Boularbah Souissi

Optimal SVM Classification for Compact Polarimetric Data Using Stokes Parameters

In this paper, our objective is twofold: first, to assess the potential of the new compact polarimetry imaging radar system called hybrid-polarimetry (CL-pol): circular transmitted polarization and coherent dual linear receive polarizations for full characterization and exploitation of the backscattered field. Useful characteristics that are unique to the hybrid-polarity architecture are invariance to geometrical orientations and minimizing on-board resource requirements. Second, to develop a classification polarimetric method based on the support vector machine (SVM) which uses full- and the compact-pol modes. We present a study of the polarimetric information content derived from the decomposition for the CL-mode using Stokes parameter data products and from Freeman-Durden-decomposition derived from the full-pol imaging mode. We compare SVM classification both among the partial polarimetric datasets and against the full quad-pol dataset. We illustrate our results by using the polarimetric SAR images of Algiers city in Algeria acquired by the RadarSAT2 (FQ19) in C-band.

Interferometric Coherence Optimization: A Comparative Study

The Interferometric coherence measures the degree of similarity or the correlation between the radar signals corresponding to complex SAR images viewed from close angles. The strong dependence of the interferometric coherence to the polarization state can consider that there is a combination of polarization to achieve maximum interferometric coherence. The methods developed in this field are based on the combination of information derived from polarimetric channels of the interferometric couple and define new projection vectors that maximize the highest possible value of coherence. In this paper we investigated two interferometric coherence optimization methods. The first one is based on the selection of the best pair of projection vectors that maximizes the coherence. The corresponding algorithm leads to an eigenvectors / eigenvalues problem, where each vector represents a scattering mechanism used for view. The second method used an identical projection vector for the two images representing a unique scattering mechanism. To evaluate this work, we used a couple of interferometric airborne polarimetric data acquired on the Tapajos in Brazil, in the P band. Results compared to classical coherence reveals that both methods give better results than single acquisition, however, the first one is the best.

Adaptive model based on polarimetric decomposition using correlation coefficient in horizontal–vertical and circular basis

Abstract. This paper presents two decomposition schemes for polarimetric synthetic aperture radar data. The proposed schemes intend to overcome the problem of scattering ambiguity and reduce the volume scattering power in oriented urban areas. The first proposed scheme uses an empirical volume model based on the correlation coefficients of the Pauli component in the horizontal–vertical basis, whereas the second one employs a volume model defined on correlation coefficients of the Pauli components expressed in the circular basis. The correlation coefficients are calculated from polarimetric interferometric synthetic aperture radar (PolInSAR) data. The characteristics adopted from these volume models are used to enhance the results of the decomposition schemes. The scattering powers estimated from the proposed methods give promising results compared to existing methods in the literature, particularly in urban areas since all the oriented built-up areas are well discriminated as double or odd bounce scattering. The methods are evaluated using the experimental airborne SAR sensor (E-SAR) PolInSAR L band data acquired on the Oberpfaffenhofen test site in Germany.

Classification of multi-look polarimetric SAR imagery based on the automated EM Gaussian clustering algorithm using the complex Wishart distribution

In statistical classification, such mixture models allow a formal approach to unsupervised clustering. Fitting mixture distributions can be handled by a wide variety of techniques. A standard method to fit finite mixture models to observed data is the Expectation-Maximization (EM) algorithm which is an iterative procedure which converges to a (local) maximum of the marginal a posteriori probability function. In this paper we provide a review on the Gaussian classification for polarimetric images. In this review work, the analysis of multi-look polarimetric covariance matrix data uses an automated statistical clustering method based upon the expectation maximization (EM) algorithm for finite mixture modeling, using the complex Wishart probability density function. This classification technique is compared to that obtained with the most known standard H/ decomposition combined to the Wishart distribution which gives 8 fixed classes identified from the H/ space, in opposite to the EM-Wishart classifiaction which automatically determines the number of statistically distinct clusters in finite mixture modeling in an image using the Goodness-of-fit (GoF) and more than eight classes can be identified. Both approaches conducted on the polarimetric images show very convincing clustering results.

