多源异构土地基础数据一体化管理检索方法研究

为了从多源异构的复杂土地基础数据中快速准确地提取用户所需信息,提出了基于元数据的一体化管理检索方法.在元数据信息提取、元数据加权索引、实体同义词扩展检索3个环节中,结合土地领域专业知识和用户实际需求,设计和开发了共享元数据表结构、加权元数据中字段相对重要性和信息熵因子,构建地名实体和专题数据层实体同义词库,并集成到包括中文分词、实体识别、同义词扩展、索引检索和相似度计算的一体化管理检索框架中,解决了多源异构土地基础数据统一管理和精确检索的问题.实践表明,该方法较传统的通用信息检索方法具有更好的适用性和更高的准确率.     

线性常微分方程参数稳定性的统计推断

对具有随机误差的观测数据, 讨论了常系数线性常微分方程参数稳定性的统计推断问题. 通过残差项的Karhunen-Loeve 分解, 给出了变点检验步骤及其在原假设下的极限分布. 在对立假设下定义了变点的估计, 证明了检验以及估计的一致性. 对常系数二阶常微分方程进行了统计模拟, 结果表明原假设下的极限分布是对真实分布非常好的近似; 对立假设下, 即使输入函数的频率存在0.75% 的变化, 上述检验也能以大概率拒绝原假设. 最后利用上述方法研究了英国中部地区的气温数据, 揭示了数据一些新的特点.     

Semi-continuous High Speed Gas Analysis of Generated Vapors of Chemical Warfare Agents

A method is presented for the continuous analysis of generated vapors of the nerve agents soman and sarin and the blistering agent sulfur mustard. By using a gas sampling valve and a very short (15 cm) column connected to an on-column injector with a “standard length” column, the system can either be calibrated or used for performing high speed gas analyses. When using a flame ionization detector, the detection limit was ca. 100 ppb (ca. 0.5–1.0 mg/m³). This technique is applied in inhalation toxicokinetic studies of nerve agents and mustard gas in the guinea pig.     

Origin Identification of Hungarian Honey Using Melissopalynology,Physicochemical Analysis,and Near Infrared Spectroscopy

The objective of the study was to check the authenticity of Hungarian honey using physicochemical analysis, near infrared spectroscopy, and melissopalynology. In the study, 87 samples from different botanical origins such as acacia, bastard indigo, rape, sunflower, linden, honeydew, milkweed, and sweet chestnut were collected. The samples were analyzed by physicochemical methods (pH, electrical conductivity, and moisture), melissopalynology (300 pollen grains counted), and near infrared spectroscopy (NIRS:740–1700 nm). During the evaluation of the data PCA-LDA models were built for the classification of different botanical and geographical origins, using the methods separately, and in combination (low-level data fusion). PC number optimization and external validation were applied for all the models. Botanical origin classification models were >90% and >55% accurate in the case of the pollen and NIR methods. Improved results were obtained with the combination of the physicochemical, melissopalynology, and NIRS techniques, which provided >99% and >81% accuracy for botanical and geographical origin classification models, respectively. The combination of these methods could be a promising tool for origin identification of honey.     

Nanoscale C‐Rich SixC1−x Bus/Ring Waveguide Based Cross‐Wavelength Data Converter

The carbon‐rich silicon carbide (C‐rich Si_xC_1?x) micro‐ring channel waveguide with asymmetric core aspect is demonstrated for all‐optical cross‐wavelength pulsed return‐to‐zero on‐off keying (PRZ‐OOK) data conversion. Enhanced nonlinear optical Kerr switching enables 12‐Gbit per second data processing with optimized modulation depth. The inverse tapered waveguide at end‐face further enlarges the edge‐coupling efficiency, and the asymmetric channel waveguide distinguishes the polarization modes. To prevent data shape distortion, the bus/ring gap spacing is adjusted to control the quality factor (Q‐factor) of the micro‐ring. Designing the waveguide cross section at 500 × 350 nm² provides the C‐rich Si_xC_1?x channel waveguide to induce strong transverse electric mode (TE‐mode) confinement with a large Kerr nonlinearity of 2.44 × 10^?12 cm² W^?1. Owing to the trade‐off between the Q‐factor and the on/off extinction ratio, the optimized bus/ring gap spacing of 1400 nm is selected to provide a coupling ratio at 5–6% for compromising the modulation depth and the switching throughput. Such a C‐rich Si_xC_1?x micro‐ring with asymmetric channel waveguide greatly enhances the cross‐wavelength data conversion efficiency to favor its on‐chip all‐optical data processing applications for future optoelectronic interconnect circuits.     

