POTENTIAL OF IKONOS AND QUICKBIRD IMAGERY FOR ACCURATE 3D POINT 
POSITIONING, ORTHOIMAGE AND DSM GENERATION 
H. Eisenbeiss , E. Baltsavias, M. Pateraki, L. Zhang 
Institute of Geodesy and Photogrammetry, ETH-Hoenggerberg, CH-8093, Zurich, Switzerland - (ehenri, manos, 
maria, zhangli)@geod.baug.ethz.ch 
Thematic Session 20 — Applications of High Resolution Data 
KEY WORDS: IKONOS, QUICKBIRD, high resolution, radiometric preprocessing, sensor models, point positioning, orthoimage, 
DSM, accuracy analysis 
ABSTRACT: 
This paper describes the processing of IKONOS and QUICKBIRD imagery of two different datasets in Switzerland for analyzing 
the geometric accuracy potential of these images for 3D point positioning, and orthoimage and DSM generation. The first dataset 
consists of panchromatic and multispectral IKONOS and QUICKBIRD images covering the region of Geneva. In the second area 
around Thun with a height range of ca. 1650 m, the dataset consisted of a triplet and a stereo pair with an overlap of 50 %. In both 
areas, laser DTM/DSM existed and in Geneva also aerial orthoimages. GCPs with an accuracy of 0.2-0.4 m have been used in both 
sites. The investigations for 3D point positioning included 4 different sensor models, different GCP measurement, variable number 
of control points and area covered by them. The results showed that the Rational Polynomial Coefficient (RPC) model compared to 
2D and 3D affine models are more general and can model sufficiently imaging modes that depart from linearity. This is particular 
so for QUICKBIRD which needs after the use of RPCs an additional affine transformation in order to reach accuracies of Im or 
less. With sufficient modeling, the planimetric accuracy was 0.4 — 0.5 m, even for few GCPs and only partly covering the images. 
Orthoimages were generated from both QUICKBIRD and IKONOS with an accuracy of 0.5-0.8 m, using a laser DTM. A 
sophisticated matching algorithm was employed in Thun. In spite of various difficult conditions like snow, long shadows, 
occlusions due to mountains etc., the achieved accuracy without any manual editing, was 1-5 m depending on the landcover type. 
while in open areas it was about 1 m. Under normal conditions, this accuracy could be pushed down to about 0.5 m. Thus, 
  
IKONOS, and to a lesser degree QUICKBIRD, could be an attractive alternative for DSM generation worldwide. 
1. INTRODUCTION 
1.1 Aims 
The topic of this paper is the analysis of the potential of 
IKONOS and secondary QUICKBIRD (QB) for 3D point 
positioning, orthoimage and DSM generation. Two test sites, 
in Geneva and Thun, were used with accurate reference data 
and partly different aims. In both projects, there was a 
cooperation with the Swiss Federal Office of Topography 
(swisstopo) and Space Imaging (SI). In, Geneva, the final aim 
was the investigation whether high-resolution satellite (HRS) 
imagery can be used for updating the Swiss national maps at 
foreign border areas, which has as prerequisite the generation 
of accurate orthoimages. Another aim was the analysis of 
accuracy of IKONOS and QB for 3D point positioning and 
orthoimage generation using Rational Polynomial Coefficients 
(RPCs) and simpler models. The HRS 
orthoimages will be compared to alternative information 
sources regarding feature interpretation and mapping by the 
swisstopo. In Thun, the main aim was accuracy investigations 
of IKONOS for point positioning and DSM generation using a 
block of images (2 strips with 5 images) over a terrain with 
large height range and very variable landcover. The whole 
processing was performed exclusively with software based on 
good quality algorithms and developed at our Institute, most of 
it part of an operational software package for processing of 
other sensor 
linear array digital imagery. 
1.2 Datasets 
In Geneva, we used two slightly overlapping IKONOS images 
(west and east, each about 10 km x 20 km) and one QB image 
covering the eastern and 60% of the western IKONOS images. 
In Thun, one stereo pair (eastern part) and a triplet (western 
part) of IKONOS images (each image 10 km x 20 km) were 
used, with each image group acquired on the same day (see 
Table 1). The two strips in Thun had a ca. 50% overlap, and 
the triplet images were covered in about 70% of the area by 
snow, while all images had long shadows. The nadir image in 
the triplet was very close to one image of the stereopair, which 
had a suboptimal base/height ratio. All IKONOS images were 
Geo, 11-bit with DRA off, with Im panchromatic (PAN) and 
4m multispectral (MS) channels (in Thun only PAN was used), 
while the QB image was Basic 1B, 11-bit, 0.63m PAN and 
2.52m MS. IKONOS and QB images had associated RPC files. 
For the measurement of GCPs in the Geneva site we used in 
the Canton of Geneva orthoimages with 0.25 m pixel size and 
ca. 0.5 m accuracy, derived from 1 m laser DTM with 0.5 m 
accuracy and outside the Canton, Swissimage orthoimages with 
0.5 m pixel size and 1 m accuracy, derived from a 25m DTM 
(DHM25) with ca. 2 m accuracy. The coordinates of the GCPs 
in Thun were measured with differential GPS. In all cases, 
GCPs were measured in the images semi-automatically using 
least squares and intersection of straight, long enough lines or 
ellipse fit. The control points have an accuracy of 0.2 - 0.4 m in 
object and image space. In Thun, a 2m laser DSM with an 
accuracy of 0.5 m - 1 m (1 sigma) for open areas and 1.5 m for 
vegetation areas was used as reference data for the DSM 
generation from IKONOS. 
1250