ALOS-PALSAR polarimetric SAR data to observe sea oil slicks

ALOS-PALSAR polarimetric SAR data to observe sea oil slicks M. Migliaccio, A. Gambardella and F. Nunziata

M. Shimada and O. Isoguchi

Università degli Studi di Napoli Parthenope Dipartimento per le Tecnologie Centro Direzionale, isola C4 - 80143 Napoli Email: ferdinando.nunziata at uniparthenope.it

Earth Observation Research and Application Center Japan Aerospace Exploration Agency (JAXA). Tsukuba, Ibaraki, Japan, 305-8505 [email protected]

Abstract—In this study an electromagnetic approach is proposed for exploiting polarimetric information for sea oil slick observation in L-band ALOS PALSAR full polarimetric SAR data. The problem is tackled form an electromagnetic viewpoint by describing the sea surface scattering mechanism with and without oil slicks.Following this rationale, a filtering technique, based on the Mueller scattering matrix, is applied to detect oil slicks in full polarimetric SAR data. Successively, the filtering results are verified by the analysis of the slick-free and slickcovered pedestal height and polarimetric entropy (H). Experiments, accomplished on a meaningful set of Level 1.1 LBand ALOS PALSAR full polarimetric data, demonstrate the effectiveness of the proposed approach.

I. I NTRODUCTION A fully polarimetric SAR transmits and receives two orthogonally polarized fields and, as result, gets the scattering matrix S for each resolution cell. Hence, this measurement process, taking into account the vectorial nature of the scattered field, allow retaining all the information in the scattered wave and describing the polarimetric properties of the observed scene. In January 2006, the Advanced Land Observing Satellite (ALOS) was launched by Japan Aerospace Exploration Agency (JAXA). It carries on board the Phased Array type Lband Synthetic Aperture Radar (PALSAR) which is the first space-borne full-polarimetric radar utilizing horizontally (h) and vertically (v) polarized microwaves both in transmission and reception. As a matter of fact, PALSAR measurements, acquired in the polarimetric mode, can serve as useful data source to provide additional information for environmental remote sensing applications. Within this context, in this study an electromagnetic approach is proposed for exploiting polarimetric information for sea oil slick observation in Lband PALSAR data. Sea oil pollution is a matter of great concern since it affects both the environment and human health. Oil slick detection is fundamental to effectively plane countermeasures and to minimize pollution effects and the SAR is unanimously recognized, under low to moderate wind conditions, as the most important imaging sensor for a synoptic and effective oil slick observation since its all-weather day and night capabilities [2]. In fact, the presence of a surface slick, reducing the signal backscattered to the radar antenna, generates a low-backscattering area which, in SAR images, appears as a dark area [3]. Following this rationale image

978-1-4244-3395-7/09/$25.00 ©2009 IEEE

processing techniques are commonly employed on singlepolarimetric SAR data to detect dark areas [3]. However, other physical phenomena, such as biogenic films, low winds..., can generate low backscattering areas (look-alikes) in SAR images. Thus, oil slick detection and classification techniques are still an open issue. First investigations on sea oil slick observation by means of polarimetric SAR data were generally unsatisfactory [4], only in recent times the usefulness of fully and partially polarized SAR measurements has been demonstrated for the C-band [5], [6], [7], [8]. The approach here proposed is based on the different sea surface scattering mechanism expected with and without surface slicks. It has been demonstrated in [6], [7], [8] that though the sea surface scattering follow a Bragg or tilted Bragg scattering mechanism, the presence of a surface slick, depending on its damping properties, may lead to a completely different and non-Bragg scattering mechanism. In this study, following this theoretical rationale, polarimetric information are exploited for sea oil slick observation. In detail, provided S, the Mueller scattering matrix (M) is constructed, and a physically based filtering technique, dubbed Mueller filtering [6], is applied in order to detect oil slicks. Successively, the filtering results are verified by the analysis of the slick-free and slick-covered copolarized signature and polarimetric entropy (H), performed by using the Kennaugh matrix (K) and the coherence matrix (T), respectively, following the guidelines developed in [5], [6], [7], [8]. Experiments, accomplished on a meaningful set of Level 1.1 L-Band PALSAR full polarimetric data, confirm the effectiveness of the proposed polarimetric approach and allow underlining the importance of fully polarimetric L-band SAR data for oil slick observation. II. P OLARIMETRIC APPROACH Polarimetric surface scattering can be described by using M, a 4 × 4 real matrix, never symmetric [9], which relates the scattered Stokes vector ss , to the incident one si : ss =

1 Msi (kr)−2

.

