Search

We present the theory, design, and experimental results obtained from a scanning passive W-band fully polarimetric imager. Passive millimeter-wave imaging offers persistent day/nighttime imaging and the ability to penetrate dust, clouds and other obscurants, including clothing and dry soil. The single-pixel scanning imager includes both far-field and near-field fore-optics for investigation of polarization phenomena. Using both fore-optics, a variety of scenes including natural and man-made objects was imaged and these results are presented showing the utility of polarimetric imaging for anomaly detection. Analysis includes conventional Stokes-parameter based approaches as well as multivariate image analysis methods.

We present experimental results obtained from a scanning passive W-band fully polarimetric imager. Passive millimeter wave imaging offers persistent day/nighttime imaging and the ability to penetrate dust, clouds and other obscurants, as well as thin layers of clothing and even dry soil. The selection of the W-band atmospheric window at 94 GHz offers a compromise as there is sufficient angular resolution for imaging applications using modestly-sized reflectors appropriate for mobile as well as fixed location applications. The imager is based upon an F/2.1 off-axis parabolic reflector that exhibits -34 dB of cross polarization suppression. The heterodyne radiometer produces a 6 GHz IF with 4 GHz of bandwidth resulting in an NEDT of < 200 mK. Polarimetric imaging reveals the presence of man-made objects due to their typically anisotropic nature and the interaction of these objects with incident millimeter wave radiation. The phenomenology studies were undertaken to determine the richest polarimetric signals to use for exploitation. In addition to a conventional approach to polarimetric image analysis in which the Stokes I, Q, U, and V images were formed and displayed, we present an alternative method for polarimetric image exploitation based upon multivariate image analysis (MIA). MIA uses principal component analysis (PCA) and 2D scatter or score plots to identify various pixel classes in the image compared with the more conventional scene-based image analysis approaches. Multivariate image decomposition provides a window into the complementary interplay between spatial and statistical correlations contained in the data.