基于主成分分析和决策级融合的高光谱图像分类方法研究

为了评价主成分分析方法和决策级数据融合方法对高光谱图像地物分类结果的影响,作者利用各分类器和主成分分量的优势,提出了两组地物分类结果的决策级融合策略,利用8种常用监督分类方法对高光谱原始图像和PCA变换后不同分量组合图分类得到的结果进行决策级融合,并应用覆盖黄河入海口新老河道交界处的CHRIS/PROBA高光谱图像开展实验研究。结果表明:直接采用每类地物分类精度最高的、空穴和缝隙采用总体分类精度最高的融合策略,在综合考虑生产者精度和用户精度的情形下,仅使用最大似然法、支持向量机和人工神经网络3种分类方法,按照分类精度从高到低的顺序进行的融合分类效果最好,总体分类精度为87.82%。与8种监督分类方法中效果最好的最大似然法相比,精度提高了2.7个百分点,同时明显减少了错分现象,尤其是对于分布面积较小的翅碱蓬和柽柳,滩涂被误分为翅碱蓬、芦苇被误分为柽柳的现象大大降低。     

基于地物光谱可分性的CHRIS 高光谱影像波段选择及其分类应用

本文以黄河口湿地为研究区,应用覆盖该区域的CHRIS高光谱遥感影像,提出了一种基于地物光谱可分性的滨海湿地高光谱影像波段选择方法。该方法利用研究区的7种典型地物的110余条现场实测地物光谱曲线,通过分析比对地物两两之间的光谱可分度,确定地物类型之间的光谱可分区间,基于此选取CHRIS高光谱影像的地物分类特征波段,应用三种经典的监督分类方法(支持向量机法SVM、人工神经网络法ANN和光谱角制图法SAM)开展利用全波段的和利用本文方法选择特征波段的分类对比实验。结果表明:(1)基于光谱可分性特征波段的方法较全波段分类精度有所提高,其中ANN分类精度最高,为82.52%,较全波段分类精度提高了约为5.1%;(2)芦苇、水体、黄河水和裸滩4种地物的识别能力高,生产者精度都在80%以上;(3)碱蓬的用户精度提升最为明显,约在7%。     

基于光谱-空间联合Transformer模型的黄河三角洲湿地高光谱影像分类

黄河三角洲湿地地物类型在光谱曲线上差异较小,且在空间上分布零散,呈破碎化特性。现有的分类方法受限于局部感受野难以捕捉到图像的长距离依赖关系,导致在黄河三角洲湿地高光谱影像中分类精度不理想,针对此问题,本文提出了一种光谱-空间联合Transformer模型。光谱和空间支路分别以光谱向量和空间邻域为输入,基于自注意力机制提取全局光谱和空间特征,在两个支路引入多阶特征交互层,实现浅层边缘信息和深层语义信息的融合,最后采用自适应相加的方式将两路特征融合,送入分类器实现最终分类。本文在黄河三角洲湿地高分五号GF-5和CHRIS两幅高光谱影像上验证方法的有效性,实验结果表明,该方法显著提高了湿地分类的精度,在选用3%的样本训练条件下总体精度分别达到了90.90%和94.17%,优于其他分类方法。研究结果可实现黄河三角洲湿地地物类型的高精度分类,为湿地的保护提供技术支持。     

联合光谱和纹理特征的滨海湿地高光谱深度学习分类—以黄河三角洲湿地为例

本文基于CHRIS高光谱遥感影像，发展了一种结合地物光谱特征和多纹理空间特征信息，采用双全链接的8层深度卷积神经网络分类算法对滨海湿地高光谱影像进行遥感地物分类，并在黄河口滨海湿地进行了应用。结果表明：1）基于测试样本数据，联合光谱特征和K-L变换的纹理特征信息，采用DCNN模型方法展现了高的分类精度，精度高达99%；2）利用光谱特征和全纹理特征的精度比仅使用光谱特征和光谱特征联合K-L变换后纹理特征的分类精度低。利用K-L变换后的光谱特征和纹理特征的DCNN分类精度达到99.38%，相比于使用全纹理特征信息的精度提高了4.15%；3）基于验证图像，发展的DCNN分类方法精度优于其他算法，DCNN方法总体分类精度为84.64%，Kappa系数为0.80；4）相比于浅层分类方法，本文发展的DCNN模型分类算法保证了所有地物类型的分类精度更加均衡，保持了主要地物类型的分类精度几乎不变，同时提高了滩涂和农田的精度。基于DCNN模型，潮滩和农田的分类精度分别达到79.26%和56.72%。比其它浅层分类方法提高了2.51%和10.6%。     

