Inverse Methods For Three Dimensional Parametric Scattering Model Of Target

Scattering model generation of targets is a key issue of model-based synthetic aperture radar(SAR)automatic target recognition(ATR)system.Three-dimensional(3D)parametric scattering model of target,especially parametric model based on canonical shapes,coupling physical relevance to predictive utility,are potentially informative for ATR.The target model inversion from scattering measurements is a challenging task.This dissertation is dedicated to target 3D parametric scattering model inversion from multiple-aspect aperture measurements.The framework and several key techniques of model inversion are studied.Chapter 2 presents a flexible framework for inversing target model based on canonical shapes,which is a formalism consisting of an initialization and a parameter optimization,for model inversion.It can aggregate a variety of techniques to facilitate the model generation.Aiming at the model initialization,this dissertation focuses on the feature extraction method of 2D/3D scattering center.According to sparse representation and compressive sensing theory,chapter 3 proposes time-domain methods for scattering center extraction from wideband measurements and SAR images.And chapter 4 proposes three tractable time-domain sparse-representation based methods,in terms of memory and computational requirements,for 3D imaging and scattering center extraction.The proposed methods improve the performances by exploiting natures of time-domain responses of target and parametric scattering model.Chapter 5 proses three methods for 3D scattering center reconstruction from 2D scattering centers extracted from large-multi-baseline synthetic aperture data.The reconstructions coincide with the geometry of target.Aiming at parameter optimization of target parametric model based on canonical scattering models,Chapter 6 developers an image-domain cost function leading to a larger convergence region,and an optimization method based on image segmentation to improve the stability and efficiency.Chapter 7 is an implementation of the framework,in which approaches and a prototype system of global model inversion are presented.In the experiment,three global-azimuth width-elevation 3D parametric scattering models of a target are inversed from data produced by electromagnetism code.The SAR signature prediction ability and feather prediction ability of the inversed models are analyzed.The results exhibit the feasibility and effectiveness of our methodology.