(1)

Note that the scattering Stokes vector and the Mueller matrix are meant in the mean sense, since they are related to random processes.

IV - 669

IGARSS 2009

In the case of sea surface scattering, as detailed in [6], it is possible to distinguish the scattering mechanism occurring with and without surface oil slicks looking at the terms of M related to the co-polarized and cross- polarized scattering amplitudes. In a nutshell, in case of oil-free sea surface the co-polarized term is expected to be greater that the crosspolarized one, while the opposite behavior is expected in case of oil-covered sea surface. This capability has been exploited to develop a filter (a.k.a. Mueller filter) capable to both observe oil slicks and to distinguish them from biogenic look-alikes. It is important to remark that, from an operational viewpoint, the filter is very important since its output is a binary image which is very much advisable for segmentation purposes. Two polarimetric analysis, accomplished on the slick-free and the slick-covered sea surface, are then applied to verify the filter output. The first one is based on the use of the copolarized polarimetric signature. Polarimetric SAR signatures, have been successfully employed to classify a wide range of terrain types, according to their different scattering properties [10] and, recently its usefulness in the context of sea oil slick observation has been demonstrated [7]. In fact, in [7] it was shown that, for low to moderate wind conditions, oil-free sea surfaces are responsible for a low unpolarized backscattered signal which calls for a co-polarized signature characterized by a small pedestal height. Conversely, in the case of oil-covered sea surface, a large unpolarized backscattered signal occurs which makes the pedestal height larger. The second polarimetric analysis was firstly proposed for oil slick observation in [5] and it is based on the use of the polarimetric entropy derived by the Cloude-Pottier decomposition theorem [11]. As a matter of fact, in the case of oil-free sea surface, a Bragg scattering mechanism is in place which is characterized by a low polarimetric entropy [5], [6], [8]. Conversely, in the case of oil covered sea surface, a non-Bragg scattering mechanism is in place which is characterized by a low backscattered return an by multiple scattering mechanism of comparable strength, i.e. a high polarimetric entropy [5], [6], [8]. III. E XPERIMENTS In this section results obtained applying the Mueller filter and performing the polarimetric analysis on a meaningful set of Level 1.1 L-Band PALSAR polarimetric data are shown and discussed. The nominal slant (ground) resolution is 9.4 (26.0) meters in range and 4.5 meters in azimuth. The Level 1.1 data, although affected by speckle, are characterized by the finest spatial resolution and, thus, have to preferred for oil slick observation purposes. The first case considered is related to the acquisition of 27 August 2006, 14:22 UTC (PALSAR, ALPSRP031440190, ascending pass) relevant to a well-known oil spill accident widely documented [12]. Fig.1 shows the module of the SLC ground projected VV SAR image relevant to a sub-image of the PALSAR data in which the oil slick is clearly visible. The filtering output is a black and white image, which clearly shows features related to the oil slick, Fig.2. This result

Fig. 1. Modulus of the SLC SAR data relevant to the acquisition of August 27, 2006

Fig. 2.

Filter output

confirms the completely different scattering mechanism which is expected in case of oil-free and oil-covered sea surface. Moreover, it must be explicitly noted that for the first time, the usefulness of L-band polarimetric SAR data is demonstrated. The analysis accomplished by considering the co-polarized signatures evaluated on the oil-free and oil-covered sea surface (not shown to save space) show that the presence of the oil slick increases the pedestal height. As a matter of fact, a tailored filtering technique to estimate the pedestal height has been developed. The estimated pedestal relevant to the first data set, is shown in gray tones in Fig.3. The result confirms the theoretical model which predicts an higher unpolarized energy (high pedestal height) within the oil-covered sea surface. A second confirm of the capability of fully polarimetric Lband SAR data for oil slick observation is provided by the analysis of the estimated entropy (see Fig.4). It can be noted that the H values, within the oil-covered area are higher than the surrounding sea. The second case considered belongs to the acquisition of 10 March 2007, in which a look-alike is present. In this case the ground truth was not available and the first guess was provided

IV - 670

Fig. 3.

