【作者】 冯德军；

【导师】 王国玉；

【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】 弹道中段是导弹防御(MD)系统目标识别最具挑战性的阶段,雷达是中段最主要的传感器,其目标识别能力在很大程度上反映了MD系统中段目标识别的总体水平。深入开展弹道中段目标雷达识别与评估研究,对于发展我国新体制雷达探测技术及增强我国弹道导弹生存和突防能力具有十分重大的军事意义和理论价值。本文源于国家安全重大基础研究973项目、全国百优博士论文专项资金资助项目、总装国防预研项目等多项项目。论文以弹道导弹突防为研究背景,以中段目标群为研究对象,深入系统地研究了弹道中段目标的雷达特性、弹道中段目标的特征提取方法、弹道中段目标雷达识别方法和评估方法等多项关键技术。论文充分利用了微波暗室测量数据并通过逼真的战情仿真,构造了弹道目标识别的动态仿真平台,在若干典型战情下进行了目标特性、特征提取、目标识别和评估研究。在研究中注重所提取特征的物理意义及所提出识别和评估方法的工程可实现性,得到了一系列有价值的结果。雷达目标特性研究是雷达特征提取的基石。为此,论文首先研究了中段目标的雷达特性。在理论上详细分析了弹道中段目标的运动特性,建立了相应的运动学模型。然后,结合已有的测量数据,对中段目标的静态散射特性进行了分析。在此基础上,提出了一种中段目标动态散射特性的仿真方法,该方法能逼真反映中段目标的雷达散射特性。最后,通过所提出的方法分析了弹头和诱饵的窄带和宽带特性。特征提取是雷达目标识别的关键环节,能否提取出物理意义清晰、可分性强的特征在一定程度上决定了识别的成功与否。论文根据中段目标群的运动特性和结构特性,提出了一系列的特征提取方法,其中包括基于窄带序列的RCS周期特征估计方法、基于单个回波和回波序列的运动特征提取方法、基于单个宽带回波的结构特征提取方法及基于非高斯模型的结构特征提取方法。通过这些方法,可以获得目标的章动(或翻滚)周期、速度、径向距离、长度等物理意义清晰的特征,为后续的识别奠定了良好的基础。对弹道目标进行ISAR成像是当前弹道目标特征提取的一个重要方向。论文根据中段目标高速飞行的特点及宽带回波全去斜处理的工作方式,详细论述了对高速目标距离像的展宽补偿方法和残余相位补偿方法,提出了对弹道目标进行ISAR超分辨成像的酉ESPRIT方法,该方法能明显改善成像质量、提高成像效率。接着,论文分析了中段目标的ISAR成像距离及成像积累时间。最后,通过对实测弹头和缩比模型的成像,分析了弹头类目标的ISAR图像特点,在此基础上提出了弹头类目标的ISAR图像特征提取方法。识别器设计与评估是目标识别的中心环节。论文根据弹道目标识别特点,设计了两种识别器:基于多特征综合的模糊识别器和基于分类树的识别器。在特征评估方面,提出了基于可分测度的评估方法和基于模糊理论的评估方法;在识别评估方面,提出了基于置信度的可靠性评估方法和基于ROC曲线的评估方法。论文紧紧围绕弹道中段目标雷达识别与评估这一主题,进行了深入、系统的理论分析,开展了一系列的理论创新和方法创新,创新点主要体现在:1.在理论上深入分析了弹道中段目标的运动特点,通过结合微波暗室测量数据,提出了一种弹道目标动态散射特性的仿真方法,该方法可构造逼真的空间目标识别电磁环境,有效解决弹道目标识别研究中的数据缺乏问题;2.多层次、多角度地研究了弹道中段目标的特征提取问题,提出了获取目标运动特性、结构特征的多种新方法,这些物理意义清晰的特征获取为后续的目标识别奠定了良好基础;3.深入、系统地研究了弹道中段目标的ISAR成像及特征提取问题,提出了酉ESPRIT超分辨成像算法,估算了成像距离和成像积累时间,通过对大量实测数据成像分析了弹头类目标的ISAR图像特点并在此基础上提出了相应的特征提取方法;4.针对弹道目标识别先验信息相对缺乏的特点,设计了两种分类器:基于多特征综合的模糊分类器和基于分类树的分类器,这两种方法均只需要相对简单的先验信息,适用于弹道真假目标识别场景;5.开展了弹道目标特征及识别方法评估研究,提出了基于可分性测度的特征评估、基于置信度的识别可靠性评估等多种方法,为分类器的设计、组合和优化提供了理论支撑和可行手段。在注重理论研究的同时,论文通过仿真手段并充分利用已有测量数据,进行了大量仿真试验,着力使所提出的方法得到验证。最后还要指出的是,本论文研究内容来源于实际项目需求,而大多数研究成果也已成功应用于工程项目中,并取得了良好效果。更多还原

