Akira MUKAIDA RESTEC Feb. 21 [email protected] MUKAIDA RESTEC Feb. 21 2012@UNESCAP Workshop on Flood Risk Reduction through Space Application in South‐East Asia Satellite‐based

Akira MUKAIDA RESTECFeb. 21 2012@UNESCAPWorkshop on Flood Risk Reduction through Space Application in South‐East Asia

Satellite‐based Digital Terrain Mapper

Mar 14, 2011 Ishinomaki cityALOS/AVNIR-2

Mar 24, 2011 Rikuzentakata cityALOS/PRISM, AVNIR-2

Nov 19, 2011 Central BangkokBy Pi-SAR-L

Contents

• What remote sensing can provide?

• Terrain observation from space.– Basics of terrain observation.

– Observation by optical sensor.• Example of ALOS/PRISM

• Example of water resource management using remote sensing.

• Summary

What remote sensing can provide?

Satellite remote sensing can provide…• High frequent observation.• Wide coverage and high resolution.• Observation in equivalence quality (objective).

What remote sensing can provide?

The information from remote sensing data will help water resource management ,

• Nowcasting(i.e. inundation monitoring..)

• Forecasting

• Together with prediction model or assimilation.

Digital Terrain Modelfrom satellite remote sensing

Digital Elevation Model (DEM)– An array of elevation points without any attributes (i.e. may represent the

terrain surface. No buildings and trees.)

Digital Surface Model (DSM) <our target>– An array of elevation points representing the surface closest to the sensor

(may be same as a DEM)

Digital Terrain Model (DTM)– An array of points representing the terrain. Also called Digital Ground

Model.– Strictly speaking, the term DTM, Digital Terrain Model, should be reserved

for those models of reality which includes information relating to surface texture, etc., in addition to information regarding elevation.

Reference : Ian Dowman, “Generating Digital Elevation Models from Satellite Imagery”, presentation materials of the special seminar on DEM extractions in JAXA/EORC, Tokyo, 2002.

Basics

Height reference– Orthometric height = Height above the reference geoid level (mean sea

level) (i.e. EGM96)– Ellipsoidal height = Height above the reference ellipsoid model (i.e. GRS80

ellipsoid) <our target>– Geoidal height = Height of geoid surface above the reference ellipsoid

model (i.e. GRS80)

Generally…

ジオイド

準拠楕円体

地面

楕円体高

ジオイド高

標高

数値表層モデル（DSM ）

数値地形モデル（DTM ）

Digital Surface Model (DSM)

Digital Elevation Model (DEM)

Ground

Reference Geoid model(i.e. EGM96)

Reference Ellipsoid model (i.e. GRS80)

Geoidal height

Orthometric heightEllipsoidal height

Basics

Types and patterns– Generally corrected as pseudo random points (TIN)– Re-sampled to grid using interpolation <our target>– Breaklines added to define special features

Data sources– Ground survey– Aerial Photography– Satellite data

– Optical sensor <our target>– Interferometric SAR (InSAR)

– LIDAR– Existing data - maps

Basics

ALOS mission objectives are to;Provide and update maps for Japan and other countries including those in the Asian-Pacific region (Cartography),Perform regional observation for “sustainable development,” harmonization between earth environment and development (Regional Observation),Conduct disaster monitoring around the world (Disaster Monitoring),Survey natural resources (Resources Surveying), andDevelop technology necessary for future earth observing satellites (Technology Development).

Launch DateSpacecraft MassGenerated PowerDesign LifeOrbitAltitudeInclinationPeriodRecurrent Cycle

: 2006.01.24: Approx. 4 tons: Approx. 7 kW (at End of Life): 3‐5 years: Sun‐Synchronous Sub‐Recurrent: 691.65 km (at Equator): 98.16 deg.: 98.7 min.: 46 days (Sub Cycle : 2 days)

ALOS Specification

Panchromatic Remote-sensing Instrument for Stereo Mapping

Observing Geometry of triplet mode

Wavelength : 0.52 ~ 0.77 μmNumber of Optics : 3 (Nadir; Forward; Backward)at +/‐23.8 deg. inclinationBase‐to‐Height ratio : 1.0 (Forward‐Backward)Spatial Resolution : 2.5mSwath Width : 35km (Triplet mode)S/N : >70MTF : >0.2Pointing Angle : ‐1.5 to +1.5 deg. (Cross Track)Quantization : 8 bits

PRISM configuration

PRISM Specification

13

Forward lookingBackward looking Nadir looking N

©JAXA

©JAXA©JAXA

Satellite path

PRISM Triplet observation

14

ForwardNadirBackward

約24°約24°

Nadir:red, Forward:blue

Backward:red, Forward:blue

©JAXA

©JAXA

2006.05.05

2006.05.05

Anaglyph image by PRISM

At least two looking images are required to generate Stereo looking image (ex. DSM).

Difference of terrain image looking from each angle.©RESTEC included JAXA

©RESTEC included JAXA

Stereo Viewing by PRISM

Conditions– Sensor geometries of input stereo products (Level 1B1) should be fixed with

orientation process• Relative geometry for generating epipolar images of stereo matching• Absolute geometry for calculating XYZ coordinates in object space if necessary

Strategies of stereo matching– Area based grid matching with cross-correlations– Exclusive triplet stereo matching algorithm on triplet epipolar images– Coarse To Fine (CTF) ortho-image base parallax search (Ortho-image pyramid)– Automatically optimizing the cross-correlation patch size depending on image

characteristics (i.e. textures, contrast, terrain, etc.)

