融合网格注意力阀门和特征金字塔结构的高分辨率遥感影像建筑物提取

在高分辨率遥感影像中提取建筑物轮廓是地区基础建设信息统计的一项重要任务。适应性较强的深度学习方法已在建筑物提取研究中取得较大进展,受网络模型对影像特征表达的局限性,存在局部建筑轮廓边缘模糊的问题。本研究提出一种基于注意力的U型特征金字塔网络（AFP-Net）可以聚焦高分遥感影像中不同形态的建筑物结构,实现建筑物轮廓的高效提取。AFP-Net模型通过基于网格的注意力阀门Attention Gates模块抑制输入影像中的无关区域,凸出影像中建筑物的显性特征;通过特征金字塔注意力Feature Pyramid Attention模块增加高维特征图的感受野,减少采样中的细节损失。基于WHU建筑物数据集训练优化AFP-Net模型,测试结果表明AFP-Net模型能够较清晰地识别出建筑物轮廓,在预测性能上有更好的目视效果,在测试结果的总体精度和交并比上较U-Net模型分别提高0.67%和1.34%。结果表明,AFP-Net模型实现了高分遥感影像中建筑物提取的结果精度及预测性能的有效提升。     

图像数据融合的地貌类型识别分类与制图

计算机遥感地貌制图是利用航空像片或者卫星影像进行识别制图;另是利用DEM数据融合计算提取。对此,本文介绍了一种对区域基本地貌形态类型进行计算机自动分类的方法。它通过识别标志在影像上对地貌分布区进行数字化,把TM影像中的地貌信息和从DEM中提取出来的地貌信息结合,以划分出详细的地貌类型:如河北省沽源县的台地、河谷平原、开阔平原、丘陵、低山和中山6大类。最后,通过一定的算法进行分类成图。     

基于PRCUnet的高分遥感影像建筑物提取

基于高分辨率遥感影像的建筑物提取具有重要的理论与实际应用价值,深度学习因其优异的深层特征提取能力,已经成为高分影像提取建筑物的主流方法之一。本文在改进深度学习网络结构的基础上,结合最小外接矩形与Hausdorff距离概念,对建筑物提取方法进行改进。本文主要改进内容为：① 基于Unet网络结构,利用金字塔池化模块 (Pyramid Pooling Module, PPM )的多尺度场景解析特点,残差模块(Residual Block, RB)的特征提取能力以及卷积块注意力模块(Convolutional Block Attention Module, CBAM)对空间信息和通道信息的平衡能力。将金字塔池化、残差结构以及卷积块注意力模块引入到Unet模型中,建立PRCUnet模型。PRCUnet模型更关注语义信息和细节信息,弥补Unet对小目标检测的欠缺;② 基于最小外接矩形与Hausdorff距离,改进建筑物轮廓优化算法,提高模型的泛化能力。实验表明,本文的建筑物提取方法在测试集上准确率、IoU、召回率均达到0.85以上,精度显著优于Unet模型,提取出的建筑物精度更高,对小尺度及不规则的建筑物有较好的提取效果,优化后的建筑物轮廓更接近真实的建筑物边界。     

基于支持向量机的遥感图像建筑物识别与分类方法研究

摘 要:高分辨率遥感影像可以更精细地描述地物目标的几何特征、空间特征和纹理特征等信息,在各个领域中都得到广泛的应用。建筑物作为地物信息分类中的主要部分,是地形图成图的主要组成元素,对建筑物的识别与提取,直接影响到地物测绘的自动化水平,对它的识别和定位可以为特征提取、特征匹配、图像理解、制图和作为其他目标的参照体有重要的意义。针对建筑物的遥感影像特征,研究了支持向量机分类器(SVM)在建筑物识别与分类中的应用,提出了一种交叉验证的方法对参数敏感度进行分析,通过使用GridSearch算法确定模型参数设置的最优方案,并对分类结果中建筑物进行轮廓提取。通过实验表明,优化后的SVM算法对建筑物的分类精度达到90％,对比随机森林算法、最近邻分类器优势非常明显。      

