项目工程风险评估云判别模型设计

针对工程项目的特点和项目风险定量评估任务,提出了基于云判别模型的风险评估方法.云理论是一种用于处理不确定性和含糊性知识的数学工具,可实现概念与数据之间的相互转换.在云理论的基础上,建立了综合能力模型以及风险评估云模型,提出了基于云理论项目工程风险评估的方法,解决了风险评估中定性语言到定量的转化问题.该方法通过分析项目生命周期中的具体风险与风险因素之间的因果关系,建立项目工程风险评价模型,拓展了在国内经济转轨时期项目工程风险模型的研究思路,并结合应用实例评价该方法对建立工程内部风险评价体系的应用价值.      

基于时空关联图模型的视频监控目标跟踪

多摄像机监控环境下的无重叠视域目标跟踪问题十分具有挑战性,其原因在于跟踪目标在网络中的转移与运动规律往往具有不确定性.目标跟踪的关键问题在于摄像机之间的目标关联以及如何依据网络拓扑结构来找到目标之间的对应关系.提出了一种图模型来对摄像机网络中的时空关联关系进行表达.图模型中的节点表示目标在摄像机视域中的出现区域和消失区域,边由时间与空间关系进行约束.提出了一种将目标外观模型与图模型相融合的跟踪方法,其中外观模型通过协方差描述子进行特征融合,同时,结合二部图匹配策略来解决多摄像头目标跟踪中的识别与匹配问题.在真实监控视频上的实验验证了该方法的有效性.      

网格计算环境中的安全信任协商系统

网格技术促进了广域网络环境下的资源共享和协同工作.然而,在网格环境中,服务的提供方和请求方往往位于不同的安全域,如何为位于不同安全区域的陌生主体间动态地建立信任关系成为一项应用难题.基于信任管理基础设施,通过协作主体间信任证、访问控制策略的交互披露逐渐为各方建立信任关系,设计并实现了一个保护协商方隐私信息的安全信任协商系统.该系统采用一种基于属性的委托授权信任证、访问控制策略及一致性验证算法,实现策略的动态调整,从而生成满足访问控制策略的所有最小可满足信任证集,能够使理论上存在的信任关系得以自动建立.通过在CROWN平台中的应用,显示该系统能够满足网格应用中信任自动建立和敏感信息保护的需求.      

基于CIDBN的战术活动识别模型及在线精确推理

战术活动识别是战场态势感知的重要研究内容。为提高战术活动识别的准确性与实时性,提出了一种基于上下文独立动态贝叶斯网络（CIDBN）的战术活动识别模型及在线精确推理。通过对战术活动机制的分析,采用动态贝叶斯网络（DBN）理论,建立了一个初始战术活动识别模型。该模型引入了威胁指数节点来影响战术活动的终止与选择,并采用模糊隶属度函数对连续变量进行离散化处理。依据上下文独立关系对该模型进行简化,获得了一个基于CIDBN的战术活动识别模型。将接口算法扩展于该模型上,提出了在线精确推理算法。仿真结果表明,所提出的战术活动识别方法,具有识别精度高、较低不确定性和实时性高的优点。      

面向工作流系统的生态组织模型研究

为了解决传统组织模型结构柔性差和缺乏主动变化的问题,总结了目前研究工作的缺陷,提出生态组织模型.描述了生态组织模型的基本要素,体现了对环境关联因素和实体变化特征的描述,建立了组织元模型和形式化描述.提出在工作流系统生态环境下描述和预测组织对象关联程度和变化倾向的研究方法论.引入序方法做为组织势态分析的数学工具,并且设计了一组预示实体变化倾向的势态参数.上述研究内容已经在开放边界工作流系统ABCflow中得到应用和验证.      

基于云模型的目标威胁等级评估

目标威胁等级评估的实时性、正确性对于后续作战仿真具有重要意义.提出了基于云模型的威胁等级评估的方法,通过对依据专家经验得到的对定性概念评估的隶属云图的贝叶斯修正,得到了标尺云.由战场态势信息对标尺云的匹配和处理,最终得到了目标威胁评估云图,实现了威胁评估.定性概念对战场态势的描述以及隶属云图体现了模糊性与随机性的较好结合,反映了空战时飞行员获取信息的不确定性的特点.通过实例验证表明:基于云模型的威胁等级评估的方法能有效地实现对目标威胁等级的确定.     

基于多信号流图的诊断贝叶斯网络建模

多信号流图与贝叶斯网络是两种不同的建模方法,对两种建模方法的异同分析表明多信号流图到贝叶斯网络的转化是可行的.为了降低诊断贝叶斯网络模型的开发费用,提出了一种基于多信号流图的诊断贝叶斯网络建模方法,从模型结构和模型参数两个方面论述了多信号流图向诊断贝叶斯网络的转化方法.依据该方法生成的诊断贝叶斯网络模型具有应对复杂系统故障诊断的证据冲突及不确定性问题的优点.实验证明所获得的诊断贝叶斯网络模型的诊断准确性与多信号流图模型一致.     

