基于超像元词包特征和主动学习的高分遥感影像变化检测     

杨进一  徐伟铭  王成军  翁谦 《地球信息科学学报》2019,21(10):1594-1607

为解决高分辨率遥感影像变化检测中存在底层特征缺乏语义信息、像元级的检测结果存在“椒盐”现象以及监督分类中样本标注自动化程度较低,本文提出一种基于超像元词包特征和主动学习的变化检测方法。首先采用熵率分割算法获取叠加影像的超像元对象;其次提取两期影像像元点对间的邻近相关影像特征（相关度、斜率和截距）和顾及邻域的纹理变化强度特征（均值、方差、同质性和相异性）,经线性组合作为像元点对的底层特征;然后基于像元点对底层特征利用BOW模型构建超像元词包特征,并采用一种改进标注策略的主动学习方法从无标记样本池中优选信息量较大的样本,且自动标注样本类别;最后训练分类器模型完成变化检测。通过选用2组不同地区的GF-2影像和Worldview-Ⅱ影像作为数据源进行实验,实验结果中2组数据集的F1分数分别为0.8714、0.8554,正确率分别为0.9148、0.9022,漏检率分别为0.1681、0.1868,误检率分别为0.0852、0.0978。结果表明,该法能有效识别变化区域、提高变化检测精度。此外,传统主动学习方法与改进标注策略的主动学习方法的学习曲线对比显示,改进的标注策略可在较低精度损失下,有效提高样本标注自动化程度。  相似文献   

气候变化对西藏湖泊变迁的影响(1973—2017)   总被引：1，自引：0，他引：1  

闫立娟 《地球学报》2020,41(4):493-503

西藏湖泊众多,其水位和面积变化对气候波动有敏感记录。本文运用RS和GIS技术,以1973—1977年、1989—1992年、1999—2001年、2008—2010年和2017年5期遥感影像为底图矢量化了西藏所有湖泊边界,建立了湖泊空间数据库。以湖泊空间数据为基础,分析了1973—2017年西藏湖泊面积动态变化特征,结论如下:从20世纪70年代至2017年,西藏湖泊总面积持续增加,共增长了47.23%;从20世纪70年代至90年代,研究区北部和中部湖泊呈萎缩的趋势,其余地区呈扩张的趋势;从2000年至2017年,西藏湖泊呈持续扩张的趋势。另外,笔者分析了研究区年平均气温、年降雨量和年蒸发量的变化特征。1981—2017年,西藏气候向暖湿方向发展,主要表现为气温升高、降雨量增加和蒸发量减少。气候变化对不同地区不同时期湖泊变迁影响显著:(1)西藏北部和中部湖泊主要以冰川、冰雪融水和地表径流为主要补给源。20世纪70年代至90年代,气温和降雨量波动较小,引起这些地区湖泊萎缩的主要原因为冰川和冰雪融水补给的减少。(2)从20世纪90年代至2017年,气温和降雨量增加、蒸发量减少,导致研究区湖泊呈现全面扩张的趋势。  相似文献   

遥感技术在“6·17”丹巴堵江泥石流灾害链灾区应急救援抢险决策中的应用     

唐尧  王立娟  赵娟  张成信 《中国地质调查》2020,7(5):114-122

2020年6月17日,四川丹巴县半扇门镇梅龙沟发生泥石流,阻断小金川河,形成堰塞湖,造成重大财产损失与人员伤亡。为掌握第一手灾情,辅助开展应急救援,以国产高分二号卫星数据为主要数据源,结合无人机航空摄影数据和灾后现场调查等资料,开展了“6·17”丹巴堵江泥石流灾害链灾区应急调查与分析。研究认为: 灾区影响人口约6.5万人; 国网丹巴县供电公司应急救援队与小金县联通通信应急抢险救援队距离最近,半扇门中学等5所学校可作为临时安置避难场所备选; 受影响矿山企业6家、重要水库1座、水电站1座; 泥石流沟总体呈NW—SE向展布,掩没面积约14.27万m²; 堰塞湖湖面面积约1.03 km²,淹没区约49.8万m²,路基路面全毁路段约2.1 km,下游疑似灾害隐患点9处,因灾受损民居25处、桥梁5处; 优选了2条灾后通达性较好的通往泥石流重灾区的救援生命线。借助遥感技术深入分析泥石流灾害链影响人口等灾情先期研判、灾情灾损应急调查及应急救援生命通道优选,对我国西南山区类似条件地质灾害隐患点的应急抢险救援、防灾减灾工作具有重要的指导意义。  相似文献   

Locality Attributes and Networks Serving to Reveal Amazonian Urbanization beyond the Cities     

