通过雷达目标回波信号的分析和处理，可实现从雷达散射截面积(Radar Cross Section，RCs)序列、径向距离、多普勒、极化等特征信号中反演出目标运动轨迹、速度、结构和表面特性等物理属性，为目标的分类、识别提供重要依据。空间非合作目标的微动特征是对目标特征的精细刻画，为目标识别提供了新的、有效的途径。本文围绕空间微动目标特征提取和重构技术，基于微动目标回波的建模分析，在窄带和宽带两种雷达体制下对微动特征曲线提取、参数估计以及目标重构方法进行了研究。论文主要创新点如下： 1)针对平动分量引起微多普勒谱混叠问题，提出了一种信号域线性调频锯齿波近似和图像域形态学处理相结合的平动补偿方法，有效提升了目标频谱混叠情况下的平动补偿性能。 微动目标回波中存在的平动分量将造成距离上的整体平移，而在相位上附加一个共有的平动相位项，进而影响到距离维和频率维观测到的微动特征，给微动特征提取带来不便。因此，在实际处理中需要研究复杂环境下有效而精确的平动补偿算法，如在相对较大的平动调制下微多普勒谱发生混叠的情况。针对微多普勒谱混叠情况，提出了一种信号域和图像域处理相结合的抗混叠平动补偿算法。先将混叠的瞬时频率峰值曲线近似为线性调频锯齿波信号，通过基频信号的参数估计可快速减少混叠周期数。进而，在频率维周期扩展的时频图上基于形态学图像处理实现了无混叠瞬时频率的有效提取和平动补偿。所提方法对目标的运动形式以及噪声有较好的鲁棒性，且参数设置简单。 2)针对低重频和小视线角观测条件下目标微动周期估计问题，提出了一种基于逆合成孔径雷达(Inverse Synthetic Aperture Radar,ISAR)像序列周期匹配特征的微动周期估计方法，提升了小幅微动目标周期估计的稳健性，同时具有较高的精度。 当目标周期性的微运动时，其回波中所有的微动调制特征都将具有相同的周期性。因此，微动周期通常作为最优先估计的参数，成为微动参数估计的起始步骤或者作为已知量出现在参数估计算法中。然而，较低的脉冲重复频率(Pulse Repetition Frequence， PRF)带来的频谱混叠以及较小的雷达视线角(Line of Sight，LOS)变化造成的散射点(区域)小幅径向走动将不利于单一的频率维或距离维处理。为此，利用ISAR像的二维分辨能力，提出了一种基于图像序列周期匹配特征的微动周期估计方法。基于最小图像熵准则的基准图像选取提高了算法精度，并通过构建基于全图像的匹配函数以及图像的动态范围压缩降低了算法对ISAR图像质量的要求。 3)针对锥体目标，提出了一种基于ISAR像序列分析的微动参数估计方法，可有效估计目标的结构参数和微运动参数，提升了无弹道等先验信息下单部雷达处理的适用性。 旋转对称的锥体是空间目标的一种典型结构，现有参数估计方法通常伴随着多视角观测、平均视线角已知等约束条件。而在实际应用中受到雷达资源和数据同步的限制，这些约束条件时常是难以满足的。为此，通过引入进动下的等效转速估计，提出了一种基于ISAR像序列分析的微动参数估计方法。首先，利用等效转速和瞬时径向速度建立二维参数搜索空间中的特征函数。由于特征函数在搜索空间中的能量主要集中在一条曲线上，因此参数搜索的维度可近似降至一维，相比于直接的高维参数搜索有效提升了运算效率。进而基于雷达视线角重构，实现了搜索空间中二参数耦合的匹配参数集解耦，最终完成了锥体长度、底面半径等结构参数和进动角、平均视线角等进动参数的估计。 4)针对基于多视角窄带雷达回波的目标重构问题，提出了一种旋转平面投影最优匹配的转动目标重构方法，以及一种基于刚体目标几何不变性的横向距离因式分解法，在窄带雷达目标重构方面具有参考价值。 窄带雷达通常拥有较强的发射能量，同时当宽带距离分辨力不满足算法要求时，有必要进行窄带处理或宽-窄带联合处理以保证算法有效性。而受限于较低的距离分辨力，窄带雷达处理通常需要引入多视角观测数据以提供额外信息。 针对转动目标，在锥形结构约束下提出了一种基于旋转平面投影最优匹配的目标重构方法。通过对任意个数的锥底散射点进行统一建模，在锥顶／底两参数模型下实现了目标重构过程的简化。针对传统的基于径向距离因式分解的目标重构方法，在窄带场景下提出了一种横向距离因式分解法。通过引入微动视线角模型，基于视线角、加速度理论值和重构值间的最优匹配，在实现横向定标的同时完成了目标的重构。同时，在不同的视线角模型下，算法还可扩展到其他微动形式中。 关键词：微动；特征提取；目标重构；平动补偿；旋转匹配；多视角观测；等效转速估计；横向距离因式分解

 Based on the radar signal analysis and processing, the physical properties such as target trajectory, velocity, structure and surface characteristics may be reconstructed from radar cross section (RCS) sequence, radial range, Doppler frequence, polarization and other characteristic signals. It provides important basis for target classification and recognition. The micro-motion feature is a fine description for the space targets, which provides a new and efficient way for the recognition of non-cooperative targets. Focuse on the feature extraction and target reconstruction of the micro-motion space targets, the dissertation studies the methods of micro-motion feature curve extraction, parameter estimation and target reconstruction under the narrow-band and wideband radar system. The main innovations of the dissertation are summarized as follows: 1)To deal with the problem of micro-Doppler spectrum aliasing, a translational compensation method which combines the approximation of linear frequency modulation (LFM) sawtooth wave in signal domain and morphological processing in image domain is proposed, which may effectively improve the compensation performance. The translational component of the micro-motion target will cause a radial range shift and add a polynomnial phase term in the phase, which may destroy the micro-motion feature in both range and frequency domains. Therefore, the effective and accurate translation compensation method in the complicated condition, such as the micro-Doppler spectrum aliasing which may caused by the overlarge translation, is still a problem. In view of this, a translational compensation method which combines signal processing and image processing is proposed. Firstly, the instantaneous frequency of the aliasing micro-Doppler is approximated as a linear frequency modulated sawtooth signal, and the number of aliasing period is rapidly reduced by the parameter estimation of base frequency signal. Then, the morphological image processing is applied to the extension time-frequence (TF) image. Finally, the translation componet may be compensated and the instantaneous frequency without aliasing may be extracted. The proposed method is robust to the micro-motion forms and noise, and the initial parameters can be set simply. 