ACCURACY OF DTM GENERATED FROM SPOT 4 AND SPOT 5 HRS DATA 
Inter! 
— 
  
CC 
I. Ewiak , R. Kaczynski 2.1 | 
Department of Photogrammetry, Institute of Geodesy and Cartography, In or 
ul. Modzelewskiego 27, 02-679 Warsaw, Poland — (rene, rom)@igik.edu.pl orien 
ofm 
Commission IV, WG IV/6 follo 
KEY WORDS: photogrammetry, DTM, SPOT, orientation, rectification 
ABSTRACT: 
SPOT 4 stereo data from two different orbits with B/H 0.83 of the area in South-East Poland with denivelations reaching 550 m was : 
elaborated in IGiK. 12 GCP’s and 15 ICP’s taken from 1: 10 000 topographic maps and orbital data have been used for orientation The 
with the use of Image Station Multisensor Triangulation SPOT. The accuracy achieved on GCP’s: RMS X = 2.2 m, RMS 
Y = 2.3 m, RMS Z = 1.1 m; and on independent check points (ICP’s) RMS X = 5.0 m, RMS Y = 2.9 m, RMS Z = 4.7 m. DTM has phot 
been generated automatically with the use of MATCH-T Inpho software and filtered with in-house software. DTM prepared on the semi 
basis of SPOT stereo data has been checked against reference data - DTM generated from aerial photographs in scale 1:26 000. detai 
The accuracy of DTM generated with SPOT 5 HRS within the HRS Scientific Assessment program (HRS-SAP) of ISPRS and CNES these 
on site 9 Chiemsee (Germany) is also presented. Geomatica OrthoEngine of PCI has been used for orientation and DTM generation. orier 
conti 
SCen 
1. TEST DATA The 
; 1. HRS SPOT 5 their 
1. HRV SPOT 4 
ident 
The two SPOT-4 panchromatic images level 1A over The following data has been used: The 
South-East Poland were acquired on 5 September and 12 - 8 bit SPOT 5 HRS image data - four scenes level 1A - (size ue p 
1987 and recorded by Spot Image France with significant 12000 x 12000 pixels = 120 km x 60 km) of the Bavarian test TRII 
radiometric differences between the two images, area from two viewing directions in TIFF format with metadata unde 
particularly in agricultural areas. The inclination of the in DIMAP format. The image data has been acquired on (Ima 
sensors optical axis was 24.8? R and 18.5? L respectively. October 1^ 2002 with a Sun elevation of 38 degrees and nearly The 
i ; no clouds, which promises sufficient image matching the 
The base-height (B/H) ratio of the model was ae ants ; bg ii: : 
t s e possibilities. HRS sensor on SPOT 5 produces image stereo adju: 
approximately 0.85, which pt omised good resection. The pairs with two optical sensors looking forward and backward adju: 
elevation range of the terrain was from 50 m - 550 m. — with £20 degrees with respect to nadir. It has a spatial D 
Modern topographic maps in the scale 1:10 000 cover the resolution of 10 meters across track and 5m ground sampling E 
test area. From these maps, 43 well-defined control points distance along track for obtaining higher accuracy of the oie: 
and 15 check points have been captured using a digitizing — parallaxes for the DTM generation. The swath of the HRS is stere 
table with accuracy better than 3 meters in the X, Y and 120 km (12000 CCD elements) and one acquisition sequence is meas 
under | meter in Z. Four-times enlargements were used, 
making it possible to mark them within 1/4 of a pixel. 
Two test areas have been selected within borders of 
stereoscopic SPOT 4 scene. Two sequential aerial 
photograph models at a scale of 1:26 000 marked range 
of every of these areas. For first test area representing 
mountainous and hilly terrain 6 control points were 
measured in field by GPS methods. Aerotriangulation 
was carried out in order to calculate ground coordinates 
for additional 15 control points with standard deviations 
6x=0.3m, 6y=0.3m, 6,=0.2m. These points, along with 
GPS points have been used as full set of control points 
for the first test area. For second test area representing 
hilly and flat terrain, 8 control points were measured in 
field by GPS methods. Similarly as for the first test area, 
21 control points were estimated with statistical accuracy 
Ox-0.2m, oy-0.1m, 07-0.2m. This set of points was 
used as full set of control points for second test area. 
600 km along track; 
- XML-files containing all additional information regarding 
time synchronization, position (DORIS), attitude (star sensors) 
and inner orientation; 
- ASCII text files containing information on the delivered data; 
- 81 ground control points (fixed points), with detailed 
description of the exact location of each point, with 
planimetric and height accuracy in terms of the RMSE ' 
values for each of the GCP's were better than + Im; 
- Interactive topographic map “Bayern 3D” in scale 1: 25 000 
for locating ground control points in the HRS data. 
This data has been delivered by CNES for realization of 
the program HRS-SAP.