A New Target Detector Based on Subspace Projections Using Polarimetric SAR Data

Most applications of radar imagery require processing techniques which achieve one fundamental goal: characterize and detect the constituent scatterers for each pixel in the scene. In this paper, we take a new look at the target detection issue in polarimetric synthetic aperture radar data and assume several canonical scattering mechanisms as our signal sources whose combination of them with appropriate weight fractions formed the scattering vector of each pixel. The presence of speckle as a consequence of coherent processing of the scattered signals is modeled as signal-dependent additive noise. The set of the scattering mechanisms is divided into two groups: objected scattering mechanism belonging to the target, and nonobjected scattering mechanisms. Then, we make use of two techniques based on subspace projections for speckle reduction and the nonobjected scattering mechanisms annihilation, followed by detecting the presence of the scattering mechanism of interest. In the problem formulation scenario, a novel feature space is proposed consisting of two subspaces-the objected subspace and the nonobjected subspace. Then, the detection approach under this scenario is derived. An orthogonal subspace projection technique is utilized for speckle reduction. Moreover, in order to annihilate the nonobjected subspace, each pixel's feature vector is obliquely projected onto the objected subspace. With the annihilation of the nonobjected subspace and using the polarimetric information of the objected subspace, the detectability of the target scattering mechanism is therefore enhanced. Finally, evaluation against C-, L-, and P-band fully polarimetric SAR data sets is provided with a significant agreement with the expected results.