Top Banner
COMPARISON OF GEOMETRY-BASED DOPPLER COMPARISON OF GEOMETRY-BASED DOPPLER AMBIGUITY RESOLVER IN SQUINT SAR AMBIGUITY RESOLVER IN SQUINT SAR & & Wenchao Li Wenchao Li University of Electronic Science and Technology of University of Electronic Science and Technology of China China AN INDIRECT DOPPLER RATE ESTIMATION SCHEME AN INDIRECT DOPPLER RATE ESTIMATION SCHEME OF SAR IN LOW-CONTRAST SCENE OF SAR IN LOW-CONTRAST SCENE
24
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Liwenchao.ppt

COMPARISON OF GEOMETRY-BASED DOPPLER COMPARISON OF GEOMETRY-BASED DOPPLER

AMBIGUITY RESOLVER IN SQUINT SARAMBIGUITY RESOLVER IN SQUINT SAR

&&

Wenchao LiWenchao Li

University of Electronic Science and Technology of ChinaUniversity of Electronic Science and Technology of China

AN INDIRECT DOPPLER RATE ESTIMATION AN INDIRECT DOPPLER RATE ESTIMATION

SCHEME OF SAR IN LOW-CONTRAST SCENESCHEME OF SAR IN LOW-CONTRAST SCENE

Page 2: Liwenchao.ppt

IntroductionIntroduction

Synthetic Aperture Radar (SAR)

An important remote sensing tool.

• Monostatic SAR

• Bistatic SAR

• Stripmap SAR

• Scan SAR

• Spotlight SAR

• Side-looking SAR

• Squint-looking SAR

……

Page 3: Liwenchao.ppt

IntroductionIntroduction

Squint SAR

z

y

O

BR

v

x

P

Geometry model

Squint angle

Page 4: Liwenchao.ppt

IntroductionIntroduction

Squint SAR

2 22

0 0 00

cos( ) ( ) sin ( ) ( )

2 ( )

vR t R t v t t t t

R t

Range History:

The linear term is range walk

The quadratic term is range

curvature

Page 5: Liwenchao.ppt

IntroductionIntroduction

21( ) exp[ 2 ( )]

2dc drs t j f t f t

• Doppler centroid ( )

• Doppler rate ( )drfdcf

After range walk correction, the azimuth data can be modeled as a linear frequency modulated (LFM) signal,

Signal model

Page 6: Liwenchao.ppt

IntroductionIntroduction

In principle, it is possible to compute Doppler

parameter from prior information provided by GPS/INS,

but measurement uncertainties on the information will

limit the accuracy.

2 22 cosdr

vf

R

2sindc

vf

Page 7: Liwenchao.ppt

IntroductionIntroduction

An incorrect Doppler centroid • a loss of SNR;

• an increase in the azimuth ambiguity level;

• a shift in the location of the target; …...

Incorrect Doppler rate Correct Doppler rate

An incorrect Doppler rate would blur the image.

Page 8: Liwenchao.ppt

IntroductionIntroduction

Therefore, Doppler parameter Estimation from echo data is an essential procedure for high quality SAR processing

Page 9: Liwenchao.ppt

Doppler centroid estimation(Resolve Doppler centroid ambiguity)

Page 10: Liwenchao.ppt

Doppler centroid estimationDoppler centroid estimation

As azimuth data are sampled by pulse repetition

frequency (PRF), there is always Doppler centroid

ambiguity, and the bigger the squint angle is, the more

serious the doppler ambiguity will be.

Doppler centroid is usually expressed in two parts: the

baseband Doppler centroid and Doppler ambiguity

number .

,dc amb dc basef M PRF f

Page 11: Liwenchao.ppt

Doppler centroid estimationDoppler centroid estimation

• Multiple PRF

• Wavelength diversity algorithm (WDA)

• Multilook algorithms (MLCC,MLBF)

• Geometry-based algorithms

Page 12: Liwenchao.ppt

Geometry-based algorithmsGeometry-based algorithms

0

2 ( ) 2dc

t t

dR tf K

dt

The relationship between Doppler centroid and

range walk slope,

Since this method is based on the geometric

characteristics of the range compressed data, it is

named as a geometry-based Doppler ambiguity

resolver.