车载激光扫描数据中建筑物立面快速提取

本文提出一种结合多种投影影像从车载激光扫描数据中提取建筑物的方法。该方法首先将点云数据投影到XOY平面,生成多种投影影像;然后结合建筑物几何语义特征,对已获取的投影影像进行几何约束与形态学运算,得到建筑物种子区域;在此基础上,通过设置高差阈值,在最高高程影像上进行建筑物种子区域的八邻域区域生长,得到建筑物区域;最后将影像上的建筑物区域反投影到三维空间,提取出建筑物目标。实验结果表明,该方法能有效提取点云数据中的建筑物立面,取得较高的正确率和完整率,且大大提高了计算效率。     

融合LiDAR数据与高分影像特征信息的建筑物提取方法

建筑物是城市环境中的主要地物类型,从高分影像等数据中自动提取建筑物对于提升土地利用变化检测、城市规划与土地执法等业务的质量与效率具有重要意义。本文针对现有建筑物提取方法存在的边界提取不精确的问题以及采用手工特征表达图像信息的局限性,融合LiDAR数据与高分影像两种数据源的特征信息,提出一种基于SegNet语义模型的建筑物提取新方法。首先,对LiDAR数据预处理得到数字表面模型（DSM）、数字地形模型（DTM）、归一化数字表面模型（nDSM）,利用高分影像NDVI值去除nDSM中部分树木点,得到结果影像nDSM_en;其次,分别获取LiDAR数据回波强度、表面曲率以及高分影像NDVI值 3个特征构建特征图像训练SegNet语义模型,利用训练得到的模型完成建筑物初始提取;最后,采用阈值法分割nDSM_en得到影像对象,利用影像对象约束建筑物初始提取结果,完成建筑物精提取。在以ISPRS 官方提供的标准数据集（数据采集的地理区域为德国Vaihingen,采集时间2008年7—8月）为样本的实验中,本文方法在像素层次的平均查全率、平均查准率和提取质量分别为96.4%、94.8%和91.7%;针对面积大于50 m ²的建筑物对象,上述3个指标均为100%。实验结果表明：本文提出与实现的建筑物提取方法更好地利用了反映建筑物与非建筑物本质差异的特征信息,有效地实现了2种数据源的相对优势互补,提高了建筑物的检测与提取精度。     

基于深度学习的城市建筑物阴影提取方法

城市建筑物阴影的存在增加了基于高分影像信息提取的复杂度,影响了高分影像的行业应用。以GF-2影像为数据源,以影像覆盖的两个典型阴影区域为实验区,设计了一种基于深度学习技术的城市建筑物阴影提取方法。借助于全卷积神经网络(FCN)对图像进行语义分割,获取阴影信息,通过建立光照方向相邻对象的空间关系,去除建筑物阴影信息中的干扰项。将提取的建筑物阴影结果与面向对象方法提取的结果进行对比,在实验区1和实验区2中,DBSE提取的总体精度分别为97.5%和98.1%,面向对象的总体精度分别为96.2%和97.2%。提升精度的同时,在视觉效果上,DBSE方法在阴影完整性与真实阴影一致性,以及减少阴影错分等方面明显优于OBSE方法。     

基于图分割的高分辨率遥感影像建筑物变化检测研究

建筑物是城市地理数据库中最容易发生变化和最需要更新的部分,其更新工作量巨大,因此开展对高分辨率遥感影像中的建筑物进行自动提取和变化检测研究具有重要的意义.本文以精确提取变化建筑物的位置和轮廓为目标,基于图分割提出一种高分辨率遥感影像建筑物变化检测方法.首先,将遥感影像中的每个像元映射成图的顶点,利用像元之间的距离阈值构造图的边,综合利用位置,灰度和边缘3种特征计算边的权值,将遥感影像的分割转化为图的分割,并用归一化图分割方法得到分割对象集合;然后,以长宽比和矩形度作为约束条件,对2期遥感影像中的分割对象集合进行筛选,提取建筑物对象;最后,根据2期影像中建筑物之间的空间,面积和格局关系识别建筑物的变化类型(包括新增,消失和改建),并对其进行可视化表达.为了验证本文方法的有效性,分别以深圳市的WorldView影像和北京市的QuickBird全色影像为数据源,从中选取13组具有代表性的子影像进行实验.结果表明,本文提出的方法对配准精度较低的影像组具有一定的适应性,容许的配准误差达到20个像元(10 m),平均查准率和平均查全率分别达到93.16%和87.90%.     

