geometric and modeling of high-resolution satellite...

Dr.Phisan Santitamnont 1Aug 2007 Dr.Phisan Santitamnont 1

Geometric and Modeling of High-resolution Satellite Imageries

Dr. Phisan SantitamnontSurvey Engineering, Chulalongkorn University

Dr.Phisan Santitamnont 2

Chronology

• 1960 The first remote sensing satellite• Cold War era, high-resolution satellite

imagery for surveillance and reconnaissance– 1960-1972 Corona Program – 800,000 images ,

12 years• 1970s Landsat Satellites

Dr.Phisan Santitamnont 3Dr.Phisan Santitamnont

Dr.Phisan Santitamnont 4

Corona 1960s

Aug 2007 4

Dr.Phisan Santitamnont 5

The US. Land Remote Sensing Policy Act of 1992

•  Landsat program returned to government operation, mandated that its data be made available at cost

• provision for the licensing of commercially operated remote sensing satellites

• 1999 commercial IKONOS satellite launched

Dr.Phisan Santitamnont 6

Chronology

• 2000s many commercial high-resolution launched

• 2005 Google Earth 3.0 becomes the major user of high-resolution satellite

• 2008 signed agreement with GeoEye-1 and prepare 2012 GeoEye-2 mission

Dr.Phisan Santitamnont 7Aug 2007 Dr.Phisan Santitamnont 7

Quickbird #0.61 m (2001)

GeoEye-1

#0.41 m (2008)

IKonos-2

#1 m (1999)

SPOT-5

#2.5 m (2003)

WorldView-1

#0.5 m (2008)

THEOS ธ�ออส#2 m (2008)

Dr.Phisan Santitamnont 8

High-resolution Satellite Imageries

• 1999 : IKONOS Pan#1.0 m• 2001 : QuickBird Pan#0.61 m• 2002 : SPOT-5 Pan#2.5 m• 2006 : ALOS Pan#2.5 m• 2007 : WorldView-1 Pan#0.4 m• 2008 : THEOS Pan#2 m• 2008 : GeoEye-1 Pan#0.43 m• 2012: GeoEye-2 Pan# 0.25 m ??

Dr.Phisan Santitamnont 9

Sensor Model• Rigorous Parametric Model (Orbital Model)

– For SPOT-5,ALOS,THEOS– perspective projection by frame e.g. aerial photography – Perspective projection line-by-line , push-broom,

moving single linear detector array on the focal plane• Replacement ( Rational Function Model : RFM)

– case : QuickBird,IKONOS– 3-D Rational function supplied with product– 3-D Rational function computed from GCPs

Dr.Phisan Santitamnont 10

ราคาแผนท�จากภาพถ�ายดาวเท�ยมรายละเอ�ยดส�ง

1,000 Baht/Sq.KM 10,000 Baht/Sq.KM.+/50-100 m. +/-1..0.4 m.

Dr.Phisan Santitamnont 11

ความส�มพ�นธ�ระหว�างพ�ก�ดอ�มเมจก�บพ�ก�ดว�ตถ�

Rational Function Model(cubic polynomials)

)3939~(:..,..,..,..

),,(:,,

),(::,

Zd Xd XYZd YZd XZd XYd Zd Yd Xd 1Zc Xc XYZc YZc XZc XYc Zc Yc Xc c

Zb Xb XYZb YZb XZb XYb Zb Yb Xb 1Za Xa XYZa YZa XZa XYa Za Ya Xa a

191190191190

319

287654321

319

2876543210

319

287654321

319

2876543210

+

+…++++++++++…+++++++++=

+…++++++++++…+++++++++=

polynomialcubictscoefficienfunctionrationalddccbbaa

elevationlongitudelatitudescoordinateobjectZYXcolumnlinescoordinateimageyx

y

x

pp

p

p

Dr.Phisan Santitamnont 1212

Normalized Rational Function

Dr.Phisan Santitamnont 13

RPC Biases

• DigitalGlobe’s WorldView-1 Satellite: Increased Image Collection Opportunity, by Philip Cheng and Chuck Chaapel

...

...

