【作者】 伍光新；

【导师】 邓维波；

【作者基本信息】 哈尔滨工业大学 ， 通信与信息系统， 2008， 博士

【摘要】 谐振区雷达目标散射特性携带了目标的形状、尺寸等本原信息,是可供雷达目标识别使用的最为有效的电磁频谱特性。开展谐振区雷达目标识别研究可对重点防御目标进行早期预警、有效跟踪和精确打击;可充分发挥现役高频地波超视距雷达的特点,提高系统的整体性能。在高频地波超视距雷达工作频率范围内,舰船目标和飞机目标的散射特性基本处于谐振区。这一特点为高频地波超视距雷达目标识别提供了重要依据。针对这样的研究背景,本文从雷达目标散射特性的获取与分析,谐振区雷达目标特征的提取,谐振区雷达目标识别等三个方面进行了逐步深入的研究。首先,改进了在以往矩量法计算目标RCS中采用的线框几何建模方法,提出了一种适合谐振区复杂目标RCS计算的基于有限元建模软件FEMAP的曲面几何建模方法,实现了通过仿真计算手段获取更准确的目标RCS。提出利用一阶海杂波获取准确的高频地波超视距雷达舰船目标实测RCS的方法。利用实测RCS验证了曲面建模方法的准确性。利用仿真计算的手段研究舰船目标和飞机目标的在高频波段的RCS的变化特点,分析了目标尺寸和典型部件对RCS的贡献。其次,深入研究了利用仿真计算的频域数据提取复杂目标极点特征需要解决的问题。提出了一种基于谐振区频域散射数据的复杂目标极点特征提取方法。分析了由于频域数据截断造成的时域入射波形展宽和由此带来的对时域瞬态响应数据获取的影响;详细分析了入射波和目标相互作用产生瞬态响应的过程,定义了更加合理的晚期响应开始时间。提出基于晚期响应恢复误差最小化原则的改进总体最小二乘矩阵束方法,实现了飞机和舰船目标极点特征的可靠提取,首次给出了飞机和舰船目标各方位一致的极点特征。实现了用较少的极点和留数准确恢复目标的时域响应和目标频域响应。然后,研究了对目标频域响应数据进行有理逼近的问题。深入研究了以往的有理逼近方法对目标频域响应进行逼近存在的问题。将正交矢量拟合方法引入目标频域响应逼近问题中,运用迭代策略克服线性化误差,用正交部分分式代替多项式作为基函数解决法方程矩阵病态问题。对简单目标和复杂目标的仿真均获得了极好的频域响应逼近效果;同时可获得目标极点特征。与改进整体最小二乘矩阵束方法的效果相比,该方法获得的简单目标极点也更加准确。方法既实现了直接对目标频域数据进行有理逼近获得目标极点特征,又得到了目标频域响应的精确逼近。最后研究了谐振区雷达目标识别问题。在宽带雷达目标识别中,实现了利用极点特征匹配方法识别复杂目标。重点研究了噪声对极点提取以及对识别的影响。针对窄带雷达多频RCS特征目标识别中存在的问题,着重开展了雷达工作频率优化选择方面的研究。提出了基于平均分类错误概率最小准则的频率优化选择方法,实现了在较少的工作频率情况下实现较高的目标识别率。提出了一种基于门限技术的近邻分类器扩展方法,实现了在对库中舰船目标可靠识别的基础上,对非库中舰船目标有效拒判。更多还原

【Abstract】 Radar target scattering characteristic in resonance region which takes the shape and size principle information of target is the most efficient spectrum characteristic for radar target identification (RTI). Ship and airplane targets locate in the resonance region for the working frequency of high frequency surface wave radar (HFSWR). This feature is the foundation of HFSWR target identification. Conducting the radar target identification research in resonance region can realize early stage warning and effective tracking and attack to major targets, exploit the advantage of HFSWR and improve the system performance of the HFSWR. According to this research background, research is carried out in three aspects which is radar target scattering characteristic analysis, aspect independence radar target feature extraction and HFSWR target identification.Firstly, surface geometry modeling method is used to replace wire-frame geometry modeling method in method of moments (MoM) based radar cross section (RCS) simulation. A modeling method based on the finite element modeling and post-processing software (FEMAP) which is fit to the resonance region targets is proposed in detail. Furthermore, a calibrating RCS approach utilizing the first order sea clutter to decrease the offset which is produced by the inaccurate of Norton surface wave attenuation and antenna gain is given to obtain more accurate measured RCS. Comparing with the measured RCS from HFSWR system, Simulation RCS results using these surface geometry modeling methods is more precise than the result of wire-frame geometry modeling method. Ship and airplane target RCS change characteristics are given and influences of target dimension and classic parts on the RCS are also analyzed based on the plenty of simulation results.Secondly, problems in complex target natural resonance frequencies (poles) extraction from MoM-based simulation frequency domain scattering data are investigated. A complex target poles extraction method exploiting simulated scattering data in frequency domain is provided. Waveform extended process due to the limited frequency band is analyzed and its influence to the transient response in time domain is also studied. More reasonable beginning time of late time response is defined. Then a modified total least square matrix pencil method (MTLS-MPM) is proposed to remove the disturbance of poles number uncertainty along with calculation errors based on minimal reconstruction error of late time response. Thus reliable poles extraction is realized, and poles location in complex plane of ship and airplane targets are first presented which is coincident in different aspects. Therefore, transient response is reconstructed by few pairs of pole and residue, and target frequency domain response is represented by poles and residues for the first time.Rational approximation to the target frequency response data is studied to obtain the target feature and accurate frequency response approximation. Least square rational approximation method and mini-max rational approximation method are investigatied. However, shortcomings of normal equation matrix ill-condition along with linearization error are discovered in theoretical analysis and numerical simulation. A rational approximation method based on orthogonal vector fitting (OVF) technique is proposed. Central to this method is that orthogonal partial fraction, instead of polynomial, is adopted as basic function of denominator and numerator to improve numerical stability, and iteration is used to decrease the linearization error. Simulations of simple and complex targets show that excellent approximations to frequency response are achieved. In addition, poles location is accurate as well as aspect independent. Particularly, Results of simple target poles are more precise than results of MTLS-MP.Finally, RTI in resonance region is studied. Target identification exploiting poles extraction and matching is investigated in broadband radar. Influence of noise in measured response on the poles extraction and target indentification are also studied. In narrow band radar target identication, research is emphasized on radar working frequency selection so as to improve multi-frequency radar target identification performance. Frequency optimal section based on minimal error rules (MER) of classification is proposed which can increase the identification probability without added frequency number. An expanded nearest neighbor (NN) classification method is given to realize the fundmation of rejection.更多还原