Principles of Modern Radar - Volume 2 1891121537

652 CHAPTER 14 Automatic Target Recognition14.4.5 Summary of Synthetic Aperture RadarThe steps in the unified framework for ATR using synthetic aperture radar imagery havebeen discussed. Samplings of available datasets, feature types, pre-screening techniques,discrimination algorithms, and classification approaches found in the literature have beenpresented with brief summaries. This synopsis is meant to provide the reader with anappreciation for the breadth of techniques present in the SAR ATR literature and anunderstanding of how various algorithms fit within the Unified Framework.14.5 INVERSE SYNTHETIC APERTURE RADARMany of the ATR challenges and algorithms associated with SAR imagery are similarto those in ISAR imagery. Hence, rather than go through all four steps again (as in Section14.4), we highlight the issues and techniques that are either unique to ISAR imagesor of particular concern when the target itself is moving, as in the ISAR imaging case.14.5.1 Image FormationThis section describes a few challenges and alternatives in the ISAR image formationprocess that can make a significant impact on the performance of ATR.14.5.1.1 Motion CompensationMotion compensation for small water craft with more than one type of rotation (roll, pitch,or yaw) is particularly difficult [67]. One algorithm for this process involves establishingand tracking an image central reference point in range and Doppler and determining therotation around the central reference point by tracking a secondary point [67]. Anotheralgorithm uses Doppler tracking to remove higher frequency line-of-sight phase errors andstabilize the Range Track Doppler beam sharpening filters, range tracking to remove theline of sight range motion through range gates, and autofocus to remove low-frequencyquadratic phase error [67]. Without adequate motion compensation, the resultant imagemay be improperly scaled or ill-focused [67], both of which present challenges to an ATRalgorithm.14.5.1.2 Estimating Angular Velocity and AccelerationBy assuming a rigid object and a low grazing angle, such that the line of sight is assumedto be in the horizontal plane, the target will exhibit angular motion about the centroid.Using this angular velocity, the instantaneous velocity in range can be calculated as thecross product of the distance between the point scatterers that make up the target and itscentroid with the angular velocity projected into the range direction. The acceleration ofthe target is then simply the derivative of the instantaneous velocity. By converting theexpression for acceleration into the coordinate system of the ISAR focus plane (range,cross-range, and displacement perpendicular to the slant plane but unobservable in ISAR)and applying the scale factor to convert cross-range to Doppler frequency, the partialderivatives of radial acceleration with respect to range and frequency can be expressed interms of physical target parameters [68]. If these physical quantities are combined withestimation techniques, such as Kalman filters, the global angular velocity and accelerationof a rigid target can be estimated [68].For ships, the yaw rate can be calculated from the Doppler difference between thebow and the stern, and the length of the ship. Similarly, an effective roll and pitch can be