基于朴素贝叶斯分类器的空中红外目标抗干扰识别方法研究

红外诱饵对抗技术的发展使得空战环境日益复杂化,对红外成像制导空空导弹抗干扰目标识别技术提出了更高的要求。红外诱饵的投放使得目标特征的完整性、显著性及稳定性遭到破坏,基于特征融合匹配的统计模式识别方法无法准确识别目标。提出了一种基于朴素贝叶斯分类器的抗干扰目标识别方法,该方法对空战对抗仿真图像数据集进行了特征挖掘,利用实验拟合方法构建了典型特征的概率密度函数模型,构造了朴素贝叶斯分类器,实现了飞机目标和干扰的分类识别。仿真实验结果表明,该方法在已测试的弹道图像数据集下的平均识别正确率达到了81.82%,且能够解决假目标、目标遮挡等抗干扰目标的识别难题。     

一种基于Gabor深度学习的无人机目标检测算法

随着智能控制技术的不断成熟,无人机给军事领域带来快速发展的同时也带来了威胁.因此针对空中飞行的无人机进行实时检测的任务需求,设计了一种基于Gabor深度学习的无人机目标检测算法.首先,搭建基于Gabor滤波器的深度神经网络,输入的图片经过该网络进行网格化划分,用以特征提取;然后,针对每个格子的特征利用回归算法计算其中物体的位置信息,并利用分类算法计算物体的类别信息,对以上得到的回归和分类结果进行筛选、融合得到最终的检测结果;最后,采集空中飞行的无人机真实数据构建数据集,在此基础上进行网络模型训练和算法验证.     

基于FRFT域特征差异的压制干扰检测与分类算法

针对干信比未知情况下有源压制干扰分类识别结果可信度较低的问题, 提出了一种基于FRFT域特征差异的压制干扰检测与分类算法。首先, 通过FRFT域峰值阶次的序贯判决算法, 进行压制干扰的存在性检测, 以保证压制干扰分类识别在较高的干信比条件下进行; 然后, 在此基础上, 分别提取回波信号在FRFT域的极值阶次标准差和峰值阶次标准差作为分类识别特征量, 同时, 为避免硬判决造成的分类错误, 采用模糊判决的方法得到基于不同特征参数的分类识别结果; 最后, 按一定准则将2种分类识别结果进行融合, 以进一步提高分类识别正确率。仿真结果表明, 与现有压制干扰分类识别算法相比, 该算法较好地解决了分类识别结果可信度较低的问题, 同时具有较高的分类识别正确率。      

面向天基监视的红外弱小飞行目标识别算法

针对当前红外弱小飞行目标特征不明显、背景干扰大等问题,提出了一种基于深度学习的红外弱小目标识别算法。检测框架以YOLOv4模型为基础,通过使用K-means++算法对训练集的候选框进行聚类处理,在初始大小的选取上放弃随机生成初始点的方式,在样本集里选取某一个样本作为初始中心使锚框（anchor）大小的选取更加合理。在模型结构中引入卷积注意力模块,使算法模型计算资源分配更合理,对红外弱小飞行目标的特征信息更加敏感。改进空间金字塔池化模块,使用平均池化可以更多保留图像的原始信息,降低天基成像中的噪点与坏点的影响。仿真实验表明采用K-means++计算Anchor大小时准确率可以达到80.13%,在加入了SPP和CBAM模块后之后在测试集上算法识别准确率达到了83.3%,经过对模型的修改有效提升了对红外弱小飞行目标识别的准确率。     

基于Faster R-CNN的精密零部件的识别方法

在航空航天领域中,惯性陀螺等精密器件装配精度要求较高,目前大多采用人工装配的方法,装配效率低、装配过程受人主观影响大。针对上述存在的问题,采用基于Faster R-CNN模型的目标识别算法,通过VGG16特征提取网络提取特征信息,在模型训练过程中利用COCO数据集的深度网络模型进行迁移训练,防止模型过拟合并加速参数的训练过程。同时,该方法还与其他深度学习模型以及传统的目标识别算法进行了对比,在自建的数据模型测试集上进行试验。结果表明,基于VGG16的Faster R-CNN目标识别模型在复杂环境及物体发生遮挡的情况下对于惯性陀螺的识别具有明显的优势,准确率可达到87.80%,召回率80.30%,识别速度可达到15FPS,能够满足实时性要求。     

改进YOLOv5s的弱光水下生物目标检测算法

针对水下光学图像目标检测过程中由于水中光线衰弱严重、图像环境复杂和拍摄设备移动等造成的生物识别精度低的问题,提出了基于改进YOLOv5s的弱光水下生物目标实时检测算法YOLOv5s-underwater。针对弱光水下光线衰弱的问题,引入了限制对比度自适应直方图均衡(CLAHE)算法对输入图像进行预处理,解决了颜色失真和图像毛糙的问题。针对复杂的弱光水下图像环境,提出了快速空间金字塔池化(SPPF)模块,解决了水下物体区分度低和特征损失严重的问题。针对拍摄设备移动带来的场景和形态变化问题,提出了一种基于旋转窗口的SwinTransformer模块,提高了模型的泛化能力。针对水下小目标,修改了网络模型结构,提高了小目标的检测能力。仿真和实验结果表明：所提算法相较于YOLOv5s检测精度提高30.7%,证明了算法的有效性。     

