投影寻踪回归与BP神经网络方法在前汛期降水预测中的比较研究

以1962—2006年粤北地区7个站4—6月前汛期降水量资料为基础,将前汛期降水量与74项环流指数资料进行灰色关联度分析,确定了影响粤北地区前汛期降水量的16个关键环流指数因子,分别应用投影寻踪回归、BP神经网络和逐步回归方法,建立前汛期降水趋势预测模型,对粤北地区前汛期降水趋势进行预测。结果表明:投影寻踪回归和BP神经网络方法的预测能力均优于传统的逐步回归模型。其中,PPR模型比BP神经网络方法的预测效果更好。     

基于多尺度小波特征的高光谱影像亚像素目标识别

提出了一种基于多分辨率小波高频特征系数的高光谱遥感影像亚像素目标识别方法。首先利用多尺度小波变换将光谱信号分解为不同尺度的高频特征信号,然后借助接收操作特性曲线(ROC)和马氏距离投影寻踪求取一维最佳识别特征,最后通过高斯最大似然决策函数求解亚像素目标的存在概率。通过38种小波函数的高光谱数据实验证明,该方法对亚像素目标的识别效果较好。     

利用相似性特征进行Web服务语义匹配

通过对传统语义相似性度量方法的分析和比较,利用相似性关系本身所具有的相对性和传递性,选取共同的领域本体作为相似性比较的基准,提出了基于相似性特征的Web服务语义匹配算法,解决了现有Web服务语义相似性度量方法中需要外在文本集、无法对相似性进行定量表达以及相似性区分程度不明显等问题.最后,选取德国人工智能研究中心发布的基于OWL-S的geography领域服务测试集进行语义相似性度量对比实验,实验结果证明了所提算法的可行性和有效性.     

洮河泥沙分布及变化分析

以洮河红旗站1956~2005年水文监测资料为依据,通过对泥沙分布特征、年内年际变化规律和水沙匹配情况进行分析,得出洮河泥沙具有含量高、输沙总量大、产沙的区域性、年际的波动性、年内的不均匀性和时间的集中性六大特点。     

地物模型与地形模型的匹配研究

由于地物与地形模型分开建模等现实原因,地物模型与地形模型集成的时候会出现不匹配问题.因此,本文通过分析地物与地形之间的空间依存关系,针对地物独立于地形和地物依赖于地形两种情况分别提出了地物模型与地形模型的匹配方案,并着重研究地物依赖于地形时的匹配方案与算法.     

基于匹配追踪和遗传算法的大地电磁噪声压制

针对匹配追踪计算量大、大地电磁数据处理效率低的问题,提出基于匹配追踪和遗传算法的大地电磁噪声压制方法.首先,利用Gabor原子构建过完备原子库,并对过完备原子库集合进行划分.然后,借助遗传算法的自适应性,快速搜寻最优匹配原子及所在位置.最后,运用最优匹配原子对待处理信号进行稀疏分解,重构有用信号.通过对计算机模拟的典型强干扰和矿集区实测大地电磁数据进行分析处理,实验结果表明,相对于匹配追踪和正交匹配追踪,文中所提方法能从过完备原子库中快速、自适应地选取最优匹配原子与不同噪声干扰类型高精度的匹配,极大地提升了计算效率;大地电磁时间域序列中的大尺度强干扰被有效剔除,视电阻率曲线更为光滑、连续,低频段的数据质量得到明显改善.

