Technical Exploitation Needs for Defenc

8/12/2019 Technical Exploitation Needs for Defenc

http://slidepdf.com/reader/full/technical-exploitation-needs-for-defenc 1/37

OSA May 30-31, 2011 Technical exploitation needs for defence applications: SAR imagery

Technical exploitation needs for defence

applications: SAR imageryImage Analysis Laboratory, Observation Systems Area

Dept. of Earth Observation, Remote Sensing and Atmosphere (DOTTA)

INTA

Severino Fernández [email protected] Support tools for IMINT exploitation

Patrick Vaughan [email protected] SIIRS and reference manual

Enrique Nicolás [email protected] SAR geometric exploitation

Presented at the “Congress on technologies and applications of synthetic aperture radar (SAR)”, 30th-31st

May, 2011, INTA, Torrejón de Ardoz, Spain.




Objective:

• Identify IMINT user needs for SAR image exploitation in the spanish MoD

• Identify techniques considering availability of Helios I (since 1995) and Helios II

(since 2005) imagery in archive, and fast access Helios II imagery (Ground

Segment in Torrejon)

• Consider availability of Pleiades imagery in archive (Pleiades launch planned for

beginning 2012) and fast access to Pleiades 1 imagery (Ground Segment in

Torrejon).

• Identify modern methods that can improve IMINT analyst productivity and

confidence on interpretation results.

• Build prototype to assess feasibility

characterize results (for which cases produces reliable results, for which

cases does not)

once robustness verified, evaluate operational feasibility (processing time)

if both OK, integrate into IMINT user workstation

Support tools for

IMINT exploitation




Support tools for

IMINT exploitation

IMINT activity/need Experiment-IntegrationComplementary use of large asset and eventual

shortly simultaneously acquired optical high

resolution imagery

• automatic or largely automatic image co-

registration of SAR and optical imagery.

• SAR/optical image data fusion after co-

registration

• Detect:

• Ships: detection seems to be feasible

already

• Ground vehicles: needs characterization

• Paked aircraft: detection seems feasible

(somehow like ships, background is

“simple”)

•

Identify vehicles• Difficult with 1 m resolution

• Evaluate visual interpretability enhancement

with Superresolution methods

• Evaluate detectability with same or similar

methods as for ships

• Evaluate subaperture processing for target

enhancement

• Evaluate identification with feature

classification+feature enhancement withSuperresolution methods

Methods must produce results in seconds for

images of size of workstation screen (somehow

1000x1000 pixels)

• Once robustnessis validated, optimize code

• Multithreading implementation for multicore

workstations




Imagery co-registration. Preliminary results and status.

• Mutual information similarity measure

• Simple local deformation (e.g. translation), global deformation derived from local

one by interpolation• Optimization: probably stochastic gradient

• Several open source libraries available

• Some provide robust optimization with adaptive parameters, but not yet

refactored for multithread

• Others already optimized some similarity measures, some optimizers and

natively multithreaded• Has been just preliminary validated (suspended because no resources

allocated)

Support tools for

IMINT exploitation




Imagery co-registration.

Some results with real data (Ikonos+Terrasar Torrejon Air Base)

Automatic co-registration with local displacement and global

spline deformation.

Pending Actions:

Exhaustive tests to verify robustness of approach

Characterize at least following aspects:

• Behaviour on buildings, infrastructure

• Timing, maximum size with acceptable processing time using multicore PC

• Identify failure cases

Support tools for

IMINT exploitation




Before

Support tools for

IMINT exploitation


http://slidepdf.com/reader/full/technical-exploitation-needs-for-defenc 7/37OSA May 30-31, 2011 Technical exploitation needs for defence applications: SAR imagery

After

Support tools for

IMINT exploitation



SAR imagery superresolution

Concept demonstration prototypes:

Adaptive Weighted Norm (spectrum) Extrapolation

• 2D version coded AR spectrum estimation

• a particular implementation coded

APES (Amplitude and Phase Estimation) and Minimum

Variance (Capon)• open source code from Prof. Erik Larsson tested with

available imagery

• Limited results due to lack of resources.

Support tools for

IMINT exploitation



SAR imagery superresolution- Adaptive Weighted Norm

Extrapolation:Minimum wighted norm, matches given data and frequency distribution.

Test: A subaperture is taken from the data and extrapolated to compare

with the original image.

.

Original image image from the lower frequency image from spectrum

components extrapolation

Support tools for

IMINT exploitation





SAR imagery superresolution - 2-D AR extrapolation

Two quadrant AR coefficients are computed from the data (the

signal history obtained from the image). The new signal history

is obtained from the AR coefficients.

Support tools for

IMINT exploitation



Prof. Larsson implementation of APES and Capon

AWNE seems visually more attractive.But both APES and Capon perform better

for single scatterers.

Original Capon APES (Amplitude and Phase)

AWNE

Support tools for

IMINT exploitation



Prof. Larsson implementation of APES and Capon

Using an ideal signal produced by three

scatterers, APES and Capon reconstruct the

image with more fidelity than AWNE. The ideal

scatterers are correctly situated.