0

0

+⋅⋅+⋅+⋅+=∆+⋅⋅+⋅+⋅+=∆

LineSampleBLineBSampleBBRLineSampleALineASampleAAP

SLLs

SLLs

Dr.Phisan Santitamnont 14

WorldView-1 Sensor Refinementafter Cheng&Chaapel

Dr.Phisan Santitamnont 15

Example RPC from IKONOSO F F S E T

LINE_OFF: +5540.00 pixelsSAMP_OFF: +5470.00 pixels

LAT_OFF: +18.74020004 degreesLONG_OFF: +98.94059753 degreesHEIGHT_OFF: +842.000 meters

S C A L E

LINE_SCALE: +6972.00 pixelsSAMP_SCALE: +7703.00 pixels

LAT_SCALE: +0.06040000 degreesLONG_SCALE: +0.05480000 degreesHEIGHT_SCALE: +600.000 meters

Dr.Phisan Santitamnont 16

Example : IKONOS RPC - LineLINE_NUM_COEFF_1: +5.126440082676709E-04LINE_NUM_COEFF_2: +8.107643723487854E-01LINE_NUM_COEFF_3: -2.017990797758102E-01LINE_NUM_COEFF_4: -7.095409557223320E-03LINE_NUM_COEFF_5: +1.357850385829806E-03LINE_NUM_COEFF_6: -1.415262522641569E-05LINE_NUM_COEFF_7: -1.002094995783409E-05LINE_NUM_COEFF_8: -1.903933589346707E-04LINE_NUM_COEFF_9: -3.566308005247265E-04LINE_NUM_COEFF_10: +8.047554160839354E-07LINE_NUM_COEFF_11: +1.648443088697604E-07LINE_NUM_COEFF_12: -5.703462875317200E-07LINE_NUM_COEFF_13: +1.151551623479463E-05LINE_NUM_COEFF_14: -3.738494172011997E-07LINE_NUM_COEFF_15: +5.723961294279434E-06LINE_NUM_COEFF_16: -3.347716074131313E-06LINE_NUM_COEFF_17: +1.011000705375409E-07LINE_NUM_COEFF_18: -1.910687615236384E-06LINE_NUM_COEFF_19: -8.052054312202017E-08LINE_NUM_COEFF_20: +3.335755671685092E-09

LINE_DEN_COEFF_1: +1.000000000000000E+00LINE_DEN_COEFF_2: -2.402324898866937E-04LINE_DEN_COEFF_3: +1.970350043848157E-03LINE_DEN_COEFF_4: -9.962612530216575E-04LINE_DEN_COEFF_5: +6.829987796663772E-06LINE_DEN_COEFF_6: -2.126053459505783E-06LINE_DEN_COEFF_7: -2.352771389269037E-06LINE_DEN_COEFF_8: +3.301248625575681E-06LINE_DEN_COEFF_9: +2.234998464700766E-05LINE_DEN_COEFF_10: -3.700658339766960E-07LINE_DEN_COEFF_11: -6.862588719513951E-09LINE_DEN_COEFF_12: +1.931878212957372E-09LINE_DEN_COEFF_13: +2.601792781220524E-09LINE_DEN_COEFF_14: +1.781106817588807E-09LINE_DEN_COEFF_15: +1.046838615792467E-08LINE_DEN_COEFF_16: +1.526291271147784E-08LINE_DEN_COEFF_17: +4.082192639476290E-10LINE_DEN_COEFF_18: +2.580622604497762E-09LINE_DEN_COEFF_19: -2.583811742340458E-08LINE_DEN_COEFF_20: +2.116114783401457E-10

Dr.Phisan Santitamnont 17

Example : IKONOS RPC - SampleSAMP_NUM_COEFF_1: -1.008798699331237E-05SAMP_NUM_COEFF_2: +1.578782796859741E-01SAMP_NUM_COEFF_3: +8.484598994255066E-01SAMP_NUM_COEFF_4: -6.666941102594137E-03SAMP_NUM_COEFF_5: +5.125267853145488E-05SAMP_NUM_COEFF_6: +5.131273610459175E-06SAMP_NUM_COEFF_7: -1.126090413890779E-03SAMP_NUM_COEFF_8: +1.051116050803103E-04SAMP_NUM_COEFF_9: +1.713753445073962E-03SAMP_NUM_COEFF_10: +7.220215138659114E-06SAMP_NUM_COEFF_11: -1.375646434098599E-06SAMP_NUM_COEFF_12: -1.728069918272013E-07SAMP_NUM_COEFF_13: +8.261971743195318E-06SAMP_NUM_COEFF_14: -1.224103698405088E-07SAMP_NUM_COEFF_15: +5.640218205371639E-06SAMP_NUM_COEFF_16: +2.094774754368700E-05SAMP_NUM_COEFF_17: +6.102967020638062E-09SAMP_NUM_COEFF_18: +2.125393166352296E-06SAMP_NUM_COEFF_19: -2.417262976450729E-06SAMP_NUM_COEFF_20: +4.360036220951713E-10

