analysis of polsar maritime data - defence …cradpdf.drdc-rddc.gc.ca/pdfs/unc71/p529331.pdf ·...

Analysis of PolSAR maritime data

Garrett Parsons, Craig Williams, and Martin St-Hilaire

The scientific or technical validity of this Contract Report is entirely the responsibility of the Contractor and the contents do not necessarily have the approval or endorsement of Defence R&D Canada.

Defence R&D Canada --- Ottawa CONTRACT REPORT

DRDC Ottawa CR 2008-067 April 2008

Analysis of PolSAR maritime data

Garrett Parsons, Craig Williams, and Martin St-Hilaire Prepared By: Vantage Point International Inc. 400 March Road, suite 210 Ottawa, Ontario, CA K2K 3H4 Contract Number: W7714-06-0989/001/SV CSA: Chen Liu, DRDC Ottawa, (613) 993-8381 Nicholas Sandirasegaram, DRDC Ottawa, (613) 998 - 2057

The scientific or technical validity of this Contract Report is entirely the responsibility of the Contractor and the contents do not necessarily have the approval or endorsement of Defence R&D Canada.

Defence R&D Canada – Ottawa Contract Report DRDC Ottawa CR 2008-067 April 2008

Contract Scientific Authority

Original signed by Nicholas Sandirasegaram

Nicholas Sandirasegaram

Defence Scientist

Approved by

Original signed by Gary Geling

Gary Geling

Head, Radar Applications and Space Technologies

Approved for release by

Original signed by Pierre Lavoie

Pierre Lavoie

Chair, Document Review Panel

© Her Majesty the Queen in Right of Canada, as represented by the Minister of National Defence, 2008

© Sa Majesté la Reine (en droit du Canada), telle que représentée par le ministre de la Défense nationale, 2008

DRDC Ottawa CR 2008-067 i

Abstract

This document summarizes the analysis of polarimetric SAR data of maritime vessels collected by the CV-580 aircraft for the Radar Application and Space Technology Section (RAST) of Defence R&D Canada – Ottawa (DRDC Ottawa). This analysis was conducted to support polarimetric SAR research in the area of target classification and target discrimination.

The work involved a search of recent literature as well as selecting suitable software tools to generate the desired polarimetric decompositions. The selected software are Polarimetric Workstation software tool, PolSARPRO software tool, DRDC Ottawa MATLAB code and VPI MATLAB code. The analysis techniques are correlation and covariance between polarimetric channels, synthesis images, mean clutter, peak of target and target peak-to-clutter ratio, Symmetric Scattering Characterization Method (SSCM), decomposition of Pauli , Cameron, Krogager, Entropy/Alpha and Freeman with Entropy/Alpha, and ratios of HH/HV, VV/VH, and HH/VV. These techniques are described in detail and a sample processed data set is presented in the report, while the CD-ROM available with the scientific authority includes results from all other analysed vessels.

Résumé

Le présent document résume l’analyse des données de radar à synthèse d'ouverture (SAR) polarimétrique sur des navires maritimes, recueillies à bord de l’avion CV-580 pour la section des applications radar et de la technologie spatiale (ARTS) de R & D pour la défense Canada - Ottawa (RDDC Ottawa). Cette analyse a été menée à l’appui de la recherche par SAR polarimétrique dans le domaine de la classification et de la discrimination des cibles.

Les travaux comportaient une recherche de documentation récente ainsi que la sélection des outils logiciels convenant aux décompositions polarimétriques désirées. Les logiciels sélectionnés étaient l’outil de poste de travail polarimétrique, l’outil logiciel PolSARPRO, le code MATLAB de RDDC Ottawa et le code MATLAB de VPI. Les techniques d’analyse faisaient appel à la corrélation et à la covariance entre canaux polarimétriques, à des images de synthèse, à la moyenne de clutter, à la crête de cible et au rapport de la crête de cible au clutter, à la méthode de caractérisation de diffusion symétrique (SSCM), aux décompositions de Pauli , Cameron et Krogager, aux techniques d’entropie/alpha et Freeman avec entropie/alpha, ainsi qu’aux rapports HH/HV, VV/VH et HH/VV. Ces techniques sont décrites en détail et un ensemble échantillon de données traitées est présenté dans le rapport, alors que le CD-ROM disponible auprès de l’autorité scientifique comprend les résultats de tous les autres navires analysés.

ii DRDC Ottawa CR 2008-067

This page intentionally left blank.

DRDC Ottawa CR 2008-067 iii

Executive Summary

Analysis of PolSAR maritime data Garrett Parsons, Craig Williams, and Martin St-Hilaire; DRDC Ottawa CR 2008-067; Defence R&D Canada – Ottawa; April 2008.

Introduction: This document summarizes the analysis of polarimetric SAR data of maritime vessels collected by the CV-580 aircraft for the Radar Application and Space Technology Section (RAST) of Defence R&D Canada – Ottawa (DRDC Ottawa). This analysis was conducted to support polarimetric SAR research in the area of target classification and target discrimination.

Results: Vantage Point International Inc. (VPI) has been contracted by DRDC Ottawa to assist in polarimetric research and data processing of CV-580 data. This work has involved conducting a short literature search and working with the DRDC Ottawa Scientific Authority to propose polarimetric processing techniques for analysis of the CV-580 vessel data set. The following set of polarimetric measurements were agreed upon:

• Correlation and covariance between polarimetric channels • Synthesis images • Mean clutter, peak of target and target peak-to-clutter ratio • Pauli, Cameron, Krogager Decompositions, and SSCM • HH/HV, VV/VH, HH/VV ratios • Entropy/Alpha and Freeman + Entropy/Alpha

VPI used the Polarimetric Workstation (PWS) developed by the Canadian Centre for Remote Sensing (CCRS), PolSAR Pro supported by the European Space Agency (ESA), as well as MATLAB code supplied by DRDC and developed by VPI to process as many vessel dataset possible and supply these results to DRDC Ottawa in an electronic format suitable for further processing such as classification and discrimination.

A total of 46 ship data were processed. This report contains a detailed description of each measurement for the 342_Toronto dataset. These measurements are also reported in a CD-ROM available with the scientific authority which contains report for the rest of the 45 datasets.

Significance: The processed dataset contains results for different polarimetric classifications for a variety of ship classes and these classes demonstrate some common dominating features or scattering mechanisms. It is believed that this type of data is essential for unsupervised classification of ships as well as discrimination between ships and icebergs. There are also very few polarimetric datasets and studies done on this type of data, thus this dataset can confirm a lot of the previous assumption about how polarimetry can aid in classification and discrimination exercises.

iv DRDC Ottawa CR 2008-067

Sommaire .....

Analysis of PolSAR maritime data Garrett Parsons, Craig Williams, and Martin St-Hilaire; DRDC Ottawa CR 2008-067; R et D pour la défense Canada – Ottawa; April 2008.

Introduction : Le présent document résume l’analyse des données de radar à synthèse d'ouverture (SAR) polarimétrique sur des navires maritimes, recueillies à bord de l’avion CV-580 pour la section des applications radar et de la technologie spatiale (ARTS) de R & D pour la défense Canada - Ottawa (RDDC Ottawa). Cette analyse a été menée à l’appui de la recherche par SAR polarimétrique dans le domaine de la classification et de la discrimination des cibles.

