基于模式识别神经网络的重力匹配算法

为了提高重力辅助惯性导航系统在重力异常明显区域内的定位精度和匹配率,用模式识别神经网络的方法进行了重力匹配.在匹配时刻,根据惯导指示位置确定在一定的网格点范围内搜索载体真实位置,以每个网格点为终点把惯导指示航迹放置到重力图上,由此提取一系列的参考重力图上数据,并把它和对应网格点的位置定义成一个模式类,把所有的模式类作为概率神经网络的样本训练一个模式识别神经网络,然后把重力仪测量数据使用该神经网络识别到某个模式类,对比模式类的定义可以确定此时的载体位置.计算仿真研究表明,该算法的重力匹配率优于通常的相关匹配算法,其组合导航系统的定位误差在1个重力图网格左右.

Gravity matching algorithm based on pattern recognition neural network

In order to improve the locating precision and matching rate of the gravity aided inertial navigation system(INS) in regions with significant gravity anomaly characteristic,the pattern recognition neural network was used to investigate the problem of gravity matching.While matching,a scope with certain grid points surrounding the INS indicating position was plotted as the matching area to search the real location of the vehicle.A set of gravity data was then drawn from reference gravity map by placing the indicating track of the INS onto the gravity map.These data were defined as some pattern classes and used to train a probabilistic neural network.The measurements of gravity meter can be recognized to one pattern class by this neural network and the vehicle can be located according to the definition of this pattern class.Simulation results show that the presented algorithm achieves better performance than ordinary correlative matching algorithm and the matching precision is approximately one grid on gravity map.