雷达成像具有远距离、全天时和全天候的特点，对军用和民用都有重大实用价值。由于雷达的工作及成像方式的不同，成像雷达可分为逆合成孔径雷达(ISAR)和合成孔径雷达(SAR)。在机动目标的ISAR成像中，经过运动补偿后的回波数据可用多分量多项式相位信号模型来描述，这是一类典型的非平稳信号。实际对目标进行成像处理时可根据目标的运动特点对这一模型进行合理的近似，而成像质量的高低在很大程度上取决于信号处理方法的优劣，即可采用信号处理的方法来提高机动目标 ISAR成像质量。本文基于时频分析原理，从参数估计与时频表示两个方面将目标回波信号分为四种情况进行研究，分别是：常幅线性调频信号、常幅多项式相位信号、变幅多项式相位信号和基于时频分布的方法直接得到信号的时频结构，分别提出新的信号处理方法以提高机动目标ISAR成像质量。本文主要工作概述如下：　　1.以飞机目标的外场实测数据为例，研究了一阶近似条件下目标做机动飞行时 ISAR成像的距离——瞬时多普勒法原理，其本质为多分量线性调频信号的参数估计问题。进而，研究两种估计线性调频信号参数的超分辨方法在 ISAR成像中的应用，在获得同样的横向分辨率条件下，减小了成像所需的观测角度。最后对目标机动运动情况下的回波信号特点做出总结性归纳，并从信号处理的角度对其进行理论分析，进而达到统领全文的目的。　　2.针对常幅线性调频信号，提出了一种新的线性调频信号参数估计方法——三次相位函数法。新方法通过一维搜索可以得到调频斜率的估计，进而通过解调频技术和快速傅里叶变换获得信号初始频率和幅度的估计，具有较小的计算量。基于一阶扰动分析原理详细推导了该方法的渐进统计特性，理论与仿真结果均表明该方法具有很高的参数估计精度。将该方法应用于机动飞行目标的ISAR成像中，可提高成像质量。　　3.针对常幅多项式相位信号，首先提出基于改进形式 Wigner-Ville分布(WVD)的三、四次多项式相位信号参数估计方法，解决了 ISAR成像中的转角估计问题。对于任意阶次的多项式相位信号，本文研究了一个重要概念——瞬时频率变化率，提出两种估计瞬时频率变化率的新方法，并将其应用于多项式相位信号的参数估计中，在保持较高参数估计精度的同时具有较小的计算量。　　4.针对变幅多项式相位信号，提出采用信号分解的方法对其进行分析与处理。信号分解是信号处理领域的重要研究内容，通过将信号分解成一系列基函数的线性组合，可以反映信号的内部结构特征。本文研究了目前比较常用的信号的自适应 Chirplet分解，首先提出一种优化分解算法，即采用RELAX技术在估计当前分量的参数时，固定其余分量的参数，逐个依次调整，以达到内部收敛，该方法可明显提高信号的参数估计精度，但计算量较大。本文提出一种自适应 Chirplet分解的快速算法，该算法利用计算信号的三次相位函数，得到其能量分布集中于信号的调频率曲线上的结论，此时通过谱峰检测可同时获得 Chirplet调频率、时间中心和幅度的估计，进而通过解线性调频技术获得其初始频率和时间宽度的估计。研究了自适应 Chirplet分解在如下两个方面的应用：机动飞行目标的ISAR成像和非刚体目标的ISAR成像。　　针对自适应 Chirplet分解的不足，提出了修正的自适应 Chirplet分解法，将 Chirplet基函数推广到三次相位信号的形式，可以很好地逼近信号中的非线性时变结构成分，给出了相应的实用算法与快速算法。最后研究了基于修正自适应Chirplet分解的机动目标ISAR成像方法。　　5.研究信号的时频分布。对于多分量多项式相位信号，提出三种时频分布设计新方案，抑制了交叉项的影响，同时保持较高的时频聚集性；提出传统多项式 Wigner-Ville分布和四阶复时间延迟型多项式 Wigner-Ville分布的频域卷积实现方法，使其适用于分析多分量信号。最后，研究了基于新型时频分布的机动目标ISAR成像方法。　　6.研究外场实测数据的舰船目标 ISAR成像。提出采用(修正)自适应Chirplet分解法或时频分布法得到目标的瞬态像，同时针对多径效应问题，提出在距离——瞬时多普勒域对目标初像进行分离，进而进行二次成像的方法，可以很好地解决多径效应情况下高机动舰船目标的瞬时成像问题。对于尺寸较大的目标，距离走动比较明显，提出采用 Keystone变换的方法来校正距离走动，然后再进行方位向压缩来获得目标ISAR像的方法。　　Radar imaging has the characteristic of all-weather, day/night and long range, and has great value in civilian and military applications. Due to the dissimilarity of the Radar’s working mechanism and imaging mode, imging Radar can be divided into two categories-inverse synthetic aperture Radar(ISAR) and synthetic aperture Radar(SAR). In ISAR imaging, the echo after motion compensation can be characterized as the model of multi-component polynomial phase signal(PPS), and this is a typical kind of nonstationary signals. When implement the ISAR image of a target, this model can be approximated logically according to the movement character of the target, and the quality of the Radar images lies on the signal processing technique. That is to say, the quality of the Radar images can be improved greatly by the signal processing methods. This paper divides the echo of the target into four categories from the aspects of parameter estimation and time frequency distribution, they are the constant amplitude linear frequency modulated(LFM) signal model, constant amplitude polynomial phase signal(PPS) model, time varying amplitude PPS model and directly obtaining the time frequency structure of the signal based on the time frequency methods. New algorithms are presented in signal processing corresponding