SAMP_DEN_COEFF_1: +1.000000000000000E+00SAMP_DEN_COEFF_2: -2.402324898866937E-04SAMP_DEN_COEFF_3: +1.970350043848157E-03SAMP_DEN_COEFF_4: -9.962612530216575E-04SAMP_DEN_COEFF_5: +6.829987796663772E-06SAMP_DEN_COEFF_6: -2.126053459505783E-06SAMP_DEN_COEFF_7: -2.352771389269037E-06SAMP_DEN_COEFF_8: +3.301248625575681E-06SAMP_DEN_COEFF_9: +2.234998464700766E-05SAMP_DEN_COEFF_10: -3.700658339766960E-07SAMP_DEN_COEFF_11: -6.862588719513951E-09SAMP_DEN_COEFF_12: +1.931878212957372E-09SAMP_DEN_COEFF_13: +2.601792781220524E-09SAMP_DEN_COEFF_14: +1.781106817588807E-09SAMP_DEN_COEFF_15: +1.046838615792467E-08SAMP_DEN_COEFF_16: +1.526291271147784E-08SAMP_DEN_COEFF_17: +4.082192639476290E-10SAMP_DEN_COEFF_18: +2.580622604497762E-09SAMP_DEN_COEFF_19: -2.583811742340458E-08SAMP_DEN_COEFF_20: +2.116114783401457E-10

Dr.Phisan Santitamnont 18Dr.Phisan Santitamnont

Example with HypercubeReading IKONOS TIFF and RPC

line,sample, h φ, λ h = N+H ;H≡0 (assuming flat area)

Dr.Phisan Santitamnont 19

IKONOS

Dr.Phisan Santitamnont 20

Geo Product

• Most affordable but lowest positioning accuracy

• Not corrected for terrain distortions , RMSExy=50m CE90%

• No RPCs !!! Need to solve 39+39 unknowns , thus need 39 GCPs

• Cheap, 500-1,000 Baht per Sq.Km.

Dr.Phisan Santitamnont 21

Geo Ortho KIT

• Geo imagery with Image Geometry Model (IGM) ‘terrain independent’ virtual sensor model RPC

• Allow generation of precise ortho using local DEM and GCPs

• Expensive ! 62-98 USD/sq.km.

Dr.Phisan Santitamnont 22

GEOMETRIC PROCESSINGOF IKONOS GEO IMAGES WITH

DEMThierry Toutin

Natural Resources Canada, Canada Centre for Remote Sensing588 Booth Street, Ottawa, Ontario, Canada, K1A 0Y7

thierry.toutin@CCRS.Nrcan.gc.ca

Dr.Phisan Santitamnont 23

Test

• Geo Product with full-scene 100 sq.km.• Two GCPs dataset

– RMSE +-4 m from map– RMSE +-1 m from GPS capture

Dr.Phisan Santitamnont 24

GCP RMSE=4m ; need 15-20 GCPsGCP RMSE=1m ; need 10 GCPs

Dr.Phisan Santitamnont 25

QuickBird

Dr.Phisan Santitamnont 26

QuickBird

• Basic Imagery Product– must order full-scene (US$ 6120)– Least processed , for later advanced processing– Including RPCs

• Ortho-ready Standard Product – Not include topography correction– can be ordered by sub-scene , B/W US$ 22 per

sq.km. (minimum 25 sq.km.)

Dr.Phisan Santitamnont 27

QuickBird End-Product

Orthorectified Product– Orthorectified with GTOPO30 and resample

into a cartographic project– Positional accuracy of 23m (CE90%)– Serveral level of refinements with more

accurate DEM and GCP– Price ranges from US$40 to US$90 per sq.km.– Cannot be further orthorectified

Dr.Phisan Santitamnont 28

Quickbird along-track viewing angle = -27.24/26.35

RPC Solution is dramatically improve starting with only one GCP!

