stanbul 2004 
ogrammetry. 
i. 54 pages. 
undamentals 
'es. 
refinements 
ls (DEM's). 
> Manual of 
immetry and 
7 pages:151- 
refinements 
els (DEM). 
for the 21st 
and Remote 
is maturity. 
w matching 
, &ccuracy, 
V, Istanbul, 
' processing, 
London. 555 
accuracy of 
ructed from 
14(82): 661- 
accuracy of 
)bservations. 
in accuracy. 
ACCURACY OF DEM GENARATION FROM TERRA-ASTER STEREO DATA 
A. Cuartero *, A.M. Felicísimo *, F.J. Ariza" 
* Dept. of Graphic Expression, Extremadura University, 10071 Cáceres , Spain. acuartero@unex.es amfeli@unex.es 
b > pn : . ; : re or 
Dept. of Cartography, Geodesy and Photogrammetry Engineering, University of Jaén, 23071, Jaén. Spain 
fariza(@ujaen.es 
Commission VI, WG VI/4 
KEY WORDS: Accuracy, DEM, stereoscopic images, satellite, softcopy. 
ABSTRACT: 
In this work we studied the accuracy of DEMs generated from ASTER stereoscopic images by automated stereo-matching 
techniques with two different softwares (OrthoBase PRO and OrthoEngine). We compare several DEMs generated for a test area of 
23 km x 28 km situated in the province of Granada (south Spain). This is an area was selected because its variable and complex 
topography with elevations ranging from 300m up to 2800 m. 
The method is based on the photogrammetric principle of collinearity in orienting the images, for which purpose 15 control 
points were used. The accuracy was studied using 315 ground check points whose coordinates were determined by differential GPS. 
Results indicate that elevation root-mean-square error (RMSE) equals 13 m, which is less than the pixel size (15 m). We think that it 
is satisfactory for many cartographic and analytical applications comparable to that of conventional topographical maps. It is 
particularly important to have abundant and accurate check control points available since this determines the reliability of the quality 
control itself. Finally, we compare accuracies between both TERRA-ASTER DEMs and DEMs elaborated from conventional 
1:50.000 topographical maps of same area. 
1. INTRODUCTION 
1.1 Instruction 
Photogrammetric techniques have been known for decades, but 
the possibility of using satellite's stereoscopic images for global 
digital clevation data producing did not arise until the launch of 
the first of the SPOT series satellites in 1986. The quality of 
digital elevation models (DEMs) elaborated from stereoscopic 
pairs is affected by the topography of the terrain and the data 
source (aerial photograms, digital satellite images), as well as 
other variables that depend on the data (aerial or spacial), on the 
algorithms used in the photogrammetric workstations, and on 
the data structure (triangulated irregular networks versus 
uniform regular grids). 
A digital elevation model (DEM) can be extracted automatically 
from stereo satellite images. Numerous applications are based 
on DEM, and their validity directly depends on the quality of 
the original elevation data. High quality DEM are seldom 
available, even though photogrammetric technology, the most 
common to work with DEM has been around for a few years. 
Dependence on analogue aerial images ended formally in 1980, 
when the American Society of Photogrammetry and Remote 
Sensing (ASPRS) included the possibility of using digital data 
from remote sensing in its definition of photogrammetry 
(Slama, 1980). 
Digital photogrammetric techniques have been known for 
decades, but the possibility of using stereoscopic images from 
satellites for global digital elevation data production did not 
arise until the launch of the SPOT series in 1986. Today several 
satellites also offer the possibility for stereoscopic acquisition: 
SPOT (Priebbenow & Clerici, 1988), MOMS (Lanzl et. al., 
1995), IRS, KOMSAT, AVNIR (Hashimoto, 2000), TERRA 
(Welch et al., 1998) and more recently, the high resolution 
pushbroom scanners IKONOS (September 1999), EROS-AI 
(December 2000) QUICKBIRD-2 (October 2001), SPOT 5 
559 
(May 2002), and ORBVIEW-3 (June 2003). Thus, some studies 
focus on constructing DEM from stereoscopic images by means 
of high resolution pushbroom scanners, IKONOS (Li et al., 
2000, Toutin, 2001), EROS A1 (Chen & Teo, 2001.), SPOT 5 
(Petrie, 2001); furthermore, it is assumed that the automatic 
generation of a DEM from remotely sensed data with a Z sub- 
pixel accuracy is possible (Krzystek, 1995). 
The accuracy of DEM elaborated from aerial stereoscopic pairs 
has been exhaustively analyzed but not all knowledge can be 
accepted in the spatial images case without a detailed analysis. 
Several factors distinguish both cases, e.g. the image spatial 
resolution, and the timing and geometric design of acquisition. 
These factors cause some common problems when using 
stereoscopic spatial images, e.g., the difficulty of identifying the 
Ground Control Points (GCP), or the existence of radiometric 
differences among the images due to acquisition at different 
dates that may make the stereo-matching process more difficult 
(Baltsavias & Stallmann, 1993). Nonetheless, it is clear that 
advantages such as wide coverage and good temporal 
resolution, give support to the general use of this data source. 
Automation allows the construction of DEM with an almost 
randomly large point density. The selection of “very important 
points”, common in manual processing, is not applicable to 
automatic photogrammetric processes. The result often entails a 
very ‘hard’ DEM where a lot of redundant or unrelevant 
information can be removed. In literature review we could find 
no references to possible optimization strategies for this phase 
of the process. 
Accuracy estimation can be carried out by comparing the DEM 
data with a set of check points measured by high precision 
methods. The basic conditions for a correct work flow are: a) 
high accuracy of check points, and b) enough points to 
guarantee error control reliability. 
We have examined that most research does not satisfy those 
conditions. The common source of check points is topographic