Electronic Military & Defense magazine was developed for engineers, program managers, project managers, and those involved in the design and development of electronic and electro-optic systems for military, defense, and aerospace applications.
Issue link: https://electronicsmilitarydefense.epubxp.com/i/546858
Technology Why Use Polarimeters For Target And Threat Identification? Hyperspectral has proven useful for TTI, as many man-made materials and targets of military interest have specific, identi- fiable spectral signatures. Spectral phenomenology depends on bulk material properties and chemical composition, but hyperspectral imaging cannot measure material orientation properties, crystal axis properties, or surface structure prop- erties (imagine stretching a film or a specific type of surface roughness). We won't go into hyperspectral imaging for TTI in detail here, but hyperspectral is very good for classifying, discriminating, or quantifying chemical composition and bulk material type. Due to hyperspectral's phenomenological properties, a mechanism exists to defeat hyperspectral TTI modalities: spec- tral background signature matching. For example, suppose a target of interest — perhaps a tank — is located in a jungle, and we have good measurements of vegetative spectral reflec- tance for the particular theater in which we are operating. We could then design paint for the tank that matches the spectral — or spatial-spectral, for imaging — signature of the jungle. However, polarimetric imaging still could potentially detect the paint because it has a different surface roughness and polari- metric signature compared to jungle vegetation. Polarimetric imaging excels in detecting: • Geometric shapes • Birefringent or other crystalline materials with anisotro- pic indices of refraction • Surface structure (roughness) and (a)symmetries of sur- face structure • Highly reflective or smooth surfaces • Aerosols when the particle shape distribution has a specific structure • Classes of polarization changing materials (optically active materials) • Surface orientation • Other manmade materials Polarimetric imaging complements hyperspectral imag- ing, and it can detect features that are invisible to pure hyperspectral sensors. In the long term, the best sensors for TTI will be hybrid polarimetric-hyperspectral sensors. However, hybrid sensors present a difficult engineering problem, and hyperspectral imaging sensors currently enjoy a greater maturity compared to polarimetric imag- ing sensors. Polarimetric sensors need both theoretical and engineering improvements before the technology is mature enough to be combined with hyperspectral imaging in a portable, deployable way. Our group already has shown that Mueller matrix mea- surements have a great deal of promise for identifying specif- ic targets of interest to the U.S. Army and the U.S. Air Force ii . Figure 5 shows an ROC curve of an Army target of interest vs. various background and clutter materials iii . This curve is computed on a per pixel basis. Note that this work shows potential for polarimetric TTI, but the underlying dataset is likely too sparse to generalize to a deployable instrument. Hindrances To Polarimetric Imaging In the optical wavelength regime (0.3-25µm), polarimetric information cannot be directly measured by optical sensors. Optical sensors only measure quantities proportional to the total irradiance at the sensor plane; they cannot directly measure the polarization information. Subsequently, optical polarimetric measurements require a scheme that can infer the polarization properties from the irradiance. Typically, this is accomplished via modulating the irradiance using special, known, polarization-altering optical elements. The modulation of the irradiance then can be inverted to obtain polarization information. For active instruments, modulation of both the polarization-controlled source and the camera viewing the object must be accomplished, increasing instru- ment complexity. Optical imaging polarimeters typically accomplish this modulation in four broad domains: temporal modulation, spatial modulation, spectral modulation and, most recently, angular modulation (a subset of spatial modulation) iv . There also are some hybrid modulation techniques v , and we have made major advances recently in spatio-temporal iv and gen- eral modulation techniques vii .These modulation techniques introduce complexity into instrument design and theoreti- cally limit combinations of instrument speed (temporal band- width), image resolution (spatial bandwidth), and spectral resolution. Electronic Military & Defense Annual Resource, 5th Edition 9 Figure 4: Rebellion Photonics hyperspectral camera sensing a methane gas leak. Image courtesy of Rebellion Photonics. Figure 5: ROC curve for a per pixel support vector machine classification on Mueller matrix data of an Army target of interest vs. background/clutter. The horizontal axis is log scaled, and the black line corresponds to the chance line. Data originally published in Proceedings of SPIE, Vol. 8364.