ICP = Independent Check Point

Dr.Phisan Santitamnont 29

Geometrical Processing of Quickbird High-Resolution Satellite Data

F.Volpe, Eurimage S.p.A., ROMA, volpe@eurimage.com

• Two type of GCPs and Terrain tested with QB Basic Product

GCPs from 1:5,000 topo map # 0.3175 meter , DEM raster 10 meter posting RMSEz< 5 m. LE90% applied to QB over 150-840m relief

GCPs from GPS collected points RMSExy<1m. RMSEz<2m applied to QB over 13-125 m. relief

Dr.Phisan Santitamnont 30Dr.Phisan Santitamnont

Geometrical Processing of Quickbird High-Resolution Satellite Data

F.VolpeEurimage S.p.A., ROMA, volpe@eurimage.com

1

2

Dr.Phisan Santitamnont 31

Processing with PCI

Dr.Phisan Santitamnont 32

Processing Quickbird with PCIToutin’s model (Satellite Orbital

Modeling)

• Minimum 6 accurate GCPs

Dr.Phisan Santitamnont 33

RPC processing with PCI

• Where GCPs is inadequate• Typical 1 to 4 GCPs are recommended• RPCs files are store outside the image file

– For QuickBird *.RBP– For IKONOS *_rpc.txt

Dr.Phisan Santitamnont 34

Processing RPCs

Dr.Phisan Santitamnont 35

Orthorectification

Dr.Phisan Santitamnont 36

Influence of DEM Accuracy and Orthorectification

อ�ทธ�พลของความละเอ�ยดถ�กต�องของแบบจ�าลองระด�บ

Dr.Phisan Santitamnont 37

แบบจ�าลองระด�บ (DEM in Thailand)

• MoAC Mapping Project : จากโครงการแผนท�#ฯของกระทรวงเกษตรและสหกรณ( (กรมพ�ฒนาท�#ด�น)– ระวางละ 2 x 2 ตร.กม (1,000 บาทต*อระวาง)– DEM GSD= 5 เมตร� σHor = 1 m ; σver = 2-4 m

• แบบจ�าลองระด�บ SRTM หร+อ DTED-2 (1:50,000)– DEM GSD =30 /90 เมตร� σHor = 20 m ; σver = 15 m

Dr.Phisan Santitamnont 38

ASTER Global Digital Elevation Model (ASTER G-DEM)

• MITI/Japan ร*วมก�บ NASA/JPL• ใช� VNIR / ASTER Stereo (Band 3N/3B) #15

m• จะได� DEM ท�#วโลก ท�#ความระยะห*างจ�ดระด�บ 30 เมตร• ความละเอ�ยดถ�กต�องค*าระด�บ 7-10 เมตร• ให�ใช�ฟร�

Dr.Phisan Santitamnont 39Aug 2007 Dr.Phisan Santitamnont 39

เส�นช�!นความส�ง SRTM-3 (1:50,000) ก�บ LDD DEM/Ortho (1:4,000)

Using Dept. of Land Development ‘s DEM at scale 1:4,000

SRTM-3(1:50,000)

LDD DEM/Ortho1:4,000

Dr.Phisan Santitamnont 40

ความคลาดเคล$อนของจ&ดภาพ Δr เน$องจากความส�งต(าของภ�ม)ประเทศ

Relief Displacement

∆H

HAGL

P

P′

O

O′

H

f

p′p

R

r′

H-∆H

r

∆r

O″

O′″

Dr.Phisan Santitamnont 41

QB Supplemental Files

Dr.Phisan Santitamnont 42

Product Levels

Ortho Kit

Dr.Phisan Santitamnont 43

Sensor Model and Minimum GCPs

Dr.Phisan Santitamnont 44

Virtual GCP 3D Registration

1,4000 LDD DEM

1:4,000 Ortho map Satellite Image

Dr.Phisan Santitamnont 45

Field of ViewAerial Camera

VSSatellite Sensor

2-4 กม

600-800 km

2-4 กม

FOV

1-2°

FOV

75-90°

10-20 กม

Dr.Phisan Santitamnont 46

DEM Accuracy (∆H), off-nadir (∆η)

η

Off-nadir

∆X

∆H

δ incident angle

and plan. Error(∆X)

Dr.Phisan Santitamnont 47

Viewing Angle Versus Incident Angle

γ

ι

Dr.Phisan Santitamnont 48Dr.Phisan Santitamnont

Planimetric accuracy on orthophoto results from DEM quality and

viewing angle