水中非合作运动磁性目标跟踪及参数估计

Study on the tracking and parameter estimating of unknown moving magnetism objects

针对水中非合作运动磁性目标的跟踪及参数估计问题,提出了一种基于Hopfield神经网络和磁梯度张量定位结合粒子滤波的方法对磁性目标进行跟踪,利用Hopfield神经网络和磁梯度张量定位为粒子滤波算法提供一部分更加优化的粒子,以增加粒子滤波算法中粒子的多样性,能够有效克服粒子退化现象,同时保证动态跟踪的速度和精度,并解决了远场磁定位中信噪比下降的问题.通过模拟仿真和磁性船模磁定位实验验证了算法的有效性,该算法可用于空中磁探潜和港口磁防御以及一般的UXO探测和定位问题,具有一定的军事和民用意义.

In aiming to examine the tracking and parameter estimating of unknown moving magnetic objects, a mag-netic target tracking method based on Hopfield neural network and magnetic gradient tensor orientation combining particle filter was presented. The Hopfield neural network and magnetic gradient tensor orientation may provide some more optimized particles for particle filter algorithm. Thus the diversity of particles is improved in a particle filter algorithm. The algorithm was found able to effectively conquer the degeneration of particles, and simultaneous-ly ensure the speed and precision of dynamic tracking, and solve the problem on declining of signal-to-noise in far field magnetic localization. The results of simulation and magnetic ship model magnetic orientation experiment show the validity of this algorithm. This algorithm will have some significance in martial and civil field such as submarine magnetic detection, magnetic defense of military port, and unexploded ordnance （UXO） detection and orientation.