Résultats : Vantage Point International Inc. (VPI) a reçu un contrat de RDDC Ottawa en vue d’aider à la recherche polarimétrique et au traitement des données du CV-580. Les travaux comportaient une courte recherche de documentation ainsi que des travaux avec l’autorité scientifique de RDDC Ottawa en vue de la proposition de techniques de traitement polarimétrique pour l’analyse de l’ensemble de données sur les navires du CV-580. Une entente a été conclue au sujet de techniques de mesure polarimétrique faisant appel à ce qui suit :

• Corrélation et covariance entre canaux polarimétriques. • Images de synthèse. • Moyenne de clutter, crête de cible et rapport de la crête de cible au clutter. • Décompositions de Pauli, Cameron et Krogager, et méthode de caractérisation de

diffusion symétrique (SSCM). • Rapports HH/HV, VV/VH et HH/VV. • Entropie/alpha et Freeman + entropie/alpha

VPI a utilisé l’outil logiciel du poste de travail polarimétrique mis au point par le Centre canadien de télédétection (CCT), PolSAR Pro pris en charge par l’Agence spatiale européenne (ASE), ainsi que le code MATLAB fourni par RDDC et développé par VPI pour traiter autant d’ensembles de données que possible sur les navires et pour fournir les résultats à RDDC Ottawa sous une forme électronique convenant au traitement ultérieur, comme la classification et la discrimination.

Des données portant sur un total de 46 navires ont été traitées. Le présent rapport renferme une description détaillée de chaque mesure relative à l’ensemble de données 342_Toronto. Ces mesures sont également disponibles sur un CD-ROM disponible auprès de l’autorité scientifique et contenant aussi les 45 autres ensembles de données.

Portée : L’ensemble des données traitées contient les résultats de différentes classifications polarimétriques pour diverses classes de navires, et ces classes démontrent quelques caractéristiques ou mécanismes de diffusion prédominants communs. On croit que ce type de données est essentiel à la classification non supervisée des navires ainsi qu’à la discrimination entre navires et icebergs. Il n’existe également que très peu d’ensembles de données polarimétriques et d’études sur ce type de données, de sorte que cet ensemble de données peut confirmer beaucoup d’hypothèses antérieures sur la façon dont la polarimétrie se prête aux exercices de classification et de discrimination.

DRDC Ottawa CR 2008-067 v

This page intentionally left blank.

vi DRDC Ottawa CR 2008-067

Table of contents

Abstract……….. .............................................................................................................................. i Résumé……….. ...........................................…………………………………………….…………i Executive Summary........................................................................................................................ iii Sommaire ....................................................................................................................................... iv Table of contents ............................................................................................................................ vi List of figures ................................................................................................................................ vii List of tables ................................................................................................................................... ix 1....Litterature Search...................................................................................................................... 1

1.1 Ship Detection and Classification.................................................................................. 1 2....Requested Measurements ......................................................................................................... 3

2.1 Propositions/Discussion ................................................................................................ 3 3....Software Recommendations ..................................................................................................... 5 4....PolSAR Data Analysis.............................................................................................................. 7

4.1 Correlation between Polarimetric Channels .................................................................. 7 4.2 Covariance between Polarimetric Channels ................................................................ 11 4.3 Synthesis Image........................................................................................................... 11 4.4 Mean Clutter, Peak of targets and Target Peak-to-Clutter .......................................... 14

4.4.1 Evaluation of Clutter Mean........................................................................... 14 4.4.2 Target Detection Threshold........................................................................... 17 4.4.3 Target Peak Amplitude ................................................................................. 18 4.4.4 Target Peak-to-Clutter (PCR) ....................................................................... 18 4.4.5 Results ........................................................................................................... 18

4.5 Pauli Decomposition ................................................................................................... 25 4.6 Cameron Decomposition ............................................................................................. 27 4.7 Krogager Decomposition............................................................................................. 29 4.8 Symmetric Scattering Characterization Method (SSCM) ........................................... 30 4.9 HH/HV, VV/VH, HH/VV Ratios ................................................................................ 33 4.10 Entropy/Alpha ............................................................................................................. 35 4.11 Freeman + Entropy/Alpha ........................................................................................... 36

5....Conclusions............................................................................................................................. 40 References ..... ............................................................................................................................... 41 Annex A .. Processing Steps .......................................................................................................... 43 List of symbols/abbreviations/acronyms/initialisms ..................................................................... 73 Distribution list.............................................................................................................................. 74

DRDC Ottawa CR 2008-067 vii

List of figures

Figure 1 HH-HV Gamma Coefficient for Toronto, 17 Oct 2005, l22p2......................................... 7 Figure 2 HH-HV Phase Difference for Toronto, 17 Oct 2005, l22p2 ............................................. 8 Figure 3 HH-VV Gamma Coefficient for Toronto, 17 Oct 2005, l22p2......................................... 8 Figure 4 HH-VV Phase Difference Image for Toronto, 17 Oct 2005, l22p2 .................................. 9 Figure 5 HV-VV Gamma Coefficient Image for Toronto, 17 Oct 2005, l22p2 .............................. 9 Figure 6 HV-VV Phase difference Image for Toronto, 17 Oct 2005, l22p2 ................................. 10 Figure 7 LV Synthesis Image ........................................................................................................ 12 Figure 8 LH Synthesis Image ........................................................................................................ 12 Figure 9 HH Synthesis Image........................................................................................................ 13 Figure 10 HV Synthesis Image...................................................................................................... 13 Figure 11 VV Synthesis Image...................................................................................................... 14 Figure 12 HH Clutter Analyses ..................................................................................................... 15 Figure 13 HV Clutter Analyses ..................................................................................................... 15 Figure 14 VH Clutter Analyses ..................................................................................................... 16 Figure 15 VV Clutter Analysis...................................................................................................... 16 Figure 16 HH Linear Polarization Image ...................................................................................... 18 Figure 17 HH Target Mask............................................................................................................ 19 Figure 18 HV Linear Polarization Image ...................................................................................... 19 Figure 19 HV Target Mask............................................................................................................ 20 Figure 20 VH Linear Polarization Image ...................................................................................... 20 Figure 21 VH Target Mask............................................................................................................ 21 Figure 22 VV Linear Polarization Image ...................................................................................... 21 Figure 23 VV Target Mask............................................................................................................ 22 Figure 24 Final Target Mask ......................................................................................................... 22 Figure 25 RR Circular Polarization Image .................................................................................... 23 Figure 26 LL Circular Polarization Image .................................................................................... 23 Figure 27 RL Circular Polarization ............................................................................................... 24 Figure 28 Pauli Decomposition with a “winner takes all” scheme ............................................... 26 Figure 29 Pauli Decomposition Histogram ................................................................................... 26 Figure 30 Cameron Decomposition Image.................................................................................... 27

viii DRDC Ottawa CR 2008-067

Figure 31Cameron Decomposition Histogram.............................................................................. 28 Figure 32 Krogager Decomposition Image ................................................................................... 29 Figure 33 Krogager Decomposition Histogram ............................................................................ 30 Figure 34 Target Poincare Latitude ............................................................................................... 31 Figure 35 Target Poincare Longitude............................................................................................ 31 Figure 36 SSCM Classifications ................................................................................................... 32 Figure 37 SSCM Classification Histogram ................................................................................... 33 Figure 38 HH/HV Ratio Image ..................................................................................................... 34 Figure 39 VV/VH Ratio Image ..................................................................................................... 34 Figure 40 HH/VV Ratio Image ..................................................................................................... 35 Figure 41 H/alpha Histogram ........................................................................................................ 36 Figure 42 Freeman Decomposition ............................................................................................... 37 Figure 43 Freeman Decomposition Histogram ............................................................................. 38 Figure 44 H/alpha analysis for each Freeman Categories ............................................................. 38 Figure 45 Combined H/alpha and Freeman Decomposition ......................................................... 39