DSM Generation of PRISM

Triplet stereo matching on epipolar images– Resample forward, nadir and backward images on UTM epipolar frame– Search forward and backward parallaxes against nadir simultaneously

FWD

NDR

BWD

sum

yx

N

i ii yyxxN

σσρ

∑ =−−

=1

))((1

Cross correlation

yx

N

i ii yyxxN

σσρ

∑ =−−

=1

))((1

Cross correlation

Parallax direction


Epipolar image (forward)

Par

alla

x di

rect

ion


Epipolar image (nadir)

Par

alla

x di

rect

ion


Epipolar image (backward)

Par

alla

x di

rect

ion


Full Ortho‐Image

Coarse To Fine (Image pyramid)– 3~5 stages (default=3)

1/4 Averaged Ortho‐Image

Calculated DSM Grid Matching Image ( Ortho‐rectified image )

Existing DEM(optional)

1/20 Averaged Ortho‐Image

1st Coarse DSM

2nd Coarse DSM

Fine DSM

Matching Point

Search Range


DSM of CTF loop 1Matching image resolution = 50m


DSM of CTF loop 2Matching image resolution = 10m


DSM of CTF loop 3Matching image resolution = 2.5m


DSM evaluation– Compare with existing DEM/DSM

• SRTM• LiDAR-DSM• Photogrammetry DSM• Etc…

PRISM-DSM

Height difference image

Difference histogram

Evaluation of PRISM/DSM

DSM evaluation– Compare height profiles


PRISM-DSM validation test sites– Six sample scenes (including 9 DSM sites)

No. Ref. DSM Sites Obs. Date No. of GCPs No. of TPs

1 Mt.Tukuba/Chiriin 03/01/07 42 9

2 Okazaki 03/23/07 17 9

3 Fukuoka 04/29/07 24 9

4 Saitama 05/03/07 213 9

5 Thun /SW/ Bern 06/24/07 54 9

6 Mt.Ibuki 09/11/07 13 9

Site Terrain Source Size HeightRange

GroundResolution

HeightAccuracy

Source Year

Saitama*1) Flat & Urban LiDAR 14.0x12.0km 100m 1m <1m 2002

Okazaki*1) Mountainous Aerial Photo 6.0x6.0km 400m 10m ~10m 2005

Thun *2) Various Aerial Photo 7.5x14.5km 500m 2.5m 0.5~2.5m 2004

SW *2) Steep Aerial Photo 7.5x14.5km 1000m 2.5m 0.5~2.5m 2004

Bern *2) Various Aerial Photo 11.0x11.5km 400m 2.5m 0.5~2.5m 2004

Fukuoka*1) Various LiDAR 12.0x9.0km 500m 1m <1m 2002

Mt.Tukuba Mountainous LiDAR 1.5x1.5km 200m 1m 0.8m 2004

Chiriin Flat LiDAR 1.5x1.5km 50m 1m 0.8m 2004

Mt.Ibuki Steep LiDAR 1.2x1.9km 700m 1m 0.8m 2005

Test scenes (Triplet stereo)

EORC/JAXA Reference DSM sites

*1) provided by GSI*2) provided by ETH

Thun/SW/Bern

SaitamaOkazak

i

Saitama

OkazakiFukuoka

Chiriin/Mt.Tsukuba

Mt.Ibuki


Scene 1 : Tsukuba / JapanDate = 03/01/2007

35km NDR image in UTM frame

Reference DSM site(Mt.Tsukuba）

PRISM-DSM in 0.3 arc-sec Lat-Lon frame

Reference DSM site(Chiriin）


Mt.Tsukuba 03/01/2007

PRISM-DSM (1.8x2.2km) Height difference from Reference-DSM

500m0 250


SRTM-3 DEM GSI50m DEM PRISM-DSM LiDAR-DSM

Height scale [m]

Various DSM/DEM visual comparison(Fukuoka: 4km x 4km)

90m 50m 10m 1m

All world land area in N60~S56

All Japan PRISM observation areas except for

clouds

N/A

InSAR from STS-99 mission

Measured from contour lines of 1/25,000 map

Triangulation with satellite stereo

images

Airborne LiDAR

Mesh size

Data area

Sources


DSM from PRISM first light image on Mt. Fuji (Feb. 14, 2006)

First light image of PRISM/DSM

PRISM DSM with Pan-sharpen image of PRISM & AVNIR-2Contributing to the research on “Great Sichuan Earthquake” in China 2008

Disaster monitoring by PRISM/DSM

Fly-thru movie (Mt. Everest)

33Gray-scaled height image on 124 tiles with mask data

: Cloud/Snow mask (01)

: Land‐water mask (02): Sea mask (03)

2000m

0m

height

Valid (00) 5,436,770,700

Mask (01) 98,921,718

Mask (02) 40,784,118

Mask (03) 12,279,523,464

Number of data

Rate of land‐data completed =((00)+(02))/((00)+(01)+(02))*100= 98.2%

Maintained PRISM/DSM on Japan

DSM mosaic

Kanto3x3 deg. Tile@10m reso.Kyusyu3x3 deg. Tile@10m reso.

• Experiment on Mekon delta.

Water resource management using remote sensing

DMC@HNI

internet

JapanVietnam

SIWRP@HCM

Image & InformationDisplay

Image analysis &communication

Terminal

Image & InformationDisplay

Distribute toSMS/FAX

ALOS

A typical End‐User

Summary

• Satellite remote sensing data can provide,– High frequent, widely covered and objectively

observed data.– Terrain information which essential input for

inundation model.• With satellite derived terrain model,

– Expect improvement of prediction by inundation model.

• It will improve both nowcast and forecast of flood situation.

Thank you so much.

Akira [email protected]://www.restec.or.jp/

Documents

Akira MUKAIDA RESTEC Feb. 21 [email protected] MUKAIDA RESTEC Feb. 21 2012@UNESCAP Workshop on Flood Risk Reduction through Space Application in South‐East Asia Satellite‐based