not necessary. The iterative solution
method to RFM provides a better accuracy than the direct solution method, but the direct solution method is usually
adequate when enough control points are available.
REFERNENCES
Chen, L.C., Lee, L.H., 1993. Rigorous generation of digital orthophotos from SPOT images. Photogrammetric
Engineering and Remote Sensing 59(5): 655-61.
Greve, C. W., 1992. Image processing on open systems, Photogrammetric Engineering and Remote Sensing, 58(1): 85-
89.
Mercer, B., Schnick, S, 1999.Comparison of DEMs from SATR-3i Interferometric SAR and Scanning Laser, ISPRS
Commission III, Workshop, La Jolla, CA, November 9-11, 1999.
Novak, K., 1992. Rectification of digital imagery. Photogrammetric Engineering and Remote Sensing, 58(3), pp. 339-
344.
OpenGIS Consortium, 1999. The OpenGIS Abstract Specification - Topic 7: The Earth Imagery Case,
http://www.opengis.org/public/abstract/99- 1 07.pdf.
Okamoto, A., Ono, T., Akamatsu, S., Fraser, C., Hattori, S., Hasegawa, H., 1999, Geometric characteristics of
alternative triangulation models for satellite imagery. Proceedings of 1999 ASPRS Annual Conference, From Image to
Information, Oregon, May 17-21.
Palà, V., Pans, X., 1995. Incorporation of relief in polynomial-based geometric corrections. PE&RS, 61(7), pp. 935-44.
Tao, C.V. and Hu, Y., 2000. Investigation of the Rational Function Model, Proceedings of ASPRS Annual Conversion,
Washington D.C, May 22-26, 2000
880 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.