基于联合分布的雷达目标检测与分类方法研究

发布时间：2018-07-29 18:28

【摘要】：本文主要研究强杂波或噪声背景下雷达目标检测与分类问题。由于隐身飞机/舰艇、低空/超低空突防武器和综合电子干扰在现代战争中的大规模应用，雷达需要具有在强杂波和噪声中检测和分类目标的能力。本文在雷达回波信号的联合分布（时频分布和时间-调频率分布）和特征提取方面展开研究，并将研究成果应用于海面微弱目标检测、空中机动距离扩展目标检测和空中直升机分类。本文的研究成果可概括为以下四部分： 1，研究了海面微弱目标检测问题。分析了海面波浪的形成过程及结构特点，由此得到海杂波的非平稳特性。进而根据海杂波和目标回波在时频域的区别提出了基于时频迭代分解的海面慢速微弱目标检测方法：基于特征值分解，我们首先提出快速信号合成方法（FSSM），FSSM可以从信号的维格纳分布（WD）中更精确和快速地恢复出信号；然后，基于遮隔WD（MWD）和FSSM，提出信号迭代分解方法（IDM）；最后应用IDM将海面回波分解成许多分量，并根据各分量的时频聚集性从中找出目标回波，实现目标检测。应用实测海面回波验证该检测方法的结果表明其不仅能够高效地检测海面慢速微弱目标还能够显示目标的瞬时运动状态。 2，研究了高斯白噪声背景下空中机动距离扩展目标检测问题。基于高分辨雷达（HRR）接收到的混频器输出，我们提出了三类距离扩展目标检测方法。(1)基于两个相邻混频器输出时频分解特性的距离扩展目标检测方法：首先，基于奇异值分解我们提出一种信号合成方法，它可以从两信号的互维格纳分布（CWD）中合成两个单位能量的信号，并且将两信号的能量集中在两个奇异值上；接着，提出了两个相邻混频器输出的互S-方法（CSM）。然后，将信号合成方法应用于两相邻混频器输出的CSM，得到奇异值；最后，，根据所得奇异值的聚集性实现目标检测。应用没有经过距离走动校正的雷达回波数据验证所提检测方法，结果表明所提方法的检测性能优于传统方法且适用于高速运动目标。另外，该方法具有恒虚警（CFAR）特性。(2)基于一个混频器输出的匹配信号的距离扩展目标检测方法：首先构造混频器输出的匹配信号为频率与混频器输出中目标信号的最大频率相同的正弦函数；然后定义一个混频器输出的匹配模糊函数（MAF）和改进匹配滤波器（MMF）；根据混频器输出在MAF和MMF中零多普勒/频率的聚集性提出了两种距离扩展目标检测方法，分别命名为MAF-D和MMF-D。该类检测器利用一个回波因而可以检测高速（包括平动速度和转动速度）运动目标。应用实测数据的检测结果表明该类方法的检测性能优于传统的检测器且对目标姿态不敏感。(3)基于一个混频器输出的调频率（FR）函数的距离扩展目标检测方法：首先根据三次相位函数（CPF）定义FR函数并确定应用其分析离散线性调频（LFM）信号时的FR范围；然后根据回波信号在零FR处的聚集性实现目标检测。该方法可以检测高速目标。应用于实测距离扩展目标的检测结果表明该方法优于基于HRRP的检测器。 3，针对高阶多项式相位信号的参数估计问题，我们提出了一种高分辨时间-调频率分布（TFRR），并分析了其在目标检测方面的潜在应用。我们推导出该TFRR的分析表达式，并证明了其相对于CPF具有较高的分辨率。该TFRR可以用来分析两个在时间-调频率（TFR）域非常靠近的信号。由于该TFRR是双线性变换，所以当两信号的瞬时频率（IF）信号相交或非常靠近时会受到交叉项的困扰。为了抑制交叉项，我们通过引入一个FR窗提出了平滑TFRR（STFRR）。应用STFRR分析噪声背景下的高阶多项式相位信号，结果表明STFRR具有检测目标的潜能。 4，研究了高脉冲重复频率（PRF）雷达下直升机分类问题。分析了直升机主旋翼的微多普勒调制特征，针对回波信号积累时间是否大于两“闪烁”之间间隔这两种情况，我们提出了两类直升机主旋翼参数估计方法，最终实现直升机分类。(1)针对回波信号积累时间大于两“闪烁”之间间隔这种情况我们提出两种直升机分类技术：第一种方法是匹配滤波器（MF）方法，包括时域匹配滤波器（TMF）和时频域匹配滤波器（TFMF）；第二种方法为时频遮隔模板（TFMs）方法。该类方法可以分类叶片数目不同但微多普勒参数相同的直升机。仿真结果表明它们对于直升机的姿态和微多普勒参数的估计误差都具有鲁棒性。(2)针对回波信号积累时间小于两“闪烁”之间间隔这种情况我们提出了基于最小均方误差（MMSE）的部分周期数据微多普勒参数估计方法：在MMSE准则下应用正弦信号拟合从时频分布中提取出的旋翼叶片微多普勒信号，从而估计出叶片的转速和半径。仿真和实测数据的微多普勒参数估计结果都验证了该方法的有效性与精确性。
[Abstract]:This paper mainly studies the detection and classification of radar targets under strong clutter or noise background. Due to the large-scale application of stealth aircraft / ship, low altitude / ultralow altitude penetration weapon and integrated electronic jamming in modern warfare, radar needs to have the ability to detect and classify the targets in strong clutter and noise. The joint distribution (time frequency distribution and time modulation frequency distribution) and feature extraction are studied, and the research results are applied to the detection of weak target in the sea, the target detection in air maneuver distance and the classification of helicopter in the air.
The research results of this paper can be summarized as follows:
1, the detection of weak target in the sea is studied. The formation process and structure characteristics of sea surface wave are analyzed, and the nonstationary characteristics of sea clutter are obtained. Based on the difference between sea clutter and target echo in time frequency domain, a slow weak target detection method based on time frequency iterative decomposition is proposed: Based on eigenvalue decomposition, we first First, the fast signal synthesis method (FSSM) is proposed, and FSSM can recover the signal more accurately and quickly from the Wigner distribution (WD) of the signal. Then, based on the WD (MWD) and FSSM, the signal iterative decomposition method (IDM) is proposed. At last, the sea surface echo is decomposed into many components with IDM, and the order of the time frequency aggregation of each component is found out. The target detection is realized by the target echo, and the result of the detection method using the measured sea surface echo shows that it can not only detect the slow and weak target of the sea surface, but also can display the instantaneous motion state of the target.
