Study Of Two-dimensional And Three-dimensional Inverse Synthetic Aperture Radar Imaging Methods

Inverse synthetic aperture radar (ISAR) imaging has found wide applications in space situation surveillance and air defense anti-missile areas since it can not only obtain the one dimensional (1-D), two dimensional (2-D) and three dimensional (3-D) high resolution images, but also provide the size and geometry of the observed target. After more than 60 years of development, the basic principles of ISAR imaging have been well formulated. However, there exist many unsolved problems in 2-D and 3-D ISAR imaging because of updating ISAR systems and increasing observation requirements. Firstly, with higher operating frequency, ISAR is more sensitive to the phase noise induced by the nonlinear property of the transmitting and receiving chains. Therefore, the final image quality can be degraded seriously due to the range phase errors. Secondly, long observation range and small radar cross section (RCS) of the target can make the signal-to-noise ratio (SNR) of the received echoes rather low, which results in conventional translational motion compensation methods failing to work. Thirdly, at specific observation scene, e.g. several planes flying in a formation, multi-targets may emerge in the same antenna beam where the echoes of different targets overlap each other seriously. Hence, how to obtain the 2-D high resolution images of the observed targets with range phase errors at different imaging scenes is a research hot topic. Finally, based on the sequential ISAR images, precise image scaling and the following target 3-D geometry reconstruction can facilitate the applications of ISAR images in target recognition and classification areas. Therefore, effective 3-D reconstruction and image scaling of ISAR target are also the challenges of ISAR imaging researches. Hence, the research of this dissertation has important theoretical significance and application value. Under the support of the National Science Foundation of China, the National High Technology Research and Development Program of China and lateral researches, this dissertation studies the range phase errors estimation and compensation, ISAR imaging under low SNR, multi-targets ISAR imaging, image scaling of ISAR images and target 3-D reconstruction etc. Well focused imaging results can be obtained to improve the information capturing ability of ISAR. The main content of this dissertation is summarized as follows.1. The first part introduces the basic principles of ISAR imaging. The ISAR imaging and signal models are formulated firstly, and then the physical meaning of translational motion compensation is illustrated and clarified. Classical translational motion compensation methods such as range alignment and autofocus are described in detail. Next, the migration through range cells (MTRC) is discussed, including the influence and one order Keystone based compensation method. In a word, this part is the basis of the following discussion.2. The second part proposes a precise and effective range phase errors compensation method based on the minimum entropy and range autofocus processing to deal with the problem that nonlinear range phase errors of high frequency band ISAR degrades the final image quality seriously. Firstly, the range phase errors are modeled as a high order polynomial. Then, the influences of different order phase errors on the quality of range compression results are analyzed in detail. Next, minimum entropy is used as the cost function and heuristic search is utilized to estimate and compensate second order phase error. After that, the third or higher orders phase errors are corrected motivated by the minimum entropy autofocus method in cross-range dimension. Finally, simulation results verify the validity of the proposed method.3. The third part proposes an adaptive ISAR translational motion compensation method under low SNR based on frequency spectrum phase differentials and particle swarm optimization (PSO), and well focused images of the real data are obtained. Firstly, based on coherent raw ISAR echoes, the translational motion is modeled as a high order polynomial which is sufficient to describe various kinds of target motion. Then, frequency spectrum phase differentials of each echo are made to estimate the corresponding polynomial coefficients coarsely. The coarse values are used as the initialization of the following precise translational motion estimation method to improve the efficiency. Next, PSO is utilized to estimate the translational motion with high precision and efficiency. After that, by analyzing the variation of final image quality with the model precision, the polynomial order of translational motion can be determined adaptively. The experimental results of simulated and real data show the validity of the proposed method.4. The fourth part proposes a novel multi-targets ISAR imaging method based on PSO and modified CLEAN (M-CLEAN) technique to deal with the problem that the echoes of different targets overlap each other at specific imaging scenes, e.g., planes flying in a formation. Traditional multi-targets imaging methods consider the motion of different targets as completely distinguishable or almost the same, which are clearly not current. In this part, multi-targets are first divided into separated group-targets, in which each target shares analogous motion. Then, the translational motion of each group-target is modeled as a polynomial and PSO is utilized to estimate the polynomial coefficients because of its high efficiency and model simplicity. After obtaining coarsely focused image of one group-target, M-CLEAN technique is proposed to extract it. The proposed M-CLEAN can keep the integrity of extracted group-target with fewer spurious points and lower noise. Finally, clustering number estimation and K-means algorithm are utilized to separate each target, of which refined image can be obtained sequentially via conventional single target ISAR imaging processing. The validity of the proposed method is validated by the simulation results.5. The fifth part proposes a novel image scaling method based on discrete polynomial-phase transform (DPT) and the analysis of the analytical expression of the rotational angular velocity and second order phase coefficients of strong scatterers. Precise scaled images of Yak42 data can be obtained. Firstly, the range cell echoes are analyzed in detail and the analytical expression of two-order phase coefficient of each strong scatterer is derived, which can be computed based on the equivalent rotational angular velocity. Then, the normalized amplitude variances of all range cell echoes are computed and sorted. Several range cells e.g.8 or 10 with the lowest normalized amplitude variances are chosen. Next, frequency windowing technique is applied to extract the frequency spectrum of dominant scatterers. After that, the second order DPT is applied to estimate the two-order phase coefficient of each range cell echo with high efficiency. Finally, equivalent rotational angular velocity can be obtained via least square error (LSE) estimation and the cross-range scaling is achieved based on the computed cross-range resolution. The experimental results of simulated and real data verify the validity of the proposed method.6. The last part proposes a novel joint 3-D geometry reconstruction and image scaling method of ISAR target based on modified factorization method with relaxed constraints to deal with the problem that conventional sequential ISAR images based 3-D geometry reconstruction methods cannot achieve image scaling of ISAR images. Firstly, the models of target’s 3-D movement and ISAR imaging are built and then the projection equations of 3-D target geometry on the imaging plane are derived. Next, wide angle echoes are divided into small sub-apertures to decrease the difficulty of directly 2-D imaging. Then, sequential ISAR images can be obtained via imaging processing of each sub-aperture. We can acquire the trajectory matrix via the extraction and correlation of each scatterer from sequential ISAR images. After that, the factorization method with relaxed constraints is applied to the trajectory matrix to obtain the initial 3-D target geometry and projection matrix. Next, the rotational motion parameters of the observed target are estimated based on the projection matrix. The rescaling of cross-range positions of the trajectory matrix can then be achieved. Repeated utilization of factorization with relaxed constraints and rotational motion parameters estimation can improve the precision. The experimental results of simulated data verify the validity of the proposed method.