DRDC Ottawa CR 2008-067 ix

List of tables

Table 1 Linear Polarization Correlation ........................................................................................ 10 Table 2 Circular Polarization Correlation ..................................................................................... 10 Table 3 Linear Polarization Covariance ........................................................................................ 11 Table 4 Circular Polarization Covariance ..................................................................................... 11 Table 5 Image Synthesis Parameters ............................................................................................. 11 Table 6 Linear Polarization Target Peak-to-Clutter ...................................................................... 24 Table 7 Circular Polarization Target Peak-to-Clutter.................................................................... 24 Table 8 Synthesised Polarization Target Peak-to-Clutter.............................................................. 25 Table 9 Pauli Decomposition Statistics ......................................................................................... 25 Table 10 Cameron Decomposition Pixel Counts .......................................................................... 28 Table 11 Krogager Decomposition Pixel Counts .......................................................................... 30 Table 12 SSCM Histogram values ................................................................................................ 32 Table 13 Freeman Decomposition Statistics ................................................................................. 37

DRDC Ottawa CR 2008-067 1

1 Litterature Search

1.1 Ship Detection and Classification

Early publications such as that by Touzi [TOUZR1999] proposed that polarimetry should improve ship detection and classification. It indicates that HH is best among classical polarizations (HH, VV and HV). It also suggests that polarization entropy is promising for detection at incidence angles lower than 60 degrees. His latter publications [TOUZR2000, TOUZR2004] will improve those observations by saying that among classical polarizations, HV provides best contrast at small incidence angles (<50 degrees) and that HH is better otherwise. The contrast as a function of incidence angle was found to behave similarly in a study by Olsen et al. [OLSER2004] using only non-calibrated ENVISAT Alternate Polarisation data. Touzi continues by presenting polarization anisotropy (closely linked to entropy) as a feature that maximizes contrast for detection at small incidence angle [TOUZR2004]. Touzi then introduces the Symmetric Scattering Characterization Method (SSCM) as an improvement over the Cameron decomposition that may allow to obtain useful ship feature such as pitch angle.

Other early studies, such as that by Ringrose and Harris [RINGR2000] used the Cameron decomposition to demonstrate that radar scattering by ship target mostly behave like dihedrals, narrow dihedrals and ¼ wave devices whereas the ocean behaved primarily like cylinder type scatterers. Their study was done using SIR-C data.

The work by Yeremy et al [YEREM2002] indicates that the Cameron decomposition offered high performance, at least for detection purposes and false target discrimination. Their measurements were applied to C/X SAR data.

Haiyan et al [HAIYL2006] used fully polarimetric simulated data as well as SIR-C data and proposed that the convolution between polarimetric channels may help at least for ship detection. Convolution (or correlation) of channels was shown to be sensitive to the size and shape of the ship and less to sea state.

The Polarimetric Gaussian White Filter was used by Liu et al [LIUC2005] and was shown to improve the detection of ship targets when compared to partially polarized (with or without coherent phase) and single polarized data. More recent [LIUC2007] work have shown that target decomposition techniques including the Pauli decomposition, Cameron decomposition and SSCM can improve ship classification.

Recent work by Staples et al [STAPG2007] aimed at evaluating the performance of polarimetric classification as a function of the aspect angle of imaging. The goal was to find scattering mechanisms that were aspect angle independent. It was found that, among SSCM fundamental mechanism, dipole and narrow diplane appeared invariant to look direction whereas trihedral and dihedral have a pronounced look direction dependency.

The modeling work by Margarit et al. [MARGG2006a, MARGG2006b] emphasizes the difficulties of classifying ship using polarimetric SAR data. In particular, Margarit explains that the ship response to polarimetric SAR is highly dependant over the measurement aspect angle.

2 DRDC Ottawa CR 2008-067

A quick review of a new emerging technique named Compact Polarimetry (also related to Hybrid Polarimetry) has revealed that it is a “simplified” polarimetric mode that is aimed at alleviating PRF and data volume limitations linked with full polarimetric SAR [SOUYJ2005, DUBOP2007]. By transmitting a single linear polarization oriented at 45 degrees (or alternatively a circular polarized wave) at all pulses and receiving using the 2 conventional H and V polarizations, most of the full polarimetric information can be retrieved under certain assumptions. There might be an interest to see how Compact Polarimetry performs for ship classification.

1.2 Iceberg/Ship Discrimination

When icebergs are comparable in size to ships, single polarized SAR platform can hardly discriminate the two targets. Full polarimetric data, with its ability to decompose the radar return into different scattering mechanism is expected to improve the discrimination. Few studies have been done using fully polarimetric data for ship/iceberg discrimination, apparently because of the lack of data. Howell et al. have done it using a significant amount of data, first with alternate polarization data and more recently using fully polarimetric decomposition along with classification schemes to separate the two types of targets.

Howell et al. [HowellC2004] explained that ships tend to have more dihedral and trihedral surface scattering whereas icebergs tend to have a combined surface and volume scattering mechanism. They have found that a combination of the HH and HV channels offer good discrimination between iceberg and ship target because of the different response in both channels. While ships and icebergs both appear with good contrast in the HH channel, only ships have significant contrast in the HV channel. In practice the arithmetic ratio between the HH and HV channel is used to separate the two targets. As well, the HV signal-to-clutter ration (SCR) offers good classification performance.

In a more recent piece of work [HoweC2006, HOWEC2007], Howell et al. have introduced a method using multivariate classification tools such as the maximum likelihood Gaussian classifier to discriminate the two target types from multiple polarimetric measurements with very promising results. In particular, classification using both the mean HV and maximum VV return from identified targets allowed perfect discrimination over their data set.


2 Requested Measurements

The list of measurements below was part of the requirements for the present project. The following are to be computed for the datasets of interest. Those that are readily implemented in commercial/free software will be listed as such in the software evaluation section.

• Correlations between polarimetric channels

• Covariance between polarimetric channels

• Synthesis Images

• Mean Clutter, Peak of Targets and Target Peak-to-Clutter ratio

• Pauli Decomposition

• Cameron Decomposition

• Krogager Decomposition

• SSCM

• HH/HV, VV/VH, HH/VV ratios

• Entropy/Alpha

• Freeman + Entropy/Alpha

2.1 Propositions/Discussion

Based on the recent literature search and more specifically on the difficulties raised by the studies by Staples et al [STAPG2007] and Margarit et al [MARGG2006a, MARGG2006b] , we recommend that the investigation be based on a more practical approach. The identification of ship targets appears highly dependant on the ship heading and resolving this issue may alleviate the problem of classification. There is a need for a measure that is aspect angle independent for reliable identification of targets.

