DUAL-POLARIZED COSMO SKYMED SAR DATA TO OBSERVE METALLIC TARGETS AT SEA F. Nunziata, M. Montuori and M. Migliaccio Università degli Studi di Napoli Parthenope Dipartimento per le Tecnologie Centro Direzionale, isola C4-80143 Napoli IEEE IGARSS 2011 Conference 24-29 July 2011, Vancouver, Canada
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
An important maritime surveillance task Ship detection is a key task in the frame of maritime surveillance. Traffic control. Fisheries. Oil pollution.
An important maritime surveillance task Ship detection is a key task in the frame of maritime surveillance. Traffic control. Fisheries. Oil pollution.
An important maritime surveillance task Ship detection is a key task in the frame of maritime surveillance. Traffic control. Fisheries. Oil pollution.
Traditional tools Traditionally ship monitoring is accomplished by:
Modern tools Nowadays, increasing interest is devoted to innovative technologies:
Remote sensing: SAR Synthetic Aperture Radar is the key remote sensing tool for ship observation. Global coverage. All weather. Day & night. Fine spatial details.
SAR ship observation: main issues SAR observation of metallic objects at sea is not an easy task: Main technical issues Speckle. False alarms vs. Missing targets. Complexity. Main technological issues Revisit time. Spatial resolution.
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
Single-pol approaches Several single-pol approaches have been proposed: Image-processing techniques (brightness anomalies... ). Physically-based techniques (CFAR, GK speckle model... ). A proper threshold is needed to achieve the best compromise between false alarms and missing targets rate
Multi-pol approaches The increasing number of satellite missions equipped with polarimetric SARs pushes toward multi-polarization approaches.
Multi-pol approaches Few approaches have been developed to take benefit of pol SAR data: Signal processing based: centralized and decentralized. Electromagnetically-based: Cloude-Pottier parameters, improved Coherent Target Decomposition (CTD) method, polarimetric GK, symmetry... Quite complex to be implemented. Threshold.
Our approach Both technical and technological issues can be tackled at the same time
Polarimetric SAR constellation Italian Space Agency (ASI) COSMO-SkyMed is a constellation of four satellites equipped with radar sensors of the SAR type, operating in the X band. Average revisit time in the worst case < 12h Dual-pol data are provided in PingPong mode
PingPong mode PingPong mode implements a strip acquisition by alternating a pair of TX/RX polarization across bursts by means of an antenna stirring The signal pol is alternated between two possible ones: VV, HH, HV and VH. The time offset between two successive bursts, τ B, varies, according to the beam type, between 0.15s and 0.25s.
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
PingPong mode & sea surface scattering PingPong mode acts like an along-track interferometer Two cases must be distinguished: Sea surface scattering. Sea surface scattering with a metallic target.
PingPong mode & sea surface scattering Sea surface scattering: Bragg or tilted-bragg model
PingPong mode & sea surface scattering Sea surface scattering: Bragg or tilted-bragg model Time offset between successive pulses (τ B ): 0.15-0.25s. X-band scene coherence time (τ c ): 0.01s. τ B τ c : HH-VV mostly uncorrelated over sea surface
PingPong mode & sea surface scattering τ C versus wind speed for CSK parameters and assuming a Pierson Moskowitz sea surface spectrum τ C is always lower that 0.035s
PingPong mode & sea surface scattering Metallic target scattering: High random mechanism Offset between successive pulses: 0.15-0.25s. Large X-band scene coherence time. HH-VV strongly correlated over metallic targets
In a nutshell The HH-VV correlation, r, is expected to allow enhancing the presence of metallic targets over the sea surface.
Data set COSMO-SkyMed full resolution SCS SAR data gathered in PingPong mode (HH/VV) have been ordered in the frame of Italian Space Agency (ASI) project ID 1221: SAR Remote Sensing for Sea Oil Spill Observation Spatial resolution 15 15m. Swath 30 30Km. Incidence angle 20 60 degree. Data size 1.2Gb - 600Mb HH and 600Mb VV. 5 5 average moving window to measure HH-VV correlation.
Data set Details on the CSK PingPong mode data set A very wide range of radar parameters, sea conditions and geographic areas is covered.
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
Introduction Our approach Experiments Conclusions April, 23rd 2010 Gulf of Mexico The CSK SAR data is relevant to the polluted area of Gulf of Mexico, the dark area can be recognized in the HH-polarized SAR data, together with few bright spots.
April, 23rd 2010 Gulf of Mexico The HH-VV correlation, as expected, emphasizes targets with respect to sea background.
April, 23rd 2010 Gulf of Mexico Free sea surface HH-VV correlation follows a Lognormal distribution...... a CFAR threshold can be set that calls for a 10 4 P fa.
April, 23rd 2010 Gulf of Mexico The whole data can be processed (4sec) by the HH-VV CFAR filter to provide a binary output
Outline Introduction Motivations State of art Our approach The model Experiments April 23rd, 2010 - Gulf of Mexico May 26th, 2010 - Strait of Singapore Conclusions
Introduction Our approach Experiments Conclusions May, 26th 2010 Strait of Singapore An excerpt of the HH-polarized SAR data where few targets are present is shown
May, 26th 2010 Strait of Singapore The r-image is shown. All the targets have been correctly highlighted with respect to the background sea.
May, 26th 2010 Strait of Singapore The CFAR r-filter output is shown... 137 targets are in place
Overview of the CFAR filter s performances Remarkable performances are obtained by employing two nested windows whose size are 100 100 and 5 5, respectively. Scenes Targets Detected targets False alarms Missed targets 9 305 303 1 3
Conclusions An electromagnetically-based approach to observe ships in PingPong mode HH/VV CSK SCS SAR data has been developed The model exploits the HH/VV time offset to distinguish targets from sea surface. The model has been successfully applied to observe ships on CSK SAR data. A CFAR approach has been conceived to provide a binary output. The filter is very computer-time effective (4s to process a full data). All data used in this study are COSMO-SkyMed products - c ASI - Agenzia Spaziale Italiana - 2010. All rights reserved.