利用人工神经网络提前1h预报电离层TEC

提出了一种利用人工神经网络提前1h预报电离层TEC的简便方法. 考虑到实际工程应用要求, 没有使用其他空间天气参数, 而是选择电离层TEC观测数据本身作为输入参数. 输入参数为当前时刻TEC、一阶差分、相对差分和时间, 输出参数为预报时刻TEC. 利用文中介绍的GPS/TEC处理方法解算厦门站2004年电离层TEC观测数据, 对预报方法进行评估, 全年平均相对误差为9.3744%, 预报结果与观测值相关性达到了0.96678. 结果表明, 利用人工神经网络方法提前1h预报电离层TEC有很好的应用前景.     

基于宇航级FPGA的YOLOv5s网络模型硬件加速

由于遥感图像具有分辨率高和背景信息复杂的特点,其对目标检测的精确性和鲁棒性要求越来越高,因此遥感图像处理领域逐渐引入了卷积神经网络算法。然而此类算法通常模型复杂且计算量庞大,难以在空间与资源受限的星上平台高效运行。针对这一问题,提出一种基于宇航级现场可编程门阵列（Filed Programmable Gate Array, FPGA）的卷积神经网络硬件加速架构,并选用YOLOv5s作为目标网络,采用输入与输出通道并行展开以及数据流水线控制的策略进行架构设计。实验结果表明,在使用该处理架构加速YOLOv5s的推理阶段,卷积模块的工作频率可以达到200 MHz,其运算性能高达394.4GOPS（Giga Operations Per Second）,FPGA的功耗为14.662 W,数字信号处理（Digital Signal Processing, DSP）计算矩阵的平均计算效率高达96.29%。     

Correlation between sunspots and interplanetary shocks measured by ACE during 1998–2014 and some estimations for the 22nd solar cycle and the years between 2015 and 2018 with artificial neural network using the Cavus 2013 model

The Advanced Composition Explorer (ACE) spacecraft has measured 235 solar-based interplanetary (IP) shock waves between the years of 1998–2014. These were composed of 203 fast forward (FF), 6 slow forward (SF), 21 fast reverse (FR) and 5 slow reverse (SR) type shocks. These data can be obtained from the Interplanetary Shock Database of Harvard-Smithsonian Centre for Astrophysics. The Solar Section of American Association of Variable Star Observers (AAVSO) is an organization that counts the number of the sunspots. The effects of interplanetary shock waves on some physical parameters can be computed using a hydrodynamical model. There should be some correlations between these effects and the sunspot variations. The major objective of this paper is twofold. The first one is to search these correlations with sunspots given in the database of AAVSO. As expected, high correlations between physical parameters and sunspots have been obtained and these are presented in tables below. The second objective is to make an estimation of these parameters for the 22nd solar cycle and the years between 2015 and 2018 using an artificial neural network. Predictions have been made for these years where no shock data is present using artificial intelligence. The correlations were observed to increase further when these prediction results were included.     

基于深度学习的航天器组合体惯性参数在轨智能辨识

针对在轨服务过程形成新组合体的动力学参数未知的问题,借助深度学习在多参数寻优上的优势,提出了一种基于卷积神经网络的智能参数辨识算法,实现在外力作用下,线动量和角动量不守恒条件下的航天器组合体多参数辨识。利用卷积神经网络权值共享的特点,设计4层卷积神经网络,通过短时间内对大量特定存储形式的状态数据的训练,实现航天器组合体多参数快速高精度辨识。利用数学仿真试验对算法的可行性进行验证,结果表明：在24s内,质量与质心位置收敛;1190s内,惯量参数收敛,辨识精度在3%以内。说明所提方法在外界随机干扰力和力矩影响下能准确快速辨识出航天器组合体质量、质心位置和惯量矩阵。     

An approach to multi-temporal MODIS image analysis using image classification and segmentation

This paper discusses an approach for river mapping and flood evaluation based on multi-temporal time series analysis of satellite images utilizing pixel spectral information for image classification and region-based segmentation for extracting water-covered regions. Analysis of MODIS satellite images is applied in three stages: before flood, during flood and after flood. Water regions are extracted from the MODIS images using image classification (based on spectral information) and image segmentation (based on spatial information). Multi-temporal MODIS images from “normal” (non-flood) and flood time-periods are processed in two steps. In the first step, image classifiers such as Support Vector Machines (SVM) and Artificial Neural Networks (ANN) separate the image pixels into water and non-water groups based on their spectral features. The classified image is then segmented using spatial features of the water pixels to remove the misclassified water. From the results obtained, we evaluate the performance of the method and conclude that the use of image classification (SVM and ANN) and region-based image segmentation is an accurate and reliable approach for the extraction of water-covered regions.     

