The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
Methodology 
Accuracy assessments conducted in this article focus mainly on 
the quality of automated Digital Elevation Model extracted 
from PRISM triplet of images. In order to conduct such 
validation, workflow using Pixel Factory software (F. Bignone, 
2003) has been investigated. 
Figure 3: PRISM Digital Elevation Model validation workflow 
The Pixel Factory software package offers a rigorous push 
broom sensor model for ALOS, without any approximation. It 
means the available sensor model take into account all the 
physics involved in line scanning acquisition. By using external 
geometric information provided in the level 1B1 of PRISM data, 
it is then possible to take most benefit of the satellite a-priori 
spatial localization and to correct images from all geometric 
distortion induced by the push-broom, line-scanning, 
acquisition. Concerning the stereo-matching algorithm used in 
this software package, it is a multi-resolution matching 
algorithm with sub-pixel accuracy. Fusion of height information 
extracted from different views is then realized in order to 
increase accuracy and robustness of all height measurements. 
By combining this matching and fusion algorithms, it is then 
possible to extract automatically dense Digital Elevation Model 
from multiple stereoscopic images with higher accuracy. 
Using previous described reference dataset, statistical analyses 
of PRISM performance for Digital Elevation Model production 
will be conducted on overlapping area. Qualitative analysis of 
PRISM imagery will also be provided on the complete coverage 
of available dataset.