基于倒置残差注意力的无人机航拍图像小目标检测

针对无人机航拍图像背景复杂、小尺寸目标较多等问题,提出了一种基于倒置残差注意力的无人机航拍图像小目标检测算法。在主干网络部分嵌入倒置残差模块与倒置残差注意力模块,利用低维向高维的特征信息映射,获得丰富的小目标空间信息和深层语义信息,提升小目标的检测精度;在特征融合部分设计多尺度特征融合模块,融合浅层空间信息和深层语义信息,并生成4个不同感受野的检测头,提升模型对小尺寸目标的识别能力,减少小目标的漏检;设计马赛克混合数据增强方法,建立数据之间的线性关系,增加图像背景复杂度,提升算法的鲁棒性。在VisDrone数据集上的实验结果表明：所提模型的平均精度均值比DSHNet模型提升了1.2%,有效改善了无人机航拍图像小目标漏检、误检的问题。     

基于迁移学习的暴恐图像自动识别

利用人工智能和深度学习技术自动化地分析互联网海量图片,快速、准确地识别有害的暴恐图像并及时处置是反恐工作的重要手段之一。研究了利用深度学习和迁移学习技术对暴恐图像进行分类识别。首先,定义了暴恐图像的主要概念特征,并针对性地构建数据集;其次,针对暴恐图像正样本较少的问题,设计深度神经网络模型和迁移学习方式;最后,基于构建的训练数据集进行模型训练和测试。结果显示:所提方法可以快速、准确地对互联网图片进行分类识别,平均分类准确率达到96.7%,从而有效降低人工检测的劳动强度,为反恐预警工作提供决策支持。      

复杂背景下机场道面细带状结构病害检测算法

机场道面裂缝、角隅断裂、接缝破碎、修补等病害宽度狭小、长短不一、图像中像素占比少,呈细带状结构,且与复杂背景对比度低,现有检测算法效果不佳。针对以上问题,提出了一种基于注意力机制与特征融合的深度神经网络模型DetMSPNet。首先,利用注意力机制模块CBAM,使得特征学习更加专注于细带状结构病害区域,抑制干扰信息;其次,构建残差空洞金字塔模块,提取不同尺度空间下的特征信息;然后,设计最大池化支路,便于之后浅、深层不同层次特征进行融合,加强模型对于病害的定位能力,并且将深层特征输入3种不同扩张率的扩张卷积和金字塔池化模块,使得病害特征包含更多全局上下文信息;最后,对所有层输出的病害特征信息进行融合,实现不同尺度、不同层次特征的信息互补。与目前3种经典的目标检测算法在机场道面病害图像数据集APD上做了对比实验,结果表明:所提算法的mAP达到78.51%,优于对比算法。所提DetMSPNet模型,提高了算法对机场道面细带状结构病害检测中宽度狭小、长短不一、图像中像素占比少、与复杂背景对比度低等情况的适应能力。      

Scaling analysis and model estimation of solar corona index

A monthly average solar green coronal index time series for the period from January 1939 to December 2008 collected from NOAA (The National Oceanic and Atmospheric Administration) has been analysed in this paper in perspective of scaling analysis and modelling. Smoothed and de-noising have been done using suitable mother wavelet as a pre-requisite. The Finite Variance Scaling Method (FVSM), Higuchi method, rescaled range (R/S) and a generalized method have been applied to calculate the scaling exponents and fractal dimensions of the time series. Autocorrelation function (ACF) is used to find autoregressive (AR) process and Partial autocorrelation function (PACF) has been used to get the order of AR model. Finally a best fit model has been proposed using Yule-Walker Method with supporting results of goodness of fit and wavelet spectrum. The results reveal an anti-persistent, Short Range Dependent (SRD), self-similar property with signatures of non-causality, non-stationarity and nonlinearity in the data series. The model shows the best fit to the data under observation.     

Stochastic particle acceleration by the forced interaction of relativistic current sheets

We address the problem of interacting relativistic current sheets in self-consistent kinetic plasma simulations within the framework of the Particle-In-Cell model. The interaction is enforced in head-on collisions of up to 10 current sheets at relativistic bulk speeds. The simulations are motivated by the general problem of Poynting flux dissipation in ‘striped wind’ configurations presumably governing the relativistic outflows pervasive in pulsar winds and gamma-ray bursts. We identify the generation of non-thermal particles and formation of a stable power-law shape in the particle energy distributions f(γ) dγ ∝ γ^−s dγ. In 1D, a spectral index s ∼ 2 is observed and attributed to a stochastic Fermi-type acceleration mechanism. In 2D, the generic index of s ∼ 3–4 is retained as in previous simulations of individual current sheets. Whereas in 2D the high energy cut-off is constrained by the limited dissipation of magnetic energy, in 1D the process converts the bulk motion of current sheets towards directed particle momentum of an exclusive class of non-thermal particles.     