Then Doppler centroid can be computed via

estimating the range walk slope.

Page 13: Liwenchao.ppt

Conventional Radon transform scheme

Radon transform:

• A well-known tool for detecting

slope.

• It integrates intensity along every

possible direction in the image and

maps this information into a

feature space.

• It involves two-dimensional

search.

Geometry-based algorithmsGeometry-based algorithms

Range compressed data

Calculate the range walk slope

Output the Doppler centroid

Radon transform

Page 14: Liwenchao.ppt

Initialize the range walk slope

Integrate energy over azimuth Update the range walk slope

Range compressed data

Output the corresponding range walk slope

Output Doppler centroid

Compute waveform entropy

Has minimum entropy been obtained ?

No

Yes

Range walk correction

Iterative schemeIterative schemeThe slope is seen as a searching variable, and entropy is used to measure the robustness of signal.

It utilizes the prior information of platform, and replaces the Radon transform with a simple integration of image intensity over azimuth.

Geometry-based algorithmsGeometry-based algorithms

Page 15: Liwenchao.ppt

Range compressed data

Edge detection

Calculate the range walk slope

Output the Doppler centroid

Set transform region and step size, and do the coarse Radon transform

Update the transform region and step size, and do the fine Radon transform

Improved Radon transform schemeImproved Radon transform scheme

This scheme utilizes

prior information of

platform, convert the

data into binary data, and

conduct two-step Radon

transform, which would

make it is sparse and

reduce the need to

search the whole feature

space .

Geometry-based algorithmsGeometry-based algorithms

Page 16: Liwenchao.ppt

ComparisonsComparisonsCOMPARISONS FOR DIFFERENT DOPPLER AMBIGUITY NUMBER

Geometry-based

Doppler resolver

True

value

Estimated

value

Computation

time

Conventional Radon

transform 6 6 14.57 s

Iterative scheme 6 6 1.48 s

Improved Radon

transform scheme 6 6 0.62 s

It is obviously that they can give the accurate Doppler

ambiguity number. While the speed of the iterative

scheme and the improved Radon transform scheme is

much faster than conventional Radon transform.

Page 17: Liwenchao.ppt

Doppler rate estimation(Autofocus)

Page 18: Liwenchao.ppt

Parametric Method • Metric-based Autofocus

(Entropy/Contrast)• Map Drift (MD)• Shift and correlate (SAC)

Doppler rate estimationDoppler rate estimation

Nonparametric Method • Prominent Point Processing (PPP)• Phase Gradient Autofocus (PGA)• Time-Frequency Analysis

High contrast scene.

Page 19: Liwenchao.ppt

Low contrast scene

• Low SNR

• Without prominent point

What can we do ?

Doppler rate estimationDoppler rate estimation

Page 20: Liwenchao.ppt

Estimation SchemeEstimation Scheme

2 2

11

2 cosdr

vf

R

2 2

22

2 cosdr

vf

R

12 1

2dr dr

Rf f

R

According to the fact that Doppler rate is inversely proportional to range,

It indicates that the Doppler rate in low- contrast scene can be estimated indirectly.

Page 21: Liwenchao.ppt

Range compressed data

Choose the range bin that contains bright line

Radon transform

Output the Doppler rate of the range bin

Wigner-Ville distribution

Fit the Doppler rate of the range bin in low-contrast scene

Range walk correction

Estimation Flowchart

Doppler rate in high-contrast scene is estimated first using Radon-Wigner transform, and then using the fact that Doppler rate is inversely proportional to range, Doppler rate in low-contrast scene is estimated indirectly.

Page 22: Liwenchao.ppt

ResultsResults

High Contrast Scene With Prominent Point

Time frequency analysis Radon transform

Low Contrast Scene

Without Prominent Point

Time frequency analysis Radon transform

Page 23: Liwenchao.ppt

ConclusionsConclusions

The geometry-based scheme can be used to resolve Doppler ambiguity of SAR with large migration.

The speed of improved Radon transform scheme and the iterative scheme is about ten times faster than conventional Radon transform scheme.

An indirect estimation scheme of Doppler rate for low contrast scene is developed.

The indirect scheme is effective, and it can reduce computation time greatly.

Page 24: Liwenchao.ppt

Thank you for your listening!Thank you for your listening!