Analysis of Orientation Angle Shifts on the Polarimetric Data Using Radarsat2 Images

Polarization orientation angle (POA) shifts induced by the variations of range and azimuth slopes cause the polarization to rotate about the radar line of sight. Their existence reduces the accuracy measurement of geophysical parameters from polarimetric synthetic aperture radar (PolSAR) images and may generate erroneous scattering characteristics that could be misinterpreted. In real situations, terrain slopes rotate the polarization basis of the polarimetric scattering matrices by an orientation angle shift, and induce significant cross-polarization power. Consequently, it is desirable to compensate the data for the orientation effect before applying scattering model-based decompositions. In this paper, we investigate the compensation effect on the elements of the coherency matrix using the orientation angle extracted from circular polarization technique and from the copolarization signature applied to building areas. The effect of this compensation is that the volume scattering power is consistently decreased, while the double-bounce power is increased. The surface and helix scattering powers are roll invariant. Comparing both methods, we find that the circular polarization algorithm gives mostly the best results except for some targets. In this way, a combined use of both algorithms has been proposed to choose an optimum orientation angle, which can be used directly to compensate POLSAR data to ensure improvement in the overall polarimetric decomposition and classification. We illustrate our results using the polarimetric SAR images acquired on the Algiers city by the RadarSAT2 (FQ19) in C-band.

Estimation of surface topography from single polarimetric SAR imagery using RADARSAT2

In this paper a practical method is demonstrated for estimating terrain slopes in azimuth and ground range directions for digital elevation model (DEM) generation without any prior knowledge on the terrain by using only one single pass of polarimetric synthetic aperture radar (PolSAR) instead of two-pass or interferometric SAR (INSAR). The basic approach is by combing the orientation angle estimation and a shape-from-shading technique (SFS) which is mostly used by the computer vision community. In particular, when limited PolSAR data are available, this technique provides an alternative way for DEM generation. The polarization orientation angle (POA) is related to both the range and azimulh angles of the tilted surface and radar viewing angle and it can be estimated from the PolSAR data by using the circular polarization method which shows the best performance in computation efficiency and accuracy with respect to the other methods. After terrain slopes in both the range and azimuth directions have been estimated initially by the combination of the POA estimation and the SFS algorithm, a least squares method similar to that used in interferometric phase unwrapping is used to generate the topography. The least squares approach to phase unwrapping obtains an unwrapped solution by minimizing the differences between the discrete partial derivatives of the (wrapped) phase data and the discrete partial derivatives of the unwrapped solution. We illustrate our results by using the polarimetric SAR images acquired in Algeria by the RadarSAT2 (FQ19) in C-band.

Analysis of orientation angle shifts on the polarimetric data using RadarSAT2

The polarimetric orientation angle (POA) shifts can be compensated by the rotation about the line of sight based on the derived orientation angle. Based on uncompensated data, surface parameter estimation may produce incorrect results, and model-based target decompositions may generate erroneous scattering characteristics that could be misinterpreted. Consequently, it is desirable to compensate the data for the orientation effect before applying scattering model based decompositions for various applications. In this paper, we investigate the compensation effect on the elements of the linear, circular and coherency matrices. For our research work, we adopt the circular polarization techniques for OA compensation, because its effectiveness has been demonstrated and verified by comparison with interferometric SAR measurements. We illustrate our results by using the polarimetric SAR images acquired in Algiers by the RadarSAT2 (FQ19) in C-band.

Yamagushi Decomposition Based on the Compact Polarimetry Mode Using RADARSAT2 Data

Recently, dual-mode partially polarimetric SAR modes (DP) called compact polarimetry have been proposed. In these polarimetric configurations, only one transmit/receive cycle is required instead of two in a full quad-pol system, reducing the pulse repetition frequency and data rates by a factor of two for a given swath width. Souyris et al. introduced the π/4 compact polarimetric mode, in which the transmitted polarization is the superposition of linear horizontal and vertical polarizations H + V, resulting in a linear polarization oriented at 45° with respect to the horizontal. The radar receives returns in horizontal and vertical polarizations. Another hybrid dual-pol mode is the circular transmit, linear receive (CTLR) mode. In these new polarimetric modes, an equivalent covariance or coherency matrix may be reconstructed to produce the so-called pseudo quad-pol data that accurately reproduces the full quad-pol data. The compact polarimetry was proposed to assess various architecture designs that could be implemented on lowcost/low-mass. In that context, the comparison between full polarimetry (fp) versus dual polarimetry (dp) is a subject of most importance. This paper provides a comparison of the information content of full quad-pol data and the pseudo quad-pol data derived from compact polarimetric SAR modes. Both the polarimetric signatures based on the kennaugh matrix and the Four component decomposition in the context of this compact polarimetry mode are explored. We illustrate our results by using the polarimetric SAR images of Algiers city in Algeria acquired by the RadarSAT2 in C-band.