Based on the work by Staples et al, data sets of ships acquired at multiple aspect angles exhibit clear fundamental scattering mechanisms at specific angles. Similar studies should help improve classification at other, less than ideal, aspect angles. If there is a possibility to obtain simulated data from wire mesh models, this avenue should be pursued.


One possible way of classifying ships is to look for major structures of the ship that repeat periodically such as equipment found on defence ships (masts, weapons, cranes etc.) Those structures should be large enough and be separated by distances greater than the image resolution.

Ship and Iceberg discrimination appears to be somewhat simpler as both target types are characterized by distinct scattering mechanisms. Recent results by Howell et al. have shown that as few as two parameters provided very promising discrimination performances.

Although such multivariate methods for supervised classification have been successful in some cases, we recommend that caution be taken when applying such methods with large amounts of inputs. Almost anything can be classified when the number of inputs becomes really high which can make the result irrelevant.

The set of polarimetric measurements listed in the previous section appears justified based on the literature. Among the listed ones, we do not expect that compact polarimetry will improve ship classification or discrimination with iceberg although we recognize the interest to investigate its performance for such applications.

Finally, we want to stress the large amount of measurement requested for the upcoming tasks. That combined with the fact that not one software package was found that performed all requested measurements (see next section) might reduce the total amount of data that we will be able to process for the time allowed. There will be a need to define precisely the method to be used for each task and to evaluate the time required to apply it.


3 Software Recommendations

VPI holds a license for the Polarimetric Workstation (PWS) obtained from CCRS in 2002. PWS is a robust piece of software written in MATLAB that has been used by VPI for various projects. Among the data formats supported by PWS, the CV-580 is of interest for the present research and the functions of interest are:

• Individual channel and RGB (HH,HV,VV) display

• Channel coherence plots (magnitude and phase difference) for pixels within selected polygons

• Polarization Synthesis (amplitude only)

• Channel Phase Difference

• SSCM

The other functions available in PWS are:

• Schwartz Contrast

• Channel Phase

• Poincare Sphere

• Phase difference

The PolSARPRO software is freely distributed and supported by the European Space Agency (ESA) that is widely used in the scientific community. It is mostly written in Python. Among the data formats supported by PolSARPRO, CV-580 and PALSAR are of interest for the present research and the functions of interest are:

• Correlation Coefficients

• Polarimetric White Filter (PWF)



• Cameron Decomposition

• Compact Polarisation Mode


Although POLSARPRO does not possess the most intuitive interface, it has proven the most complete of the three programs considered. Its stability is comparable to that of the polarimetric workstation and much better than the RAT package (see below).

RAT is a radar data analysis package developed at the University of BERLIN that is freely available. It is written in IDL and thus requires the freely available IDL virtual machine. RAT supports both PALSAR and CV-580 data. RAT provides a different set of tools when compared to the other two applications. Its interface is much more intuitive to use but its installation and use appeared buggy, at least with the CV-580 test data. The tools of interest available in RAT are:

• Channel Phase Difference

• Channel Correlation

• Channel Amplitude Ratio



POLSARPRO was found to be the most complete software package of the three. The very versatile RAT package was somewhat simpler to use but appeared buggy. The polarimetric workstation appears to be the most robust tool but with limited capabilities. The analysis was conducted using Polarimetric Workstation and PolSARPRO software tools. Other software will be needed to complement the work.

DRDC will supply some MATLAB code with specific capabilities to augment the above applications. The DRDC software will provide the correlation a covariance values. This software will also use the SSCM results of PWS and do a classification of the SSCM results.

Some custom MATLAB code was also written to compliment the above software packages. The code generates a target mask to apply to the results of the decompositions done in PWS and PolSARPRO. A Pauli decomposition was also written using a “winner takes all” approach to classify a pixel.


4 PolSAR Data Analysis

In this section the results of the analysis on the 342_Toronto a data set is presented.

4.1 Correlation between Polarimetric Channels

The correlation is done using the PWS using the “square sample” method. In this case, the window size has roughly 10 azimuth pixels by 1 range pixel. The result of the correlation is shown in Figure 1 - Figure 6 showing the amplitude and the phase of the correlation operation.

Figure 1 HH-HV Gamma Coefficient for Toronto, 17 Oct 2005, l22p2


Figure 2 HH-HV Phase Difference for Toronto, 17 Oct 2005, l22p2

Figure 3 HH-VV Gamma Coefficient for Toronto, 17 Oct 2005, l22p2


Figure 4 HH-VV Phase Difference Image for Toronto, 17 Oct 2005, l22p2

Figure 5 HV-VV Gamma Coefficient Image for Toronto, 17 Oct 2005, l22p2


Figure 6 HV-VV Phase difference Image for Toronto, 17 Oct 2005, l22p2

DRDC software was used to generate the correlations between the linear channels. The results are shown in Table 1 and Table 2.

Table 1 Linear Polarization Correlation

HH HV VH VV

HH 1 0.20437 0.19582 0.43459HV 0.20437 1 0.97664 0.15645VH 0.19582 0.97664 1 0.1477VV 0.43459 0.15645 0.1477 1

Table 2 Circular Polarization Correlation

LL RR LR

LL 1 0.71427 0.44595RR 0.71427 1 0.50673LR 0.44595 0.50673 1


4.2 Covariance between Polarimetric Channels

DRDC software was used to generate the covariance between the linear channels and between the circular polarizations, and is shown in Table 3 and Table 4.

Table 3 Linear Polarization Covariance

HH HV VH VV

HH 1.5412 0.066075 0.057994 0.40099 HV 0.066075 0.067825 0.060677 0.030283 VH 0.057994 0.060677 0.05691 0.026188 VV 0.40099 0.030283 0.026188 0.55239

Table 4 Circular Polarization Covariance

RR LL LR

RR 0.38882 0.27771 0.23583LL 0.27771 0.38878 0.26796LR 0.23583 0.26796 0.71925

4.3 Synthesis Image

Using PWS the HH header file for L22P2 was opened. The image synthesis was applied using the full resolution. Images were synthesized for modes specified in Table 5. The synthesized data was read by a MATLAB program and plotted in the figures below. Figure 7 - Figure 11, show the results of the image synthesis.

Table 5 Image Synthesis Parameters

LH LV VV HH HV

Transmission Ellipticity 45 45 0 0 0

Transmission Orientation 0 0 90 0 0

Reception Ellipticity 0 0 0 0 0

Reception Orientation 0 90 90 0 90


Figure 7 LV Synthesis Image

Figure 8 LH Synthesis Image


Figure 9 HH Synthesis Image

Figure 10 HV Synthesis Image


Figure 11 VV Synthesis Image

4.4 Mean Clutter, Peak of targets and Target Peak-to-Clutter

4.4.1 Evaluation of Clutter Mean

The clutter mean is evaluated by averaging the amplitude or pixels that are known to be clutter. This is done by selecting pixels from the four corners of each dataset, which is far form the target (located in the middle). Figure 12 - Figure 15 illustrates the statistical significance of the average as a function of the number of pixels selected from the corners for each of the HH, HV, VH and VV channels. The results from this exercise using 3600 points indicate that this amount of data provides for sufficient statistical stability. One obvious method of selecting that particular number of pixels is to compute the average from rectangular samples of size 30 range samples by 300 azimuth samples in the four corners. Considering the size of the datasets (128 range samples by 2048 azimuth samples) our selection will be safely far enough of the target pixels. This analysis was performed using MATLAB code developed by VPI.


