Principles of Modern Radar - Volume 2 1891121537

14.5 Inverse Synthetic Aperture Radar 653calculating using the Doppler extent of the height features beyond the edge of the deck.Multiple image frames are necessary to calculate the real roll and pitch of a ship [69].14.5.1.3 Using High Range Resolution ProfilesMuch of the formation and utility of HRR profiles will be discussed later in this chapter,but this section will briefly describe their extraction from SAR images. HRR profilesbecome useful with ground target identification when irregular motion of a maneuveringtarget distorts SAR imagery. While HRR profiles processed directly from raw radar datawill include an abundance of ground clutter, Doppler processing in the SAR image formationprocess removes much of the ground clutter from HRR profiles extracted along thedown-range axis of SAR images. The down-range axis can be determined by the shadoworientation in the image. A cross-correlation template matching technique can be appliedto classify targets based on HRR profiles; the performance of this method can be improvedby fusing data from multiple looks after classifying each individually [21].14.5.1.3 Using Different Frequency SubspacesThe bandwidth and center frequency of pulses influence the ATR performance on theresulting ISAR imagery. Linear chirped pulses with equal bandwidths but different centerfrequencies yield different recognition rates. However, information from chirps at differentcenter frequencies can be combined to give a higher ATR response than the full bandwidthreconstruction with less processing and target movement knowledge required [2].14.5.1.4 Time Frequency ProcessingThe short-time Fourier transform or the time-frequency distribution series can be used toform continuous range-Doppler images from overlapped ISAR images [67]. This techniqueallows high-resolution images to be formed at different times so that the rotationof the image through time can be monitored [67]. Radar frequency – fast time imagescan also be formed using the fast Gabor transform. This technique yields several otheradvantages, including: better target dimension estimation accuracy, better target rotationestimation, and the possible development of additional features in Doppler-time or radarfrequency – fast time images [67].14.5.2 Scattering ModelsStatistically independent noise fluctuations exist between images of a moving target inISAR imagery [70]. Additionally, the low rotational speed and weight of the target canlead to irregular sampling in azimuth and, thus, speckle noise and ambiguities [24]. Todeal with these error sources, scattering centers are often extracted from ISAR images andused in subsequent recognition steps.14.5.2.1 Filtering and Phase AnalysisScattering centers can be extracted from ISAR and SAR imagery using filtering operations.For example, spatial variant apodization (SVA) seeks to achieve maximum spatialresolution while reducing sidelobes. By applying this algorithm to an ISAR image followedby phase analysis to extract the most robust scattering centers, a binary image ofthe dominant scatterers in a target can be constructed [24].