ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
appropriate for the methodology proposed. In order to simulate
an accurate SAR image orientation, shifts in range and time
were introduced in the object-to-image projection.
4. INTEGRATION OF SAR AND SPOT MODELS
4.1 SAR-SPOT intersection
Both SAR and SPOT images are significantly affected by the
terrain height. Relief displacements (AX) of a height ^ depend
on the image incidence angle (o) on the imaged point and are
expressed in the following ways:
AX, =hcotay,, (6)
AX spor =htan og;
A pair composed of a SAR and a SPOT image provides a
parallax effect, from which heights can be determined. Three-
dimensional coordinates of SAR-SPOT conjugate points can be
calculated using the sensor equations. In geometric terms this
corresponds to determining the intersection of a straight line
and a circle in space. Alternatively, heights can be calculated
from parallaxes (Gongalves, 2001).
SAR-SPOT stereopairs can be acquired with both satellites on
the same side relative to the point or on opposite sides. Figure 6
represents these two situations, where a point with height H
above a reference datum is being observed. In the first case the
two displacements, with respect to the reference datum, occur in
opposite directions, resulting in an additive parallax. In the
second case the parallax effect is subtractive, corresponding in
principle to a less favourable situation for height determination.
However, as the main contribution to the parallax comes from
the SAR displacement, the accuracy is not significantly affected
(Toutin, 2000).
Spot Spot
SAR®,
/\H
Pr PS PR : Ps
(a) (b)
Figure 6 — Same side (a) and opposite side (b) configurations in
a SAR-SPOT stereopair.
The SPOT and Radarsat image used in this work form a same-
side stereopair.
4.2 Image registration
As the SPOT and the SAR images have different resolutions,
registration is needed in order to measure parallaxes. In this
work it was decided to register the SAR image to the SPOT
image space using a set of points generated with the image
orientation data. À point can be projected from the SPOT image
onto the ellipsoid (equation 3) using the approximate
orientation parameters and from there onto the SAR image
(equation 3, with SAR model). This was done for a set of points
on the SPOT image (grid of 7 by 7 points). Figure 7 shows the
points in the SPOT and the SAR image.
A - 128
Figure 7 — Location, on both image spaces, of the points used
for the registration
A 3" degree polynomial was then fitted to these points, with
residuals smaller than 0.1 pixels in absolute value. This
polynomial function (later designated as R) was used in the
registration of the SAR image to the SPOT image space. Figure
8 shows the SPOT image (a) and the SAR image registered to
the SPOT image space.
(a) per
Figure 8 — Full SPOT image (a) and registered SAR image (b)
4.3 SAR-SPOT tie-point measurement
Tie points had to be measured in the SAR-SPOT pair.
Identification of well-defined individual points in the SAR
image is extremely difficult, especially because the area does
not have large cities or roads. However many polygonal features
such as water bodies can be clearly identified on both images.
The manual method used to generate tie-points from these
features was the following: first the boundary is digitised, in
vector format, on the SPOT image. Then, on the SAR image
layer, the line is manually moved until matching the
corresponding feature on the SAR image. Figure 8 represents
the boundaries of a small lake and the vector line digitised on
the SPOT image (a) and then dragged on the SAR image (b).
Es M
Figure 9 — Determination of tie-points using area features.
The centroid of the polygon, or even any vertex of the line, can
be used as a tie-point. Figure 10 represents the location of the
tie-points on the SPOT image. Most of them were derived from
lakes, reservoirs or river margins. If original SAR image
coordinates are required, the registration function, R, should be
used.
