高光谱影像概率分类向量机分类方法研究

从分析基于支持向量机和相关向量机的高光谱影像分类方法的优势和不足出发,将基于概率分类向量机的方法用于高光谱影像分类试验。在贝叶斯理论框架下,概率分类向量机为基函数权值引入截断Gauss先验概率分布,使得不同类别的基函数权值具有不同符号的先验分布,并利用EM算法进行参数推断,得到足够稀疏的概率模型,弥补了相关向量机选取错误类别的样本作为相关向量的不足,从而有效地提高了模型的分类精度和稳定性。OMIS和PHI影像分类试验表明,概率分类向量机能够很好地应用在高光谱影像分类。     

耦合长期预报的跨流域引水受水水库调度模型

为提高跨流域引水工程受水水库引水有效性,研究了耦合长期径流预报信息的跨流域引水受水水库调度模型。首先选取汛期径流预报信息,采用径流预报概率修正先验概率来描述径流的不确定性,建立了贝叶斯随机动态规划模型(BSDP-LTF)。然后将模型应用于碧流河水库,并与仅考虑径流相关的随机动态规划模型(SDP-I)、仅考虑长期预报信息的随机动态规划模型(SDP-LTF)进行比较。比较结果得出在供水保证率基本一致且不增加调度风险的情况下,BSDP-LTF模型相比SDP-I、SDP-LTF模型,可分别减少引水8.2%、4.1%。表明贝叶斯随机动态规划模型BSDP-LTF有效改进了径流描述,提高了跨流域引水的有效性。     

基于IBUNE方法的水文模型不确定性分析

在贝叶斯理论框架下,根据一种可结合多个水文模型给出模拟或预报结果的IBUNE方法探讨了水文模型的输入、参数以及结构的不确定性问题。将SCEM-UA算法和EM算法嵌入新安江和TOPMODEL水文模型用于参数优化和模型平均,进而将输入与参数的综合不确定性处理后得到的预报量后验分布进行多模型综合,据此对水文模型的不确定性及其对水文模拟结果的影响进行评价。以湖南洣水流域龙家山水文站以上集水区域为例进行了应用研究,结果表明,IBUNE方法能够有效估计水文模型的不确定性,并能给出合理的概率预报区间。     

热稳定κ-卡拉胶酶产生菌Bacillus sp. Car19的发酵条件优化

对一株分泌热稳定κ-卡拉胶酶印尼热泉菌进行了种属鉴定, 并采用响应面法对该菌发酵产酶条件进行了优化。鉴定结果表明, 该菌株属于芽孢杆菌属(Bacillus), 命名为Bacillus sp. Car19(GeneBank: KT865196)。发酵条件优化结果显示, 9个环境因子影响Bacillus sp. Car19产酶量。其中影响Bacillus sp. Car19产酶量的三个主要因素分别为培养温度、培养基中Cu²⁺浓度和培养基中NaCl浓度。综合次要因素对Bacillus sp. Car19产酶影响, Bacillus sp. Car19最佳产酶发酵条件为: 培养温度52.31℃、Cu²⁺浓度6.93 mmol/L、NaCl浓度37.03 g/L, 培养基pH为6, 接种量1%, 培养时间36h, 半乳糖浓度0.3 g/L,硝酸铵浓度7g/L, 卡拉胶浓度0.5g/L。优化后发酵上清液酶活力达到15.21 U/mL, 与优化前相比提高了1.5 倍。     

基于SMOS卫星数据的海表面盐度模型

海表面盐度是研究海洋对全球气候影响以及大洋环流的重要参量之一,而卫星遥感技术是获取海表面盐度数据的最有效方法.目前,L波段的SMOS和Aquarius/SAC-D遥感卫星正在用于探测海表面盐度,并根据卫星观测数据和物理机制反演出海表面盐度的产品.但在某些近陆地区域,由于淡水流入及陆地射频(RFI)等因素影响,卫星反演盐度的产品精度较低.文中利用“东方红2号”科学考察船的实测数据、SMOS卫星数据,首次针对中国南海海域提出了用贝叶斯网络模型计算海表面盐度,并用验证数据集(实测Argo盐度)对模型进行适应性评估.经过计算,模型误差和验证误差分别为0.47 psu和0.45 psu,而相应的SMOS Level 2产品的精度分别为1.90 psu和1.82 psu.此模型为海表面盐度的计算提供了一个新方法.     