Figure 12 HH Clutter Analyses

Figure 13 HV Clutter Analyses


Figure 14 VH Clutter Analyses

Figure 15 VV Clutter Analysis


Note that the amplitude associated with a single pixel is evaluating from the real and imaginary part of each sample using the following equation:

22imagreal HHHHHH +=

and the associated amplitude value expressed in dB can be obtained using:

( )HHHHdB 10log20=

The clutter mean found using 30 range by 300 azimuth samples from each corner, thus giving 36000 samples, are listed in Table 6, Table 7, and Table 8.

4.4.2 Target Detection Threshold

A threshold was established based on the mean clutter and a constant number of standard deviations to select pixels. Pixels with values above the threshold are determined to be part of the target. Through trial and error a constant was found for each image. The constant is listed in Table 6. Note that this threshold is evaluated in absolute amplitude values and not in dBs. All pixels whose amplitude is above that threshold are identified as target return. A mask is thus created using these pixels. The target mask obtained for each image is shown in this section for the 342 Toronto (17 Oct 2005, l22p2) dataset. This analysis was performed using MATLAB code developed by VPI.

Clutter Threshold = Clutter Mean + Threshold Constant * Clutter Standard Deviation

A global target mask is then created by combining each of the masks generated from the linear polarizations (HH, HV, VH, VV). The global mask is then use in the analysis.

Global Mask = mask (HH) | mask (HV) | mask (VH) | mask(VV)


4.4.3 Target Peak Amplitude

The peak target amplitude for each image was done by finding the maximum value within the target mask, using MATLAB code developed by VPI.

4.4.4 Target Peak-to-Clutter (PCR)

A mean clutter value is determined for each image using the above method. An area in each corner of 20 range by 200 azimuth samples is used to determine the clutter mean. The target peak-to-clutter ratio is calculated using the formula

PCR = 20 * log(Target Peak) – 20 * log (Clutter Mean)

4.4.5 Results

Figure 16 HH Linear Polarization Image


Figure 17 HH Target Mask

Figure 18 HV Linear Polarization Image


Figure 19 HV Target Mask

Figure 20 VH Linear Polarization Image


Figure 21 VH Target Mask

Figure 22 VV Linear Polarization Image


Figure 23 VV Target Mask

Figure 24 Final Target Mask


Figure 25 RR Circular Polarization Image

Figure 26 LL Circular Polarization Image


Figure 27 RL Circular Polarization

Table 6 Linear Polarization Target Peak-to-Clutter

Threshold

Constant

Threshold

(dB)

Target Peak

(dB)

Clutter Mean

(dB)

Peak-to-

Clutter (dB)

HH 15 -14.4396 16.8021 -33.4188 50.2209HV 10 -23.4113 4.7422 -39.3233 44.0655VH 10 -24.5751 4.1898 -40.4484 44.6382VV 10 -13.3272 13.3025 -29.2822 42.5846

Table 7 Circular Polarization Target Peak-to-Clutter

Target Peak (dB) Clutter Mean (dB) Peak-to-Clutter (dB)

RR 23.26 -67.6805 90.9405LL 22.9028 -68.0042 90.9071RL 29.5543 -62.7461 92.3004


Table 8 Synthesised Polarization Target Peak-to-Clutter

Target Peak (dB) Clutter Mean (dB) Peak-to-Clutter (dB)

LH 27.6964 -69.4448 97.1412LV 20.618 -62.0501 82.668HH 33.6042 -64.7155 98.3197HV 8.3249 -80.9376 89.2625VV 26.6049 -56.4306 83.0355

4.5 Pauli Decomposition

The Pauli decomposition was calculated by applying the following equations to generate the red, green and blue image channels.

R = |HH - VV| Double bounce scattering G = |HV + VH| Volume scattering B = |HH + VV| Single bounce scattering

A “winner takes all” scheme is applied to each pixel. The scattering mechanism with the highest coefficient is associated with the pixel, and applying the target mask, Figure 28 is generated. A mask is applied to the image, which is a combination of all the masks generated from the HH, HV, VH and VV channels. A histogram of the Pauli decomposition can be seen in Figure 29. The Pauli decompositions statistics can be found in Table 9. This analysis was performed using MATLAB code developed by VPI.

Table 9 Pauli Decomposition Statistics

Double Volume Single

1609 423 3284


Figure 28 Pauli Decomposition with a “winner takes all” scheme

Figure 29 Pauli Decomposition Histogram


4.6 Cameron Decomposition

The Cameron decomposition is made using the dedicated function found in PolSARPRO. The PolSARPRO function outputs the data in a binary format file that is read by a MATLAB script along with the colour palette used by PolSARPRO. The script first applies the mask found in section 4.4 over the data and then displays the result by associated the correct colour to each of the pixels indices. Figure 30 presents the result of the Cameron decomposition over the 342 Toronto dataset. The figure shows the different scattering mechanism associated with each of the target pixel. Figure 31 shows a histogram of occurrence of each scattering mechanism for all of the target pixels. The pixel counts can be found in Table 10.

Figure 30 Cameron Decomposition Image


Figure 31Cameron Decomposition Histogram

Table 10 Cameron Decomposition Pixel Counts

Scattering Type Pixel CountTrihedral 395Diplane 208Dipole 1165Cylinder 668Narrow Diplane 1503¼ Wave Device 1053Left Helix 166Right Helix 158


4.7 Krogager Decomposition

The Krogager decomposition was applied with the PolSARPRO application. The CV-580 data was read by the application. The Krogager decomposition was done with a window size of 1. The resulting decomposition was read by a MATLAB function written for this analysis. The MATLAB function generates a target global mask using the method discussed in section 4.4. The target mask is applied to the Krogager decomposition. The resulting image can be seen in Figure 32. A histogram, Figure 33, is generated from the target pixels. The pixel counts are shown in Table 11.

Figure 32 Krogager Decomposition Image


Figure 33 Krogager Decomposition Histogram

Table 11 Krogager Decomposition Pixel Counts

Diplane Helix Sphere

1481 117 3718

4.8 Symmetric Scattering Characterization Method (SSCM)

The SSCM analysis was done with the PWS. Due to a bug in the application the image area was reduced by 10 rows and 1 column. The offset was applied to the start of the image. Other than the offset the analysis was done on the full resolution image. The SSCM analysis was performed with a coherence threshold of 0.86, a square window size of 9 and a radiometric Rician threshold of 18dB. PWS generate two images as a result of the analysis, an image of the Poincare latitude, Figure 34, and of the Poincare longitude, Figure 35.


Figure 34 Target Poincare Latitude

Figure 35 Target Poincare Longitude


The images output from PWS, are then read by the DRDC software to generate a classification for the SSCM decomposition, shown in Figure 36. A histogram of the classifications is shown in Figure 37. The histogram values are listed in Table 12.

