Validation of a high-resolution (400m) SAR motion tracking system near the APLIS’07 Ice Camp M. Thomas, C. A. Geiger and C. Kambhamettu.

Validation of a high-resolution (400m) SAR motion tracking system near the APLIS’07 Ice Camp

M. Thomas, C. A. Geiger and C. Kambhamettu

Roadmap We successfully implemented a “near real time” sea ice

motion tracking system

Measure motion dynamics at ~400m resolution

Logistical aid to deploy 12 GPS (strain-rate) buoys, 1 ice mass balance buoy and 5 stress buoys.

High regions of activity in the 200km x 200km region

Used to setup the data collection location for AUVs and human divers

http://vims.cis.udel.edu/~mani/SEDNA - The “Map of Moving Topography” (MMT)

This talk discusses the initial validation of the system against GPS and ARGOS.

VIMS 12

VIMS 11

National Ice Center

GPS Buoys

Field Scientists

Analysis @ UD vims.cis.udel.edu

X

©CSA1998

Alaska Satellite Facility

System Overview

Toggle

Toggle

Toggle

GIS Data Layering Scheme

Data visualization done in Matlab

Instrument measurements (Buoy 74357)

Buoy data from Jenny Hutchings’ SEDNA buoy array

Validation against GPSNorth-South displacementEast-West displacement

Error in North-South displacementError in East-West displacement

Validation against raw ARGOSEast-West displacement North-South displacement

Error in East-West displacementError in North-South displacement

Validation against filtered ARGOSEast-West displacement North-South displacement

Error in East-West displacementError in North-South displacement

One-way repeated measure ANOVA Dependent variable: Sea ice motion

Independent variables: Motion Estimation, GPS, ARGOS, Filtered ARGOS

Cross validations at the 12 GPS Buoy locations

Results indicate no significant difference between our algorithm with GPS, ARGOS and filtered ARGOS.

Scatter plot shows the high correlation between our estimated algorithm and GPS measurements

Significance tests

Location F dF p F dF p1 0.203 1.802 0.795 1.615 1.636 0.2132 0.110 1.238 0.795 0.444 1.789 0.6223 0.866 1.634 0.408 0.939 1.167 0.3554 0.559 1.558 0.534 0.132 1.221 0.7695 0.518 1.411 0.538 0.318 1.689 0.6936 0.114 1.691 0.861 1.590 1.278 0.2207 1.316 2.260 0.277 0.277 1.895 0.7478 0.654 2.290 0.544 0.387 1.270 0.5889 0.348 1.989 0.706 1.550 1.522 0.226

10 2.762 1.608 0.085 0.300 1.224 0.63411 0.287 1.137 0.626 0.078 1.336 0.85012 0.625 1.497 0.496 0.272 1.943 0.757

Horizontal displacement Vertical displacement

F: statistic dF: degrees of Freedom P: – significance level (0.050)

Sample East-West Directions

Sample North-South Directions

Conclusions Motion tracking algorithm is within the

accuracy of the GPS data (mean vector difference 0.06 cm/s)

Specifically, ANOVA technique cross validates the accuracy of the repeated measurements at the 12 locations.

Estimation Resolution: Motion can be estimated at ~400 meters resolution,

an order of magnitude higher than typically available techniques.

Computational Requirements: Images pairs of 200km x 200km (4096 x 4096 pixels) can

be processed in under 20 minutes on a 2.93 GHz Core 2 Duo Processor PC.

Future directions Developing mechanisms to handle

discontinuities from within the motion estimation framework.

Development of a open source code initiative that could be used by researchers for sea ice monitoring.

Web based framework for remote analysis of imagery for researchers.

Acknowledgements We would like to thank our sponsor

without whom this would not have been achieved NSF Arctic Natural Science within the Office

of Polar Programs (ARC-0612105 (UD), ARC-0612527 (UAF), and ARC-0611991 (CRREL)).

We would also like to thank the Canadian Space Agency (CSA), who provided the RADARSAT-1 imagery through ASF.

Thank You!!!