Nonparametric Estimation of Undrained Shear Strength for Normally Consolidated Clays

For decades, various well-known empirical correlations have been established and commonly used to estimate the undrained shear strength of normally consolidated clays in geotechnical practice. However, there are still many contradictory cases regarding the validity of these correlations, so their applicability and reliability are questionable. In this study, an extensive geotechnical database for normally consolidated clays has been compiled. Next, a Bayesian nonparametric method is exploited to identify the most critical soil index properties in inferring the undrained shear strength and a new predictive model is proposed. Finally, the performance of the proposed predictive model is assessed and compared with two well-known correlations using an independent testing database. It confirms that the proposed model not only possesses high predictability, but also provides superior performance over other existing empirical correlations.     

Probabilistic calibration of a coupled hydro-mechanical slope stability model with integration of multiple observations

Multiple types of responses, such as displacements, ground water level, pore water pressures, water contents, etc., are usually measured in comprehensive monitoring programmes for rainfall-induced landslide prevention. In this study, a probabilistic calibration method for coupled hydro-mechanical modelling of slope stability is presented with integration of multiple types of measurements. A numerical example of a soil slope under rainfall infiltration is illustrated to compare the effects of single and multiple types of responses on parameter estimation and model calibration. The results show that the soil parameters can be estimated with less uncertainty and total uncertainty bounds are narrower with multiple types of responses than with a single type of response. Model calibration based on multiple types of responses can compromise different responses and hence the means and standard deviations of model error are the smallest. A feasible correlation coefficient between soil modulus and permeability can be obtained from model calibration with multiple types of responses and single type of response as long as the responses include displacement data.     

Effect of loading duration on uncertainty in creep analysis of clay

Some studies suggest that creep parameters should be determined using a greater quantity of creep test data to provide more reliable prediction regarding the deformation of soft soils. This study aims to investigate the effect of loading duration on model updating. One‐dimensional consolidation data of intact Vanttila clay under different loading durations collected from the literature is used for demonstration. The Bayesian probabilistic method is used to identify all unknown parameters based on the consolidation data during the entire consolidation process, and their uncertainty can be quantified through the obtained posterior probability density functions. Additionally, the optimal models are also determined from among 9 model candidates. The analyses indicate that the optimal models can describe the creep behavior of intact soft soils under different loading durations, and the adopted method can evaluate the effect of loading duration on uncertainty in the creep analysis. The uncertainty of a specific model and its model parameters decreases as more creep data are involved in the updating process, and the updated models that use more creep data can better capture the deformation behavior of an intact sample. The proposed method can provide quantified uncertainty in the process of model updating and assist engineers to decide whether the creep test data are sufficient for the creep analysis.     

基于神经网络的页岩微纳米孔隙微结构分析的正则化和最优化方法

页岩气成藏机理与页岩内部孔隙结构紧密相关,对页岩孔隙结构的研究成为页岩气勘探开发技术中至关重要的一环。页岩内部不同结构体组分对X-射线的吸收能谱不一样,这样就导致观测数据是由不同页岩组分衰减不同波段的X-射线构成的。经过对CT图像分割,能够获得页岩微孔结构的图像,尤其是获得有机质中孔隙类别、形状、尺寸、空间分布、连通特性。本文利用同步辐射X射线扫描重构的页岩CT数据,研究并设计基于多能CT图像的神经网络图像分割技术和算法,以期得到页岩体三维结构特征及空间分布,可以为建立有机质种类和无机矿物组成与微纳孔隙特征的联系以及最终实现页岩气的资源储量评估和勘探开发提供技术支持。     

VGI and crowdsourced data credibility analysis using spam email detection techniques

Volunteered geographic information (VGI) can be considered a subset of crowdsourced data (CSD) and its popularity has recently increased in a number of application areas. Disaster management is one of its key application areas in which the benefits of VGI and CSD are potentially very high. However, quality issues such as credibility, reliability and relevance are limiting many of the advantages of utilising CSD. Credibility issues arise as CSD come from a variety of heterogeneous sources including both professionals and untrained citizens. VGI and CSD are also highly unstructured and the quality and metadata are often undocumented. In the 2011 Australian floods, the general public and disaster management administrators used the Ushahidi Crowd-mapping platform to extensively communicate flood-related information including hazards, evacuations, emergency services, road closures and property damage. This study assessed the credibility of the Australian Broadcasting Corporation’s Ushahidi CrowdMap dataset using a Naïve Bayesian network approach based on models commonly used in spam email detection systems. The results of the study reveal that the spam email detection approach is potentially useful for CSD credibility detection with an accuracy of over 90% using a forced classification methodology.