Model-based Parametric SAR Image Formation Technique
|School||Shanghai Jiaotong University|
|Course||Signal and Information Processing|
|Keywords||synthetic aperture radar (SAR) SAR image formationalgorithm parameter estimation atomic decomposition Cramer-Raobound|
Synthetic aperture radar (SAR) is a high-resolution imaging radarsystem. With the use of the relative motion between the radar platform andthe target region, it can generate high-resolution SAR image in both rangeand azimuth dimensions. It can also achieve all-weather all-time earthobservation and imaging.Conventional SAR image formation algorithms are developed based onthe data collection geometry between the radar platform and the target.The signal backscattered from the target can be focused by analyzing thesignal phase, furthermore, by using the signal processing techniques suchas matched filtering. With the development of the SAR technique, theapplications of the SAR become wider, and it also results in greatchallenges to the SAR signal processing. First, precise motion compensation algorithms and motion measurementmethods are needed. Conventional SAR image formation algorithms arebased on the geometry between the radar platform and the target, thequality of the SAR image is directly affected by the deviation between thepresumed geometry and the real one. This geometry deviation can becompensated by the precise motion compensation algorithms and motionmeasurement methods. And then, high-quality well-focused images can beobtained. These two requirements are important, especially in the case ofthe image formation of the high-resolution SAR system or themaneuvering platform.Second, image formation algorithms are needed for high-resolutionimaging. In general, the radar resolution is determined by the signalbandwidth. The range resolution is determined by the transmitted signalbandwidth, that is, large transmitted signal bandwidth can obtain highrange resolution. The high azimuth resolution can be obtained byincreasing Doppler bandwidth. For the high-resolution SAR system, theazimuth resolution is more important. However, larger azimuth bandwidthwill result in larger range migration, which must be corrected by thehigh-resolution image formation algorithms.Third, variable motion, orbit curvature and earth curvature must be considered in the image formation. Different from the common SARplatform, the trajectory of the compact platform is more complex, since itis susceptible to external disturbation. It is impossible to provide precisemotion measurement, because it is difficult to mount motion measurementsystem on a compact platform. It is also a great challenge to the imageformation of the maneuvering platform. The orbit curvature and earthcurvature must be considered in spaceborne SAR image formation,especially in the case of the geosynchronous SAR.Finally, the study on target feature is needed. Here, the study on targetfeature includes the analysis of the scattering mechanism and the targetmotion. The SAR image can be viewed as a projection of the RCS of thetargets in the observed region. That is, the SAR image can be formed byletting the RCS of the observed region through the impulse response of theSAR system. However, target scattering mechanism is overlooked by theconventional SAR image formation algorithms, the effect of the scatteringmechanism to the image formation is not considered. Thus, the SARimage cannot provide the information of the target scattering mechanism,this is adverse to SAR image interpretation. In addition, the focusingparameters of the stationary targets and the moving targets are different,uniform signal processing cannot focus the stationary targets and the moving targets simultaneously.High maneuvering results in great challenges to imaging processing,especially in the case of high-resolution imaging. Meanwhile, targetfeature is overlooked by the conventional imaging algorithms, this willrequire more efforts to image interpretation, and will also result in themissing of target details. This paper concentrates on the model of thebackscattered signal. With the help of the signal model, the signalcomponents can be distinguished by the parameter estimation techniques,each component can then be focused separately according to its ownparameter(s). The achievements are described as follows.First, the model-based parametric SAR image formation technique isproposed. This technique does not require explicit knowledge of the datacollection geometry, it can focus the target even when the platform motionis unknown. The target that can be described by the model is treated as theminimum imaging unit, each unit in the imaging region can becharacterized by the model parameters. Thus, each unit can be focusedprogressively. This technique also provides an approach to target featureextraction and recognition from a signal-level point of view.Second, the atomic decomposition-based image formation algorithm isproposed. It is a realization of the model-based parametric SAR image formation technique. Atomic decomposition-based image formationalgorithm can focus the scatterers, and it can be also useful in the case ofbackground noise suppression and highlighting the target of interest.Third, the model-based adaptive SAR image formation algorithm isproposed. The model parameters can be estimated along the rangemigration curve adaptively, the parameters of each scatterer can beobtained by maximizing the cost function. The final SAR image isgenerated by projecting the estimated parameters to the grid formed by theiso-range and iso-Doppler lines. The scatterers can be well focused byusing this algorithm, and the sidelobe of the impulse response can also bereduced.Finally, the effect of the target-induced azimuth envelope to the imageformation is discussed, an improved CCT-based envelope estimationalgorithm and an atomic decomposition-based envelope estimationalgorithm are proposed.