2, the problem of aerial mobile range expansion target detection under the background of Gauss white noise is studied. Based on the output of the mixer received by the high resolution radar (HRR), we propose three kinds of distance extended target detection methods. (1) a distance expansion target detection method based on the time frequency decomposition characteristics of two adjacent mixers: first, the singular value is based on the singular value. We propose a signal synthesis method that can synthesize two unit energy signals from the mutual Wigner distribution (CWD) of two signals and concentrate the energy of the two signal on two singular values. Then, the mutual S- method (CSM) for the output of two adjacent mixers is proposed. Then, the signal synthesis method is applied to the two adjacent mixers. The output CSM obtains the singular value; finally, the target detection is realized based on the aggregation of the singular value. The proposed detection method is applied without the radar echo data of the distance moving correction. The results show that the proposed method is superior to the traditional method and is suitable for the high-speed moving target. In addition, the method has the constant false alarm (CFAR). Characteristics. (2) a range expansion target detection method based on a matched signal of a mixer output: first, the matching signal of the mixer output is constructed as the sinusoidal function of the maximum frequency of the frequency and the target signal in the mixer output; then a matching fuzzy function (MAF) and an improved matching filter (MMF) are defined. According to the aggregation of mixer output in MAF and MMF, two range expansion target detection methods are proposed, named MAF-D and MMF-D., which can detect high speed (including translational speed and rotation speed) by using an echo. The detection performance of the method is superior to that of the traditional detector and is insensitive to the target attitude. (3) a range expansion target detection method based on the frequency modulation rate (FR) function of a mixer output: firstly, the FR function is defined according to the three phase function (CPF) and the FR range of the application for the analysis of the discrete linear frequency modulation (LFM) signal is determined; and then the echo signal is based on the echo signal. The target detection at zero FR is realized. The method can detect the high speed target. The detection results applied to the measured range expansion target show that the method is superior to the HRRP based detector.
3, in view of the parameter estimation of high order polynomial phase signals, we propose a high resolution time modulation frequency distribution (TFRR), and analyze its potential application in target detection. We deduce the analytical expression of the TFRR and prove that it has a higher resolution relative to the CPF. This TFRR can be used to analyze two in time. The TFR domain is very close to the signal. Since the TFRR is a bilinear transformation, it is troubled when the instantaneous frequency (IF) signal of the two signal is intersected or very close. In order to suppress the cross term, we introduce a FR window to put forward a smooth TFRR (STFRR). The high order polynomial in the noise background of the STFRR is applied to the STFRR analysis. Phase signals show that STFRR has the potential to detect targets.
4, the problem of helicopter classification under high pulse repetition frequency (PRF) radar is studied. The characteristics of the micro Doppler modulation of the helicopter main rotor are analyzed. For the two cases of whether the echo signal accumulation time is more than two "scintillation", we propose a method to estimate the parameters of the main rotor wing parameters of the two types of helicopters, and finally realize the helicopter classification. (1) Two kinds of helicopter classification techniques are proposed for the case of echo signal accumulation time greater than two "scintillation". The first method is a matched filter (MF) method, including time domain matched filter (TMF) and time domain matched filter (TFMF); the second square method is a time frequency blocking template (TFMs) method. This method can be divided into two methods. The helicopter with different number of blades but with the same micro Doppler parameters. The simulation results show that they are robust to the estimation error of the helicopter's attitude and the micro Doppler parameters. (2) we propose a partial week based on the minimum mean square error (MMSE) for the interval between the echo signal accumulation time less than two "scintillation". The method of estimating the phase data micro Doppler parameter: using the sinusoidal signal to fit the rotor blade micro Doppler signal extracted from the time frequency distribution under the MMSE criterion, the speed and radius of the blade are estimated. The results of the micro Doppler parameter estimation of the simulated and measured data all verify the validity and accuracy of the method.
【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】：2014
【分类号】：TN957.51

【参考文献】