基于深度强化学习与扩展卡尔曼滤波相结合的交通信号灯配时方法

深度Q学习网络(DQN)因具有强大的感知能力和决策能力而成为解决交通信号灯配时问题的有效方法，然而外部环境扰动和内部参数波动等原因导致的参数不确定性问题限制了其在交通信号灯配时系统领域的进一步发展。基于此，提出了一种DQN与扩展卡尔曼滤波(EKF)相结合(DQN-EKF)的交通信号灯配时方法。以估计网络的不确定性参数值作为状态变量，包含不确定性参数的目标网络值作为观测变量，结合过程噪声、包含不确定性参数的估计网络值和系统观测噪声构造EKF系统方程，通过EKF的迭代更新求解，得到DQN模型中的最优真实参数估计值，解决DQN模型中的参数不确定性问题。实验结果表明:DQN-EKF配时方法适用于不同的交通环境，并能够有效提高车辆的通行效率。      

基于特征建模的空间非合作目标姿态智能测量方法

为实现空间非合作目标姿态的测量,基于特征建模的思想,提出了一种根据目标激光点云数据进行高效处理的智能测姿态方法.首先,针对空间目标姿态测量的需求,寻找并实现能够高效表征目标姿态的点云数据特征.接着,应用神经网络的方法对目标点云姿态特征进行学习,通过建立合理的神经网络模型和训练数据,实现目标点云特征与姿态间非线性映射关系的建立.最后,利用目标点云仿真数据集,对方法的测量精度和测量实时性进行了评估.实验结果表明：利用特征建模的思想,提出并建立目标点云姿态协方差矩阵特征,实现了点云数据在表征目标姿态方面的信息高度压缩,为神经网络模型的轻量化设计和计算资源严重受限的在轨应用,提供了可行的智能化工程实现方案.     

不确定性下的低轨卫星激光星间链路误差模型修正方法

在工程应用中,激光链路构建过程不可避免地会受到各种不确定性的影响,导致激光建链在扫描过程中对指向区域估计不准,降低星间激光建链的准确性.因此,应多方面考虑不确定因素来修正低轨激光星间链路的动力学指向误差模型,为后期的扫描捕获提供基础.提出一种考虑不确定性因素的低轨卫星激光星间链路动力学指向误差模型修正方法,激光星间链路系统以及不确定参数进行描述,建立系统动力学模型;对不确定参数进行了灵敏度分析,并基于灵敏度指标,使用Kriging方法建立了原有系统动力学的代理模型,并验证模型预测精度;确定不确定性量化指标,使用贝叶斯随机模型修正技术对系统动力学模型不确定参数进行修正;通过数值仿真验证了不确定性下的星间链路指向误差动力学模型修正方法的有效性和可行性.     

International comparison of satellite winds — An update

Winds obtained from geostationary satellites are compared with each other and with rawinsondes. These comparisons serve as a periodic quality check of satellite cloud motions (or winds) set up by the CGMS (Coordination for Geostationary Meteorological Satellites). Differences are taken between colocated cloud motions observed by adjacent satellites in areas of overlapping coverage (Type 1) and between colocated rawinsondes and cloud motions within the field of view of each individual satellite (Type 2).Among colocated satellite winds (Type 1) RMS vector difference of high clouds rarely exceed 10 mps and of low clouds, 6 mps. But, among colocated cloud and balloon vectors (Type 2), RMS vector differences are large. At high levels, differences range from 12 to 40 mps for GMS (Geostationary Meteorological Satellite) winds and from 10 to 18 mps for GOES (Geostationary Operational Environmental Satellite) winds. The greater disagreement of satellite winds with rawinsonde winds than with each other is attributed in large part to error in the assignment of cloud height especially in the presence of strong vertical shear and to a lesser extent on time differences between cloud and balloon measurements. Both Type 1 and 2 comparisons suffer from separations in distance (tolerated for purposes of establishing “colocation”) between cloud and balloon in the presence of strong horizontal shear. The discrepancy existing between GMS and GOES differences with rawinsondes is attributed primarily to differing techniques of height assignment.At low levels satellite winds departed from balloon winds by a RMS vector difference of about 6 to 9 mps which approached or exceeded the mean wind speed itself. This problem is attributed chiefly to the uncertainty of wind levels controlling the motion of the various low cloud types.     

The absorption of solar energy and the heating rate in the atmosphere of Venus

The Solar Flux Radiometer (LSFR) experiment on the large probe of the Pioneer Venus (PV) mission made detailed measurements of the vertical profile of the upward and downward broadband flux of sunlight at a solar zenith angle of 65.7°. These data have been combined with cloud particle size distribution measurements on the PV mission to produce a forward-scattering model of the Venus clouds. The distribution of clouds at high altitudes is constrained by measurements from the PV orbiter. Below the clouds the visible spectrum and flux levels are consistent with Venera measurements at other solar zenith angles. The variations in the optical parameters with height and with wavelength are summarized in several figures. The model is used to evaluate the solar heating rate at cloud levels as a function of altitude, solar longitude, and latitude for use in dynamical studies.     

