基于非参估计的局部滑窗双参CFAR目标检测

针对特定杂波概率模型不能有效的描述SAR图像背景杂波这一问题,提出了一种基于非参估计的局部滑窗双参CFAR目标检测算法。该方法首先用非参估计方法逼近SAR图像局部背景,完成对局部背景的精确建模;在此基础上,理论推导了局部双参CFAR检测算法的阈值,设计了阈值求解的数值算法。对典型目标图像进行实验,结果表明,该方法检测速度较快、精度较高。     

基于深度学习的防空武器红外目标识别技术应用研究

针对防空武器的作战使用需求,提出了一种基于深度学习的防空武器红外目标识别流程,通过大视场进行目标检测,小视场进行目标跟踪识别.在目标检测阶段,采用YOLO网络模型实现全图多目标识别定位;在目标跟踪阶段,采用超分辨率重建算法提升目标局部图像分辨率,利用深度残差网络模型实现跟踪目标的识别分类.试验结果表明,基于深度学习的空...     

红外目标分割方法研究

红外目标分割算法对红外目标检测、跟踪具有非常重要的价值。本文利用背景和目标灰度特征,提出一种实现红外目标有效分割的方法,克服红外目标内部温度不稳定造成的误分割问题。本文方法首先采用基于灰度-显著度最大相关准则的二维直方图分割算法进行图像分割;然后,在分割后二值图上进行基于随机种子点选取的区域增长,提取背景;最后,采用形态学方法优化分割结果。相对传统的红外目标检测算法,这种算法具有更好的抗干扰能力,更强的鲁棒性。不仅可以应用于红外图像的目标分割,而且可以应用于其他类似的目标分割问题。     

智能监视系统中红外运动目标检测

本文使用一种基于Cauchy分布统计背景模型的背景剔除（background subtraction）技术成功地实现了对低分辨率红外视频目标的检测。提出以柯西分布作为斧景图像像素强度比值的统计分布模型,并融合图像局部线性相关技术实现背景剔除。该方法同时利用了单个像素点的时序统计特性和邻近像素点所蕴涵的空间区域特性,实现了精确的目标检测。最后的实验表明,所提供的算法可以对室外场景中的红外视频目标实现精确地检测,可以有效地抑制背景中杂波的干扰。     

红外空中小目标小波域检测方法

针对空中远距离红外小目标检测的实际问题,提出了一种基于小波变换的检测算法.该方法首先对小目标图像进行小波分解并在考虑高频系数能量的基础上对噪声和背景边缘系数进行抑制,然后将遗留下来的高频系数通过线性映射变换成灰度图像.其次对3个方向的高频图像按照一维最大熵法进行二值化处理并通过形态开算子进一步滤除噪声,随后将高频图像两两相与关联生成单帧检测结果,并进一步利用帧间目标位置的相关性完成小目标检测过程.最后,在原图像中以检测结果图像的质心为中心生成跟踪窗口.试验结果表明,相对于通常的小目标检测算法,提出的算法在背景抑制、检测准确度以及速度方面都具有一定的优势.     

基于卷积神经网络的SAR图像目标检测综述

针对合成孔径雷达(SAR)图像传统检测算法精度低、鲁棒性差等特点,对基于卷积神经网络的SAR图像目标检测算法进行归纳总结.归纳了基于CNN的目标检测框架的发展现状,并综述了其在SAR图像目标检测中的应用,分析了其应用难点与关键技术.最后结合深度学习方法在目标检测领域中的发展,对SAR图像目标检测算法的发展进行了展望.     

