PolSAR-Ap: the use of Polarimetric SAR to Improve Detection of Targets at Sea
Marino, Armando¹; Nunziata, Ferdinando²; Migliaccio, Maurizio²; Ouchi, Kazuo³; Sugimoto, Mitsunobu³; Hajnsek, Irena^4
1ETH Zurich, SWITZERLAND; ²Universita' di Napoli Parthenope, ITALY; ³National Defense Academy, JAPAN; ⁴ETH Zurich & DLR, GERMANY

INTRODUCTION:
Ship detection is an important topic for security and surveillance of maritime and coastal areas. A solution exploiting satellite SAR sensors is particularly interesting, because it offers wide scale surveillance capabilities, which are not reliant on solar illumination and are independent of the weather conditions ([1, 2, 3]). SAR polarimetry has a valuable contribution in improving ship detection capabilities ([2, 4]). The ship detection contribution in PolSAR-Ap aims at demonstrating the benefits of exploiting polarimetry compared to algorithms making use of one single channel. For this reason, several detectors were compared over surveyed data and the probabilities of detection and false alarm were estimated.

SHIP DETECTION WITH SAR POLARIMETRY:
In SAR images, the main feature of ships is a relatively large backscattering signal, which is usually brighter in comparison with the sea background. This led to the idea of performing a statistical test on the intensities of target and clutter background setting a fixed probability of false alarms (CFAR) [1, 3]. Regarding the benefits of polarimetry, it can readily be observed that the simple exploitation of the cross-polarised channel (HV) rather than the co-polarised ones (HH or VV) may increase substantially the detection performance. A first approach to exploit polarimetric information is statistical ([1, 2]). A second type of polarimetric ship detectors is based on the physical scattering properties of targets and ships ([5, 7]).

CFAR FOR HV INTENSITY:
This detector sets a threshold on the intensity of the cross-polarisation channel (HV), since many ships are expected to have a strong backscattering in this component due to orientation effects, while the sea should have a low or null return (i.e. Bragg surface). The threshold is set exploiting a CFAR methodology with the K-distribution to model the statistical behaviour of the sea clutter [1].

LIU ET AL.:
Liu et al [2] proposed this methodology which is an approximation of the Generalised Likelihood Ratio Test between targets and sea clutter where the statistics for the sea are considered Normal distributed and the covariance matrix of the target is unknown. Even though this is only an approximation for real data, it is presented here because it shows adequate results and has an interesting physical interpretation of its final formulation.

SYMMETRY DETECTOR:
This detector is based on the concept of scattering symmetry [5] and the detection rule sets a threshold on the magnitude of the C12 element of the covariance matrix (complex inner product of HH and HV), which is expected to be particularly small (theoretically zero) in case of reflection symmetry. The sea should show reflection symmetry, since it is assumed to be a surface, while complex scatterers as ships are likely to not be symmetric which returns higher values of C12.

GEOMETRICAL PERTURBATION POLARIMETRIC NOTCH FILTER (GP-PNF):
This methodology was proposed by Marino et al. [6] and is based on the idea of isolating the return coming from the sea and detecting anything else is in the scene (i.e. it works as a Notch Filter in the space of the partial targets, where the null is located on the signature of the sea). The sea clutter can be completely characterized with a vector in a six dimensional complex space. On the other hand, the targets of interest can have a large variety of polarimetric signatures depending on orientation, material and structure of the vessel. The GP-PNF approach is focusing on targets lying in the complement orthogonal subspace of the sea vector (5 dimensional complex subspace).

POLARIMETRIC MATCH FILTER (PMF):
This algorithm firstly proposed by Novak ([7]) is based on the optimisation of the contrast between the covariance matrices of sea and ships over different scattering mechanisms. In other words, the algorithm returns the scattering mechanism that optimises the diversity providing the highest contrast possible. A threshold is then set fitting a distribution (the intensity ratio) to the obtained contrast enhanced image.

DATASETS: 1) ALOS-PALSAR:
This first dataset covers the Tokyo Bay area (Japan), which is renowned to have a large traffic of vessels. The acquisition was performed on the 9th of October 2008, (10:19 am local time). The resolution in ground range is about 27 m, while in azimuth is about 4.9 m. The incidence angle of these acquisitions is about 24 degrees. During the acquisitions in Tokyo Bay a ground survey was carried out with a video camera and a ground radar.

2) RADARSAT-2:
Two RADARSAT-2 Single Look Complex (SLC) quad-pol SAR scenes were acquired on the 15th of May 2010 (under the SOAR project) in the Gulf of Mexico where ground truthed metallic targets are present.

RESULTS:
From the analysis of the ALOS-PALSAR detection masks it is evident that the quad polarimetric detectors are able to provide detection capabilities higher than the single polarimetric ones. All the vessels listed in the ground truth or visible in the Pauli RGB image were detected by the quad-polarimetric detectors. On the other hand, detectors using only dual or single polarimetric information miss several vessels.
Regarding the RADARSAT-2 data, the experimental results witnesses that the extra information inherently carried on polarimetric detectors allows a better discrimination of the ''anomalies'' in the polarimetric backscattering of sea surface related to targets. This results in a better target/sea discrimination, even when very challenging scenes are considered.

REFERENCES:
[1] D. J. Crisp, ''The State-of-the-Art in ship detection in Synthetic Aperture Radar imagery,'' Australiane Government Department of Defence, 2004.
[2] C. Liu, P. W. Vachon, and G. W. Geling, ''Improved ship detection using polarimetric SAR data,,'' IGARSS Geoscience and Remote Sensing Symposium, vol. 3, pp. 1800-1803,, 2004.
[3] P. W Vachon, ''Ship detection in synthetic aperture radar imagery.,'' Proceedings OceanSAR, St. John s, NL, Canada, 2006.
[4] M. Yeremy, G. Geling, M. Rey, B. Plache, and M. Henschel, ''Results from the crusade ship detection trial: polarimetric sar.,'' Proceeding on IGARSS 2002, 2002.
[5] F. Nunziata, A. Montuori, and M. Migliaccio, ''Dual-polarized cosmo skymed sar data to observe metallic targets at sea,'' IGARSS , Geoscience and Remote Sensing IEEE International Symposium, pp. 2270-2273, 2011.
[6] A. Marino, N. Walker, and I. H. Woodhouse, ''Ship detection using SAR polarimetry. the development of a new algorithm designed to exploit new satellite SAR capabilities for maritime surveillance,'' Proceedings on SEASAR, Frascati, Italy, January, 2010.
[7] L.M. Novak, M.B. Sechtin, and M.J. Cardullo, ''Studies of target detection algorithms that use polarimetric radar data,'' IEEE Transactions on Aerospace and Electronic Systems,, vol. 25, pp. 150-165, 1989.