Table 12 SSCM Histogram values

Trihedral 0.10908Diplane 0.013004Dipole 0.22399Cylinder 0.29671Narrow Diplane 0.0918261/4 wave device 0.26539

Figure 36 SSCM Classifications


Figure 37 SSCM Classification Histogram

4.9 HH/HV, VV/VH, HH/VV Ratios

Ratio images were generated within MATLAB. First the CV-580 data for each channel (HH, HV, VH, VV) was read. An element-by-element division of the complex values for each pixel was performed to generate the ratio data sets. The following figures were generated using the log of the absolute value for each pixel.


Figure 38 HH/HV Ratio Image

Figure 39 VV/VH Ratio Image


Figure 40 HH/VV Ratio Image

4.10 Entropy/Alpha

The Entropy/Alpha (H/α) decomposition was performed in PWS using a window size of 7x7 (range and azimuth). The result is plotted in the H/α plane shown in Figure 41.


Figure 41 H/alpha Histogram

4.11 Freeman + Entropy/Alpha

Both the Freeman decomposition and the H/α decomposition were processed in PolSARPRO using a window size of 7 in both cases. Then, both results were loaded in MATLAB. The result of the Freeman decomposition of all targets pixels found using the target mask from section 4.4 is shown in Figure 42 along with a histogram, Figure 43, of the decomposition. For each of the Freeman categories, a H/α plot was generated in Figure 44. Finally, an H/α plot for all target pixels is show in Figure 45, where the RGB colour scheme represents the Freeman classes. The Freeman decomposition statistics are shown in Table 13.

In Figure 45, odd bounce pixels appear as expected at low alpha, which represents single bounce mechanisms. The other two classes appear at higher alpha, which is also as expected although


they are not very well separated. The entropy of the two latter classes is quite high; this may mean that the Freeman decomposition is not significant for these pixels [CLOUS1996].

Table 13 Freeman Decomposition Statistics

Double Volume Odd

1129 1592 2595

Figure 42 Freeman Decomposition


Figure 43 Freeman Decomposition Histogram

Figure 44 H/alpha analysis for each Freeman Categories


Figure 45 Combined H/alpha and Freeman Decomposition


5 Conclusions

Based on a search of relevant literature, a set of polarimetric SAR measurements have been proposed for the discrimination and classification of ships. These measurements were carried out on CV-580 fully polarimetric datasets using a combination of Polarimetric Workstation, PolSARPRO, and MATLAB code.

A total of 46 ship datasets were processed in this report. The processed 342_Toronto L22P2 dataset is presented in this report, and the remaining 45 datasets can be found in the CD-ROM, which is submitted to the Scientific Authority. The Annex A gives detailed steps for processing a dataset.

The next logical step would be to complete the data processing of the remaining ships and import these results into the third party software tool HNeT to see if, in fact, these data will allow for vessel classification.


References .....

Cloude, Shane R., and Pottier, E. 1996. A review of target decomposition theorems in radar polarimetry. IEEE Transactions on Geoscience and Remote Sensing, vol 34, No 2, pp. 498 – 518.

Dubois-Fernandez, P. et al 2007. Compact polarimetry SAR for natural surface characterization: an attractive compromise, presentation material from http://marte.dpi.inpe.br/col/dpi.inpe.br/marte@80/2007/08.24.16.43/doc/62_11h05-PascaleDubois.pdf (as of 12 October 2007).

Haiyan, L., Yijun, H., and Wenguang, W. 2006. A New Approach for Ship Detection in SAR Imagery Based on Convolution between Different Polarization Channels. 7th International Symposium on Antennas, Propagation & EM Theory, (ISAPE), pp.1-4.

Howell, C. et al. 2004. Iceberg and ship discrimination with ENVISAT multipolarization ASAR, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Anchorage, Alaska, Volume: 1, pp. 113 – 116.

_____. 2006. A Multivariate Approach to Iceberg and Ship Classification in HH/HV ASAR Data, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Denver USA, pp. 3583-3586.

_____. 2007. Ship-Iceberg Discrimination Using Polarimetric SAR Data: a Quadratic Discriminant Approach, Presentation Material for the 7th advanced SAR Workshop, Vancouver Canada.

Liu, C., Vachon, P.W., and Geling, G.W. 2005. Improved Ship Detection with Airborne Polarimetirc SAR Data. Can. J. Remote Sensing, Vol. 31, No. 1, pp. 121-131.

Liu, C., and Vachon, P.W. 2007. Analysis of PolSAR Maritime Data, Presentation material for the 7th Advanced SAR Workshop (ASAR2007), Vancouver Canada.

Margarit, G. et al. 2006. On the Usage of GRECOSAR, an Orbital Polarimetric SAR Simulator of Complex Targets, to Vessel Classification Studies, IEEE transactions on Geoscience and Remote Sensing, Dec. Volume: 44, Issue: 12, pp. 3517-3526

_____. 2006b. Study of the Influence of Vessel Motions and Sea-Ship Interaction on Classification Algorithms Based on Single-Pass Polarimetric SAR Interferometry, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Denver, USA, pp. 75-78.

Olsen, R.B., Arnesen, T.N., and Eldhuset, K. 2004. Signatures of Vessels in ENVISAT AP-Mode Imagery, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Volume 6, Issue , 20-24 pp. 3895 – 3897.

Ringrose, R., and Harris, N. 2000. Ship Detection using Polarimetric SAR Data, Proceedings of the CEOS SAR Workshop, Toulouse, ESA SP-450, March 2000, pp. 687-691.


Souyris, J.C. et al. 2005. Compact polarimetry based on symmetry properties of geophysical media: the �/4 mode, IEEE Transactions on Geoscience and Remote SensingVolume 43, Issue 3, pp. 634 – 646.

Staples, G. et al. 2007. Ship-Iceberg Discrimination Using Polarimetric SAR Data, Presentation material for the 7th Advanced SAR Workshop (ASAR2007), Vancouver Canada.

Touzi, R. 1999. On The Use Of Polarimetric SAR Data For Ship Detection, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 1999, Hamburg, Germany, Volume 6, pp. 3895 – 3897.

_____. 2000. Calibrated Polarimetric SAR Data for Ship Dectection, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2000, Volume 1, pp. 144 – 146.

Touzi, R., Charbonneau, F.J., Hawkins, R.K., and Vachon, P.W. 2004. Ship detection and characterization using polarimetric SAR, Canadian J. Remote Sensing, Vol.30, No. 3, pp. 552-559.

Yeremy, M.L. et al. 2002. Results from the Crusade ship detection trial : polarimetric SAR. Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Volume 2, pp. 711 – 713.


Annex A Processing Steps

POLSAR Processing: In the data set directory, create a work directory and results directory. Copy the CV-580 files (properly named) into the work directory along with the mycameronmap.pal (which can be found in the work directory of a processed dataset) file. Polarimetric Workstation (PWS) PWS assumes that the CV-580 files have the proper naming convention, and all output filenames must be entered as indicated in each step. The MATLAB processing will look for specific file names.

1. Open PWS, Click Load Image

2. Navigate to the dataset /work directory (ie. C:\306_Grenfell\work) and choose one of the

four .hdr files (eg. .6p8exthh.hdr)


3. Select OK

4. The image window will then open


5. Perform Image Synthesis

a. Click Define sub-image, Full image, and Full resolution

b. In the drop down box, choose Sigma naught


c. Click Apply, then select User choice

Refer to the table for the input for the next window

LH LV VV HH HV Tx Ellipticity 45 45 0 0 0 Tx Rotation 0 0 90 0 0 Rx Ellipticity 0 0 0 0 0 Rx Rotation 0 90 90 0 90


Click OK

Use the table column header as the output name (ie, lh, lv, vv, hh, hv) Repeat step c for remaining four combinations. d. Again under image synthesis click Apply, then choose Circular Right-Right


Click Ok and enter rr as output file name

Repeat d for Circular Left-Left and Circular Right-Left, using ll and rl respectively as the output file name. It is a good idea to close the 8 output windows.

