Institute of Photogrammetry and GeoInformation

TUTORIAL Information extraction, with emphasis on DSM generation, from high resolution optical satellite sensors

Karsten Jacobsen1, Emmanuel Baltsavias2, Nicolas Paparoditis3, Peter Reinartz4

1 University 2 Institute

3 Institut

of Hannover, Nienburger Strasse 1, D-30167 Hannover, Germany, [email protected]

of Geodesy and Photogrammetry, ETH Zurich, Wolfgang Pauli Str. 15, CH-8093 Zurich, Switzerland, [email protected] Géographique National, 4 avenue Pasteur, 94165 Saint-Mande, France, [email protected]

4 DLR

(German Aerospace Centre), Institut für Methodik der Fernerkundung, Bildwissenschaften, D-82234 Oberpfaffenhofen, Post Wessling, Germany, [email protected]

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

TUTORIAL Information extraction, with emphasis on DSM generation, from high resolution optical satellite sensors Section 8 SRTM height model, Conclusion, outlook Karsten Jacobsen University of Hannover, Nienburger Strasse 1, D-30167 Hannover, Germany, [email protected]

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Shuttle Radar Topographic Mission (SRTM) Interferometric Synthetic Aperture Radar (INSAR) C-band: USA X-band: Germany + Italy SRTM mission: February 2000, 11 days computations finished 2003

ΔD

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

SRTM C-band + X-band DEM from Shuttle Radar Topography Mission (SRTM) Cband InSAR DEMs covering 99.96% of land area from 56° south – 60.25° north available 0.15% gaps, mainly in steep areas, water surface + dry sand desert outside USA spacing 3” (~90m), in USA 1” (~30m) X-band data can be bought from German DLR with spacing 1“, but with large gaps between covered strips, width of strip 45km (C-band 225km) red: covered by X-band

SRTM Cband DEMs Arizona 10km x 10km spacing 1“

spacing 3”

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

SRTM height model Paris 1° x 1°

dark spots = voids Æ

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Voids in C-band DEM Mount Everest

grey = no data -problems with InSAR in steep parts globally 0.15% voids

N20W010

sand dunes in desert poor return signal in dry sand desert distribution of voids

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

No penetration of vegetation Æ DSM

Z

cross section through SRTM-X-SAR-profile

250

forest

200

systematic error of ~3m Accuracy in open areas after absolute orientation: ~ +/- 3m

150

100

0

1km

2 km

3 km

Influence of vegetation – X-band and C-band cannot penetrating the vegetation ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Radar – penetration of forest X-band 2.4 – 3,75cm reflection close to top

C-band 3.75 – 7.5 cm reflection close to top

P-band 77 – 136 cm close to surface

dense conifers ~ 6m above ground young trees ~ surface clear cut – penetration of mud ~0.5m

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

SRTM C-band and X-band DSM Zonguldak

SRTM C-band DEM

SRTM X-band DEM

coverage by X-band DEM

16 14 12 10 8 6 4 2 0 .00

.05

.10

.15

.20 .25

.30

.35

.40 .45

.50

.55

.60

.65

RMSZ X-band DSM open area as F(slope)

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Height error map SRTM X-band 95m

Black Sea

3m

7m

0m predicted poor Z-accuracy in locations with lay-over (steep mountain parts and partially vertical shore line) + in Black Sea Area Zonguldak, Turkey ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Filtering SRTM C-band DSM Æ DEM

Bavaria

forest not filtered open area filtered

open area not filtered

forest filtered

frequency distribution Z-differences SRTM C-band height model – LIDAR DEM

RMSZ

bias

RMSZ F(slope)

open, not filtered

5.4m

-2.3m

4.4 + 2.5∗tan α

forest, not filtered

16.5m

-13.8m

15.0 + 3.8∗tan α

open, filtered

4.0m

-2.1m

3.4 + 1.9∗tan α

forest, filtered

11.8m

-9.1m

10.8 + 3.5∗tan α

Influence of filtering SRTM C-band height model with program RASCOR

DEM

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

SRTM C-band RMSZ [m]

