基于地基雷达仿真系统的目标识别技术研究

【作者】 郑小亮；

【导师】 汪学刚；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2006， 硕士

【摘要】 弹道导弹自问世以来,以其射程远、威力大、精度高、机动性好和生存能力强等优越性,已成为攻防对抗中一种有效的打击手段。作为其对立面,弹道导弹防御系统应运而生,它为己方的基础设施、军事要点和部队提供全方位和多层次的防御,免遭敌方弹道导弹袭击。地基雷达(GBR)是导弹防御系统中探测、跟踪和识别目标的重要工具,在导弹防御系统中有举足轻重的地位。因此研究GBR的工作原理,有着重要的国防意义。本文是某GBR仿真系统的合作项目,负责雷达目标识别系统的研究和开发。雷达目标识别是GBR的一个主要功能,它利用雷达获得的目标信息,通过综合处理,得到目标类别的描述和威胁等级的判断,使防御系统能对目标作出准确的攻防对抗。本文在总结当前主流雷达目标识别算法的基础上,建立起基于GBR的雷达目标识别系统,并且对弹道导弹的各种特识别方法进行了研究,主要进行的工作和创新有:1.研究和总结了当前常用的分类识别方法,针对雷达目标识别的特点,对近邻分类器、多层感知器和径向基函数网络(RBFN)分类器的算法和性能进行了研究。仿真结果显示这三类分类器适合应用于雷达目标的分类识别,且具有很高的识别率,是一类简单、高效的非线性分类器。2.分析了弹道导弹中段飞行的特点,建立起导弹飞行的数学模型。推导出了根据一系列雷达测得的目标点迹数据拟合弹道并估计目标落点的方法。测试结果表明该方法可以准确的估计出导弹的落点。3.研究了导弹目标一维距离像和ISAR像的特性。针对ISAR像边缘模糊、灰度动态范围大等特征,对图像预处理方法提出了改进。通过样本库的建立,消除角度变化对识别的影响;利用基于傅里叶变换的特征提取方法得到移位不变的识别特征向量,建立起宽带识别系统。通过实验结果可知,该方法有很高的识别率,且对噪声有抑制作用。4.针对导弹的多目标突防,以合理利用雷达资源为出发点,建立了一种雷达目标的层次化融合识别系统。该系统将多目标管理、目标威胁等级评估和融合处理有机地结合起来,在有效地对目标进行识别的同时也为雷达资源的管理提供有力的依据。更多还原

【Abstract】 Since it was been made, ballistic missile has been an efficient weapon in warfare because of its long range, strong power, high precision, good maneuverability and strong viability. As its contrary, missile defense system emerges as the times require. It provides all-orientation and multi-hierarchy defense for our basic facilities, military establishment and armies. Ground based radar (GBR) is playing a key role in missile defense because it’s an indispensable tool for detecting, tracking and recognizing targets.Radar targets recognition (RTR) is one of the main functions of GBR. Depending on it, we can get description of type and estimation of threaten level for targets by dealing with the signals and information received by radar, then, provide evidence and advice for missile defense system to make the right decision. In this dissertation, base on the review and analysis of current mainstream algorithms and techniques, we build up whole RTR system, and study some efficient recognition methods for different radar characters. The major research work and contributions in this dissertation are summarized as below:1.Summarized current popular pattern recognition methods, this dissertation researched in the algorithms and performances of nearest neighbor (NN) classifier, multi-layer perceptron and RBFN (Radial Basis Function Network). Experimental results show that these three classifiers can get high recognition rate. They are simple, efficient, nonlinear and suitable for RTR.2. After analyzing the characteristic of ballistic trajectory in midcourse, we established spatial objects’kinetic model, and introduced a method for trajectory fitting and impact point estimation using a series of targets’plot detected by radar. Test results show that this method can get high precision.3.We studied the feature of targets’range profile and ISAR profile. In order to eliminate the illegibility of boundary and large dynamic range in amplitude, this dissertation presented a preprocess method for ISAR profile using adaptive threshold segment. In the stage of classification, typical simples is chosen in each scale of 17°更多还原