ISAR高分辨成像和参数估计算法研究

发布时间：2018-01-12 10:44

本文关键词：ISAR高分辨成像和参数估计算法研究　出处：《西安电子科技大学》2016年博士论文　论文类型：学位论文

【摘要】：随着国民经济的提升和现代战争形式的转变,雷达成像技术凭借其远距离、全天时、全天候高分辨成像的独特优势,在民用和国防遥感领域发挥着不可替代的作用。作为对空间、空中及海洋观测最重要的手段之一,逆合成孔径雷达(Inverse Synthetic Aperture Radar,ISAR)成像是非合作目标识别的关键技术。为了满足日益增加的应用需求,ISAR正朝着多功能、多维度和协同网络等方向发展。工作模式和数据获取方式的多样化,以及目标运动的复杂性,使得现有的ISAR成像体制面临着高分辨成像和目标参数提取等挑战。在国家“973”计划课题等多个项目支持下,本文针对现有ISAR成像中存在的短孔径低分辨成像、非相干的稀疏孔径以及平稳或机动目标定标等问题开展研究,旨在增强ISAR图像分辨率和探讨稳定的参数估计方法,以提高雷达自动目标识别能力。研究内容主要包括以下几方面:(1)ISAR短孔径数据的稀疏高分辨成像由于雷达的多模式工作状态或目标的机动性,短孔径数据在ISAR成像中普遍存在。尽管成像方法简单高效,但同时限制了图像分辨率,影响目标识别性能。在分析ISAR回波模型和典型运动补偿算法的基础上,本文第二章基于压缩感知理论介绍了稀疏ISAR信号的重构方法。该方法重点从统计的角度提出了一种基于数据的稀疏约束参数估计方法。通过推导重构高分辨图像的正则化问题,稀疏约束参数可由最大似然估计得到解析形式。其中,噪声方差由粗图像中大量的噪声单元估计,权值由预处理的降噪图像进行初始化。利用优化求解的高分辨图像,再对稀疏约束参数进行更新并重新估计图像,多次循环提升算法性能。结果表明,迭代估计的稀疏约束参数能在重构高分辨图像过程中,较好地权衡信号逼真度和稀疏性。(2)稀疏孔径的相干化处理和高分辨成像除了雷达多样化的工作模式外,外界或系统的干扰也会造成数据的缺损,形成稀疏孔径。对于块状稀疏分布的子孔径,经过独立的包络对齐和自聚焦处理后,残余的线性相位和复幅度将在观测孔径之间存在差异。针对这种稀疏孔径之间的非相干性,本文第三章提出了一种相干化处理方法。将子孔径的包络按质心对齐后,对各子图像中的特显点单元建立全极点模型,并由求根MUSIC算法估计极点,计算子孔径间的多普勒偏移。以其中一个子孔径作为基准校正线性相位后,通过最小二乘(Least Square,LS)方法求解各子孔径的模型系数。由估计的极点和系数,将频偏校准后的子孔径分别进行前向和后向外推至整个孔径长度。再利用LS估计复幅度偏差并进行校正。在相干化处理后,构造部分FFT基作为观测矩阵,通过稀疏信号处理的方法对缺失孔径进行恢复。从实验结果可看出,经过相干化处理和空缺孔径重构后,成像效果得到了明显提升。(3)匀速转动目标的定标方法目标识别除了需要高分辨图像之外,还需要精确的目标尺寸参数。因此,本文的第四章和第五章针对匀速转动目标提出了两种不同的定标方法。第四章先分析了平动补偿过程引入的残余平动相位对不同定标方法的潜在影响。重点提出了一种利用散射点调频率相消的有效转动速度(Effective Rotaional Velocity,ERV)估计方法,该方法几乎不受残余平动的影响。通过选取散射点单元,并将其时域信号看作时变自回归模型。对短时数据段估计瞬时极点后,滑窗获得整个孔径的多普勒历程。根据关系式,ERV可以从不同距离单元的散射点调频率解算。转动相位补偿后,再进行自聚焦处理提高图像聚焦度。由于该方法依赖于提取的散射点质量,在信噪比(Signal-to-noise ratio,SNR)较低时估计性能会下降。为了提高定标算法的稳定性,第五章针对匀速转动目标提出了基于图像整体性能的ERV估计方法。该方法考虑残余平动的影响,将其作为参数联合ERV估计。通过迭代补偿二次相位误差,直到补偿后的图像幅度平方锐化度(Intensity-squared Sharpness,ISS)达到最大值,同时获得聚焦图像和ERV估计值。在此基础上,再对聚焦图像进行自聚焦处理,进一步聚焦图像。其中,ISS最大化问题是典型的非线性最小二乘问题,利用高斯牛顿的方法可实现高效求解。与多种方法的对比结果说明,这种从图像性能角度估计ERV的方法虽然效率中等,但具有较高的精度和鲁棒性。(4)非匀速转动目标的定标和高分辨成像方法由于匀加速转动引起的距离-方位二维耦合,增加了转动参数估计的难度。针对此类目标,本文第六章首先提出了一种基于匹配傅里叶变换(Matched Fourier Transform,MFT)的图像ISS最大化的定标方法。该方法将RD成像中FFT线性变换替换为参数化的MFT变换。通过对二维耦合相位迭代补偿,使得MFT图像的ISS最大,估计得到MFT调频率和ERV参数。该定标方法能在信噪比较低的条件下保持较好的估计性能。但强噪声或低分辨率仍影响目标识别,因此,利用部分MFT基和稀疏信号处理的方法进一步完成了高分辨图像的重构。
[Abstract]:With the promotion of national economy and the transformation of modern warfare, radar imaging technology with its unique advantages of long distance, all day long, all-weather, high resolution imaging, plays an irreplaceable role in civilian and military fields. As for space remote sensing, air and ocean observation is one of the most important means of inverse synthetic aperture radar (Inverse Synthetic Aperture Radar, ISAR) imaging is the key technology of non cooperative target recognition. In order to meet the application demand increasing, ISAR is moving towards multi function, multi dimension and collaborative network development. The diversification of working mode and data acquisition, and the complexity of target movement, the ISAR imaging system of existing face a high resolution imaging and target extraction and other challenges. In the national "973" project and other projects supported by the existing short aperture ISAR imaging in low Sparse aperture resolution imaging, non coherent and stable or maneuvering target calibration and so on, in order to enhance the resolution of ISAR image parameters and discuss the stability estimation method to improve radar automatic target recognition. The main research contents include the following aspects: (1) ISAR sparse short aperture data of high resolution imaging radar due to mobility the multi mode state of target, short aperture data exists in ISAR imaging. Although the imaging method is simple and efficient, but also limits the resolution of the image, affect the performance of target recognition. Based on analyzing the ISAR echo model and typical motion compensation algorithm, the second chapter introduces the theory of compressed sensing sparse signal reconstruction method ISAR based on this method. The key from a statistical point of view presents a method for estimating sparse constraint parameters based on data. By deducing the reconstruction of high resolution map As the regularization problem, we obtain the closed form estimation by maximum likelihood parameter sparsity constraint. The noise variance from coarse image noise estimation unit is large, denoising image pre-processing by weight initialization. Using high resolution image optimization, then the parameters are updated about sparse beam and re estimation of the image, many times to raise the performance of the proposed algorithm. The results show that the sparse constraint parameter iterative estimation to high resolution in image reconstruction process, better balanced signal fidelity and sparsity. (2) coherent