基于局部色调映射的异源图像匹配方法

色调映射是一种适用于亮度和对比度变化的快速图像匹配方法。由于异源图像间存在复杂的灰度变换关系,直接采用色调映射方法进行匹配的成功率通常难以满足应用需求。为了提高匹配成功率,本文提出基于局部色调映射的异源图像匹配方法。首先将实时图划分为不重叠的子区,对每个子区进行直方图均衡化和弱切片变换,通过局部色调映射计算子区与基准图的距离系数图。融合全部距离系数图得到图像匹配结果。实验结果表明,该方法匹配成功率高于现有的色调映射方法,且计算时间仅略高于现有方法,优于异源图像匹配中常用的互信息方法。     

舰用计算机局部网络通信协议的设计

介绍了舰用计算机局部网络通信协议的设计,重点描述了设计中为提高实时性和可靠性所采用的一些具体技术措施,其中包括:简化ISO/OSI—RM 参考模型,并与IEEE8024标准协议兼容,采用VLSI 技术设计网络接口卡。同时,对其中一些关键技术,如逻辑环的建立与维护、信息的优先级发送与实时响应、数据缓冲区的结构与管理都作了说明。     

The effects of local elections on national military spending: A cross-country study

In this paper, we study the domestic political determinants of military spending. Our conceptual framework suggests that power distribution over local and central governments influences the government provision of national public goods, in our context, military expenditure. Drawing on a large cross-country panel, we demonstrate that having local elections will decrease a country’s military expenditure markedly, controlling for other political and economic variables. According to our preferred estimates, a country’s military expenditure is on average 20% lower if its state government officials are locally elected, which is consistent with our theoretical prediction.     

Close Encounters of the Tribal Kind: the Implementation of Co-option as a Tool for De-escalation of Conflict – The Case of the Netherlands in Afghanistan's Uruzgan Province

Abstract

In contemporary counter-insurgency campaigns de-escalation is typically achieved by obtaining the local population's collaboration. The rationale underlying this approach is that such collaboration spawns control over the population and therefore weakens the insurgent's position. This article seeks to understand how this approach is practised in the reality of contemporary counter-insurgency warfare. In Iraq as well as Afghanistan counter-insurgents were confronted with fragmented indigenous societies in which the various local segments were dominated by local power-holders. Consequently, obtaining the local population's collaboration required co-option of these agents. As such a strategy of co-option is a highly localized approach, this article focuses on Afghanistan's Uruzgan province and analyzes the implementation of co-option by the Netherlands' Task Force Uruzgan (TFU), which operated there from 2006 to 2010. The analysis discusses how the counter-insurgents obtained a profound understanding of Uruzgan's societal landscape and how co-option of local power-holders became part of the counter-insurgency routine. In the end, co-option was successfully implemented as TFU even succeeded in obtaining the collaboration of previously marginalized elements of the population. However, as the methods employed for co-option were mainly of a persuasive nature, the TFU co-option strategy was less effective in containing the influence of notorious spoilers. Thus this case study provides an insight into the dynamics and limitations of co-option as a tool for de-escalation in contemporary counter-insurgency warfare.     

基于改进粒子群优化RBF神经网络的算法

针对粒子群算法易陷入局部极小的缺陷,提出了一种改进的粒子群优化算法,并将改进后的算法应用到RBF神经网络核函数参数的选取中。依照文中提出的编码方式、迭代公式和适应度函数,在全局空间中搜索具有最优适应值的参数向量。实例仿真表明,基于改进粒子群算法优化的RBF神经网络不仅收敛速度快,且误差精度高。     

一种基于局部不变特征的图像特定场景检测方法

图像场景的自动检测,对于图像的标注以及语义检索具有非常重要的作用。本文研究根据实际应用的需要,围绕会晤、集会、海滩等八类特定场景图像的检测问题展开。首先对图像进行局部关键点的检测以及SIFT特征描述子的计算,从而提取图像的局部特征,在此基础上基于支撑向量机构建多分类器,进行特征训练,最终获得较为准确的检测结果。实验重点针对分类器核函数的确定以及特征选取策略等问题展开,实验结果表明,采用径向基核函数构建多分类器以及特征点按尺度大小排序取前n位的选取策略可以获得较为准确和鲁棒的特定场景检测结果。本方法在保证满足一定程度场景检测准确率的前提下,具有简单快速的特点,能够满足实际应用的需要。