基于并行卷积神经网络的地磁方向适配性分析

发布时间：2018-04-17 16:08

本文选题：地磁导航 + 方向适配性　；参考：《中国惯性技术学报》2017年03期

【摘要】：针对地磁方向适配性分析时人工特征提取主观性较强、所取特征难以表达深层的结构性特征的问题,并为了进一步提高方向适配性分析的准确率,提出了一种基于并行卷积神经网络的地磁方向适配性分析方法。首先,从不同角度建立了地磁场在6个代表方向上的适配性分析图;然后,从同一磁场的不同角度出发,利用卷积神经网络自动完成了特征学习,得到了更为全面的方向适配性特征描述;最后,在并行卷积神经网络所得特征的基础上,利用BP网络建立了地磁方向适配性的分析模型。仿真结果证明,该方法可以有效避免人工特征提取和计算等复杂步骤,实现了地磁方向适配性分析的自动化,而且可以获得优于传统网络和单路卷积神经网络的准确率。
[Abstract]:In order to improve the accuracy of directional adaptation analysis, the artificial feature extraction is more subjective, and it is difficult to express the deep structural features in the geomagnetic orientation adaptation analysis.A method of geomagnetic orientation adaptation analysis based on parallel convolution neural network is proposed.Firstly, the matching analysis map of geomagnetic field in six representative directions is established from different angles, and then, from different angles of the same magnetic field, the feature learning is accomplished automatically by using convolution neural network.Finally, on the basis of the characteristics obtained by parallel convolution neural networks, an analytical model of geomagnetic directional fitness is established by using BP neural network.The simulation results show that this method can effectively avoid the complicated steps such as artificial feature extraction and calculation, and realize the automation of geomagnetic orientation adaptability analysis, and the accuracy of this method is better than that of traditional neural networks and single-channel convolution neural networks.
【作者单位】：军械工程学院;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项目(2014551)
【分类号】：TN966;TP183

【参考文献】