     

RESEARCH ON EARTHQUAKE EMERGENCY MAPPING METHOD BASED ON TEMPLATE MATCH

After an earthquake, earthquake emergency response and rescue are important ways to mitigate earthquake-induced losses. Various earthquake emergency maps can provide effective references and guidance to those actions. Currently, related studies include the investigation on symbols of emergency maps, remote sensing emergency mapping and GIS-based mapping methods. However, the existing studies overlook the characteristics of rapidity, dynamicity and variety of presentation methods in making earthquake emergency maps. In this paper, a map template matching method is used to quickly make earthquake emergency maps considering their characteristics. We take investigations on the service objects(users)of the earthquake emergency maps to understand the needs of making earthquake emergency maps. The audience theory in mass media field and map information transmission theory are adopted to classify the users of the earthquake emergency maps into four categories: earthquake emergency commanders, technical staffs for decision-making, earthquake emergency rescuers, and the public. The components of different types of users are described and then their diverse demands in earthquake emergency maps are analyzed, such as the needs of on-field disaster information maps, earthquake information maps, physical geography and social economic maps. Following those needs, we introduce the representation methods of the earthquake emergency maps according to their formats(vector or raster)and contents, such as point symbolization method, kilometer grid method, line symbolization method and range method. Then, we study the rapid plotting method of earthquake emergency map based on map template matching method. The core steps of the method include: 1)before earthquake, the templates of different earthquake emergency maps are designed, prepared and connect the earthquake emergency features with their related spatial database. The map layout and map elements are stored in the templates. 2)After earthquake, the earthquake emergency features will be generated from seismic models(such as attenuation model of earthquake magnitude and seismic intensity)or the information obtained from field investigation. 3)Corresponding earthquake emergency map template is selected in accordance with the generated seismic features. And the features are used to update related features inside the selected template. 4)Minor adjustments are made such as to the map scale and some map annotations to finally generate the formal earthquake emergency map. Architecture of template system of the earthquake emergency maps is designed, including map user level, map template level, template layer level and map element level. Regrading to the architecture, the general map template of earthquake emergency is presented which includes four main regions: title region, main picture region, auxiliary region and annotation region. The main picture region is the essential, which lays geographic background maps and earthquake emergency features. Finally, an earthquake emergency mapping system is developed. Based on the system, a case study is presented, which demonstrates making a simulated seismic intensity influence map. From three aspects, the case presents the application of the template-matching method including: generating earthquake emergency features, substituting the features inside the template with the generated features, and revising map annotations. Therefore, the map template matching method is verified so that it can be used to quickly generate various earthquake emergency maps.     

高维PP时间序列分析在地震预报中的应用

将投影寻踪(PP)与高维时间序列分析结合起来，建立了地震PP综合预测模型。并选取祁连山地区作研究区，做了未来三个月内最大震级的短期预测，经展望式检验，合格率(预测误差≤0．5或≤10％)≥80％。中根据实际需要提出建立一定震级门限(M≥4．0)的预测具有更好的效果。本预测模型还可增加自变量、维数和改变时间尺度，并运用到其他领域中。     

物质平衡法对定容煤层气藏生产动态的预测

由于煤层气藏特殊的赋存及渗流方式,使其有别于常规气藏,而煤层气藏开采初期的多相渗流问题使其产能预测更为复杂。在定容煤层气藏物质平衡法基础上,利用历史生产数据完成线性拟合,并用拟合得到的参数对未来生产动态进行预测。通过将模型预测的平均地层压力、平均含水饱和度及产气、产水量随时间变化与CMG数值模拟结果进行对比分析,结果显示,本文方法预测结果较为准确可靠,尤其是在煤层气峰值产量之后。     

基于非稳态匹配的角度域叠前道集去调谐方法

地层调谐效应改变了不同偏移距处反射波的干涉模式,使得远偏移距数据振幅和频率信息发生畸变,将会降低AVO/AVA分析及反演的置信度和分辨率.文中首先给出了调谐作用下反射波合成记录,通过正演模拟说明了调谐对叠前道集的影响.考虑到不同地层厚度下调谐效应的差异性,借助局域Lamoureux窗实现地震数据的自适应分解,利用角度数据之间的差异,构建了角度域地震数据的非平稳匹配目标函数,形成了非平稳匹配去调谐方法,实现了叠前道集的振幅和波形拉伸校正.本文方法在实际应用中取得了较好的效果,能够有效地改善大角度地震数据品质,为储层预测与流体识别奠定了数据基础.