Pedestal height Fig. 6.

Fig. 4.

Filter output

Polarimetric entropy Fig. 7.

by expert SAR image analysts. The module of the SLC ground projected VV PALSAR sub-image in which the look-alike is clearly visible, is shown in Fig.5. The filtering output (Fig.6) does not show any remarkable feature related to the dark area of Fig.5. This result allows classifying the dark area as a lookalike. To further confirm filter result, the pedestal height and the polarimetric entropy have been estimated, see Figs.7-8.

Pedestal height

IV. C ONCLUSION In this study ALOS-PALSAR polarimetric SAR data has been firstly exploited to observe sea oil slicks. The working hypothesis was that the extra-information provided by the polarimetric measurements can be exploited to distinguish sea surface scattering mechanism with and without surface slicks. The data set, which concerns both oil slicks - due to a tank accident - and oil look-alikes, has been processed and

Fig. 5. Modulus of the SLC SAR data relevant to the acquisition of March 10, 2007

IV - 671

Fig. 8.

Polarimetric entropy

analyzed. Experimental results show the effectiveness and the usefulness of polarimetric L-band SAR data for sea oil slick observation. ACKNOWLEDGMENT Authors kindly acknowledge the Earth Observation Research and Application Center (EORC) of the Japan Aerospace Exploration Agency (JAXA) for providing the data used in this study. R EFERENCES [1] J. J. van Zyl, “Unsupervised classification of scattering behavior using radar polarimetry data,” IEEE Trans. Geosci. Remote Sens., vol. 27, n. 1, pp. 36-45, 1989. [2] C. E. Brown and M. F. Fingas, “Synthetic aperture radar sensors: Viable for marine oil spill response ?,” Proc. 26th AMOP, Ottawa, ON, Canada, Jun. 10-12, 2003, pp. 299-310, 2003. [3] M.F. Fingas and C.E. Brown, “Review of oil spill remote sensing,” Spill Sci. Technology Bull., vol. 4, no. 4, pp. 199-208, 1997. [4] M. Gade, W. Alpers, H. Huhnerfuss, H. Masuko, and T. Kobayashi, “Imaging of biogenic and anthropogenic ocean surface films by the multifrequency/multipolarization SIR-C/X-SAR,” J. Geophys. Res., vol. 103, no. C9, pp. 18851-18866, 1998.

[5] M. Migliaccio, A. Gambardella, M. Tranfaglia, “SAR Polarimetry to Observe Oil Spills,” IEEE Trans. Geosci. Remote Sens., vol.45 , no.2 , pp. 506-511, 2007. [6] F. Nunziata, A. Gambardella and M. Migliaccio, “On the Mueller Scattering Matrix for SAR Sea Oil Slick Observation,” IEEE Geosci. and Remote Sensing Letters, vol. 5, n. 4, pp. 691-695, 2008. [7] M. Migliaccio, F. Nunziata, A. Gambardella, “Polarimetric Signature for Oil Spill Observation,” Proc. of US/EU-Baltic Int. Symposium, Tallin, Lithuania, May 27-29, 2008. [8] M. Migliaccio, F. Nunziata, A. Gambardella, “On The Copolarised Phase Difference for Oil Spill Observation,” Int. Journal of Remote Sensing, vol. 30, n. 6, pp. 1587-1602, 2009. [9] A. Guissard, “Mueller and Kennaugh matrices in radar polarimetry,” IEEE Trans. Geosci. Remote Sens., vol. 32, no. 3, pp. 590-597, 1994. [10] D. L. Evans, T. G. Farr, J. J. van Zyl, and H. A. Zebker, “Radar polarimetry: analysis tools and applications,” IEEE Trans. Geosci. Remote Sens., vol.26, no.6, pp 774-789, 1988. [11] S. R. Cloude and E. Pottier, “A review of target decomposition theorems in radar polarimetry,” IEEE Trans. Geosci. Remote Sens., vol. 34, no. 2, pp. 498-518, 1996. [12] EORC/JAXA website, “Detection of oil spill caused by a sunken tanker by using PALSAR,” available at: http://www.eorc.jaxa.jp/ALOS/img_up/.

IV - 672