Theory And Methods Of Extraction Based On The Optical Zone Radar Target 2d Image Of The Target Scattering Characteristics

Under the high-resolution imaging radar background, extracting electromagnetic scattering features of target from 2-D radar data, is a task valuable for both theoretical research and practical usage in field of radar target recognition. In this paper, to probe into key problems of feature extraction, a systematic study is performed not only about understanding and processing radar target image, but also parametric modeling of 2-D target scattering centers and parameter estimation.According to scattering phenomenology, we point out that the commonly used point scattering models fails to accurately describe the complex scattering behavior of real target. For that reason, based on the theory of effective scattering centers, researches are taken on more accurately physical-relevant description of target scattering and on approach that is more effective to characterizing and extracting target scattering features by means of time-frequency analysis and scattering centers modeling.Setting forth the concept of scattering model mismatch, we make a detailed analysis of their effects on radar imaging, that is, resulting in one- and two-dimensional artifacts or dispersion in radar image which we view as a kind of focus-out effect on the focusing process in radar imaging. Thus, by inferring and deducing the 2-D effective focus-out phase error function, a framework of analyzing and describing the scattering model mismatch in frequency domain is formed. On the other hand, in image domain, spatially varying point-spread function (SVPSF) is defined to demonstrate how the effects of complex scattering centers which cause scattering model mismatch can be interpreted as being due to a SVPSF acting on a ideal scatterer. According to the concept and basic form of SVPSF, the cavity and duct scattering is examined, and a method is developed to remove the duct-induce dispersion effect.Using the joint time-frequency (JTF) analysis, a method of understanding and processing ISAR image based on adaptive Gauss Representation is presented. First ,VM-AGR algorithm and variance-correcting method are proposed to greatly improve the performance of original AGR algorithm. Then, a conceptual framework of applying JTF processing engine AGR to radar target image is formed, which illustrates the ways to automatically extract physics-based features of both point scatterers and nonpoint-scattering mechanisms from the target image, and to clean ISAR image by removing the nonpoint-scattering artifacts. The method is demonstrated by using real data of airplane. The airplane ISAR image is evidently enhanced, and the physics-based feartures of scattering centers is gained by the method.Since the problem of feature extraction from 2-D radar data can be approached as a parametric modeling and parameter estimation problem, the 2-D Type-Discriminated Geometric Theory of Diffraction based (TD-GTD) model that discriminate between the localized and distributed scattering centers of the target is presented. This scattering model balances physical fidelity with simplicity in functional form to yield both smaller modeling error and a more physical-relevant and richer description of complex scattering behavior compared to either point scattering models or Prony models. To estimate the TD-GTD model parameters, Maximum Likelihood Estimation (MLE) algorithm is developed. An effective method of parameter initialization and the Decoupled Parameter MLE algorithm that solves the problem of mixed parameter estimation are proposed. Since estimating parameters of all scattering centers at the same time is the cause of high computational cost and complexity, RELAX technique is introduced into MLE algorithms to get around that difficulty. The efficiency and feasibility of the method is confirmed by the test based on the measured data.