.

Signal of 3 scatterers DFT APES

AWNE

Support tools for

IMINT exploitation



Present state:

The usefulness for support of visual image interpretation of the

various methods for SR from single images seems very doubtful .

At least for the resolution of Paz (TerrasarX).

Several authors state that these methods improve the results of

automatic target recognition.

We would have to check this on Paz (TerrasarX) data. This wouldimply implementing a target recognition prototype.

Another question arises: do these methods have any influence on

the results of automatic co-registration of SAR and optical data?

Support tools for

IMINT exploitation



Ground target detection

Simple approach using global-local (CFAR) detection withWeibull and K distributions + morphology operation on results

Support tools for

IMINT exploitation



Support tools for

IMINT exploitationDetection of faint tracks in coherence image. Denoising.

Method based on Poissonstatistics for pixel intensity (from

fluorescence imagery application).

Visibility improvement for extended

regions. ¿can we apparently see features

in noisiy image just because we see more

in denoised one?




Thank you

Support tools for

IMINT exploitation

SIIRS & R f M l




SIIRS & Reference Manual:

BackgroundSIIRS (SAR Image Interpretability Rating Scale) & Reference Manual

Scope: military applications operational SAR image analysis

Exploitation need: to provide operational users with answers to:

a) PAZ imagery request parameters, e.g.:

•What image mode ?

•What polarisation ?

•How many images do I need ?

b) PAZ Capability & task planning, e.g.:

•What image processing ?

•What targets can I detect, recognise or identify?

•Expected time frame?

Solution: establish straightforward reference guidelines per mission type, based onreference documents: NIIRS and STANAG 3596

Framework: Strategic Plan of OSA-DOTTA (INTA)

Milestones: 3+ (2011-2013+)

Platform and outsourcing: TBD (Industry, tech specs.)

SIIRS & R f M l





Outline and workflowSIIRS (SAR Image Interpretability Rating Scale) & ReferenceManual

NIIRS

Target categories

Reporting fields

Recce requesting

reporting guide

ADAPT

Interpretability (analysis capability)

Per spatial resolutionPer wavelength range (VIS,NIR, MIR,SAR)

Per target examplesEXPAND

SIIRS

(phase 1) VALIDATE

TOOLS ANDALGORYTHMS

FIELD TESTINGReference Manual

(phase 2)

USERFEEDBACK

SIIRS & R f M l





Basic contents

ReferenceManual

What can we discriminate:

List of targets per STANAG 3596

category that can be:

•Detected

•Distinguished

•Recognised

•Identified

How to do it:

Data requirements:• Number of images & acquisition modes

• spatial & time resolution• Recommended polarisation

Estimated time (acquisition + processing)

Level of certainty/feasibility

Processing methodology

• Alternative methodology

• Processing tool requirements

• Level of complexity

Examples

SIIRS

SIIRS & R f M l





Case Scenario

Reference

Manual

CAT 010B+2b.

Is this included in the list ?

Do we have the minimum

requirements?

Any examples ?

SIIRS

Recce request for:

Target CAT 010B+2b.

Occupation: disturbed

earth

Disturbed earth detection

using Coherent Change

Detection (CCD)

High endLow end

VALIDATION

REQUIRED?

SIIRS & Reference Man al





CCD examples

TerraSAR-X SM ~1.5m pixel 11 day CCDGoogle Earth

Preiss and Stacey, DSTO 2008

Airborne SAR-X ~25cm resolution

CCD

2nd date (I) image

?

SIIRS & Reference Manual





Validation & FieldworkREAL SCENARIO CAMPAIGN 2009

Airborne SAR – X (1m pixel) fromRadar Laboratory (Dpt. OfRadiofrequencies and ElectronicTechnology-INTA)

Validation and information on:

• Signal penetration on cammo nets• Target discrimination vs. clustering• Target size and pose vs. detection• Environment /clutter interference

COMING UP...

Validation studies of Tandem-X &TerraSAR-X “Data applications forsecurity and defence”

DLR Scientific Project Proposal

Validation aims:

• Pointing & DEM accuracy• Pixel X,Y,Z location accuracy• Trafficability• Disturbed earth & CCD






Summary

NIIRS

SIIRS

VALIDATION &

FIELDWORK

BENCHMARKING

&

STATE OF THE ART

Funnel

& filter




Patrick J. Vaughan Martín-Mateo [email protected]

Thank you very much !




SAR Geometric Exploitation

Objective:

• Identify IMINT user needs for SAR image geometric exploitation in the Spanish MoD.

• Analyze PAZ satellite capabilities to cover the user needs. Training with TerraSAR-X

and TanDEM-X.

• Analyze the different techniques for DEM generation from SAR images:

•

Cross-track single pass interferometry.• Cross-track repeat pass interferometry.

• Repeat pass radargrammetry.

• Define a methodology to characterize the influence of the observation geometry in:

• the location accuracy of the SAR images

•

the DEM height accuracy generated.