基于改进CNN-BiLSTM模型和地磁监测数据的多时间长度GIC预测

太阳风暴在电力系统网络中驱动产生的GIC会影响电力设备和系统的安全运行, 严重时还会引发大面积停电事件. 预测电网GIC水平能够为电力系统保护措施提供重要参考, 然而对这方面的研究仍显不足. 为了解决该问题, 将卷积神经网络(CNN)与双向长短时记忆(BiLSTM)以及注意力机制相结合, 利用空间天气的相关监测信息, 提出了大规模电网GIC多时间长度的预测方法. 本文在分析太阳风暴驱动产生电网地磁感应电流(Geomagnetically Induced Current, GIC)基础上, 构建了GIC预测模型; 提出了基于多头注意力机制的CNN-BiLSTM改进模型, 对GIC进行预测, 并给出了预测流程. 采用CNN捕获地磁扰动局部信息, 根据BiLSTM综合地磁暴扰动信息的全局特征, 综合利用多头注意力机制评估对GIC关键作用的地磁信息片段, 实现电网GIC的预测. 利用2004年11月8日00:00 LT－20:00 LT巨型磁暴期间DED地磁台站和QGZH地磁台监测数据, 应用所提方法对岭澳500 kV变电站GIC进行回归预测. 经过训练后, GIC预测相对误差均在12%以内, 精度高于其他模型的预测结果.     

神经网络在下颌运动轨迹图测量中的应用

介绍了采用小型磁石测量下颌运动轨迹图Electrognathography(EGN)的方法。给出利用磁偶极子模型代替小磁石模型的理论及实际依据。利用神经网络解决由磁石产生的磁感应强度推定其空间参数的反演问题 ,并且采用了分割网络及修正网络的方法来提高磁石运动测量的推定准确度。     

北斗MEO卫星辐射剂量探测数据缺失值处理方法

针对北斗MEO卫星辐射剂量探测数据出现连续性缺失的问题,开展缺失值处理方法研究.提出一种叠加正弦波的线性样条回归方法,即引入样条函数,对各数据连续缺失的时间段进行分段处理,每段样条采用叠加正弦波的线性方程填充缺失值.结果表明：利用该方法处理缺失值,每段样条中填充曲线与探测曲线在增长趋势、周期性波动等方面具有较高的一致性;相比前向插值法和线性插值法,其填充值与真实值误差更小,关联性更高.该方法较好地解决了数据连续缺失的问题,形成了完整性好、准确性高的北斗MEO卫星辐射剂量数据集,为后续数据的发布、建模和可视化展示等奠定了基础.     

人工神经网络在磁情预报中的应用

介绍了人工神经元网络在磁情预报中的应用研究，并对具体应用于C9指数预报的MLP的拓扑结构，预报效率及在预报时所表现出的特点和问题进行了讨论。     

基于人工智能技术的雷达目标检测

强杂波背景下的目标检测是各国争相研究的热点,传统的雷达目标检测技术无法实现,人工智能技术的出现,其在目标识别和特征提取方面的独特优势,使得强杂波背景下的目标检测成为可能,利用人工智能神经网络的强大非线性预测能力对海杂波的非线性动力方程进行逼近,实现了对海杂波的非线性预测,提出了一种基于神经网络技术的强海杂波背景下的脉冲多普勒雷达导引头小目标检测方法,使得强海杂波背景下的目标检测能力有2 dB到6 dB的提升。     

Using TensorFlow-based Neural Network to estimate GNSS single frequency ionospheric delay (IONONet)

