MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

Improved Processing of the CASIE SAR Data

Craig Stringham and David Long Microwave Earth Remote Sensing Laboratory

Brigham Young University

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY About CASIE

Characterization of Arctic Sea Ice Experiment

When: Summer of 2009

Where: Fram Strait region

Why: Investigate how well remote sensing can detect changes in sea ice

How: Using satellite and Unmanned Aircraft System(UAS) observations

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY CASIE UAS

NASA SIERRA UAS equipped with: �  High-Res Video Camera �  Laser altimeter �  Temperature Sensors �  Pyranometers �  Spectrometers �  MicroASAR Synthetic

Aperture Radar

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

About the microASAR •  LFM-CW SAR •  Size: 22.1x18.5x4.6cm

weight: 2.5kg power: <35W

•  Pseudo-monostatic •  C-Band •  80-200 MHz Bandwidth •  90-1000m Operational

Altitude •  300-2500m Swath width

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

Image Source http://rst.gsfc.nasa.gov/Sect14/Sect14_14.html

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

Previous Images •  Processed using RDA •  1m Resolution •  ~1km ground swath •  Sparse motion data

collected

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Backprojection Introduction

�  Time domain matched filter

�  Accounts for all flight conditions

�  Inherently creates georectified images

�  Allows for sub-aperture processing

�  Computationally intensive

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Introduction to CUDA

�  Massively parallel processing

�  30 streaming multiprocessors @ 1.45 GHz �  8 single precision

processors

�  2 special function units �  1 double precision

�  16 KB shared memory

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Initial Results RDA BP

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

Motion Measurement Alignment

�  Recorded GPS synchronized by a software interrupt

�  High-precision GPS aligned to SAR data by �  Interpolating GPS data to

match the PRF

�  Fine tune using minimum entropy of a small image

time(s)

rang

e(m

)

range compressed data + gps altitude

250 300 350 400

0

50

100

150

200

250

300

350

400

450

20

40

60

80

100

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Results with Aligned GPS

RDA BP

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Estimating the System Delay

�  System Delay �  Cable delay �  RF component delay

�  Feed-through appears in dechirped data as a single sinusoid

�  Estimating the System Delay �  Isolate the feed-through

component

�  Estimate feed-through using MUSIC algorithm

time(s)

rang

e(m

)

range compressed data + gps altitude

250 300 350 400

0

50

100

150

200

250

300

350

400

450

20

40

60

80

100

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Estimating the System Delay

�  System Delay �  Cable delay �  RF component delay

�  Feed-through appears in dechirped data as a single sinusoid

�  Estimating the System Delay �  Isolate the feed-through

component

�  Estimate feed-through using MUSIC algorithm

time(s)

rang

e(m

)

range compressed data + gps altitude

250 300 350 400

0

50

100

150

200

250

300

350

400

450

20

40

60

80

100

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Altitude Offset

�  The GPS altitude measurement were highly biased

�  Altitude bias varies with altitude

�  Surface height can be estimated from nadir return

time(s)

rang

e(m

)

range compressed data + gps altitude

250 300 350 400

0

50

100

150

200

250

300

350

400

450

20

40

60

80

100

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Correcting altitude using Nadir

�  Using an initial subjective estimate of the bias select a window of RC data

Time(s)

Ran

ge(m

)

Range compressed data

250 300 350 400

100

150

200

250

300

350

400

450

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Correcting altitude using Nadir

�  Using an initial subjective estimate of the bias select a window of RC data

�  Find the maximum in that window

Time(s)

Ran

ge(m

)

Range compressed data

250 300 350 400

100

150

200

250

300

350

400

450

Windowed maximum

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Correcting altitude using Nadir

�  Using an initial subjective estimate of the bias select a window of RC data

�  Find the maximum in that window

�  Median filter

Time(s)

Ran

ge(m

)

Range compressed data

250 300 350 400

100

150

200

250

300

350

400

450

Windowed maximumMedian Filtered maximum

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Correcting altitude using Nadir

�  Using an initial subjective estimate of the bias select a window of RC data

�  Find the maximum in that window

�  Median filter

�  Correct GPS altitude using linear error model

Time(s)

Ran

ge(m

)

range compressed data + gps altitude

250 300 350 400

100

150

200

250

300

350

400

450

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY

Results with Altitude and System Delay corrections

BP without altitide correction

BP with altitude correction

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY RDA Image (old)

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Back-projected Image

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Conclusions

�  Well focused images for the CASIE SAR data were obtained using an external GPS record

�  Processing of the full data set was made possible by the GPU backprojection implementation

�  Future work should be made to make images using attitude information in the backprojection processing

MICR

OWAVE

EARTH

REMOTE SENSI N

G

BRIGHAM YOUNG UNIVERSITY Special Thanks to:

NASA

University of Colorado

Artemis

Brigham Young University