利用带无标签数据的双支持向量机对恒星光谱分类

恒星光谱分类是天文技术与方法领域一直关注的热点问题之一。随着观测设备持续运行和不断改进,人类获得的光谱数量与日俱增。这些海量光谱为人工处理带来了极大挑战。鉴于此,研究人员开始关注数据挖掘算法,并尝试对这些光谱进行数据挖掘。近年来,神经网络、自组织映射、关联规则等数据挖掘方法广泛应用于恒星光谱分类。在这些方法中,支持向量机(SVM)以其强大的学习能力和高效的分类性能而备受推崇。SVM的基本思想是试图在两类样本之间找到一个最优分类面将两类分开。SVM在求解时,通过将其最优化问题转化为具有(QP)形式的凸问题,进而得到全局最优解。尽管该方法在实际应用中表现优良,但为了进一步提高其分类能力,有的学者提出双支持向量机(TSVM)。该方法通过构造两个非平行的分类面将两类分开,每一类靠近某个分类面,而远离另一个分类面。TSVM的计算效率较之传统SVM提高近4倍,因此,自TSVM提出后便受到研究人员的持续关注,并出现若干改进算法。在恒星光谱分类中,一般分类算法都是根据历史观测光谱来建立分类模型,其中最关键的是对光谱进行人工标注,这项工作极为繁琐,且容易犯错。如何利用已标记的光谱以及部分无标签的光谱来建立分类模型显得尤为重要。因此,提出带无标签数据的双支持向量机(TSVMUD)用以实现对恒星光谱智能分类的目的。该方法首先将光谱分为训练数据集和测试数据集两部分;然后,在训练集上进行学习,得到分类依据;最后利用分类依据对测试集上的光谱进行验证。继承了双支持向量机的优势,更重要的是,在训练集上学习分类模型过程中,不仅考虑有标记的训练样本,也考虑部分未标记的样本。一方面提高了学习效率,另一方面得到更优的分类模型。在SDSS DR8恒星光谱数据集上的比较实验表明,与支持向量机SVM、双支持向量机TSVM以及K近邻(KNN)等传统分类方法相比,带无标签数据的双支持向量机TSVMUD具有更优的分类能力。然而,该方法亦存在一定的局限性,其中一大难题是其无法处理海量光谱数据。该工作将借鉴海量数据随机采样思想,利用大数据处理技术,来对所提方法在大数据环境下的适应性展开进一步研究。     

超导Tl2Ba2CaCu2O8+x 薄膜的太赫兹复电导率的提取

采用低温太赫兹时域光谱系统, 测试了高温超导 Tl2Ba2CaCu2O8+x 薄膜的太赫兹透射谱, 并提取了它在不同温度下的太赫兹电导率. 研究过程中发现提取后的参数存在电导率随频率波动大、与理论值偏差较大等问题.通过对基片厚度和太赫兹波入射角度的误差对高温超导薄膜电导率的影响进行了分析, 结果表明导致数据波动大是由于基片厚度的偏差引起的. 针对厚度差的影响, 一种矫正方法被提出, 通过对厚度的修正, 提高了数据提取的质量.     

