Two adaptive detectors for range-spread targets in non-Gaussian clutter

Abstract

This paper addresses adaptive detection of range-spread target in spherically invariant random vector clutter. Based on the nonadaptive detectors of NSDD-GLRT and SDD-GLRT, two adaptive detectors named ANSDD-GLRT and ASDD-GLRT are devised by replacing the unknown normalized clutter covariance matrix with the sample covariance matrix based on the secondary data. The formulas of detection probability and false alarm probability are deduced. Moreover, the constant false alarm rate properties of both adaptive detectors are assessed and the influence of mismatched steering vector on detection probability is analyzed. The experimental results show that the adaptive detectors have an acceptable loss with respect to the nonadaptive counterparts. Furthermore, the ANSDD-GLRT converges much faster than the ASDD-GLRT, however, the latter rejects the misaligned signals more effectively. Both detectors perform well for the uniformly-distributedenergy target, but degenerate for the fluctuating targets. In addition, the correlation between target scatterers results in an additional gain for low signal-to-clutter power ratio.Keywords high-resolution radar, range-spread target detection, constant false alarm rate, generalized likelihood ratio test, spherically invariant random vector CitationHe Y, Jian T, Su F, et al. Two adaptive detectors for range-spread targets in non-Gaussian clutter.

CFAR assessment of covariance matrix estimators for non-Gaussian clutter

“…In addition, by (21), the NSCM-RE is independent of the texture components of secondary data. Under the H 0 hypothesis, τ 0 factors out between the numerator and the denominator of the decision statistic in (5). Hence, the NSCM-RE-ANMF with the unique solution of (8) is fully CFAR to the clutter power level and the NCCM.…”

“…However, in situations such as high-resolution radars or low grazing angles, the background clutter may no longer be modeled accurately as a Gaussian random variable (RV) [3]. On-field measurements have shown that the high-resolution radar system [4] receives non-Gaussian observations, which can be suitably modeled as a spherically invariant random vector (SIRV) [5]. In most radar applications, the statistics of SIRV are not known and need to be estimated, especially the normalized clutter covariance matrix (NCCM).…”

Adaptive Detection Application of Covariance Matrix Estimator for Correlated Non-Gaussian Clutter

“…Adaptive target detection against Gaussian clutter has been investigated in the past [1]. In situations such as low grazing angles or high-resolution radars, the background clutter may no longer be modeled accurately as a Gaussian random variable [2]. An increasing interest has been directed toward design and analysis of detection schemes optimized under nonhomogeneous [3] or even non-Gaussian [4] clutter-dominated disturbance.…”

“…t , c (1) t , and´( 1) t (z(2) t , c(2) t , and´( 2) t ) denote the real (imaginary) parts of the vectors z t , c t , and…”

Research on Sparse Targets Detection Methods Based on GLRT in Non-Gaussian Clutter

“…顾新锋 等：非高斯杂波背景中稀疏距离扩展目标检测方法研究 关键词：非高斯杂波，距离扩展目标，恒虚警率，检测 1．引言 对于低分辨率雷达，由于雷达的距离分辨率远大于目 标长度，目标回波表现为单一散射点形式 [1]，关于点目 标的检测问题已经进行了详细的讨论 [1][2][3]。而脉冲压缩 技术的使用，使得雷达具有较高的分辨率，目标回波在雷 达径向上的多个强散射点分布在不同的距离单元中,形成 距离扩展目标 [4,5]。提高雷达分辨率并采用有效的方法 可以极大的提高目标检测概率 [6], 但如果仍采用传统点 目标的检测方法，由于部分目标能量泄漏到参考单元中， 检测性能将大大下降，甚至完全失效 [7]。文献 [8][9][10]研 究了高斯背景中扩展目标的检测方法，然而，随着雷达分 辨率的提高，雷达会接受到类似于目标的尖峰，这种尖峰 可以用球不变随机向量(spherically invariant random vector, SIRV)来建模 [11,12]，文献 [13]利用广义似然比 检验(generalized likelihood ratio test, GLRT)的方 法得到了SIRV杂波背景下不依赖于散射点密度的GLRT检 测 器 (non-scatterer density dependent-GLRT, NSDD-GLRT)。NSDD-GLRT是一种被检测距离窗内各距离单 元回波能量的非相干积累检测器， 在实际应用时， 一方面， 由于不同的目标所占距离单元数不同，为了能够使目标完 全包含在某个被检测距离窗内，通常需要使窗口长度远大 于目标所占距离单元数；另一方面，由于目标的闪烁，目 标所占距离单元中只有部分单元存在较强的回波，称之为 强散射点。对于这种稀疏散射点目标，NSDD-GLRT由于积 累了不含目标的距离单元回波，会出现一定的检测性能损 失，这种损失称为"坍塌损失" [14,15]。为了提高检测 器对这种稀疏散射点目标的检测性能，文献 [13]利用散射 点 的 密 度 信 息 ， 进 一 步 提 出 了 SDD-GLRT(scatterer density dependent-GLRT)检测器。文献 [15] …”