196 
Figure 2: GCP identification - Homologous points on the orthophoto (left) and the amplitude image (right). 
Figure 3: GCP identification - Homologous points on the orthophoto (left) and the coherence image (right). 
The identification of natural GCPs on amplitude SAR 
images is very demanding because such images are 
considerably more difficult to interpret than optical images. 
Their resolution in azimuth is greater than in slant range. 
The images are highly distorted in the slant range 
direction (e.g. foreshortening, layover) and that makes the 
GCP recognition very difficult. Furthermore, speckle 
handicaps the image interpretation very much. To bypass 
this last problem, the images are usually azimuth- 
compressed. The compression reduces the noise and 
improves the image interpretation. 
In many cases it is difficult to measure a sufficient set of 
3D GCPs. On the other hand, it is easier to recover height 
GCPs, i.e. points whose height is known accurately but 
which are not very well defined in planimetry (e.g. points 
chosen in the centre of flat fields or flat homogeneous 
areas). Our calibration accepts either full or height GCPs: 
the height GCPs are introduced in the same way as the 
full ones. However, their contribution to the LS adjustment 
is differentiated through the stochastic model, i.e. they are 
distinguished from the full GCPs through their variance- 
covariance matrix. 
5. ANALYSIS OF THE RESULTS 
In the last three years we were involved in a EU 
Concerted-Action called ORFEAS (Optical-Radar sensor 
Fusion for Environmental Applications), joining five 
European research groups (University of Thessaloniki. 
Cartographic Institute of Catalonia. ETH Zurich, Technical 
University of Graz and Polytechnic of Milan). A 
comprehensive data set. covering South Catalonia - Spain