2)Under the low pulse repetition frequence (PRF) and small radar line of sight (LOS) change, a micro-motion period estimation method is proposed, which is based on the period matching characteristics of inverse synthetic aperture radar (ISAR) image sequence. It improves the robustness and precision of period estimation with the small micro-motion target. With the periodically moving target, all the micro-motion modulation characteristics in the echo will have a same micro-motion period. Therefore, the period is often estimated in the first step or to be a known quantity appears in other parameter estimation algorithm. But the low PRF and small radar LOS change may cause spectrum aliasing and small radial migration, which may not conducive to simplex range or frequency process. Take advantage of the two-dimensional resolution of ISAR image, a micro-motion period estimation method is proposed, which is based on the period matching characteristics of ISAR image sequence. Based on the reference image selection with the minimum image entropy, the period estimation precision may be improved, and the algorithm requirement of the ISAR image quality is reduced by a full image based matching function and the image dynamical compressing process. 3)For cone targets, a parameter estimation algorithm which is based on the analysis of the micro-motion features in ISAR image sequences is proposed. It may estimate the structure and motion parameters effectively, and is suitable for single-station radar processing under unpredicted conditions. Rotational symmetric cone is a typical structure of space target, and the corresponding parameter estimation methods are mainly accompanied by multiple viewing angles, known average LOS and other constraints. In practice, these constraints are often difficult to meet due to the limitations of radar resources and data synchronization. Aiming to this, the rotational speed estimation is introduced into the parameter estimation procedure, and an estimation algorithm is proposed which is based on the analysis of the micro-motion features in ISAR image sequences. Firstly, an eigenfunction which is composed of the rotational speed and the radial velocity is built in the two-dimensional searching plane. Since the energy of the eigenfunction in the searching plane is mainly concentrated on a curve, the parameter searching in the algorithm may be reduced to a one-dimensional process, which may greatly reduces the computation burden compared to the traditional high-dimensional searching. Then, based on the matching degree of the reconstructed radar LOS, the matching parameter set which is two-parameter coupled in the searching plane may be decoupled, and the structure and motion parameters, such as the radial length, bottom radius, precession angle and average LOS, are estimated precisely. 4)Aiming to the problem of target reconstruction based on multi-view narrowband radar echo, a method of rotating target reconstruction based on the optimal matching of rotational plane projections and a method of cross range factorization based on the geometric invariance of rigid body are proposed, which have reference value in the aspect of target reconstruction in narrowband radar system. Narrowband radar system usually has strong transmitting energy, and it is necessary to conduct narrowband or narrowband-wideband processing to ensure the effectiveness of the target reconstruction algorithm when the range resolution does not meet the requirements. Due to the low range resolution, narrowband radar usually needs to introduce multi-view observation data to provide additional information. Under the constraint of cone shape target, a rotating target reconstruction method which is based on the optimal matching of rotational plane projections is proposed. By uniformly expressing and solving the locations of the scattering points on the cone bottom, the reconstruction procedure may be simplified under a cone top/bottom two-parameter model. To extend the traditional target reconstruction method with radial range factorization to the narrowband scence, cross range factorization method is proposed. Under the radar LOS model of micro-motion target, the target reconstruction and the cross range calibaration may be completed at the same time, based on the optimal matching between the theoretical and reconstructed value of the LOS and acceleration. Meanwhile, the method may be extended to other micro-motion forms based on the different radar LOS model. Keywords: Micro-motion; Feature Extraction; Target Reconstruction; Translation Compensation; Rotational Matching; Multi-view Observation; Equivalent Rotation Velocity Estimation; Cross Range Sigular Value Decomposition