星地协同光学遥感影像目标识别技术验证研究

为了提高光学遥感卫星信息处理的时效性,文章提出了星地协同光学遥感影像在轨目标智能识别技术框架.将基于遥感大数据的深度学习、神经网络模型训练等数据量大、运算量大、计算复杂、要求较高的处理任务部署在地面服务器,将深度学习训练得到的模型进行压缩并上注至卫星,卫星在轨利用轻量化模型对影像进行推理计算,最后把目标识别结果下传至用...     

光学遥感图像目标检测技术综述

目标检测是光学遥感图像分析的重要内容,是将图像数据转化为应用成果的关键一环。文章重点针对光学遥感图像中常见的飞机、舰船等目标,对目标检测技术进行了系统总结,并对未来重点发展方向进行了展望。重点论述的内容包括目标检测方法、目标检测性能评价准则和数据集,其中对于目标检测方法,从候选区域选择、特征学习、分类和后处理等四个方面进行了详细介绍。期望这些技术成果的总结分析和展望对目标检测技术的应用提供一定的借鉴和参考。     

基于全局—局部特征联合的光学卫星遥感图像舰船目标细粒度识别

针对光学遥感图像舰船目标类间差异小,需要丰富舰船目标的特征表示能力提高其细粒度识别准确率的问题,文章提出了一种基于全局—局部特征联合的舰船目标细粒度识别方法,设计了双分支特征提取与融合模型.首先,全局特征提取分支通过卷积神经网络提取图像的全局特征;其次,局部特征增强分支将浅层特征图打乱并重构,加入对抗性损失函数,训练网...     

合成孔径雷达图像目标识别问题研究

首先对合成孔径雷达(SAR)图像目标识别的研究背景和国内外发展现状作了简单介绍，然后分析了SAR图像目标识别问题的复杂性，讨论了SAR目标识别系统的一般方案与方法，提出了基于多源信息融合和人机协作的识别框架，进而展望了SAR图像目标识别技术今后的发展方向。     

基于深度卷积神经网络的遥感影像目标检测

随着遥感影像数据规模的快速扩张,如何高效准确地识别遥感影像中的典型目标成为当前的研究热点。为解决传统遥感影像目标检测方法准确率低的问题,用基于深度卷积神经网络进行遥感影像目标检测,在遥感影像数据集上用基于Faster-RCNN的神经网络模型对VGG16卷积网络进行训练,对输入的遥感影像通过区域推荐网络标注出待检目标的包围框和置信度,实现对遥感影像的目标检测。以飞机和油罐为例,在TensorFlow深度学习框架下实现了数据预处理、网络训练、目标检测等功能,并在当前测试数据集上取得了较高的检测准确率和置信度。该研究成果可应用于遥感影像解译和处理等相关领域。     

Orbital remote sensing: Space technology applications in South-east Asia

The changing context of space remote sensing applications is described in terms of evolving needs and capabilities in South-east Asia and in terms of recent policy decisions in spacefaring nations. Emphasis is placed on the need for a technology which allows the monitoring of, eg the exploitation and degradation of natural resources, crop development, climatic effects and the like. The implications of policy adjustments in the US LANDSAT programme, including its proposed commercialization, are discussed, and the potential role of other existing and future satellites is reviewed.     

The land remote sensing commercialization Act of 1984

This article examines the effect that private-sector remote sensing activities might have on the international community and on the future of international cooperation in space. It describes in detail The Land Remote Sensing Commercialization Act of 1984, recently passed by the US Congress and signed by President Reagan. The article concludes that it is possible to develop domestic regulations that balance the economic interests of the private sector with the legal and political concerns of the international community.     

多源遥感影像信息融合研究现状与展望

遥感影像信息融合是近几年来国际上遥感领域的研究热点，也是解决多源遥感海量数据富集表示的有效途径。它可以增强多重数据分析处理能力，为大规模遥感应用研究提供一个良好的基础。本文首先介绍了国内外进行多源遥感影像信息融合的研究发展状况，提出存在的关键技术问题和解决办法。最后，对其未来发展前景做了展望。     

某遥感卫星微振动对成像质量影响分析

微振动对成像质量影响问题成为制约某型号任务成败的关键因素。为预示微振动对成像质量的影响,首先对微振动源进行分析,然后针对性建立微振动传递路径数学模型,包括结构传递和姿控回路补偿,再根据光学系统设计建立光学元件运动与视轴晃动关系的数学模型,最后进行微振动对相机视轴影响的仿真分析。同时,为演示微振动对成像质量的影响,进行了图像质量仿真。分析结果表明,CMG的微振动对成像质量的影响无法忽略,需采取必要的控制措施。     

一种遥感影像快速计算水面率的方法

为了有效去除阴影噪声对遥感影像水体提取的干扰,快速准确计算大区域范围内的水面率,本文提出基于GF6（高分六号卫星）影像,利用UWI（Urban Water Index,城市水体指数）并结合FROM-GLC10（Finer Resolution Observa-tion and Monitoring of Global Land Cover 10,全球陆地覆盖精细化观测与监测10）数据,通过求交计算和小碎斑删除实现水体提取结果的精细化,有效消除了阴影等噪声对水体提取的干扰,实现水面率的快速计算。本文选择中山市作为试验区,试验区域范围广、地物背景复杂,综合对比本文算法、NDWI（Normalized Water Index,归一化水体指数）算法和TSUWI（Two-Sted Urban Water Index,城市二类水体指数）算法,从定性与定量角度分析,结果表明：本文算法提取的水体连续完整、漏提和误提较少,可有效去除阴影、建筑、道路等噪声干扰,在总体精度、Kappa系数、错分误差、漏分误差等指标均取得了较好结果。利用UWI与FROM-GLC10数据相结合计算水面率具有可行性,为工程化、大面积、快速计算水面率提供了一种新的思路。     

