基于深度卷积神经网络的地磁导航方向适配性分析

发布时间：2018-02-17 06:03

本文关键词： 地磁导航适配性分析方向适配性卷积神经网络 Gabor滤波器　出处：《工程科学学报》2017年10期 　论文类型：期刊论文

【摘要】：针对地磁导航方向适配性分析时人工提取的特征主观性较强且难以表达深层的结构性特征的问题,提出一种基于深度卷积神经网络(convolutional neural network,CNN)的地磁导航方向适配性分析方法.首先,利用Gabor滤波器的方向选择特性建立了6个典型方向的适配特征图;然后,设计了卷积神经网络对深层次的方向适配特征进行提取,并通过混和粒子群算法(hybrid particle swarm optimization,HPSO)对卷积神经网络的训练参数进行优选;最后,通过仿真实验对所提方法进行了验证.结果表明,该方法可有效避免复杂的计算以及人工特征提取的盲目性,实现了地磁导航方向适配性分析的自动化,且所提方法的准确率高于传统的BP网络和支持向量机,对地磁导航和航迹规划具有指导意义.
[Abstract]:In order to solve the problem that the artificial extracted features are subjective and difficult to express the deep structural features in the analysis of geomagnetic navigation orientation suitability, In this paper, a method of direction adaptation analysis for geomagnetic navigation based on convolutional neural network is proposed. Firstly, six typical orientation adaptive feature maps are established by using the direction selection characteristics of Gabor filters. A convolutional neural network is designed to extract the directional adaptive features of the deep layer, and the hybrid particle swarm optimization algorithm (HPSO) is used to optimize the training parameters of the convolutional neural network. Finally, a hybrid particle swarm optimization algorithm is used to select the training parameters of the convolutional neural network. The simulation results show that the proposed method can effectively avoid the blindness of complex calculation and artificial feature extraction, and realize the automation of orientation adaptability analysis of geomagnetic navigation. The accuracy of the proposed method is higher than that of the traditional BP neural network and support vector machine, which is of guiding significance for geomagnetic navigation and route planning.
【作者单位】：陆军工程大学;中国人民解放军驻西北工业大学军事代表室;
【基金】：武器装备军内科研重点资助项目(2014551)
【分类号】：TN966;TP183

【相似文献】