【作者】 夏宇垠；

【导师】 冯大政；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2011， 博士

【摘要】 宽带雷达与窄带雷达相比,优势主要体现在:可提高雷达对杂波中目标的检测能力;可更灵活地设计发射波形和雷达系统;具有更高的距离分辨力;具有更好的电磁兼容性能;具有更低的截获性能等。宽带雷达的潜能使它成为雷达发展的一个重要方向。但目前对宽带雷达的应用主要集中在目标成像、识别和分类等方面,如何在宽带雷达体制下完成目标检测,跟踪等雷达传统任务具有重要的理论意义和工程价值。本论文围绕宽带雷达目标时域检测问题展开,主要内容分为五部分:(1).从目标散射点模型出发,对宽带雷达目标回波的物理特性进行了深入研究。详细分析了宽带雷达检测中的两个难点:由目标转动引起的回波相位变化,和由平动引起的相参处理期间目标的越距离单元走动,并介绍了目标回波平动补偿的方法:包络对齐和初相校正,为后续的工作奠定了基础。(2).研究了均匀高斯噪声中基于单次脉冲回波的宽带雷达目标检测问题。首先分析了宽带雷达对目标进行检测的优势和存在的问题,而后介绍了几种已有的单次回波宽带雷达目标检测算法,并分析了它们的优缺点。指出实际中提高单次回波雷达目标检测性能的难点在于无法获取目标散射中心分布的先验信息。从解决这一问题入手,提出了一种基于多量化门限的二进制积累检测算法,理论分析和实测数据的仿真实验证明,在无法获取目标散射中心分布先验信息条件下,对于散射中心分布稀疏的目标,该算法与能量积累、基于散射点密度的广义似然比等检测算法相比,计算简单,易于硬件实现,并具有更稳健的检测性能。(3).研究了均匀高斯噪声中,基于多次脉冲回波的宽带雷达运动目标特征检测问题。首先根据宽带雷达目标回波的特点,指出匀速直线运动的宽带雷达目标,其距离-慢时间域的轨迹由若干条斜率相等的直线组成。由这一特征入手,设计一种基于哈夫变换的检测算法,对其性能进行了完整的理论分析,最后通过实测数据验证了该检测算法可对宽带雷达目标回波进行有效的非相参积累,与各种已有算法相比,显著提高了宽带雷达的检测性能。(4).研究了均匀高斯噪声中,基于多次脉冲回波的宽带雷达目标相参检测问题。首先针对低信噪比环境中包络对齐精度低的难点,提出一种基于概率积累函数映射和约束时延平移量的包络对齐新方法。在此基础上,从脉间和脉内信号能量积累的思路出发,设计了两种检测算法。前者从观测数据中估计初相,用估计出的导向矢量对脉间信号进行相参积累;后者采用目标数据库的基于主分量分析的子空间对脉内信号进行次优匹配接收。理论说明了两种算法的恒虚警率性。实验结果表明,与非相参积累检测算法相比,这两种算法显著地提高了宽带雷达的检测性能。(5).研究了宽带雷达目标的稳健检测问题。分析了目标在相参处理时间内导向矢量误差给相参积累带来的影响,针对这一问题,提出了一种基于凸规划的改进稳健广义似然比检测算法。详细的理论分析和实测数据仿真实验证明,在保证检测性能的前提下,该检测算法可有效地降低稳健广义似然比检测算法的计算复杂度。在导向矢量严重失配的情况下,该算法检测性能相对于传统的广义似然比算法有明显的提高。更多还原

【Abstract】 Compared with narrowband radars, wideband radars can provide many advantages, such as, better target detection performance in clutter, more flexible design for waveforms and systems, better spatial resolution, better electromagnetic compatibility and lower probability of interception. Wideband radars become the trend of radar development for its potential applications. However, nowadays the applications of wideband radar are mainly for target imaging, recognition and classification, how to use wideband radar to fulfill the conventional radar task (such as target detection and tracking etc) is of significance in both the theory and application.The work of this dissertation mainly focuses on issues of the wideband radar target detection in the time domain. The main content can be summarized as the following five aspects:(1). Based on a scattering center model, the physical property of wideband radar target return is discussed. We point out that how to deal with the target-scattering function changing caused by rotational motion and scatters’migration through range cells caused by translational motion is a challenging task for wideband radar target detection. The translational motion compensation can be obtained by envelope alignment and initial phase correction. This forms the basis for the following study.(2). The problem of wideband radar target detection in Gaussian noise by one pulse is researched. Firstly, the advantage and problem of detecting targets by wideband radar are discussed; several existing algorithms are reviewed with their advantages and disadvantages analyzed in detail; then we point out that the difficulty of wideband radar detection is the absence of the knowledge about the spatial distribution of the target scattering centers in real applications; To overcome this difficulty we design a novel binary integrator using multiple quantization thresholds, which we term as the modified binary integrator (MBI). Theoretical analyses and simulation results show that MBI is easily implemented and can achieve a more robust performance than the present algorithms(such as energy integration detector and scatter density dependent GLRT detector etc) for sparse scattering target when the prior knowledge of the targets’scattering centers distribution is absent.(3). The problem of wideband radar moving target detection in Gaussian noise based on feature of multiple adjacent pulses is researched. Based on the characteristics of the wideband target return, we point out that the tracks of target with a constant velocity in range-time data space are composed of several straight lines with the same slope. Based on the target and noise models we design a wideband radar Hough detector using Hough transform. The statistical properties of the detector are discussed completely and rigorously. Experimental results show that the detector can integrate the target returns from multiple pulses efficiently, and has a much better performance than the existing noncoherent detectors.(4). The problem of wideband radar target coherent integration detection in Gaussian noise by the returns of multiple adjacent pulses is researched. The conventional envelope alignment method suffers severe performance degradation under low signal-to-noise ratio. To overcome this problem we design a novel envelope alignment scheme based on cumulative distribution function mapping and constrained misalignment. After envelope alignment we propose two detection schemes. In the first one, we integrate the signal from every pulse by the initial phase steering vector estimated from the observed data. In the latter one, we integrate the signal from every range cells using prior training target data based on principal component analysis. Theoretical analyses show that the proposed detection schemes achieve constant false alarm rate (CFAR). Furthermore, in the detection experiments based on measured data, the proposed schemes can obtain better detection performance than the existing coherent detectors.(5). The problem of robust detection in wideband radar is investigated. The steering mismatches during the coherent processing interval and its impact on coherent integration are introduced firstly; then we propose a new detector based on convex programming, which we refer to as improved robust generalized likelihood ratio test (IRGLRT).A detailed analysis on IRGLRT shows that compare with the existing robust generalized likelihood ratio test (RGLRT) detector, it can greatly reduce the computation complexity without loss of detection performance. Finally, experimental results for measured data of three planes show that the proposed algorithm achieves a visible performance improvement with respect to the conventional GLRT, especially in the presence of severe steering vector mismatches.更多还原