神经网络校正的EKF在水下被动目标跟踪中的应用研究

Application of Nerual Network-aided Extend Kalman Filtering Technique in Underwater Passive Target Tracking

对于水下目标被动跟踪,通常采用扩展卡尔曼滤波算法进行目标状态估计,但在目标跟踪过程中,由于目标运动的不确定性及系统噪声的影响,此时对目标的状态估计通常难以获得较高的精度.针对以上问题,本文提出一种由BP神经网络来校正扩展卡尔曼滤波的被动目标跟踪算法.利用BP神经网络的学习能力,将卡尔曼滤波过程中的滤波增益、滤波值与预测值之差、滤波值与量测值之差作为BP神经网络的输入,学习得出卡尔曼滤波的滤波误差,并利用此误差值对滤波过程进行在线校正.仿真表明,BP神经网络辅助校正扩展卡尔曼滤波的方法,对滤波过程的可靠性和精确度都有了提升.

In underwater target passive tracking field,Extend Kalman filter is usually adopted to estimate the target state.However,due to the uncertainty of the target motion and the system noise during the target tracking process,it is difficult to obtain a high accurate estimation of the states of the target.To solve these problems,in this paper we propose a passive target tracking algorithm that uses BP neural network to calibrate the results of the extend Kalman filter.The filter gain,the difference between filtered value and predicted value,and the difference between filtered value and measured value derived from the Kalman filter are employed as the input to the BP neural network.The filter error can then be obtained by training neural network.Finally,the filter process is corrected online by using this error value.The simulation results show that the method improves the reliability and accuracy of the filtering process.