一种基于众数赋值的高光谱图像地物分类方法

提出了一种融合监督分类与非监督分类结果的高光谱遥感影像分类新方法——众数赋值分类法。采用ISODATA非监督分类方法对高光谱遥感影像进行分类,并对非监督分类结果的图斑进行标记,同时用最大似然法(ML)和支持向量机(SVM)法进行监督分类,然后以监督分类结果对非监督分类后各斑块进行类别赋值。方法是:统计每个非监督分类斑块中由监督分类所获得的各类别像元数及所占比例,将非监督分类斑块的类别赋予所占比例最高的监督分类结果的类别,最终获得高光谱图像分类结果。研究表明:(1)非监督分类类别数量大于10时,其与ML分类结果融合的总体分类精度和Kappa系数均较监督分类法的分类结果好;(2)ML和20个类别的ISODATA分类结果融合的总体精度最高,为87.35%,比单独ML的总体精度高约2个百分点;(3)SVM和10个类别的ISODATA分类结果融合的总体精度提高最大,较SVM的总体精度提高近3个百分点;(4)随着非监督分类类别数量的增多,分类结果的总体精度呈现由低到高再到低的变化过程。     

联合Bi-LSTM和注意力加强多尺度卷积的黄河三角洲湿地高光谱图像分类方法研究

近年来,面向黄河口的监测需求日益增大,如黄河入海流路改道至清水沟路以来,在新老河道的交汇处存在着丰富的地物类别,对于这些地物类别的检测识别研究有助于掌握生态环境状态,对于黄河口的湿地保护以及国家改善环境的战略支持具有重要意义。因此,本文提出一种新的湿地高光谱图像分类方法,分双路分别提取图像的空谱特征并融合分类。光谱维采用分组预处理的双向长短期记忆网络(Bi-LSTM)有效学习光谱特征;空间维采用注意力加强的多尺度卷积网络有效增强所提取的空谱特征,使得分类结果更具准确性。本文实验应用覆盖黄河入海口新老河道交界处的成像光谱仪(Compact High Resolution Imaging Spectrometer,CHRIS)所采集的数据和黄河三角洲自然保护区滨海湿地高分5号传感器(GF-5)所采集的高光谱图像开展。结果表明:分组与双向长短期记忆网络(Bi-LSTM)的有效结合显著提升了网络性能,同其他监督分类方法相比提升约3%~8%,此外注意力机制的加入同比增加约3%,在使用1%的极少训练集下数据集CHRIS和GF-5的总体分类精度分别达到92.3%和86.11%。     

Sentinel-1与Sentinel-2影像联合的黄河三角洲湿地信息提取

针对目前湿地信息提取中存在的“同物异谱”、“同谱异物”问题,以黄河三角洲自然保护区为实验区,采用在信息提取方面具有优势的典型相关森林算法,将Sentinel-1A雷达影像和Sentinel-2A多光谱影像作为基础数据,考察分别应用多光谱影像、合成孔径雷达(Synthetic Aperture Radar,SAR)和多光谱二者综合、SAR纹理特征以及植被指数对湿地信息提取的效果和适用性。研究结果表明:(1)基于Sentinel-2A多光谱影像,在分类方法上,典型相关森林(canonical correlation forest,CCF)的总体精度最高,达到94.32%,与支持向量机和随机森林分类算法相比分别提高了6.55%和5.47%;(2)基于Sentinel-2A多光谱影像和Sentinel-1A后向散射系数的CCF总体精度达到了94.89%,与只利用多光谱影像相比,3种算法的总体精度和Kappa系数均得到了提升;(3)在SAR和光学联合的基础上加入SAR纹理特征后总体精度和Kappa系数均略有下降,分别为94.72%和0.935 3;(4)在SAR和光学联合的基础上加入归一化差分植被指数(NDVI)、比值植被指数(RVI)、差值植被指数(DVI)、归一化差分红外指数(NDII)和差分红外指数(DII)5种植被指数后,总体精度达到了最高为95.35%,7种地物的生产者精度有所提高,有效提高了对黄河三角洲湿地信息的提取能力。实验结果可为黄河三角洲湿地的合理开发和有效保护提供科学支持。     

