U-Net深度卷积神经网络在沙脊线提取中的应用     

高博钰  杨波  张德国 《中国沙漠》2021,41(5):21-32

沙丘形态演变过程记录着近地表风况与环境演化的历史,然而对其特征研究一直受限于大范围沙脊线提取效率低和成本高等问题。本文基于深度卷积神经网络搭建U-Net模型,实现批量、高精度沙脊线的提取。将数据增强技术、随机失活神经元、批标准化以及迁移学习技术应用于模型训练和参数更新,使得模型的精度更高。结果表明：U-Net模型以及各种策略能够高效、精确地识别遥感影像中的沙脊线;沙脊线走向的偏移与近地表风况变化有着很好的对应关系, U-Net模型可以有效地用于区域性的沙脊线走向分析。  相似文献   

基于卷积神经网络和高分辨率影像的湿地群落遥感分类——以洪河湿地为例   总被引：2，自引：0，他引：2  

孟祥锐  张树清  臧淑英 《地理科学》2018,38(11):1914-1923

以洪河国家级自然保护区为研究对象,应用卷积神经网络（CNN）方法进行高分辨率湿地遥感影像的分类研究,并与基于光谱支持向量机（SP-SVM）的方法和基于纹理及光谱的支持向量机（TSP-SVM）的方法进行了对比。结果显示,对于所选取的2个研究区域,CNN分类方法的全局精度高于SP-SVM方法5.61%和5%,高于TSP-SVM方法4.18%和4.15%。尤其对于部分湿地植被的分类精度明显高于SP-SVM和TSP-SVM方法。研究表明,卷积神经网络为湿地识别的精细划分提供了有利的手段。  相似文献   

Fluids migration and dynamics of a blocks-and-faults system     

A.M. Gabrielov  V.I. Keilis-Borok  V. Pinsky  O.M. Podvigina  A. Shapira  V.A. Zheligovsky   《Tectonophysics》2007,429(3-4):229-251

We explore the impact of fluids migrating through a fault network on the dynamics of lithosphere, both on slow movements and seismicity. For that purpose fluids in the fault zones are incorporated into modelling of blocks-and-faults systems, which takes into account driving forces and the system's geometry. Simulations have been performed for two-dimensional models: an idealised “brick wall” structure, and a coarse image of Sinai Subplate. Migrating fluids originating in different locations are considered, as well as fluids trapped in closed pockets. Basic features of the modelled and observed seismicity are in good accord, as shown by comparison with the earthquake catalog compiled by Geophysical Institute of Israel.  相似文献   

Rock material from the Weddell Sea Floor of the South Ocean     

N. A. Kurentsova  T. I. Frolova  G. B. Udintsev  I. A. Roshchina 《Russian Journal of Pacific Geology》2007,1(5):435-443

This work is devoted to the results of the joint Russian-German geodynamic research carried out in the Weddell Sea and West Antarctica during cruise ANT-XXII/3 of the R/V Polarstern in 2005. The study of rock samples collected from the sea floor showed that a heterogeneous structure of the Weddell Sea was formed by spatiotemporal combination of the destruction of continental crust, progressive thalassogenesis (oceanization-taphrogenesis), and rifting, as opposed to a spreading origin. High postconsolidation mobility during the destruction stage led to the areal dismembering and high permeability of the continental crust, as well as to tectonomagmatic activation. The main mechanism of reworking of the continental crust is recognized to be the magmatic replacement by basic-ultrabasic mantle material with formation of a secondary oceanic crust and preservation of relics of the continental crust. The Earth’s endogenous activity was driven by transmagmatic fluid flows, which ascended from the melted core and caused transformation of the Earth’s crust and mantle.  相似文献   

AutoCAD三维模型在大型洞室块体稳定性评价中的应用     

崔银祥  岳良彬  聂德新 《水文地质工程地质》2005,32(3):19-22

以某大型水电站为例,阐明了AutoCAD三维实体模型在地下洞室块体稳定性分析中的应用。首先采用AutoCAD软件建立地下洞室群与周边断裂的三维实体模型,通过沿洞室拱顶和边墙的剖切,即可获得各断裂在洞室拱顶及边墙的确定位置信息,在此基础上进行可能滑动块体的检索,进行了确定性块体稳定性评价,为设计、施工提供依据。  相似文献   

浙皖地区地壳-上地幔结构和华南与扬子块体边界   总被引：12，自引：0，他引：12  

熊绍柏  刘宏兵 《地球物理学进展》2000,15(4):3-17

基于屯溪－温州剖面宽角反射地震资料,拟合反演了地壳浅层和深部结构,并利用非纵剖面记录构制了上幔顶部结构图象。结果表明,江绍断裂带是本区最明显的构造分界,其西北一侧的扬子块体和东南侧的华南块体地壳和上地幔顶部结构以及地球物理特征明显不同。金（华）衢（州）盆地下方地幔上隆,上地幔岩石层中有低速体,可能是一个复杂的岩浆房构造。地幔岩石层呈现出由ＳＥ至ＮＷ向仰冲的叠瓦式片状构造,提示该区明显受水剂压应力控制。  相似文献   