In the last 20?years, and in particular in the last decade, the availability of propagation data for GNSS has increased substantially. In this sense, the ionosphere has been sounded with a large number of receivers that provide an enormous amount of ionospheric data. Moreover, the maturity of the models has also been increased in the same period of time. As an example, IGS has ionospheric maps from GNSS data back to 1998, which would allow for the correlation of these data with other quantities relevant for the user and space weather (such as Solar Flux and Kp). These large datasets would account for almost half a billion points to be analyzed. With the advent and explosion of Big Data algorithms to analyze large databases and find correlations with different kinds of data, and the availability of open source code libraries (for example, the TensorFlow libraries from Google that are used in this paper), the possibility of merging these two worlds has been widely opened. In this paper, a proof of concept for a single frequency correction algorithm based in GNSS GIM vTEC and Fully Connected Neural Networks is provided. Different Neural Network architectures have been tested, including shallow (one hidden layer) and deep (up to five hidden layers) Neural Network models. The error in training data of such models ranges from 50% to 1% depending on the architecture used. Moreover, it is shown that by adjusting a Neural Network with data from 2005 to 2009 but tested with data from 2016 to 2017, Neural Network models could be suitable for the forecast of vTEC for single frequency users. The results indicate that this kind of model can be used in combination with the Galileo Signal-in-Space (SiS) NeQuick G parameters. This combination provides a broadcast model with equivalent performances to NeQuick G and better than GPS ICA for the years 2016 and 2017, showing a 3D position Root Mean Squared (RMS) error of approximately 2?m.     

基于深度神经网络的空间目标结构识别

利用空间目标雷达散射截面(Radar Cross Section, RCS)序列开展空间目标结构识别是空间态势感知的重要组成部分。文章针对 RCS序列受目标物理特性、姿态特性影响大,序列信号非平稳特征明显的问题,利用深度神经网络（Deep Neural Network,DNN）算法解决空间目标结构特征识别的问题;针对特征提取不具区分度的问题,提出利用分形分析提取RCS序列的分数维特征,并利用Fisher判决率对传统特征进行选取;介绍了DNN算法以及数据处理过程;最后,利用一组仿真测试数据对算法进行了仿真验证。分析结果表明,DNN算法在解决利用RCS序列进行目标结构识别这一问题中具有鲁棒性强、识别准确的特点。     

Comparison of ANN and MLR models for estimating solar radiation in Turkey using NOAA/AVHRR data

In this paper, the estimation capacities of MLR and ANN are investigated to estimate monthly-average daily SR over Turkey. The satellite data are used for 73 different locations over Turkey. Land surface temperature, altitude, latitude, longitude and month are offered as the input variables for modeling ANN and MLR to get SR. Estimations of SR are evaluated with the meteorological values by using the statistical bases. The obtained results indicated that the ANN model could achieve a satisfactory performance when compared to the MLR model. Moreover, it is understood that more accurate results in estimation of SR are obtained in the use of satellite data, rather than the use of meteorological station data. Finally, the built ANN model is used to estimate the yearly average of daily SR over Turkey. As a result, satellite-based SR map for Turkey is generated.     

Optimal station-keeping near Earth–Moon collinear libration points using continuous and impulsive maneuvers

In this study the gravitational perturbations of the Sun and other planets are modeled on the dynamics near the Earth–Moon Lagrange points and optimal continuous and discrete station-keeping maneuvers are found to maintain spacecraft about these points. The most critical perturbation effect near the L₁ and L₂ Lagrange points of the Earth–Moon is the ellipticity of the Moon’s orbit and the Sun’s gravity, respectively. These perturbations deviate the spacecraft from its nominal orbit and have been modeled through a restricted five-body problem (R5BP) formulation compatible with circular restricted three-body problem (CR3BP). The continuous control or impulsive maneuvers can compensate the deviation and keep the spacecraft on the closed orbit about the Lagrange point. The continuous control has been computed using linear quadratic regulator (LQR) and is compared with nonlinear programming (NP). The multiple shooting (MS) has been used for the computation of impulsive maneuvers to keep the trajectory closed and subsequently an optimized MS (OMS) method and multiple impulses optimization (MIO) method have been introduced, which minimize the summation of multiple impulses. In these two methods the spacecraft is allowed to deviate from the nominal orbit; however, the spacecraft trajectory should close itself. In this manner, some closed or nearly closed trajectories around the Earth–Moon Lagrange points are found that need almost zero station-keeping maneuver.