基于空谱融合与AlexNet算法的滨海湿地植被分类研究

为提高遥感影像融合质量,提升资源一号(ZY-1 02D)高光谱遥感影像滨海湿地植被分类精度,提出将ZY-1 02D高光谱影像与空间分辨率为10 m的哨兵2号(Sentinel-2)影像进行Brovey融合,并通过搭建AlexNet卷积神经网络对ZY-1 02D高光谱影像和Brovey融合影像的滨海湿地植被进行分类,与支持向量机、随机森林和BP神经网络分类算法进行精度对比。研究结果表明:经Brovey融合后,AlexNet、支持向量机、随机森林和BP神经网络算法的植被分类总体精度分别提高15.60%、7.00%、14.80%和10.00%,Kappa系数提高了21.35%、9.93%、18.97%、12.85%;基于Brovey影像融合与AlexNet算法的植被分类精度最高,总体精度为92.40%,Kappa系数为89.42%。空谱融合配合AlexNet卷积神经网络有效解决了高光谱遥感影像在滨海湿地植被分类应用中精度较低的问题,为滨海湿地植被资源动态监测提供技术和方法支撑。     

融合浅层特征的深度卷积神经网络互花米草遥感监测方法

基于2018年10月份黄河口入海两侧的LANDSAT8 OLI影像,提取植被指数和缨帽变换分量共9维光谱特征,构建融合浅层特征的8层深度卷积神经网络(deepconvolutionalneuralnetwork,DCNN)分类模型,开展互花米草(SpartinaalternifloraLoisel)遥感监测的方法研究,并从不同的浅层特征来具体分析互花米草的监测结果。结果表明:(1)在分类方法上,DCNN模型的总体分类精度最高,达到90.33%,与支持向量机(support vector machine, SVM)、随机森林(random forest, RF)分类器相比,精度分别提高4.78%、2.7%,互花米草的生产者精度分别提高了2.56%、0.47%,说明在滨海湿地遥感影像分类中, DCNN有着更好的应用潜力;(2)融合浅层特征后, DCNN的总体分类精度和互花米草的识别精度分别提高了0.34%和3.25%,有效提高了对互花米草的监测能力。其中,融合归一化植被水分指数(NDII)浅层特征的DCNN分类方法中,互花米草的识别精度提高最多,为2.56%,比值植被指数(RVI)次之,为2.32%。研究结果可为互花米草的监测与管理提供技术与数据支撑。     

基于点云与影像融合的黄河三角洲互花米草提取方法

针对卫星遥感技术在对滨海湿地互花米草监测时受分辨率、气候条件等多种因素限制存在一定局限性且通过单一的影像数据提取互花米草时精度不稳定的问题,提出了基于无人机点云与影像融合的面向对象互花米草提取方法。以黄河三角洲自然保护区为研究对象,获取了该区域的点云和多光谱影像。先将地面滤波后提取的植被点云与多光谱影像进行特征组合优化,然后对融合影像采用FNEA算法进行多尺度分割后采用基于改进的最近邻算法进行面向对象分类,最终得到的互花米草生产者精度和用户精度分别达到了82.53%和86.43%,较未融合点云的提取精度分别提高了22.34%和7.66%,分类结果的总体精度从89.54%提升至92.61%,且融合点云后影像能够有效区分两种生长状态的互花米草,表明本文提出的方法能够有效提高互花米草的提取精度。     

Microbial Biomass and Organic Nutrients in the Deep-Sea Sediments of the Central Indian Ocean Basin