Development of a gas chromatography compatible Sample Processing System (SPS) for the in-situ analysis of refractory organic matter in martian soil: preliminary results

In the frame of the 2009 Mars Science Laboratory (MSL) mission a new sample preparation system (SPS) compatible with gas chromatography–mass spectrometry (GC–MS) has been developed for the in situ analysis of complex organic molecules in the Martian soil. The goal is to detect, if they exist, some of the key compounds that play an important role in life on Earth including carboxylic acids, amino acids and nucleobases.     

Learning in an immersive digital theater

The Houston Museum of Natural Science, in collaboration with Rice University has an outreach program taking portable digital theaters to schools and community sites for over five years and has conducted research on student learning in this immersive environment. By using an external independent evaluator, the effectiveness of NASA-funded Education and Public Outreach (EPO) projects can be assessed. This paper documents interactive techniques and learning strategies in full-dome digital theaters. The presentation is divided into Evaluation Strategies and Results and Interactivity Strategies and Results. All learners from grades 3–12 showed statistically significant short-term increase in knowledge of basic Earth science concepts after a single 22-min show. Improvements were more significant on items that were taught using more than one modality of instruction: hearing, seeing, discussion, and immersion. Thus immersive theater can be an effective as well as engaging teaching method for Earth and Space science concepts, particularly those that are intrinsically three-dimensional and thus most effectively taught in an immersive environment. The portable system allows taking the educational experience to rural and tribal sites where the underserved students could not afford the time or expense to travel to museums.     

基于LuGre模型的电液加载系统摩擦补偿

为了提高电液加载系统控制精度,针对摩擦问题提出了基于LuGre摩擦模型的前馈补偿方法.建立了用于摩擦仿真分析和补偿器设计的电液加载系统数学模型;通过实验获取并分析了相关的摩擦数据;基于实验数据进行LuGre模型参数辨识,把LuGre模型和辨识结果引入电液加载系统数学模型,并进行仿真结果与实际摩擦数据的对照,证明了LuGre摩擦模型的准确性.设计前馈补偿器,进行了实验对比,实验结果表明前馈补偿器可将摩擦产生的控制误差有效地降至未补偿时的30%左右.      

CoRoT and Kepler results: Solar-like oscillators

The space-borne observatories CoRoT (Convection Rotation and planetary Transits) and Kepler have provided photometric time series data of unprecedented precision for large numbers of stars. These data have revolutionized the fields of transiting exoplanets and asteroseismology. In this review some important asteroseismic results obtained using data from the CoRoT and Kepler space missions concerning stars that show solar-like oscillations are discussed. These results comprise, among others, measurements of the location of the base of the convection zone and helium second-ionization zone in main-sequence stars, the presence (or not) of core-helium burning in red-giant stars, as well as differential rotation in these stars.     

Return to Europa: Overview of the Jupiter Europa orbiter mission

Missions to explore Europa have been imagined ever since the Voyager mission first suggested that Europa was geologically very young. Subsequently, the Galileo spacecraft supplied fascinating new insights into this satellite of Jupiter. Now, an international team is proposing a return to the Jupiter system and Europa with the Europa Jupiter System Mission (EJSM). Currently, NASA and ESA are designing two orbiters that would explore the Jovian system and then each would settle into orbit around one of Jupiter’s icy satellites, Europa and Ganymede. In addition, the Japanese Aerospace eXploration Agency (JAXA) is considering a Jupiter magnetospheric orbiter and the Russian Space Agency is investigating a Europa lander.     

Radioactivity of the Moon and planets

In this report the main results of the study of radioactivity of the solar sistem bodies are considered. The radioactivity of the Moon and planets was measured by means of vehicles in situ. The radioactivity of the lunar samples, brought to the Earth was studied with laboratory equipment.     

基于最小值原理的冗余度机器人动力学控制

为了解冗余度机器人全局法优化中数值求解的困难,本文讨论了动力学方程的建立,无约束和有约束最优控制问题之间的内在联系,重点分析了求解最优控制问题的数值方法,文中提出了双向异步积分迭代的求解正则方程组的直接迭代法,较好解决了状态方程和协态方程稳定相逆给求解两点这值问题带来的困难。     

基于GPS载波相位的微卫星姿态算法研究

GPS姿态系统是利用GPS载波相位测量来确定载体的航向和姿态角。本文对微卫星的GPS姿态系统进行了研究。重点解决短基线的GPS状态算法，讨论了姿态价格函数的了小化方法。测试结果表明，所提的算法对小于1m的基线是有效的。