【Abstract】 In missile defense program, the midcourse of trajectory is the most difficult phase for alleging fraud. As the main sensor, the ground-based radar represents the classification capability of the missile defence system to a great extent. Making researche on radar target recognition and evaluation in midcourse is provided with great significance not only for developing new-type radar, but also for promoting the surviving and penetrating ability for our ballistic missile.This dissertation is sponsored by several funds, including the fund of national excellent dissertation, the 973 momentous foundation fund of nation security and the advance research fund of general equipment office. Under the background of ballistic missiles’ penetration, a series of key technique of radar target recognition are invested systematically, including the radar signature of midcourse objects, the method of feature extraction, the method of pattern recognition and the method of evaluation. By using static measurement data and simulating the real scene of recognition, the dissertation establishes a platform of dynamic recognition, on which recognition and evaluation for typical scenario of ballistic missile penetration in midcourse are investigated and valuable results are obtained.The study on signature of radar target is a foundation for feature extraction. So, the radar signatures of midcourse objects are studied at first. After the movement feature of midcourse objects analyzed in theory, a novel approach is proposed, which can silmulate the dynamic scattering feature of midcourse objects. Lastly, the narrowband signatures and wideband signatures of warhead and decoy are analyzed through the proposed method.Feature extraction plays a key role in radar target recognition, and is thoroughly analyzed in this dissertation. A series of feature extraction methods are put forward, including the RCS periodicity estimation algorithms based on narrowband echoes, the moving feature extraction algorithms based on narrowband and wideband echoes, the shape feature extraction algorithms based on wideband echoes. Through those methods, many features of midcourse objects are obtained, such as the length, the nutation rate, the velocity and range of object relative to radar, and so on, which facilitate subsequent alleging fraud.ISAR imaging of ballistic target is an important means of distinguishing decoy. Based on the signature analysis of wideband echo of moving objects with high velocity in midcourse, the compensation methods for high-resolution rangeprofile (HRRP) and for residual video phase (RVP) are presented detailedly. And then, a new ISAR imaging algorithm is proposed based on the unitary ESPRIT technique, which improves ISAR imaging evidently with a reduced computational burden. After estimating the range and rotation angle of ISAR imaging, the dissertation analyzes the ISAR image characteristic of warhead, by using measurement experimental data. Lastly, the approach to extraction ISAR image feature of warhead is presented.The design and evaluation on classifier are pivotal steps in target recognition. Two classifiers, the fuzzy classifier based on multi-feature fusion and the classifier based on decision tree, are proposed for alleging fraud in view of the absence of apriori information.On the other hand, two evaluation approaches, the evaluation approach based on separability criterion and the evaluation approach based on fuzzy theory are put forward to evaluate the extracted features. Furthermore, the dissertation also makes contributions towards the evaluation of radar target recognition classification systems.The work of the dissertation is focused on research of radar target rcognition and evaluation in midcourse. Some valuable results which bring forth new ideas are achieved. The main creativeness is listed as the following:1. After the kinetic characteristic of target in midcourse anlyzed thoroughly, by using static measurement data, a novel simulation approach to dynamic scattering signature of radar target is proposed, which provides a solution to the absence of real recognition data;2. On the basis of feature extraction study of midcourse objects in the round, many new feature extraction methods are presented, including the methods of extracting movement feature and the methods of extracting shape feature of objects, which facilitate later alleging fraud;3. The dissertation makes a deep research on ISAR imaging of ballistic objects in midcourse. After the range and coherent accumulating time for imaging are discussed, the. dissertation analyzes the peculiarity of warhead image and presentes the corresponding feature extraction method. In addition, a new ISAR imaging algorithm is proposed based on the unitary ESPRIT;4. Considering the absence of apriori information, the dissertation designs the fuzzy classifer based on multi-feature fusion and the classifier based on decision tree, both of which are suitable for discriminating decoy because they only need simple apriori information;5. Multiple evaluation methods are demonstrated for feature and classifier evaluation, such as the feature evaluation method based on separability criterion and the classifer evaluation method based on confidence. Those methods provide theoretic support and feasible means for designing, choosing and combining classifiers. The dissertation not only attaches lots of importance to theory research, at the same time, by taking the advantage of measurement data and simulation means, it also performs lots of simulation experimentation to test the proposed theories and methods.At last, we must point out that the research in the dissertation stems from practical projects, and its conclusion and methods are mostly applied in practice with fairly good effects.更多还原