6. Entropy/Alpha a. Under Local Area Analysis, choose Define polygons


Define an area around the ship by left clicking, and double click at the last point to finish the polygon selection.

Choose yes to accept the area, or no to redraw. In the output file box, enter the number and ship name. For example, enter 306Grenfell for dataset 306_Grenfell.


b. In the dropdown menu, choose Cloude Decomposition, then Apply

Choose Cloude Histogram

Enter 7 for number of LINE and 7 for number of COLUMN. Click OK.

A Cloude Decomposition Histogram window will open. Click File -> Save


First change the Save as type to TIFF image, and then change the work directory to results. Finally name the file halpha.tif (Close the Cloude window)

7. Channel Correlation a. Click Define sub-image and choose no

Then Full image


and Square samples

In the drop down box choose correlation

b. Click Apply, then choose HH-HV, and click Ok


Name the output ch_corr, and click Ok.

Repeat b for the HH-VV and HV-VV, each time using ch_corr as the output file name. Once finished, close the 6 windows.

8. SSCM a. Under Image synthesis choose Define sub-image, then select no


Select Coordinates

Enter 10 as first line, 1 as first column, then enter the number of lines and the number of samples. (This information found in the CV-580 .hdr file)

Choose Square samples

b. In the drop down box choose SSCM, then Apply


Choose Yes

Click Ok

Use the mouse to define a clutter polygon (left click for points, double click to end)


Enter the file name as ship then number (ie Grenfell306 for 306_Grenfell), and enter the other values as shown (should be the default values). Click Ok.

On the Target Poincare Longitude click File -> Save enter longitude.tif as the file name (in results directory) and click save.


On the Target Poincare Latitude click File -> Save enter latitude.tif as the file name and click save.

Using the left click on the mouse, click to zoom in on the ship in the latitude window. (This will resize longitude as well). Click File -> Save and save as latitude_zoom.tif. Save the longitude as longitude_zoom.tif. (It is a good idea to close PWS and reopen to process the next dataset, as this will reset all the file paths)


PolSAR Pro (PSP) PolSAR Pro v3.3 beta was used for the processing.

1. Open PSP (click enter) and then choose the top left button next to PolSARPro, and choose single data set.

Click Environment

Navigate to the data set work directory (eg. 306_Grenfell/work)


Click Import, Airborne Sensor, Convair


Choose the CV580 *hh.img file as s11, hv as s12, vh as s21 and vv as s22. Make sure to enter the .img files. Enter the number of Rows (lines) and Cols (samples) from the .hdr file. Finally check the box to Convert Input IEEE binary Format. Click Ok.

Click Ok

Click Import, Extract, Full Resolution


Choose the Full Resolution Button, then click Run


2. Choose Processing, S2, Single Data Set, Polarimetric Decompositions, CAM: Cameron

Decomposition


Click Run


3. Choose Processing, S2, Single Data Set, Polarimetric Decompositions, KRO: Krogager

Decomposition


Check the TgtG TgtG TgtG box, then click Run


Click Yes

4. Choose Processing, S2, Single Data Set, Polarimetric Decompositions, FRE3: Freeman 3

Components Decomposition (In previous versions of PSP, this was the Freeman Decomposition)


Check the TgtG TgtG TgtG box, enter 7 as the window size, then click Run


5. Choose Processing, Single Data Set, Polarimetric Segmentation, H/A/Alpha

Classification


Check the Anisotropy Entropy Alpha box, check the Entropy/Alpha Planes (BMP) + Classifier (Bin + BMP) box and set the window size to 7. Click run.


Close PSP

MATLAB: The MATLAB portion of the processing assumes that all the file created with PWS and PSP have been named as instructed above.

1. Open MATLAB (Make sure all appropriate paths have been added) 2. cd to the data set directory (ie, 306_Grenfell) 3. type run_polsar

a. When Figure 9753 opens, use the mouse to zoom in on the ship. This view will be used for all saved images, but the .mat files will store the full data. If you wish


to reprocess the data set with a different zoom, delete the zoom_axis.mat file in the /work directory.

b. Press any key to continue processing. c. When the processing is finished, copy all the output from the command window

into the results*.txt file in the /results directory. (eg. results_306Grenfell.txt). This file is then used to fill in all the tables in the dataset annex.

d. This script will call each function in the same order as used in the final report. It will also close the figures when it calls a new function.

Notes: All the images in the report are in the /results directory, although not all the images in the directory are used. Using the 306_Grenfell dataset as an example, the following images from the /results directory are used in the report Annex: (Grouped by report section) 1. Correlation between polarimetric channels

channel_corr_gamma_hhhv.tif channel_corr_gamma_hhvv.tif channel_corr_gamma_hvvv.tif channel_corr_phase_hhhv.tif channel_corr_phase_hhvv.tif channel_corr_phase_hvvv.tif

3. Synthesis Image Grenfell_24-MAR-2004_l6p8_LV_Synthesis.jpg Grenfell_24-MAR-2004_l6p8_LH_Synthesis.jpg Grenfell_24-MAR-2004_l6p8_HH_Synthesis.jpg Grenfell_24-MAR-2004_l6p8_HV_Synthesis.jpg Grenfell_24-MAR-2004_l6p8_VV_Synthesis.jpg

4.1 Evaluation of Clutter Mean Grenfell_24-MAR-2004_l6p8_HH_Linear_Clutter_Average.jpg

Grenfell_24-MAR-2004_l6p8_HV_Linear_Clutter_Average.jpg Grenfell_24-MAR-2004_l6p8_VH_Linear_Clutter_Average.jpg Grenfell_24-MAR-2004_l6p8_VV_Linear_Clutter_Average.jpg

4.2 Results Grenfell_24-MAR-2004_l6p8_HH_Linear.jpg Grenfell_24-MAR-2004_l6p8_HV_Linear.jpg Grenfell_24-MAR-2004_l6p8_VH_Linear.jpg Grenfell_24-MAR-2004_l6p8_VV_Linear.jpg Grenfell_24-MAR-2004_l6p8_HH_Target_Mask.jpg Grenfell_24-MAR-2004_l6p8_HV_Target_Mask.jpg Grenfell_24-MAR-2004_l6p8_VH_Target_Mask.jpg Grenfell_24-MAR-2004_l6p8_VV_Target_Mask.jpg


Grenfell_24-MAR-2004_l6p8_Final_Target_Mask.jpg Grenfell_24-MAR-2004_l6p8_RR_Circular.jpg Grenfell_24-MAR-2004_l6p8_RL_Circular.jpg Grenfell_24-MAR-2004_l6p8_LL_Circular.jpg

5. Pauli Decomposition Grenfell_24-MAR-2004_l6p8_Pauli_Decomposition.jpg

Grenfell_24-MAR-2004_l6p8_Pauli_Decomposition_Statistics.jpg 6. Cameron Decomposition

