【作者】 姜斌；

【导师】 郭桂蓉； 黎湘；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2006， 博士

【摘要】 本文研究了地、海杂波建模与仿真及目标检测若干技术。首先回顾了杂波建模与仿真技术、海杂波混沌特性分析与目标检测的发展历史和研究现状,分析了目前存在的主要问题,介绍了论文研究的内容。第二章首先研究了几种典型杂波模型的统计特性及零记忆非线性变换法(ZMNL,Zero Memory Nonlinearity)仿真方法,然后,针对以往的杂波模拟仿真中,都没有考虑线性滤波器的物理可实现性问题,引入最小相位特性与复倒谱技术,给出了一种物理可实现的滤波器产生方法,在此基础上,详细阐述了典型杂波随机序列产生的方法。广义复合杂波模型(Generalized Compound Probability Density Function,GC-PDF)适用范围比较广泛,论文第二章给出了该杂波模型取不同参数时与经典杂波模型的联系,深入分析了其统计特性,研究了最小二乘参数估计方法,进而以参数解耦和充分利用杂波序列信息为突破点,提出了一种新的参数解耦估计算法。该算法将一个四维非线性最优化问题转化为一个一维线性最优化问题,降低了计算量和所需样本数,提高了估计性能。仿真结果验证了上述算法的准确性与有效性。第三章首先研究了近海岸环境的特点,提出了一种新的近海岸复合杂波幅度分布模型,随后研究了其统计特性,推导得到了相关高斯序列与近海岸杂波分布序列相关系数间的非线性关系,这是基于ZMNL方法进行杂波仿真的重要前提。在此基础上,提出了一种近海岸复合杂波随机序列产生的方法。最后,分析了对LFM信号具有时频聚集性这一特点,给出了近海岸复合杂波背景下基于WVD (Wigner Ville Distribution)变换的LFM (Linear Frequency Modulation)信号目标检测方法。基于实测数据的仿真结果验证了杂波模型准确、检测算法对杂波具有较强的抑制能力。第四章首先研究了混沌基本理论,分析了混沌时间序列判决方法。然后研究了基于RBF神经网络与基于记忆库的预测方法,并对各自性能进行了比较,得出了有意义的结论。其后,引入分形布朗运动模型,基于实测的S波段雷达数据,计算了该分形模型的海杂波Hurst指数,随后求得了其分形维数。并计算得到了其Lyapunov指数,从而证实了S波段实测海杂波的混沌分形特性。随后,提出了一种基于最小二乘支持向量机(LS-SVM Least Squares Support Vector Machines)的海杂波混沌时间序列预测方法。最后,基于实测数据,进行了计算机仿真实验,验证了算法与结论的有效性。第五章给出了一种改进径向基核函数(RBF,Radial Basic Function)的支持向量机分类方法,并比较了该方法与模糊K-近邻法(FKNN,Fuzzy K-Nearest Neighbor)、模糊最小最大神经网络方法(FMMNN,Fuzzy Min-Max Neural Network)的分类性能。进而分析了实测海杂波与目标回波信号的盒维数的差异,提出了一种基于海杂波分形特性的目标检测方法。然后,基于目标回波信号与海杂波预测信息的差异,引入改进核函数的SVM分类器,提出了一种海杂波背景下的目标检测方法。基于实测数据的仿真结果证明了上述方法具有较高的检测精度与抗噪性能。第六章首先对扩展目标的一维距离像进行了离散建模,然后给出了扩展目标调频步进雷达回波信号的数学模型。通过深入分析径向运动对一维距离像的影响,推导得到了一次相位项与二次相位项的运动补偿精度要求。在此基础上,研究了时域相关法与最小脉组误差函数准则,给出了适合工程需要的运动补偿算法。最后,进行了计算机仿真实验,证明了本章的结论。第七章是结束语,总结了本文主要工作,指出需进一步研究和解决的问题。更多还原

【Abstract】 The dissertation is focused on the modeling and simulation of coastal clutter as well as radar target detection.Chapter 1 introduces the main problem of current modeling and simulation of the sea clutter and radar target detection techniques and proposes the solution.In chapter 2, a physical realizable filter that adopts minimum phase feature and complex cepstrum technology is presented. The method of generating general random clutters sequences is described in detail. The relationship between the classical clutter models and GC-PDF model is also analyzed which deduces the method of least mean squares parameter estimation. Furthermore, a new model parameter estimation algorithm is brought out through the parameter decoupling and fully using the information contained in the clutter sequences. This new estimation method translates four dimension nonlinear optimization problem to one dimension linear optimization problem, which can not only improve the estimation performance, but also depress the calculation quantity and the necessarily clutter sample numbers. The simulation is processed and the results prove the validity and veracity of the method proposed in this chapter.Chapter 3 proposes a novel coastal compound distributed model of the clutter amplitude. Then the technique of generating the random coastal compound clutter sequences is presented with deduction of the nonlinear relation of autocorrelation function between the coastal compound clutter random sequences and the correlated Gaussian random sequences based on the research on the statistic characteristic, which is the important precondition for the clutter simulation based on the ZMNL. In the end, a new approach is presented with WVD based on the signal detection in the ocean whereas WVD has the time-frequency local characteristic for LFM signal. The simulation result with the observed data shows veracity and validity for the clutter mode and significant clutter suppression with the detection algorithm.In chapter 4, the chaos basic theories are studied and the methods to define chaotic time series are analyzed at first. Then, two prediction technologies based on the RBF neural networks and memory-based predictor are analyzed with their performance comparation, which gives the analytical results. By adopting the model of fractional Brownian motion (FBM), the Hurst exponent is deduced based on the observed data of the S band radar. Furthermore, the fractal dimension of the S-band sea clutter is also deduced. Then the largest Lyapunov exponent is obtained by Rosenstein method, which proves its chaos and fractal characteristic. It turns out to be a novel method of chaotic time series prediction based on LS-SVM. Based on the observed data of S-band radar, the computer simulation is processed and the results prove the validity and veracity of the conclusion proposed in chapter 4.In chapter 5, the pattern classification method of SVM is analyzed with an ameliorated RBF kernel function, contrasted whose classification performance with those of FKNN and FMMNN. By analyzing the box dimension difference between the observed sea clutter and the target echo signal, the method of target detection is proposed based the fractal property of the sea clutter. Also by adopting the SVM classifiers of the ameliorated RBF kernel function, a novel method of target detection in the background of sea clutter is presented based on the predictive information difference between the observed sea clutter and the target echo signal. Based on the observed data, the computer simulation is processed and the results prove the effective detection performance and noise tolerance.In chapter 6, a discrete model is set up for high resolution profile of an extended target, and then the math model of echo signal is presented from an extended target produced by stepped chirp radar. The effect of target motion on a range profile is analyzed in deep, and then precision necessitation is derived for motion compensation. Studies on the method of time field correlation and the rule based on the least burst error; the motion compensation arithmetic is presented which suits the project requirement. Finally the result of a cyber-emulation proves the conclusion.Chapter 7 is the summary of the paper and points out the problems which need further research.更多还原