to the four models above in order to improve the quality of the Radar images. The following is the summarization of the main work:　　1. The principle of the Range-Instantaneous-Doppler(RID) algorithm for ISAR imaging of maneuvering targets under the condition of one order approximation based on the real data of plane is studied. The essence of RID is the parameters estimation of multi-component LFM signals. Two kinds of super-resolution techniques for estimate the parameters of LFM signals are studies and are used in the ISAR imaging of maneuvering targets, the observation angle for ISAR imaging is decreased whereas the cross resolution is invariant. Finally, the characteristics of the echo for the target which have a maneuvering flight are summarized and analyzed in the signal processing point of view with the aim of commanding the whole paper.　　2. The model of constant amplitude LFM signal is considered, and a new algorithm for estimate the parameters of LFM signal based on the cubic phase function(CPF) are presented. The chirp rate can be estimated by a one-dimensional maximization of the CPF, and the initial frequency and amplitude can be obtained by the dechirp technique combining with the fast Fourier transform. This method has a low computation complexity, and the asymptotic statistical performance is analyzed by the first order perturbation analysis of maxima of random functions. The simulated and theoretical results demonstrate the high precision of this algorithm. This new method is used in the ISAR imaging of maneuvering targets and improves the images quality greatly.　　3. The constant amplitude PPS model is considered, and a new algorithm for estimate the parameters of a cubic or quartic phase PPS based on the modified version of the WVD is proposed, this algorithm is used for the estimate of the rotation angle in ISAR imaging. For an arbitrary order PPS, a new concept-Instantaneous Frequency Rate(IFR) is studied and two new methods for estimate the IFR are presented, this notion is used for the parameters estimation of a PPS with high precision and low computation burden.　　4. The time varying amplitude PPS model is considered, and the signal decomposition technique is adopted to deal with it. Signal decomposition is a main topic in the signal processing domain, by decomposing the signal into a linear combination of a series of atoms, the inner structure of the signal can be revealed. The commonly used adaptive Chirplet decomposition is studied in this chapter, and an optimum algorithm based on the RELAX technique is presented. This algorithm fixes up the parameters of the rest components while estimating the parameters of the current component, and adjusts the estimated parameters one by one until getting the inner convergence. This method can improve the parameters estimation precision but with heavy computation burden. Then a fast algorithm for the adaptive Chirplet decomposition based on the CPF is proposed. It is concluded that the energy of the CPF of the Chirplet peaks along the chirp rate curve, and the chirp rate, time center, the amplitude can be obtained simultaneously by th