In order to assess the impact of deep-sea mining on the in situ benthic life, we measured the microbial standing stock and concentration of organic nutrients in the deep-sea sediments of the Central Indian Ocean Basin in the Indian pioneer area. Sediments were collected using box core and grab samples during September 1996. The total bacterial numbers ranged from 10 10 -10 11 cells per g -1 dry weight sediment. There was a marginal decrease in the number of bacteria from surface to 30 cm depth, though the subsurface section registered a higher number than did the surface. The highest numbers were encountered at depths of 4-8 cm. The retrievable number of bacteria were two orders less in comparison with the direct total counts of bacteria. An almost homogeneous distribution of bacteria, total organic carbon, living biomass, and lipids throughout the depth of cores indicates active microbial and benthic processes in the deep sea sediments. On the other hand, a uniform distribution of total counts of bacteria, carbohydrates, and total organic carbon in all the cores indicates their stable nature and suggests that they can serve as useful parameters for long-term monitoring of the area after the benthic disturbance. Further studies on temporal variability in this region would not only verify the observed norms of distribution of these variables but would also help to understand restabilization processes after the simulated benthic disturbance.     

Assessing the Validity of Seismo-Acoustic Predictor Equations for Obtaining Seabed and Sub-Surface Sediment Physical Properties

The interdependence between the seismo-acoustic properties of a marine sediment and its geotechnical/physical parameters has been known for many years, and it has been postulated that this should allow the extraction of geotechnical information from seismic data. Though in the literature many correlations have been published for the surficial layer, there is a lack of information for greater sediment depths. In this article, a desktop study on a synthetic seafloor model illustrates how the application of published near-surface prediction equations to subsurface sediments (up to several tens of meters burial depth) can lead to spurious predictions. To test this further, acoustic and geotechnical properties were measured on a number of sediment core samples, some of which were subjected to loading in acoustically-equipped consolidation cells (oedometers) to simulate greater burial depth conditions. For low effective pressures (representing small burial depths extending to around 10 meters subsurface), the general applicability of established relationships was confirmed: the prediction of porosity, bulk density, and mean grain size from acoustic velocity and impedance appears generally possible for the investigated sedimentary environments. As effective pressure increases through, the observed relationships deviate more and more from the established ones for the near-surface area. For the samples tested in this study, in some instances increasing pressure even resulted in decreasing velocities. There are several possible explanations for this abnormal behavior, including the presence of gas, overconsolidation, or bimodal grain size distribution. The results indicate that an appropriate depth correction must be introduced into the published prediction equations in order to obtain reliable estimates of physical sediment properties for greater subsurface depths.     

海上大直径钢管桩打桩过程中桩周土强度弱化研究

海上大直径钢管桩打桩过程中,桩周土体受到强烈扰动而发生强度弱化,掌握桩周土体强度弱化规律对于准确预测打桩过程、保证工程安全具有重要意义。为研究土体强度弱化规律,开展了环剪试验模拟打桩对桩周土体的扰动,测试土体强度随剪切速率的变化规律,建立了描述土体强度弱化规律的拟合公式,引入到打桩分析软件中。研究结果表明：土体的强度折减程度不仅与土体本身的性质有关还受到土体的埋深和剪切速率的影响,埋深越深土体强度折减程度越低,剪切速率越高土体强度折减越高,在打桩分析中可采用这里推荐的线性折减方法来模拟不同深度处土体强度的折减规律。     

Geoacoustic Inversion via Genetic Algorithm and Its Application to Manganese Sediment Identification

An acoustic inversion method using a wide-band signal and two near field receivers is proposed and applied to multiple layered seabed models including a manganese sediment. The inversion problem can be formulated into a probabilistic model comprised of signals, a forward model, and additive noise. The forward model simulates wide-band signals, such as chirp signals, and is chosen to be the source-waveletconvolution plane wave modeling method. The wavelet matching technique, using weighted least-squares fitting, estimates the sediment sound-speed and thickness on which determination of the possible numerical ranges for a priori uniform distribution is based. The genetic algorithm is applied to a global optimization problem to find a maximum a posteriori solution for determined a priori search space. Here the object function is defined by an L 2 norm of the difference between measured and modeled signals. Not only the marginal pdf but also its statistics are calculated by numerical evaluation of integrals using the samples selected during importance sampling process of the genetic algorithm.     

Interannual Sea Level Variations and Annual Tides in the Northwestern Pacific

Specific properties of the interannual sea level variations and annual tides in the Northwestern Pacific were studied. Several tide stations were monitored. The monthly mean sea level for the year of 1995 was analyzed at each tide station. A seismic event in 1995, some tectonic activity around the subject area, and the Kuroshio (the oceanic western boundary current) may possibly contaminate results which would have occurred from the astronomical annual tide alone.     

