基于神经网络的干涉仪测向方法

针对干涉仪测向系统中采用传统算法难以克服系统误差的问题,提出了一种基于神经网络的干涉仪测向方法。通过对干涉仪测向系统进行建模,分析了测向误差来源和解相位模糊算法,建立了基于相位干涉仪测向系统的BP神经网络模型,并采用了Levenberg-Marquardt算法对BP神经网络进行改进。以微波暗室的试验数据为训练数据,利用Matlab工具箱对神经网络进行了验证性的仿真试验。仿真结果表明:与传统的测向算法相比,该算法能克服系统误差,进一步提高干涉仪测向精度,改进后的神经网络的收敛速度得到大大提高。

An Interferometer Direction Finding Method Based on Neural Network

To improve the direction finding precision of interferometer,the multi-baseline interferometer is modeled and expressed with mathematic model,and the direction-finding error is shown with theoretical analysis.A new direction finding algorithm for multi-baseline interferometer is presented,which is based on neural network.After that the BP neural network is improved by Levenberg-Marquardt algorithm.Finally,a simulation experiment is made.The experiment results show that the algorithm based on the BP neural network improves the precision of airborne interferometer,and the improved BP network is faster and more efficient than the BP neural network.