强海杂波背景条件下海上目标检测方法

为了减少展宽的一阶海杂波谱对空间超分辨谱估计技术的影响，提高MUSIC算法的检测性能，用径向基神经网络实现海杂波预测和对消．利用神经网络的可对任意非线性函数模拟的特性，对展宽的海杂波进行模拟，用模拟后的结果实现一阶海杂波的对消，来满足MUSIC算法的应用条件．最后用MUSIC算法分辨海上目标的方位信息．实验结果表明，对消前后目标背景噪声子空间特征值发生改变，对消后更接近于MUSIC算法的假设条件，提高了MUSIC算法的检测性能，扩大了MUSIC算法的应用范围，实现在非高斯噪声背景条件下应用MUSIC算法检测目标．确定最小的相空间维数和时间间隔，采用更先进的神经网络算法，可以提高神经网络的预测速度．

The method to detect signal on the sea with strong sea clutter

In order to reduce the influence of the widened first-order sea clutter echo on the spatial super-resolution estimation and improve the testing performance of MUSIC algorithm, radial artificial neural network is used to realize the forecast and cancellation of sea clutter. By using the characteristic of neural network which can simulate any nonlinear function, simulation is made to widened sea clutter. The simulated result can be used to cancel the first order sea clutter and satisfy the application condition of MUSIC algorithm. Finally, MUSIC algorithm is employed to distinguish the azimuth of the signals. The experimental result shows that the eigenvalues of noise subspace change before and after sea clutter cancellation, which further approaches the assumed conditions of MUSIC algorithm, improves the detection performance, expands the application range of MUSIC algorithm and detects signals under the condition of non-Gauss noise background. The forecast speed can be increased when the minimal phase space dimensions and time interval are determined and more advanced neural network is adopted.