遥感技术在战场侦察与效果评估中的应用

从战场侦察与环境监视应用需求出发,结合当前遥感技术的发展方向,探讨了遥感技术在战场侦察与环境监视中的六个应用方向,即应用高分辨率影像进行战场打击效果评估、应用正射影像提高鉴别能力、应用热红外影像进行地表温度制图、应用雷达影像进行移动目标监测、天基红外预警系统和应用遥感影像的军事地理信息系统,对六个方向的应用思路和国内外...     

基于条件生成对抗模型的遥感卫星影像重建技术研究

为了快速侦察未知区域的地貌信息,遥感卫星可对特定区域进行扫描以获取遥感卫星影像。当卫星经过国外未知区域时,部分卫星无法针对某特定区域进行长时间的驻留扫描,本文提出一种基于条件生成对抗网络模型（Conditional Generative Adversarial Network,CGAN）进行网络训练,前期将某方法获取的区域轮廓地形信息作为CGAN网络的生成网络和鉴别网络中的条件约束信息,通过网络生成器与判别器在训练过程中互相博弈产生特定的输出集,有效地实现由单张电子轮廓图像到对应卫星遥感图像的端到端的非线性映射。本文通过原真实卫星遥感图像与生成卫星遥感图像进行四种对比误差计算,平均误差、均方误差与结构相似度均高于99%,峰值信噪比高于30 dB,生成的图像与原图像之间具备高相似度,实现了在获取坐标定位轮廓信息的先验条件下,对特定区域进行遥感卫星影像内容重建技术。     

高分辨率卫星遥感图像的偏流角及其补偿研究

卫星遥感采用高分辨率TDICCD相机时，由于地球的旋转产生图像运动的偏流角，偏流角对这种遥感图像的质量影响较大，需要进行补偿，并通过遥感视线的校正实现。补偿的首要工作是确定图像运动的偏流角和遥感视线的修正量。基于航天器飞行动力学原理，本文求出了各种近地轨道的偏流角、偏流角角速度，把遥感视线的校正分成粗校正和精校正，并给出了修正量的确定方法。针对椭圆轨道和圆轨道情况的偏流角及修正量的计算，验证了本文算法的有效性。     

Cost-driven design of small satellite remote sensing systems*

Earth remote sensing (alongside communications) is one of the key application of Earth-orbiting satellites. Civilian satellites in the LANDSAT and SPOT series provide Earth images which have been used for a vast spectrum of applications in agriculture, meteorology, hydrology, urban planning and geology, to name but a few. In the defence sector, satellite remote sensing systems are a critical tool in strategic and tactical planning – for the countries which can afford them. To date, remote sensing satellites have fallen into one of these two categories: military missions driven by the requirement for very high resolution and orbital agility; and multipurpose civil satellites using general purpose sensors to serve a diverse community of end users. For military-style missions, the drive to high resolution sets the requirements for optics, attitude control and downlink data bandwidth. For civil missions, the requirement to satisfy multiple, diverse user applications forces compromises on spectral band and orbit selection. Although there are exceptions, many small satellite remote sensing missions carry on in this tradition, concentrating on ultra high resolution products for multiple user communities. This results in satellites costing on the order of US $100 M, not optimised for any particular application. This paper explores an alternative path to satellite remote sensing, aiming simultaneously to reduce cost and to optimise imaging products for specific applications. By decreasing the cost of the remote sensing satellite system to a critical point, it becomes appropriate to optimise the sensor's spectral and temporal characteristics to fit the requirements of a small, specialised user base. The critical engineering trade-off faced in a cost driven mission is how to reduce mission cost while still delivering a useful product to the selected user. At the Surrey Space Centre, we have pursued an engineering path using two dimensional CCD array sensors, commercial off-the-shelf lenses and gravity-gradient stabilised microsatellites. In spite of the inherent limitations of such systems, recent successes with the Thai Microsatellite Company's Thai-Phutt satellite show that a system costing in the region of US $3 million, can approach the spectral and spatial characteristics of LANDSAT. Surrey's UoSAT-12 minisatellite (to be launched April, 1999) will further develop this cost-driven approach to provide 10 m panchromatic resolution and 30 m multi-spectral resolution. This paper describes the Thai-Phutt and UoSAT-12 imaging systems, explaining the engineering methods and trade-offs. Although Surrey is presently the only centre presently pursuing such implementations, our paper shows that they deserve wider consideration.     

Remote sensing of natural resources : Legal and political considerations for the Third World

A state's posture on remote sensing of the Earth by orbiting satellites varies depending upon whether it is a ‘sensing’ or a ‘sensed’ state, upon its present economic status, and upon its economic and political history. This article considers the international legal questions related to remote sensing and discusses the political aspects with special emphasis upon the views of the Third World. The author concludes that, unless Third World states and others who support them alter their views somewhat, it is probable that remote sensing of natural resources will continue without specific legal guidelines.