Monitoring spatio-temporal variations in aerosols and aerosol–cloud interactions over Pakistan using MODIS data

Clouds are important elements in climatic processes and interactions between aerosols and clouds are therefore a hot topic for scientific research. Aerosols show both spatial and temporal variations, which can lead to variations in the microphysics of clouds. In this research, we have examined the spatial and temporal variations in aerosol particles over Pakistan and the impact of these variations on various optical properties of clouds, using Moderate Resolution Imaging Spectroradiometer (MODIS) data from the Terra satellite. We used the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model for trajectory analysis to reveal the origins of air masses, with the aim of understanding these spatial and temporal variabilities in aerosol concentrations. We also documented seasonal variations in patterns of aerosol optical depth (AOD) over Pakistan, for which the highest values occur during the monsoon season (June–August). We then analyzed the relationships between AOD and four other cloud parameters, namely water vapour (WV), cloud fraction (CF), cloud top temperature (CTT) and cloud top pressure (CTP). Regional correlation maps and time series plots for aerosol (AOD) and cloud parameters were produced to provide a better understanding of aerosol–cloud interaction. The analyses showed strong positive correlations between AOD and WV for all of the eight cities investigated. The correlation between AOD and CF was positive for those cities where the air masses were predominantly humid, but negative for those cities where the air masses were relatively dry and carried a low aerosol abundance. These correlations were clearly dependent on the meteorological conditions for all of the eight cities investigated. Because of the observed AOD–CF relationship, the co-variation of AOD with CTP and CTT may be attributable to large-scale meteorological variations: AOD showed a positive correlation with CTP and CTT in northern regions of Pakistan and a negative correlation in southern regions.     

不确定理论下竞争失效系统的可靠性分析

针对有新研发产品,故障数据较少的复杂系统,提出了不确定环境下自然退化和外部冲击相互独立的竞争失效模型。考虑了系统同时遭受自然退化和外部冲击,连续的自然退化用一个不确定过程刻画,冲击到达的时间间隔和每次冲击对系统造成的损坏量分别用2个不同的不确定变量来刻画。运用不确定理论,分别在极端冲击模型、累积冲击模型、δ冲击模型下,研究了系统的确信可靠度,结果表明:在有新研发的产品、故障数据较少的复杂系统,用不确定理论的方法来描述模型更合适,并通过数值分析显示了模型的有效性。      

Cloud statistics using VAS

Statistics of cloud characteristics over North America during winter of 1985–86 have been calculated. The frequency of cloud cover with associated heights and infrared attenuation were charted using the CO₂ channel radiometric data from the geostationary VISSR Atmospheric Sounder (VAS). Cloud top pressures were determined from the ratio of VAS CO₂ channel radiances in a radiative transfer equation formulation. Cloud emissivities were then calculated from infrared window channel observations. CO₂ technique derived height and emissivity assignments have been found to be reliable in all cloud types, including thin cirrus clouds where other techniques have been inconsistent. Observations during November 1985 revealed that 25% of the United States was covered with thin clouds (radiative attenuation was less than 80%), 50% was covered with thick opaque clouds, and 25% had clear sky conditions. Geographical distribution of cloud cover over the United States for the winter of 1985–86 shows a strong latitudinal dependence. Cirrus reports in frequencies of 15% in the south increase to 30% in the northwest.     

基于优化信任评估的Ad Hoc安全路由协议

针对Ad Hoc网络中自私节点造成的攻击问题,提出了一种基于信任评估的Ad Hoc安全路由协议TEAR。该协议在节点行为信任评估方案的基础上,通过使用链路信任值和路径信任值表示节点和路径的可信度,同时,在链路信任值计算时,引入了基于公共邻居节点的间接信任评估方法和可变时间窗机制,保证了信任值计算的准确性和实效性;协议中每个节点维护一张信任表,并重新定义了路由请求和路由回复消息格式,实现了安全路径的快速选择。使用NS2进行模拟验证,又与AODV以及基于固定时间窗信任评估的安全路由协议TA-AODV进行性能对比。结果表明,改进的信任评估方案具有更好的性能,采用该方案的安全路由协议能较快地发现自私节点,并准确选择安全路径,提高了Ad Hoc网络性能。      

基于免疫抗体编码机理的卫星图像云探测

针对卫星遥感图像中云与下垫面的复杂性和多样性,提出了一种新的在复杂背景下常用目标图像描述方法.通过类比生物免疫抗体特异性与其构成单元氨基酸性质的相关性,得到目标图像的免疫基元集合形式及其分类方法.借鉴生物免疫抗体编码顺序中氨基酸结合能量最小原则,统计分析出训练样本的免疫基元亲和度计算公式,实现了目标图像描述的有限多特征优化组合.对于云检测问题,最终提出云的免疫抗体编码方法,定义了云的免疫亲和度计算公式,并成功构建了云免疫抗体.经200幅IKONOS卫星图像测试,验证了该方法在识别率和运行时间效率方面的有效性.