Grenfell_24-MAR-2004_l6p8_Cameron_Decomposition.jpg Grenfell_24-MAR-2004_l6p8_Cameron_Decomposition_statistics.jpg

7. Krogager Decomposition Grenfell_24-MAR-2004_l6p8_Krogager_Decomposition.jpg

Grenfell_24-MAR-2004_l6p8_Krogager_Decomposition_Statistics.jpg 8. SSCM latitude_zoom.tif longitude_zoom.tif l6p8ext_SSCM_Classification.tif l6p8ext_SSCM_Histogram.tif 9. HH/HV, VV/VH, HH/VV Ratios Grenfell_24-MAR-2004_l6p8_HH_HV_(dB).jpg Grenfell_24-MAR-2004_l6p8_VV_VH_(dB).jpg Grenfell_24-MAR-2004_l6p8_HH_VV_(dB).jpg 10. H/Alpha halpha.tif 11. Freeman + H/Alpha Grenfell_24-MAR-2004_l6p8_Freeman_Decomposition.jpg

Grenfell_24-MAR-2004_l6p8_Freeman_Decomposition_Statistics.jpg Grenfell_24-MAR-2004_l6p8_Freeman_H_alpha.jpg Grenfell_24-MAR-2004_l6p8_Freemand_H_alpha_combined.jpg


List of symbols/abbreviations/acronyms/initialisms

DRDC Defence Research & Development Canada

R&D Research & Development

RAST Radar Application and Space Technology Section

VPI Vantage Point International

SAR Synthetic Aperture Radar

CV-580 Convair 580

HH Horizontal transmit, Horizontal received electromagnetic wave

VV Vertical transmit, Vertical received electromagnetic wave

HV Horizontal transmit, Vertical received electromagnetic wave

VH Vertical transmit, Horizontal received electromagnetic wave

SSCM Symmetric Scattering Characterization Method

RGB Red Green Blue

SCR Signal to clutter ratio

H/α Entropy/Alpha

PWS Polarimetric Workstation

RAT Radar Toolkit


Distribution list

Document No.: DRDC Ottawa CR 2008-067

Internal Distribution by Centre

1 Gary Geling 1 Paris Vachon 1 David Schlingmeier 1 Ramin Sabry 1 Chen Liu 1 Nicholas Sandirasegaram 2 Library External Distribution

1 DRDKIM 3 Library and archives Canada

12 TOTAL

DOCUMENT CONTROL DATA (Security classification of title, body of abstract and indexing annotation must be entered when the overall document is classified)

1. ORIGINATOR (The name and address of the organization preparing the document. Organizations for whom the document was prepared, e.g. Centre sponsoring a contractor's report, or tasking agency, are entered in section 8.) Vantage Point International Inc. 400 March Road, suite 210 Ottawa, Ontario, CA K2K 3H4

2. SECURITY CLASSIFICATION (Overall security classification of the document including special warning terms if applicable.)

UNCLASSIFIED

3. TITLE (The complete document title as indicated on the title page. Its classification should be indicated by the appropriate abbreviation (S, C or U) in parentheses after the title.) Analysis of PolSAR maritime data:

4. AUTHORS (last name, followed by initials – ranks, titles, etc. not to be used) Garrett Parsons, Craig Williams, and Martin St-Hilaire

5. DATE OF PUBLICATION (Month and year of publication of document.) April 2008

6a. NO. OF PAGES (Total containing information, including Annexes, Appendices, etc.)

88

6b. NO. OF REFS (Total cited in document.)

17 7. DESCRIPTIVE NOTES (The category of the document, e.g. technical report, technical note or memorandum. If appropriate, enter the type of report,

e.g. interim, progress, summary, annual or final. Give the inclusive dates when a specific reporting period is covered.) Contract Report

8. SPONSORING ACTIVITY (The name of the department project office or laboratory sponsoring the research and development – include address.) Defence R&D Canada – Ottawa 3701 Carling Avenue Ottawa, Ontario K1A 0Z4

9a. PROJECT OR GRANT NO. (If appropriate, the applicable research and development project or grant number under which the document was written. Please specify whether project or grant.)

15ec05

9b. CONTRACT NO. (If appropriate, the applicable number under which the document was written.)

W7714-06-0989/001/SV

10a. ORIGINATOR'S DOCUMENT NUMBER (The official document number by which the document is identified by the originating activity. This number must be unique to this document.)

10b. OTHER DOCUMENT NO(s). (Any other numbers which may be assigned this document either by the originator or by the sponsor.) DRDC Ottawa CR 2008-067

11. DOCUMENT AVAILABILITY (Any limitations on further dissemination of the document, other than those imposed by security classification.)

UNLIMITED

12. DOCUMENT ANNOUNCEMENT (Any limitation to the bibliographic announcement of this document. This will normally correspond to the Document Availability (11). However, where further distribution (beyond the audience specified in (11) is possible, a wider announcement audience may be selected.)) UNLIMITED

13. ABSTRACT (A brief and factual summary of the document. It may also appear elsewhere in the body of the document itself. It is highly desirable that the abstract of classified documents be unclassified. Each paragraph of the abstract shall begin with an indication of the security classification of the information in the paragraph (unless the document itself is unclassified) represented as (S), (C), (R), or (U). It is not necessary to include here abstracts in both official languages unless the text is bilingual.)

This document summarizes the analysis of polarimetric SAR data of maritime vessels collected by the CV-580 aircraft for the Radar Application and Space Technology Section (RAST) of Defence R&D – Ottawa (DRDC Ottawa). This analysis was conducted to support polarimetric SAR research in the area of target classification and target discrimination.

The work involved a search of recent literature as well as selecting suitable software tools to generate the desired polarimetric decompositions. The selected software are Polarimetric Workstation software tool, PolSARPRO software tool, DRDC Ottawa MATLAB code and VPI MATLAB code. The analysis techniques are correlation and covariance between polarimetric channels, synthesis images, mean clutter, peak of target and target peak-to-clutter ratio, Symmetric Scattering Characterization Method (SSCM), decomposition of Pauli , Cameron, Krogager, Entropy/Alpha and Freeman with Entropy/Alpha, and ratios of HH/HV, VV/VH, and HH/VV. These techniques are described in detail and a sample processed data set is presented in the report, while the CD-ROM available with the scientific authority includes results from all other analysed vessels.

14. KEYWORDS, DESCRIPTORS or IDENTIFIERS (Technically meaningful terms or short phrases that characterize a document and could be helpful in cataloguing the document. They should be selected so that no security classification is required. Identifiers, such as equipment model designation, trade name, military project code name, geographic location may also be included. If possible keywords should be selected from a published thesaurus, e.g. Thesaurus of Engineering and Scientific Terms (TEST) and that thesaurus identified. If it is not possible to select indexing terms which are Unclassified, the classification of each should be indicated as with the title.)

PolSARPRO, correlation, covariance, synthesis images, mean clutter, peak-to-clutter ratio, Symmetric Scattering Characterization Method, SSCM, Pauli , Cameron, Krogager, Entropy, Alpha, Freeman,, Polarimetry, target discrimination, target classification

analysis of polsar maritime data - defence …cradpdf.drdc-rddc.gc.ca/pdfs/unc71/p529331.pdf ·...

Documents