基于谱范数的矩阵CFAR检测器

基于信息几何的矩阵恒虚警率（Constant False Alarm Rate，CFAR）检测器为雷达目标检测问题提供了新的解决思路，其主要组成部分是均值矩阵的估计和检测统计量的计算，且检测性能与矩阵流形上的几何测度有紧密关系.现有的信息几何测度都是从Frobenius范数考虑，本文则基于谱范数考虑了矩阵流形上的几何测度和均值矩阵估计问题.将均值矩阵估计问题转化为矩阵流形上的优化问题，根据目标函数的性质得到了计算简便的近似均值矩阵.利用不同方法得到的均值矩阵，提出了几种新的基于谱范数的矩阵CFAR检测器.通过检测势分析和仿真实验验证了其检测性能优于现有的其他矩阵CFAR检测器，复杂度分析也表明了其计算复杂度低于现有的其他矩阵CFAR检测器，为海杂波背景下的雷达目标检测提供了新的有效技术手段.

Matrix CFAR Detector Based on Matrix Spectral Norm

Information geometry based matrix CFAR (Constant False Alarm Rate) detector provides a way to the problem of radar target detection,which mainly consists of the estimation of mean matrix and the calculation of test statistics.The detection performance is closely related to the geometric measures on the matrix manifold.The existing geometric measures are considered from Frobenius norm.By contrast,this paper considers the geometric measure and the estimation of mean matrix by utilizing matrix spectral norm on matrix manifold.The mean matrix estimation is transformed into the optimization problem on the matrix manifold.The approximate mean matrix with low computational complexity is obtained according to the properties of the objective function.In addition,we propose several matrix CFAR detectors based on different mean matrix estimation methods.Finally,the detection power analysis and simulation results show that the detection performance of the proposed methods with lower computational complexity are superior to other existing matrix CFAR detectors.It provides a new effective technique for radar target detection under sea clutter background.