TUTORIAL Information extraction, with emphasis on DSM generation

Jul 6, 2006 - ... Jacobsen1, Emmanuel Baltsavias2, Nicolas Paparoditis3, Peter Reinartz4 .... SPOT 5 Zonguldak: DZ: SZ = 15,14m SZ = 12,8 + 6,12 ∗ tan α.
Télécharger le PDF
2MB taille 9 téléchargements 286 vues
commentaire
Report
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 6

Reduction of DSM to DEM and Quality Analysis 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

DEM – digital surface model (DSM)

Digital surface model (DSM)

Digital elevation model (DEM) Åelevation of visible surface including vegetation and buildings

elevation of bare soil Æ

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

Institute of Photogrammetry and GeoInformation

filtering of DSM

matched points

surface determined by simple mean value filter

simple filter will generate smooth surface, but not a DEM with points belonging to ground, median filter may cause elimination of ground points instead of buildings and vegetation ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM with program RASCOR Input parameters: 1. type of terrain: flat, rolling or mountainous 2. same type of terrain or varying type

Option: break lines

typical varying type

3. mode change up / down – for identification of large buildings 4. 1 or 2 iterations - 1 iteration = usual, 2 iterations if generation of contour lines – will be more smooth All required limits determined by automatic analysis of 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

Filtering DSM Æ DEM with program RASCOR Program RASCOR – sequence of tests, starting with Z-range

max. Höhe

Z-limit

Step 1 not for mountains

Z-range

step 1

percentage histogram

< 40.36 40.36 - 43.46 43.46 - 46.57 46.57 - 49.68 49.68 - 52.79 52.79 - 55.89 55.89 - 59.00 59.00 - 62.11 62.11 - 65.22 65.22 - 68.32 68.32 - 71.43 71.43 - 74.54 74.54 - 77.64 77.64 - 80.75 > 80.75

1.83 % 5.11 % 9.53 % 10.24 % 11.83 % 18.41 % 14.13 % 7.42 % 5.53 % 5.77 % 3.82 % 2.93 % .83 % .58 % 2.07 %

** ****** ************ ************* **************** ************************* ******************* ********** ******* ******* ***** *** *

eliminate **

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

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM with program RASCOR

Influence of Z-range

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

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM with program RASCOR

Step 2 dz

dx (dy)

Höhensprung (hoch)

************************* ********************* *************** ************ ********* ******* ***** **** *** Histogram of *** ** ** neighboured ** * Z-differences * * * * *****************

DZ neighbored points exceeding limit

Step Höhensprung (runter)

eliminate

3 (option)

elimination of large objects by height change up / down ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM with program RASCOR

step 4

Z-profile in X- and Y-direction 6

Höhe [m]

5 4 3 2 1 0 1

2

3

surface Geländeoberfläche

4

5

6 7 Profil [m]

linear lineare Interpolation

8

9

10

11

polynomialInterpolation polynomische

Type of reference line depending upon terrain type flat: horizontal line

rolling: inclined line

mountainous: polynomial

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

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM with program RASCOR step 4 DZ against sub-area – horizontal plane, tilted plane, polynomial surface, step 5 least squares interpolation (prediction)

surface of least squares interpolation

real heights

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: Result of filtering DSM Æ DEM

DSM

DEM after filtering reference DEM Z-discrepancies before filtering

open area

forest after filtering

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-DEM Bangkok

color coded DEM without after filtering Zmax = 44m Zmax=6.1m

in Bangkok terrain height < 4m, SRTM-DEM includes Z-values up to 44m Filtering digital surface model (DSM) Æ DEM – only successful if noise < influence of vegetation and buildings + available values on the bare ground In Bangkok-DEM by filtering limitation of Zmax to 6.1m 59% of points in city area removed by filtering

3D-view to original SRTM-DEM 1° elevation (like skyline of Bangkok)

3D-view to filtered SRTM-DEM 1° elevation ISPRS Technical Commission I Symposium “From Sensors To Imagery”, Paris – Marne la Vallée, France, 3-6 July 2006

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM

forest

after filtering with Hannover program RASCOR Identification and removal of points not belonging to bare ground

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

Institute of Photogrammetry and GeoInformation

Filtering DSM Æ DEM grey value coded DEM determined by automatic matching - buildings can be recognized

matched DEM

filtered DEM (RASCOR)

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

Institute of Photogrammetry and GeoInformation

Effect of filtering to generated contour lines contour interval: 4 ft left: original data set from image matching right: after filtering

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

Institute of Photogrammetry and GeoInformation

Break lines

Original Data

filtered without

filtered with break lines

Bridge

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

Institute of Photogrammetry and GeoInformation

Part 2: Analysis of DEMs

shift of DEM KOMPSAT-1 DSM Reference DEM

KOMPSAT DSM shifted against reference DEM – main reason: datum of national net – shift determined by adjustment (Hannover program DEMSHIFT) – shift ~ 200m in X, 40m in Y Æ RMSZ from 50m Æ 15.8m

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

Institute of Photogrammetry and GeoInformation

RMSZ as function of terrain inclination 30

RMSZ [m] 25

For open areas:

20

RMSZ = 15.81m

15

RMSZ = 13.0m + 10.9 * tan α

10

bias 0.72m

For all data dependency of vertical accuracy depending upon tan (slope)

5

tangent of terrain inclination Æ

1. 00

.9 0

.8 0

.7 0

.6 0

.5 0

.4 0

.3 0

.2 0

.1 0

.0 0

0

In forest influence of vegetation – separate analysis in forest and open areas based on forest layer - or even other classification layers e.g. build up area

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

Institute of Photogrammetry and GeoInformation

Euklidian Distance

DEuklid = DZ ∗ cos α Difference Euklidian distance – DZ small – limited to error component as function of terrain inclination example SPOT HRS Prien: DZ: SZ = 7,98m Euklidian: SZ = 7,92m

SZ = 7,47 + 1,55 ∗ tan α SZ = 7,47 + 1,27 ∗ tan α

SPOT 5 Zonguldak: DZ: SZ = 15,14m Euklidian: SZ = 15,08m

SZ = 12,8 + 6,12 ∗ tan α SZ = 12,8 + 5,99 ∗ tan α

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 KOMPSAT height model RMSZ

RMSZ F(α)

RMSpx [GSD] for flat terrain

Open area

15.8m

13.0m + 10.9m ∗ tan α

0.9

Forest

15.8m

12.2m + 12.7m ∗ tan α

0.9

Histogram with changed sign

Histogram with changed sign

2m bias

influence of buildings

histogram of DZ for open area

influence of forest KOMPSAT DEM above reference histogram of DZ for forest

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

Institute of Photogrammetry and GeoInformation

Z-accuracy as function of aspects Shuttle Radar Topography Mission (SRTM) - X-band DEM

RMSZ for: terrain inclination 0.0 over all points For average terrain inclination Factor B ( RMSZ = A + B ∗ tan α )

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

Institute of Photogrammetry and GeoInformation

conclusion Filtering of DSM to DEM required, remarkable improvement limitation: if no point on ground (e.g. forest) DEM cannot be generated limitation: influence of vegetation and buildings must be larger than terrain roughness and accuracy of Z-values Analysis of DSM / DEM: shift of height model has to be checked / respected it is not possible to describe the accuracy of a DEM just by one figure - terrain inclination has to be respected, different values for different types of terrain e.g. forest, open areas, build up areas

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