CENOZOIC LARGE-SCALE STRIKE-SLIP FAULT IN WEST YUNNAN, CHINA     

Ji Jianqing  Zhong Dalai  Sang Haiqing  Zhang Liansheng WT  ”BX 《地学前缘》2000,(Z1)

CENOZOIC LARGE-SCALE STRIKE-SLIP FAULT IN WEST YUNNAN, CHINA1　ZhongDalaiTapponnierP ,WuHaiwei,etal.Large scalesstrikeslipfault:themajorstructureofintracontinentaldeformationaftercollision[J] .ChineseScienceBulletin ,1990 ,35:77～ 97. 2　TapponnierP ,LacassinR ,LeloupPH ,etal.TheAilaoShan/RedRivermetamorphicbelt:tertiaryleft lateralshearbetweenIndochinaandSouthChina[J] .Nature,1990 ,343:4 31～ 4 37. 3　LiKang ,ZhongDalai.Themicrostructurecharacteristicso…  相似文献   

Estimation of potential failure risks in a mine slope using indicator kriging     

M. S. Lana  I. E. Cabral  A. H. Gripp  M. F. A. Gripp 《国际地质力学数值与分析法杂志》2010,34(16):1725-1742

Indicator kriging has been applied to the study of failure mechanisms in a mine slope in Minas Gerais, Brazil, to estimate potential failure risks in limited areas along this slope. Timbopeba Mine, Vale Company, is an open pit iron mine situated in the Quadrilátero Ferrífero, a very important mining district in Minas Gerais. A slope excavated in quartzite with a maximum height of 200 m at the time of this study, has presented many failure problems involving the sliding of blocks formed by discontinuities. These blocks are of limited size in comparison to the dimensions of the overall slope. They appear along the entire slope, wherever discontinuity orientations have led to the kinematic feasibility of these blocks. Geostatistics permits the estimation of local failure probability distributions associated to these local failures, which would not be possible with traditional statistical models. The geostatistical method employed in this study, indicator kriging, is quite suitable because it is unnecessary to assume a particular global distribution of the phenomena being modeled. The model was used for locating areas with a great tendency for sliding failure, as it considers the local spatial variability of discontinuity orientations. Copyright © 2010 John Wiley & Sons, Ltd.  相似文献   

基于深度学习的高分四号卫星图像超分辨率重建     

贺智  贺丹 《遥感学报》2020,24(12):1500-1510

针对高分四号卫星图像中波红外谱段空间分辨率远低于相应的可见光近红外谱段，提出一种基于深度学习的高分四号卫星图像中波红外谱段超分辨率重建方法。首先，设计卷积回归网络进行初步重建，利用8倍下采样的可见光近红外谱段和原始中波红外谱段训练回归网络。然后，设计卷积重建网络进行进一步重建，利用低分辨率初步重建结果和原始中波红外谱段训练重建网络。最后，将原始可见光近红外谱段依次输入到训练好的回归网络和重建网络，得到最终的中波红外谱段超分辨率重建结果。以湖北和江西部分地区真实数据进行试验，结果表明该方法能有效提高中波红外谱段空间分辨率，与其他方法对比均方根误差至少下降了7.54，且目视效果更清晰自然，有利于扩展高分四号卫星的应用范围。  相似文献   

基于编解码网络的航空影像像素级建筑物提取   总被引：2，自引：2，他引：0  

陈凯强  高鑫  闫梦龙  张跃  孙显 《遥感学报》2020,24(9):1134-1142

建筑物提取在城市规划等土地利用分析中发挥着重要作用。用于提取建筑物的传统方法通常基于手工特征和分类器，导致精度较低。本文基于编解码结构的卷积神经网络CNN(Convolutional Neural Networks),自主学习多级的和具有区分度的特征来更好地辨识建筑物和背景，实现航空影像中的像素级建筑物提取。该网络由编码子网络和解码子网络两部分组成，编码子网络对输入图像进行空间分辨率压缩，完成特征提取；解码子网络从特征中提升空间分辨率，完成像素级的建筑物提取。此外，本文使用视野增强FoVE(Field-of-View Enhancement)方法减轻边缘现象(切片边缘附近的建筑物提取精度通常低于中心区域附近的精度)的影响，并分别在两个建筑物提取标准数据集上的实验表明，编解码卷积神经网络能有效实现像素级建筑物提取，FoVE能有效提高建筑物提取准确率；通过改变预测时切片大小和重叠度，分析其对建筑物提取结果的影响，揭示了FoVE的饱和性。  相似文献