Deep-Sea Sediment Compression Curves: Some Controlling Factors, Spurious Overconsolidation, Predictions, and Geophysical Reproduction

Oedometer tests have been carried out on 70 undisturbed surficial clays (at approximately 250 mm below the mudline), mostly collected by free-fall corers from sites widely scattered throughout the deep-sea North Atlantic. Acoustic measurements were also made, initially on contiguous samples and ultimately on the same sample using a geophysically instrumented oedometer which also collected electrical resistivity data. Apart from those quiescent areas below the carbonate compensation depth, such as north of the West Indies where very fine clays exist, most of the samples are silty clays whose geotechnical-geophysical properties are dependent on the type of clay minerals present (and their ability to take in moisture), the sand-size fraction, and the quantity of carbonate present. Thus the pure clays have high compressibilities which decrease on the addition of coarse particles, while the converse is true for the acoustic parameters, these increasing with the sand fraction. Using the notion of the intrinsic compression line for all samples, and comparison to it of the measured compression curves, it is clear that, contrary to some previously held ideas, most deep-sea clays are normally consolidated; the addition of carbonate has the effect of creating an open, stronger sediment skeleton. Interestingly, where information is available, the variation with depth of a sample's acoustic velocity follows the void ratio pressure relationship of the compression curve. This allows the construction of an in-situ sediment compression curve using the in-situ geophysical observations.     

Flow Cytometric Analysis of Coastal Lagoon Bacterioplankton and Picophytoplankton: Fixation and Storage Effects

The effect of different fixation and storage protocols on the flow cytometric (FCM) simultaneous analysis of bacterioplankton and phytoplankton in coastal seawater samples (Mediterranean coastal lagoons) was investigated. FCM measurements (cell number, fluorescence and scatter characteristics) were obtained through DAPI staining. Three fixatives [glutaraldehyde (GA), formaldehyde (FA) and paraformaldehyde (PFA)] and two storage (3 months duration) methods (5 °C and −196 °C) were tested. Two dominant populations were detected in studied samples: bacteria and eukaryotic picophytoplankton. Adding fixatives (2% final concentration) appears necessary to obtain FCM exhaustive counts of all the bacteria and phytoplanktonic cells. This was related to the permeation effect of fixatives which allowed a better DAPI staining of the cells. Maximum fluorescence, i.e. optimal staining of the cells was obtained with FA or PFA, and significant lower fluorescences with GA. Fixed samples stored at 5 °C induced rapid cell loss. Only storage in liquid nitrogen of samples fixed with FA or PFA, allows mid-term (≥4 months) preservation of bacteria or picophytoplankton cell numbers, and limited evolution of DAPI-induced fluorescence and scatter characteristics.     

Phylogenetic diversity and phenotypic characterization of cultivable bacterioplankton isolated from polar oceans

A set of 27 marine planktonic bacteria isolated from the polar regions was characterized by 16S rDNA sequencing and physiological and biochemical testing. More than half of these bacteria were positive for caseinase, gelatinase and 13-glucosidase, and could utilize glucose, maltose or malic acid as carbon source for cell growth. Twelve isolates expressed nitrate reduction activities. Except for one antarctic isolate BSwlO175 belonging to Actinobacteria phylum, these isolates were classified as γ-Proteobacteria, suggesting that γ-Proteobacteria dominated in cultivable marine bacterioplankton at both poles. Genus Pseudoalteromonas was the predominant group in the Chukchi Sea and the Bering Sea, and genus ShewaneUa dominated in cultivable bacterioplankton in the Prydz Bay. With sequence similarities above 97%, genus Psychrobacter was found at both poles. These 27 isolates were psychrotolerant, and significant 16S rDNA sequence similarities were found not only between arctic and antarctic marine bacteria （ 〉 99% ）, but also between polar marine bacteria and bacteria from other aquatic environments （ ≥ 98.8% ）, including temperate ocean, deep sea, pond and lake, suggesting that in the polar oceans less temperature-sensitive bacteria may be cosmopolitan and have a bipolar, even global, distribution at the species level.