高斯混合分布下雷达目标距离检测方法

Distance detection method to radar target under Gaussian mixture distribution

在非高斯相关杂波背景下,传统检测方法难以检测到目标,基于球不变随机过程杂波的似然比检测方法通过对杂波准确建模,可以取得较好的检测效果,但由于其分布形式往往较为复杂,一般情况下很难得到检测统计量的闭合形式.针对该问题,基于信息几何理论,通过计算高斯混合统计流形上两分布间的距离,定义了一种距离检测器,该检测器通过计算估计分布与目标有无两种假设分布间的距离差,来实现目标检测,将检测问题转化为统计流形上的几何问题.仿真和实测数据验证结果表明:在复杂杂波背景下,与传统方法相比,该方法具有更好的检测性能,且易于实现.

It is hard for traditional methods to detect targets in correlated non-Gaussian clutter back- grounds. The likelihood ratio test based on spherically invariant random process clutter could get good per- formance by modeling clutter accurately. However, it is very hard to get closed form of test statistical magnitude for that the probability distribution function form is usually complicated. Based on information geometry theory, the distance detector is defined by comparing the distance between the distribution esti- mated by observation data and two hypothetical distributions, translating detecting problem into geometry problem on statistical manifold. Simulation results show that detection performance of distance detector under correlated non-Gaussian clutter backgrounds outperforms traditional methods,simultaneously,it can be easily operated.