Ana Paula Dal'Asta  Silvana Amaral 《Geographical review》2019,109(2):199-223

Beyond the Amazonian cities, distinct spatial units arranged in networks compose an extended urban space, usually invisible to government agencies and public polities. This article empirically reveals this space by studying the localities’ characteristics and connections. In southwestern Pará, Brazil, 236 communities were sampled using field expeditions. Variables related to the locality organization, history, health and education services, urban infrastructure, and land use were explored in a fuzzy cluster analysis. Local production and consumption flows were considered in the locality network analysis. The fuzzy analysis allowed identifying a hierarchy of localities for which the geographical location influences the spatial distribution patterns of five groups. On the other hand, consumption and production networks identified different locality's articulations, with persistent dependence on cities. These results characterize extensive urbanization at the local level and highlights the participation of the local populations configuring the Amazonian territory.  相似文献   

Estimation of Grassland Production in Central and Eastern Mongolia from 2006 to 2015 via Remote Sensing     

LI Ge  WANG Juanle  WANG Yanjie  WEI Haishuo 《资源与生态学报(英文版)》2019,10(6):676-684

Mongolia is an important part of the Belt and Road Initiative “China-Mongolia-Russia Economic Corridor” and a region that has been severely affected by global climate change. Changes in grassland production have had a profound impact on the sustainable development of the region. Our study explored an optimal model for estimating grassland production in Mongolia and discovered its temporal and spatial distributions. Three estimation models were established using a statistical analysis method based on EVI, MSAVI, NDVI, and PsnNet from Moderate Resolution Imaging Spectroradiometer (MODIS) remote sensing data and measured data. A model evaluation and accuracy comparison showed that an exponential model based on MSAVI was the best simulation (model accuracy 78%). This was selected to estimate the grassland production in central and eastern Mongolia from 2006 to 2015. The results show that the grassland production in the study area had a significantly fluctuating trend for the decade study; a slight overall increasing trend was observed. For the first five years, the grassland production decreased slowly, whereas in the latter five years, significant fluctuations were observed. The grassland production (per unit yield) gradually increased from the southwest to northeast. In most provinces of the study area, the production was above 1000 kg ha ^-1, with the largest production in Hentiy, at 3944.35 kg ha ^-1. The grassland production (total yield) varied greatly among the provinces, with Kent showing the highest production, 2341.76×10 ⁴ t. Results also indicate that the trend in grassland production along the China-Mongolia railway was generally consistent with that of the six provinces studied.  相似文献   

Facies and sequence stratigraphic architecture of the Mural Limestone (Albian), Arizona: Carbonate response to global and local factors and implications for reservoir characterization     

Rachel Aisner-Williams  Charles Kerans 《Sedimentology》2020,67(7):3735-3768

This study documents the detailed facies and sequence stratigraphic architecture of a multi-cyclic patch-reef and its associated ramp interior facies that formed during Oceanic Anoxic Event 1b in the Mural Limestone, Arizona, USA. Ramp interior facies are comprised of bedded wackestone/packstone, rudist build-up and coral–algal patch-reef facies located north of Bisbee, Arizona, at the Grassy Hill locality. The larger multi-cyclic patch-reef that developed coevally ca 5 km to the south of Grassy Hill consists of a high-angle windward margin with a narrow ca 70 m long reef frame containing vertically zonated Microsolena, Actinastrea, diverse branching coral and rudist assemblages, and an 870 m long low-angle leeward margin comprised of reef debris rudstone and grainstone shoal facies. Similar reef geomorphology and orientation is documented across the Gulf of Mexico and reflects the shelf-wide north to north-east-trending prevailing wind and current energies. Controls affecting reef formation and growth patterns include changes in accommodation space associated with low-amplitude global sea-level rise and regional thermotectonic subsidence, local accommodation space and nutrient fluctuations associated with the inner shelf depositional setting within a humid and siliciclastic-rich environment. Four aggradational to retrogradational high-frequency sequences are documented in Arizona: High-frequency sequences 1 and 2 represent the first pulse of patch-reef development in an overall second-order marine transgression over the Sonora/Bisbee Shelf. These sequences correlate to δ¹³C signatures associated with Oceanic Anoxic Event 1b across the Gulf of Mexico and suggest that carbonate reefs persisted on the ramp interior during this time. High-frequency sequences 3 and 4 record a second brief transgression and backstepping of reef facies followed by the final regression of shallow shelf carbonates that correlates to more robust patch-reef development in Sonora, Mexico. The patch-reef at Paul Spur is an excellent outcrop analogue for productive patch-reefs in the Maverick Basin (Comanche Shelf) of Texas. Detailed facies mapping of this outcrop analogue shows that the greatest reservoir potential is contained within the backreef grainstone shoals where primary porosity of up to 15% is observed.  相似文献   

The spatial local accuracy of land cover datasets over the Qiangtang Plateau,High Asia     

Liu  Qionghuan  Zhang  Yili  Liu  Linshan  Li  Lanhui  Qi  Wei 《地理学报(英文版)》2019,29(11):1841-1858

