一种基于卷积神经网络的地磁基准图构建方法

地磁匹配导航技术是一种重要的辅助导航制导方法,地磁基准图的构建精度对地磁匹配制导的精准度起着决定性作用。针对现有地磁基准图构建精度难以满足实际地磁匹配导航需求的问题,提出了一种基于卷积神经网络的地磁基准图构建方法。首先,利用卷积层提取低分辨率基准图中的特征图像块;然后,利用基于学习的阈值收缩算法（LISTA）实现图像块的稀疏表示;最后,利用三通道的地磁信息得到重建后的高分辨率基准图。实验结果表明:所提方法对地磁基准图具有更高的构建精度,同时对噪声有更好的鲁棒性,各种客观评价指标均高于现有的超分辨率重建方法。 

A geomagnetic reference map construction method based on convolutional neural network

Geomagnetic matching navigation technology is an important auxiliary navigation guidance method. The construction accuracy of geomagnetic reference map plays a decisive role in the accuracy of geomagnetic matching guidance. Aimed at the problem that the construction accuracy of the existing geomagnetic reference maps is difficult to meet the actual requirements of geomagnetic matching navigation, a construction method of geomagnetic reference maps based on convolutional neural network is proposed. First, the convolutional layer is used to extract the feature image patches in the low-resolution reference image. Then, the Learned Iterative Shrinkage and Thresholding Algorithm (LISTA) is used to realize the sparse representation of the low-resolution image patches. Finally, the three-channel geomagnetic information is used to obtain the final reconstructed high-resolution reference map. The experimental results show that the proposed method has a higher construction accuracy for geomagnetic reference map and better robustness to noise. Various objective evaluation indexes of the proposed method are higher than those of the existing super-resolution reconstruction algorithms.