非平稳逆高斯纹理估计的两种协方差矩阵GLRT检测器

GLRT detector of two covariance matrix estimators in the non-stataionary inverse Gaussian texture

文章研究了两种海面目标检测算法。由于海杂波的非高斯、非平稳和非均匀特性,传统的分布模型不能对海杂波进行有效的拟合,海杂波的概率密度函数会产生严重的拖尾现象。海杂波的非平稳性主要体现在纹理上,通过对纹理进行逆高斯建模,并估计纹理的最大后验值和最小均方误差值,对非平稳纹理进行归一化处理,实现海杂波协方差矩阵的平稳化。在广义似然比检测模型的基础上,提出了最大后验逆高斯广义似然比检测器(MAP-IGD-GLRT)和最小均方误差逆高斯广义似然比检测器(MMSE-IGD-GLRT)。经过仿真和实测数据实验验证,该文所提两种检测器均优于对比算法。

In this paper, two algorithms of target detection on sea surface are studied. Due to that sea clutter is non-Gaussian,non-stationary and heterogeneous, the traditional distribution models can’t fit the sea clutter effectively, and there is a serious tail of the probability density function of sea clutter. The texture is main component of the non-stationary sea clutter. Model the non-stationary texture with inverse Gaussian distribution(IGD), estimate the maximum a posteriori(MAP) and minimum mean square error(MMSE) of texture, normalize the non-stationary texture, and realize the stationary covariance matrix of sea clutter.Based on the generalized likelihood ratio(GLRT) model, two detectors, that is, MAP-IGD-GLRT detector and MMSE-IGD-GLRT detector, are proposed in the paper. Simulated and experimental results show that the two proposed detectors are superior to the comparisons.