一种基于卷积神经网络的恒星光谱快速分类法

恒星光谱数据的分类是天体光谱自动识别的最基本任务之一,光谱分类的研究能够为恒星的演化提供线索。随着科技的发展,天文数据也向大数据时代迈进,需要处理的恒星光谱数量越来越多,如何对其进行自动而精准地分类成为了天文学家要解决的难题之一。当前恒星光谱自动分类问题的解决方法相对较少,为此本文使用了一种基于卷积神经网络的方法对恒星光谱MK系统进行分类。该网络由数据输入层、四个卷积层、四个池化层、全连接层、输出层构成,与传统网络相比具有局部感知、参数共享等优点实验。在Python3.5的环境下编程,利用Tensorflow构建了一个简单高效的具有四个卷积层的卷积神经网络,并将Dropout作用于全连接层之后以防止过度拟合。Dropout的基本思想：当网络模型进行训练时,把一些神经网络节点按一定的比例丢弃,使其暂时不发挥作用。Dropout可以理解成是一种十分高效的神经网络模型平均方法,由于它不依赖于某些局部特征所以能够让网络模型更加鲁棒。实验中使用的一维恒星光谱图是取自LAMOST DR3数据库,首先进行预处理截取光谱3 600~7 300 Å的部分,均匀采样后使用min-max标准化法对其进行初始化。实验包括两部分：第一部分为依据恒星光谱MK系统对光谱进行分类,每一类的训练样本包含1 000条光谱数据,测试样本为400条光谱数据,首先通过训练样本对CNN网络进行训练,进行3 000次的迭代,用训练后的网络将测试样本进行分类以验证网络的准确性;第二部分为相邻两类的恒星光谱的分类,其中O型星数据集样本为250条光谱,其余类别恒星样本数据集均为4 000条光谱,将数据5等分,每次选取当中的一份当作测试集,其余部分当作训练集,采用5折交叉验证法求得模型准确率,用BP神经网络进行对比实验。选择对网络模型进行评估的指标包括精确率P、召回率R、F-score、准确率A。实验结果显示CNN在对六类恒星光谱进行分类时其准确率都在95%以上,在对相邻类别的恒星进行分类时,由于O型星样本量较少,所以得到的分类结果不太理想,对其余类别的恒星分类准确率都高于98%,以上结果都证明了CNN算法能够很好地解决恒星光谱的分类问题。     

Bayesian fusion of GNSS,ITS-G5 and IR–UWB data for robust cooperative vehicular localization

In the automotive domain, Cooperative Localization (CLoc) is a new promising paradigm that aims at outperforming conventional Global Navigation Satellite Systems (GNSS) in terms of positioning accuracy, robustness, and service continuity, by relying on Vehicle-to-Vehicle (V2V) communications and hybrid data fusion. However, the growing number and the variety of the sensors aboard vehicles raise unprecedented challenges, especially in the context of distributed fusion approaches. This paper thus compares parametric and nonparametric Bayesian data fusion engines (e.g., based on cooperative variants of the Extended Kalman Filter (EKF) and Particle Filter (PF), respectively), while validating a CLoc scheme suitable to Vehicular Ad Hoc Networks (VANETs). More particularly, absolute position information from both onboard GNSS receiver and ITS-G5 V2V messages, as well as relative distance measurements based on the Impulse Radio–Ultra-Wideband (IR–UWB) technology, are combined into a single location solution that is hopefully more robust and more accurate than that of standalone GNSS. First, we investigate V2V ranging accuracy on a highway under real mobility conditions. In the same environment, we then provide offline validations of CLoc positioning, confirming significant performance gains through cooperation over conventional GNSS, even in case of poor initialization. In this specific context, the PF solution is thus shown to yield even better accuracy in comparison with EKF, thanks to its fine robustness against faced non-linear dynamics and non-Gaussian noise processes. Finally, we illustrate the resilience of the proposed solution under temporary GNSS denial.     

基于深圳市出租车轨迹数据的高效益寻客策略研究

出租车作为城市公共交通的重要补充形式,在居民日常出行中发挥着重要作用.然而,客源时空分布不均衡导致了出租车空载率高、寻客难度大等问题,降低了出租车的运营效率.随着ITS（intelligent transport system）技术的发展,越来越多的学者将出租车寻客模式作为其研究点,目前尚缺乏对基于寻客效益指标的寻客策略方法的研究.为了提高出租车寻客效益,提出了基于寻客效益指标模型的时空分析方法,将出租车载客态单位时间收入E及相邻空载态寻客时间T相结合作为寻客效益评价指标,并采用数理统计与地理统计相结合的方法,对高效益客源的时空分布特征进行研究.采用深圳市2011年4月18-24日一周内13 798辆出租车的GPS轨迹数据,分析了深圳市出租车的运营特点及人们出行需求的时空分布特征.结果表明：该方法能够较好地反映高效益客源的时空分布特征,为出租车高效益寻客提供辅助决策支持,也为出租车高效运营提供了新的视角.