Bias [m]

RMSZ F(slope)

for flat part without bias

Arizona

3.9

1.3

2.9 + 22.5 * tan α

2.6 m

Williamsburg NJ

4.7

-3.2

4.7 + 2.4 * tan α

3.4 m

Atlantic City

4.7

-3.6

4.9 + 7.6 * tan α

3.3 m

Bavaria, rolling

4.6

-1.1

2.7 + 8.8 * tan α

2.5 m

Bavaria mountainous Zonguldak (mountainous)

8.0 7.4

-2.4 -4.4

4.4 + 33.4 * tan α 5.9 + 6.5 * tan α

3.7 m 3.9m

RMS average

3.3 m

Accuracy achieved with accurate reference DEMs in open areas Systematic error (bias) caused by SRTM orientation

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Morphologic quality of DEMs – contour lines contour interval 20m

SRTM C-band 70/92m

SRTM X-band 23/31m

map 1 : 25 000

Loss of morphologic details in mountainous area in case of 3 arcsec spacing ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Conclusion to SRTM DEM SRTM C-band and X-band as well as height models from matched space images = digital surface models – on top of vegetation and buildings filtering required, accuracy better in open areas, accuracy dependent terrain inclination – accuracy cannot be expressed just with one figure SRTM C-band DEM useful for several applications, accuracy of available points in range of +/- 4m for open and flat areas spacing of 3” Æ loss of details especially in mountainous areas Even with quite less accurate DEM more morphologic details possible STRM X-band DEM ~ same accuracy of available height points like for SRTM C-band, advantage of spacing 1 arcsec, but DEM not always available

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Conclusion 1 Qualified digital elevation models can be generated by automatic matching images taken from space and by InSAR With the free of charge available SRTM C-band DSM a homogenous height model is available for main part of world but mostly only with a spacing of 3 arcsec causing a loss of morphologic information in mountainous areas Even with lower accuracy of height models generated by stereo models of space images more morphologic information may be available Optimal condition for matching if images from same orbit - if from different orbits, problems may occur if large time interval and change of vegetation and shadows Advantage of near infrared or at least extended pan for matching in forest area With very high resolution space images even 3D-city models can be generated but optimal results not with the largest convergence angle Matched height models as well as SRTM data are DSMs with height values of the visible surface – reduction of DSM to DEM required by filtering

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Conclusion 2 Bottle neck for matching: stereo models taken from same orbit ASTER with 15m GSD, SPOT HRS with 5m GSD in orbit direction and the new Cartosat-1 with 2.5m GSD guarantee with 2 views always a stereo coverage – ASTER has limited resolution, SPOT HRS-images are not available - only the height models, Cartosat-1 is just starting with operation Very high resolution space image models (IKONOS, QuickBird and OrbView-3) are not so often available because of limited satellite rotation speed causing a loss of capacity if stereo models are taken Situation will be improved soon by more flexible very high resolution space images, more images from stereo systems (Cartosat-1 and ALOS) and high resolution InSAR

ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Outlook The number of high resolution space imaging system will grow very soon: -CartoSat-2, Kompsat-2, RazakSat, WorldView 1 and 2, OrbView-5, Pleiades, THEOS, EROS-C, CBERS and small satellites with high resolution Higher number of systems will improve the situation of very high resolution stereo models taken from same orbit Compact components will reduce the satellite cost, enabling also low cost high resolution small satellites Æ reduced price for images, better selection The resolution will be improved to 42cm GSD by WorldView 1 and 2 as well as OrbView-5, Worldview 1 and 2 with quite higher agility will allow fast stereo imaging SAR-systems TandemX and SAR-X Cosmo-Skymed-1 with up to 1m GSD will allow generation of detailed height models ALOS PALSAR with L-band will allow also the generation of DEMs in forest areas For several purposes sufficient accuracy by direct sensor orientation will avoid problems of classified coordinates as well as classified DEMs ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006