processing and high resolution imaging radar in addition to the work mode of diversification outside the sparse aperture, outside interference or system will defect data by forming a sparse aperture. For the block sparse sub aperture, the envelope alignment independent and self focusing treatment, residual linear phase and amplitude in complex observation aperture There are differences between. For the non coherence between the sparse aperture, the third chapter of this paper presents a coherent processing method. The envelope sub aperture according to centroid alignment after the establishment of the all pole model of each sub image in the display unit, and by the root MUSIC algorithm to estimate the pole, calculation of Doppler shift between the sub apertures in one of the sub aperture. As the benchmark calibration of linear phase, by least squares (Least Square LS) method for solving the model coefficients of each subaperture. By pole and coefficient estimation, frequency offset will be calibrated separately to the sub aperture before and after outwards to the whole aperture length. Then using LS to estimate complex the amplitude deviation is corrected. In coherent processing, construction part of FFT base as the observation matrix, the missing aperture can be restored by method of sparse signal processing. Can be seen from the experimental results, the coherent processing and Vacant aperture reconstruction imaging effect is improved obviously. (3) the uniform rotation target calibration method for target recognition in addition to the high resolution image, but also need to target size parameters accurately. Therefore, the fourth chapter and the fifth chapter according to the uniform rotation target presents two different calibration methods. The fourth chapter the first analysis of the potential impact of different calibration methods of residual translational phase motion compensation process introduced. Put forward an effective rotation speed using scattering point frequency cancellation (Effective Rotaional, Velocity, ERV) estimation method, this method is almost not affected by the residual translation. By selecting scatter point unit, and the the time domain signal as a time-varying autoregressive model. To estimate the short-term data segment instantaneous pole, Doppler won the whole course of sliding window aperture. According to the relationship, ERV from different distance unit Scattering point adjustable frequency calculation. The rotational phase compensation, then the autofocus processing to improve image focusing degree. Because of the scattering method depends on the quality of extracting, the signal-to-noise ratio (Signal-to-noise ratio SNR) low estimation performance will decrease. In order to improve the stability of the calibration algorithm, the fifth chapter for the uniform the overall performance of image rotation target ERV estimation method based on this method. Considering the influence of residual motion, the parameters of joint ERV estimation. Through iterative compensation two phase error, until after the compensation of the image sharpening square amplitude (Intensity-squared Sharpness, ISS) reached the maximum value at the same time, to obtain a focused image and ERV estimation. On this basis, the focused image autofocus processing, further focus image. Among them, ISS maximization problem is a typical nonlinear least squares problem, using Gauss Newton's method can To achieve efficient solution. Compared with other methods. The results show that the performance of the image from the angle estimation method of ERV although the efficiency of medium, but has high precision and robustness. (4) calibration of non uniform rotation of targets and high resolution imaging method with uniform acceleration caused by rotation of the distance and azimuth of two-dimensional coupled, increased the rotation parameter estimation difficult. For this goal, the sixth chapter puts forward a matching based on Fourier transform (Matched Fourier Transform, MFT) the calibration method of ISS image maximum. This method combines MFT transform RD imaging FFT linear transformation to replace parameters. Based on two-dimensional coupled iterative phase compensation the MFT image, the maximum ISS, the estimated MFT modulation frequency and ERV parameters. The calibration method can maintain a good estimation performance in the low SNR condition. But the strong noise or low resolution still affect the target Therefore, the reconstruction of high resolution images is further completed by using the method of partial MFT based and sparse signal processing.

【学位授予单位】：西安电子科技大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN957.52

【相似文献】