• Define a ground truth area for geometric CAL/VAL of SAR/Optical images and other

applications.

• User support for baseline tasks concerning orthorectification, virtual scenarios

generation and other applications.




IMINT activity/need Experiment-Integration

-Techniques to generate altimetric information to

use as input in other applications.

- Location/height accuracy knowledge

• Effects on data due to variations in geometric

conditions.

• To learn geometric behaviour in TSX and TDX

for the future PAZ satellite.

• Produce:

• DSMs (Digital Surface Models).

• DTMs (Digital Terrain Models).

• GDMs (Ground Deformation Models).

• GCPs (Ground Control Points).

• Course training material.

• To:

• Block adjustment.

• Orthorectify.

• Virtual scenarios generation.

•

Underground structure detection.• 3D feature extraction.

• User training.

• Other….

• Analyze the influence of parameters on the

pointing and height accuracy:

• the post-processed orbit type.

• processed product type.

• elevation beam angle.

• baseline size.

• orbit direction.

• polarization.

• single pass vs. repeat pass.

• influence of software processing

algorithms.• Characterize the differences between

techniques for DEM generation:

• Limitations.

• Best choice.

• Effort in the generation.

Define protocols and training for users. Define protocols and training for users.





Experiments: Validation studies of TanDEM-X data applications for securityand defence (proposal submitted to the DLR)

- Test site experiments (TBC):-Image Pairs (TSX/TDX):

-3 SL mode.

-3 HS mode.

-5 HS300 mode.

-10 SM mode.

-1 SC mode.

- Interferometric single pass:

-23DEMs.

-Interferometric repeat pass:

-36DEMs.

-Radargrammetric repeat pass:

-55DEMs.

-CCD analysis: 36 products.-Border control.

-Trafficability.

-Disturbed soil.

-DEM fusion techniques.

-SAR focussing.





- Test site definition(TSAs):

- Terrain type

scenarios: Flat area,

smooth/medium

relief, high relief.

- Cover type

scenario: bared soil,

shrubland, forest,

manmade.

10x10km

52x30km

25x30km

Ground truth work:





- Planimetric accuracy control:

- (SAR) Corner reflectors:

trihedral / offset trihedral /

luneberg lenses (TBC).

Ground truth work:

Trihedral

Offset

Trihedral

Luneberg

Lenses









- (SAR) Identified

Permanent Scatterer Points

and Intensity Change Points

(IPSPs - ICPs)

Ground truth works:









- (SAR) Identified


and Intensity Change Points

(IPSPs - ICPs)

Correct match:

Street lamp base

Measurable IPSP

No match:

Probably a tennisumpire chair

Ground truth works:


l








- (SAR) Identified


and intensity change points

(IPSPs - ICPs).

- (Optical) GPS Rapid Static

observations (GCPs grid).

- (Optical) Spanish geodetic

reference network vertex

(REGENTE, ROI).

Legend

GCPs_grid

TSAs

REGENTE vertex

ROI vertex

GCPs observed

Ground truth work:


l




SRTM

ASTER-GDEM

- DEM accuracy control:

- Continuous:

- Other well-known DEM sources.

(LiDAR, airborne interferometry,

stereo-photogrammetric,

spaceborne interferometry.

-QA of the continuous sources

with discrete data will be made.

DTED2

IGN25

Intermap NEXTmap

And others: PNOA, LiDAR spots

Ground truth work:


SAR G i E l i i





Continuous:

- Other well-known DEM

sources. (LiDAR, airborne

interferometry,

stereophotogrammetric,

spaceborne interferometry)

Discrete:

- GCPs.

- Spanish geodetic reference

network vertex and marks

(REGENTE, ROI, REDNAP).

- DGPS Z-tracks.

- Vehicle tracks.

- Walk tracks.

- Slope tracks.

- Tree height measurement.

Legend

GCPs_grid

TSAs

REGENTE vertex

ROI vertex

GCPs observed

Z_TRACKS

REDNAP XY PRECISION_m

0,05

0,1

1

Ground truth work:


SAR G i E l i i





Continuous:

- Other well-known DEM

sources. (LiDAR, airborne

interferometry,

stereophotogrammetric,

spaceborne interferometry)

Discrete:

- GCPs.

- Spanish geodetic reference

network vertex and marks

(REGENTE, ROI, REDNAP).

- DGPS Z-tracks.

- Vehicle tracks.

- Walk tracks.

- Slope tracks.

- Tree height measurement.

Legend

GCPs_grid

TSAs

REGENTE vertex

ROI vertex

GCPs observed

Z_TRACKS

REDNAP XY PRECISION_m

0,05

0,1

1

Ground truth work:




Thanks for your attention.

Severino Fernández [email protected] Support tools for IMINT exploitationPatrick Vaughan [email protected] SIIRS and reference manual

Enrique Nicolás [email protected] SAR geometric exploitation

Documents

Technical Exploitation Needs for Defenc