基于神经网络的无源多传感器属性数据关联

Multiple Passive Sensors Feature Data Association Based on Neural Networks

采用引入动量项、自适应调整步长、Levenberg-Marquardt优化方法对基本的BP神经网络进行改进,以提高学习速度,改进的BP神经网络学习算法用于对无源多传感器获得的雷达辐射源参数进行属性数据关联,能够自适应地调整阈值,即根据训练数据调整关联的门限值,与确定门限的属性关联算法相比,有着很高的关联正确率。

An introduced momentum item, adaptive step adjust and Levenberg-Marquardt optimal methods are used to improve the basic BP neural networks, and training speed is highly developed as well. The improved BP neural networks learning algorithm is presented to associate the feature data of radar emitters received by multiple passive sensors. It can adapively adjust the threshold, i.e. adjust the feature associate threshold according to training data. Compared with fixed threshold feature association algorithm, it shows a high association rate in feature association.