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
Section 5 DSM generation using SPOT-5 HRS and other space data Peter Reinartz, Karsten Jacobsen DLR (German Aerospace Centre), Institut für Methodik der Fernerkundung, Bildwissenschaften, D-82234 Oberpfaffenhofen, Post Wessling, Germany,
[email protected] D-82234 Oberpfaffenhofen, Post Wessling, Germany,
[email protected] 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
SPOT 5 HRS •
Image matching
•
Test Areas and ground reference data
•
Properties of SPOT- HRS Data
•
Orthoimage generation without ground control, accuracy
•
Image matching (image pyramid, interest operator, dense matching)
•
DEM generation without ground control
•
DEM comparison and accuracy analysis with reference DEM
•
Improvement of exterior orientation
•
DEM comparison and fusion with SRTM DEM
•
Conclusion ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
SPOT 5 HRS (high resolution stereo)
view d r a w for
iew v d war k c ba
View direction: +20°, -20°, incidence angle: +23°, -23° pixel size in orbit direction: 5m
h/b = 1.2
across orbit: 10m
advantage of configuration: no change of imaged object within 90 sec. no change of the object, no change of the illumination, no change of the atmosphere ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Methods of image matching 1. Feature based matching – corresponding well determined points – as start information for other matching methods 2. Area based matching - correlation – finding optimal position of pattern matrix in a search matrix, start values required, problems if area not flat 3. Image correlation in epipolar lines – at least relative orientation required, search only in one direction
4.
Vertical line locus – correlation of corresponding sub-matrixes in object space – exterior orientation required
5. Least square matching – respects object inclination and linear changes of grey values, precise but very good approximations required
dx1 inclin ation
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
dx2
Institute of Photogrammetry and GeoInformation
Problems with matching in build up areas first imaging
convergence angle
height
second imaging
ba se
it orb
left
top
top
right
imaged area
IKONOS Maras h/b = 7.5
OrbView-3 Zonguldak h/b = 1.4
shadow right
shadow left
= points which can be determined ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Problems with matching in build up areas sub-images IKONOS h/b=7.5
OrbView-3 h/b=1.4
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Vertical ground accuracy
hg SZ = sn ∗ ∗ Spx formula for photos b SZ =
hg ∗ a 2 ∗ GSD b
SZ = standard deviation of ground Z b = base (distance of projection centers Spx = standard deviation of x-parallax (x‘-x“)
a2 = factor
formula for digital images base
height
base
height
In case of automatic image matching: factor a2 smaller for larger h/b relation Æ SZ not so much depending upon h/b-relation
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Test area Bavaria
south
Z-range: 350m – 1850m ~ 20% forest,
lakes
view to south – DEM generated by SPOT HRS ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Orthoimage generation for accuracy analysis Test area Barcelona: for/back orthoimage overlay (Orthoimage generation with reference DEM)
The accuracy compared with 24 ground control points has been found in the range of one/two pixel for forward and backward looking channel
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Attitude data
1 scene = 60km in orbit direction = 8 sec effect on ground ~ 2m
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Attitude data
1 scene = 60km in orbit direction = 8 sec effect on ground ~ 1.5 m
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Attitude data
1 scene = 60km in orbit direction = 8 sec effect on ground ~ 0.1 m
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Absolute accuracy of orthoimages influence of sensor orientation + DEM Mean values and standard deviations for the difference to the orthoimages of 20 ground control points in meter in Transverse Mercator coordinate system (Bavaria) Forward looking x2 – x1 y2 – y1
Backward looking x3 – x1 y3 – y1
MEAN
-4,3
5.0
-14.3
11.5
Std. dev
5.89
7.35
6.23
8.64
Mean values and standard deviations for the difference to the orthoimages of 24 ground control points in meter in ED50 coordinate system (Catalonia) Forward looking Backward looking Nadir looking x2 – x1 y2 – y1 x3 – x1 y3 – y1 x4 – x1 y4 – y1 MEAN
-9,90
-16,59
-0,36
-11,16
-24,22
-6,22
STDV.
4,64
8,48
5,72
5,23
5,96
5,71
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Orthoimage generation for accuracy analysis
influence of sensor orientation + DEM Image Matching result between the two orthoimages (for/back)
Resulting mean vector length: 14 meters, RMS: 2 meters ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Image matching Matching strategy •
First image matching uses an image pyramid, restricting matching to interest operator points for each level (maximum of correlation coefficient)
•
Local least squares matching to sub-pixel accuracy in all levels of the pyramid generates tie-points for subsequent dense matching = start values for next step
•
Dense matching with region growing (Otto-Chau-algorithm) (with different window sizes and backward matching for blunder reduction) leads to mass points (here 60 to 80%) in image space
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Calculation of object coordinates •
Forward intersection using least squares adjustment of the two image rays
•
Rejection criterion: intersection of image rays exceeding one pixel
•
Interpolation of irregular distributed points into a grid of 15 x 15 meter pixel size (Catalonia), interpolation process is performed by a moving plane algorithm
distribution of matched points (about 2/3 of area covered)
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
DEM and orthoimage of Bavaria Image draped over generated height model
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Test area Bavaria
Accurate reference data: laser + photogrammetry
Colour coded height model from SPOT 5 HRS
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006
Institute of Photogrammetry and GeoInformation
Accuracy of DEM (single height points without GCP) • Comparison of height for high quality homologous points in SPOT-DEM and reference DEM Reference area
Size, Accuracy of Ref-DEM
bias [m]
Std. Dev. [m]
Points [#]
DEM-01, Prien
5 x 5 km, 0.5 m
6.8
2.0
240
DEM-02, Gars
5 x 5 km, 0.5 m
6.2
2.2
184
DEM-03, Peterskirchen
5 x 5 km, 0.5 m
5.6
1.8
261
DEM-04, Taching
5 x 5 km, 0.5 m
4.9
2.0
254
10 x 10 km, up to 5 m
5.7
3.5
458
50 x 30 km, 2-3 m
6.1
3.6
15177
DEM-05, Inzell (mountainous) DEM-06, Vilsbiburg
ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006