基于多尺度特征聚类算法的不确定目标检测

提出一种基于无样本的SAR图像目标检测分类方法。针对空战过程中难以获得大量SAR图像目标样本问题,采用基于全局CFAR的多尺度SIFT特征进行目标纹理描述,并针对特征维度较高及特征描述之间存在的冗余问题,通过PCA算法对其进行降维处理后,采用DBSCAN算法对潜在目标区域的多尺度SIFT特征进行分类实现目标检测。通过单一目标和多类目标图像进行实验验证,实验结果表明该方法具有一定的有效性和可行性。     

基于深度学习的SAR图像目标检测实验

针对目前SAR图像目标检测算法只能进行单一目标检测和检测精度不高的问题,对深度学习目标检测框架在SAR图像目标检测的应用进行了实验研究,并结合SAR图像特点进行了优化。比较了基于区域建议的Faster-RCNN和无需区域建议的SSD目标检测框架在SAR图像上的目标检测精度和速度,分析优缺点;研究了预训练模型对SAR图像目标检测精度的影响;最后通过零均值规整化提高收敛速度和检测精度。实验结果表明优化后的目标检测框架,实现了SAR图像多目标识别并提高了检测精度,可以有效地应用于SAR图像多目标检测。     

基于Hausdorff距离的目标跟踪方法研究

序列图像中的运动目标跟踪是计算机视觉的一个重要组成部分,跟踪算法的鲁棒性和计算量是算法的关键。对上述问题进行研究,提出了一种基于Hausdorff距离的目标跟踪方法。该算法结合运动检测和多分辨率技术,极大减少了计算量,并利用有效的模板更新方法,加强了跟踪的鲁棒性。实验表明,该算法能实现快速有效的目标跟踪。     

基于位置预测的靶场图像实时判读方法

在靶场经纬仪对目标实时跟踪测量时,会发生相机随机抖动的情况,引起目标在图像中大幅度运动。应对大幅度运动时,基于搜索窗口的跟踪方法容易丢失目标,而基于全图搜索的跟踪方法时效性差。针对以上问题,提出一种结合核相关滤波算法(Kernelized Correlation Filter, KCF)和目标位置预测的改进的跟踪学习检测算法(Tracking-Learning-Detection, TLD)跟踪框架。利用正交多项式最优线性滤波器及相机角度信息预测目标下一帧位置,在此区域利用KCF进行快速跟踪,可以提高跟踪的成功率和时效性,跟踪失败时再进行检测。仿真实验表明,最优线性滤波器能较准确预测目标位置,给KCF提供较准确的搜索位置,算法每帧耗时仅为1.1 ms,且定位精度优于TLD和KCF,能有效应对相机抖动的问题。靶场实际试验证明该方法可提高靶场自动判读水平,减少人工干预。     

Reviews

Douglas S. Derrer, We Are All the Target: A Handbook of Terrorism Avoidance and Hostage Survival. Annapolis, MD: US Naval Institute Press, 1992. Pp. x+135, notes, index. $14.95. ISBN 01–55750–150–5

Ian Knight, Zulu: Isandlwana and Rorke's Drift, 22–23 January 1879. London: Windrow & Greene, 1992. Pp.136, 150 illus., incl 8 colour plates, maps, biblio. £35. ISBN 1–872004–23–7

Ian Knight, By the Orders of the Great White Queen: Campaigning in Zululand through the Eyes of the British Soldiers, 1879. London: Greenhill Books and Novato: Presidio Press, 1992. Pp. 272, 17 illus., 1 map. £18.95. ISBN 1–85367–122–3

Manfried Rauchensteiner and Erwin A. Schmidl (eds.), Formen des Krieges: vom Mittelalter zum ‘Low‐intensity’ Conflict’. Graz : Verlag Styria, 1991. Pp.208. DM35. ISBN 3–22–12139–7

Harold J. Kearsley, Maritime Power and the Twenty‐First Century. Dartmouth: Dartmouth Publishing Company, Limited, 1992. Pp.xv + 203, 13 diagrams, index. £32.50. ISBN 1–85521–288–9     

Panjwai: a tale of two COINs in Afghanistan

The US Army has two approaches to counterinsurgency operations in Afghanistan. One is hard, or combat-focused, and the other is soft, or development-focused. This study examines two US Army task forces deployed to Panjwai District, Afghanistan from 2012 to 2013. CTF 4-9 and 1-38 offer a meaningful comparison because they pursued these contrasting approaches among the same population and against the same enemy at the same time and place. The study compares each unit’s approach and finds that neither approach was successful absent the other. The article concludes by recommending further research into combining the approaches at the operational level.     