We analyzed the spatial local accuracy of land cover (LC) datasets for the Qiangtang Plateau, High Asia, incorporating 923 field sampling points and seven LC compilations including the International Geosphere Biosphere Programme Data and Information System (IGBPDIS), Global Land cover mapping at 30 m resolution (GlobeLand30), MODIS Land Cover Type product (MCD12Q1), Climate Change Initiative Land Cover (CCI-LC), Global Land Cover 2000 (GLC2000), University of Maryland (UMD), and GlobCover 2009 (Glob-Cover). We initially compared resultant similarities and differences in both area and spatial patterns and analyzed inherent relationships with data sources. We then applied a geographically weighted regression (GWR) approach to predict local accuracy variation. The results of this study reveal that distinct differences, even inverse time series trends, in LC data between CCI-LC and MCD12Q1 were present between 2001 and 2015, with the exception of category areal discordance between the seven datasets. We also show a series of evident discrepancies amongst the LC datasets sampled here in terms of spatial patterns, that is, high spatial congruence is mainly seen in the homogeneous southeastern region of the study area while a low degree of spatial congruence is widely distributed across heterogeneous northwestern and northeastern regions. The overall combined spatial accuracy of the seven LC datasets considered here is less than 70%, and the GlobeLand30 and CCI-LC datasets exhibit higher local accuracy than their counterparts, yielding maximum overall accuracy (OA) values of 77.39% and 61.43%, respectively. Finally, 5.63% of this area is characterized by both high assessment and accuracy (HH) values, mainly located in central and eastern regions of the Qiangtang Plateau, while most low accuracy regions are found in northern, northeastern, and western regions.

  相似文献   

基于正交匹配追踪的声层析方法   总被引：2，自引：0，他引：2  

李倩倩  史娟  李整林  罗宇  阳凡林  张凯 《海洋学报(英文版)》2019,38(11):149-157

声速剖面的变化会对声传播产生较大的影响，经验正交函数模型经常用来实现对声速剖面数据的简化描述。然而在内波、湍流等海水不均匀性存在时，这种正则化操作会造成声速重构精度的大幅降低。本文利用字典学习生成声速剖面的非正交原子，在稀疏编码时采用正交匹配追踪（OMP，Orthogonal Matching Pursuit）算法，更新字典则使用KSVD （Kernel Singular Value Decomposition）的字典更新算法。由于字典学习不需要强制使用正交条件，对于训练数据更加灵活，从而可以使用少数的原子组合达到更高的重构精度。利用一次浅海声学实验多次测量的声速剖面研究了海水声速剖面的经验正交函数表示和字典学习，研究表明：相比于正交函数表示，学习字典可以利用少数原子（甚至一个原子）更好的解释声速剖面扰动。字典学习可以提高声速剖面的稀疏性，从而提高声速剖面的反演精度。  相似文献   

基于卷积神经网络与条件随机场方法提取乡镇非正规固体废弃物     

刘懿兰  黄晓霞  李红旮  柳泽  陈崇  王新歌 《地球信息科学学报》2019,21(2):259-268

随着村镇经济建设发展,生活垃圾和工业固体废弃物造成的污染问题日益突出,已经成为制约新农村建设发展和生态文明建设的关键问题,而目前针对乡镇非正规固体废弃物的调查与统计主要依赖全国各乡镇相关部门逐级调查上报,工作量较大。本文基于高分辨率遥感影像,将深度学习模型和条件随机场模型相结合引入到乡镇固体废弃物的提取研究中,探索一种基于深度卷积神经网络的乡镇固体废弃物提取模型。由于固体废弃物在影像上表现为面积小,分布破碎等特点,为了提高工作效率,将模型特分为识别和提取2个部分：① 通过全连接卷积网络（CNN）对固体废弃物进行快速识别判断,筛选感兴趣区域影像块;② 在传统的全卷积神经网络(FCN)的基础上加入条件随机场模型(CRF)提取固体废弃物边界,提高整体分割精度。根据安徽、山西等地区相关部门上报固体废弃物堆放点以及住房与城乡建设部城乡规划管理中心进行野外检查的结果,实验最终识别精度达到86.87%以上;形状提取精度为89.84%,Kappa系数为0.7851,识别与提取精度均优于传统分类方法。同时,该方法已经逐步应用于住房和城乡建设部有关成都、兰州、河北等部分乡镇非正规固体废弃物的核查工作,取得了较为满意的结果。  相似文献   

非等间距GM(1,1)模型的总体最小二乘算法及其病态问题     

陶武勇  花向红  鲁铁定  陈西江  张伟 《大地测量与地球动力学》2019,39(1):45-50

在非等间距GM(1,1)模型中，系数矩阵中有无误差的常数项和有误差的随机项，并且系数矩阵与观测向量误差同源，即系数矩阵与观测向量中有相同的元素存在，这些相同元素应该有相同的改正数，为此本文推导了一种适合非等间距GM(1,1)模型求解的总体最小二乘算法。同时，考虑到非等间距GM(1,1)模型中存在病态问题时影响总体最小二乘计算结果的稳定性，提出对系数矩阵常数列乘以某一常数的方法，以改善病态问题。  相似文献