线性核时频信号表示及其仿真

根据ｃｈｉｒｐ信号在模糊域的特点,本文提出了线性核时频表示方法。这种方法在提高分辨率、消除交叉项以及抑制噪声等方面都具有较高的性能。理论分析和实验结果都证实了这种方法的有效性     

Forecasting a Hurricane: Israeli and American Estimations of the Khomeini Revolution

The surprising ‘Arab Spring’ raises the question as to what would enable national intelligence to provide high quality warnings prior to the eruption of popular revolutions. This article uses new sources of evidence to trace and explain Israel's success in comparison to US failure at correctly estimating the course of the Iranian Revolution in 1977–79. In explaining this variance, the article shows that it was mainly the result of the intimate acquaintance of Israel's representatives in Iran with the local language, history and culture, as well as the ability to communicate with locals – tools which the Americans completely lacked.     

关于教师继续教育培训模式的思考

继续教育作为终身教育的重要组成部分和基本形式,对教师的发展成长意义重大。继续教育工作开展十多年来,取得了一定的成绩,但还存在不少问题,如学习动机较为被动,管理认识不到位,教学内容陈旧空泛,教学方法呆板机械等。随着课程改革的不断深入,继续教育必须采取相应的改进措施,树立以人为本,服务至上的观点,教学内容的前瞻性与实用性并重,教学人员进行整合,教学方法灵活多样,评价考核规范科学等,以此激活受训教师内在的学习动机,更大程度地提高继续教育培训的实际效果。     

用DirectShow实现网络视频

目前市场上的网络监控技术一般用昂贵的硬件设备来支撑,投资较大。在深入研究微软的DirectX9.0的基础上,提出如何用DirectShow软件技术和普通的数字摄像头来实现网络视频的监控和实现技术,内容涉及通信协议、视频的编码与解码技术、视频的编码格式和COM组件编程技术等,最终的实验结果令人满意,且费用低廉,技术可靠,为在高速局域网内构建视频监控系统提供有效的技术支持。     

Interpreting the Ribchester Helmet

In 1796 a 13-year-old boy playing on wasteland behind his father’s house in Ribchester, Lancashire discovered an assemblage of over 30 Roman artefacts, including a decorated brass helmet. Although partially corroded the Ribchester helmet is a fine example of a first-century AD cavalry sports helmet. This article considers how the interpretation of the helmet has evolved since its discovery. The multi-faceted iconography of the helmet with its mural crown diadem is that of an elite soldier and is paralleled in other examples of cavalry sports helmets of the Ribchester type. The long-neglected imagery on the helmet bowl is reinterpreted and seen not as a haphazard collection of randomly placed combatants but as a highly structured, symmetrical composition specifically designed to be viewed from different directions.     

Putin – the masked nemesis of the strategy of ambiguity

Recent events demonstrate the complex and adaptive approach employed by Russia to reassert influence in Europe. The changing face of Russia’s strategy commenced in 2007 when it launched a crippling cyber-attack against Estonia. This was followed by a large Russian conventional attack against Georgia in 2008, occupying two large areas of the nation. 2014 witnessed the Russian annexation of Crimea where in just a week, Russia seized control of Crimea “without firing a shot.” The annexation of Crimea was rapidly followed by a Russian inspired and led subversive war in eastern Ukraine. The common thread among these diverse Russian operations is its use of ambiguity to confound and confuse decision makers in the West.     

Desert insurgency: lessons from the third Tuareg rebellion

This article provides an in-depth examination and analysis of the 2006–2009 Tuareg rebellion in Mali and Niger. It identifies the underlying reasons behind the rebellion, explores contrasting counter-insurgency (COIN) strategies employed by the two governments, and presents some lessons learned. While both COIN approaches ultimately produced similar peace settlements, the article argues that the Malian strategy of reconciliation combined with the selective use of force was far more effective than the Nigerien iron fist approach at limiting the size and scope of the insurgency and producing a more sustainable peace. It concludes by looking at the role of external actors, particularly the United States, and how the failure to internationalize the conflict was actually more beneficial to the local COIN effort, as well as to the longer strategic interests of the United States in the region.