基于改进主动轮廓模型的无人机影像矿区地裂缝提取

矿区地裂缝精准识别对防灾、减灾和生态环境修复具有重要意义。针对高分辨率无人机影像较难自动精确提取地裂缝的问题,本文提出了一种基于改进主动轮廓模型的无人机影像矿区地裂缝提取方法。首先,采用Otsu算法计算背景和地裂缝初值作为先验知识;其次,构建背景和地裂缝初值的提取能量函数,并引入到传统CV主动轮廓模型,增强地裂缝提取的针对性;最后,通过轮廓的不断演化实现地裂缝的提取。以内蒙古扎赉诺尔矿区为研究区、无人机影像为数据源,采用改进主动轮廓模型方法进行地裂缝提取,并与传统的Canny边缘检测算法、支持向量机（SVM）、最大似然（MLM）和传统CV主动轮廓模型方法进行对比分析。结果表明：在地物类型较为单一的小范围区域,传统的Canny边缘检测算法和传统CV主动轮廓模型提取效果较差,改进主动轮廓模型、SVM和MLM共3种方法均可以取得较好的效果,其中,改进主动轮廓模型方法精度最高;在地物类型相对复杂的大范围区域,传统的Canny边缘检测算法、SVM、MLM和传统CV主动轮廓模型方法存在较多的漏提和误提,Kappa系数均低于0.7,而本文改进主动轮廓方法依然可以取得较好的效果,Kappa系数达到0.9左右。因此,本文提出的方法通过引入先验知识可有效提高地裂缝提取的精度和稳定性。     

融合高度特征的高分遥感影像多尺度城市建筑类型分类

城市区域建筑类型信息在城市功能区识别、城市环境变量反演等应用领域具有重要作用。本文提出一种融合高分辨率遥感影像高度特征的多尺度城市建筑类型分类方法。首先利用语义分割模型识别高分辨影像中建筑和阴影对象;然后借助建筑对象及其阴影信息在卫星成像时的几何关系估算建筑高度;最后基于多尺度图像分析思想,提取一系列表征建筑对象的高度、空间结构、几何等多尺度特征,利用机器学习方法进行建筑类型分类,并进一步分析不同粒度的建筑类型分析单元对分类结果的影响。选取福州市主城区国产高分二号高分辨率影像进行实验验证。结果表明：① 基于所提方法的建筑类型分类总体精度达到82.98%, kappa系数为0.77,分类精度优于本文中未加入高度信息的分类方法和单一尺度分类方法;② 引入高度特征有效提高了中低层居民楼和高层商住两用建筑类型的分类精度,较未加入高度特征的分类结果,总体精度提高了11.28%;③ 融合多个尺度的图像特征可有效减少粘连建筑误分为密集型建筑的情况,较单一尺度分类方法,总体精度提高了2.77%。在精细的数字表面模型数据缺失下,利用高分辨影像阴影信息可为建筑物高度估计提供一种有效的策略,提高城市建筑类型分类精度。此外,融合多粒度图像特征可提升城市区域复杂建筑类型的表征能力,进而提高分类精度。     

Preliminary study on plasma membrane fluidity of Psychrophilic Yeast Rhodotorula sp.NJ298 in low temperature

The ability of cell to modulate the fluidity of plasma membrane was crucial to the survival of microorganism at low temperature. Plasma membrane proteins, fatty acids and carotenoids profiles of Antarctic psychrophilc yeast Rhodotorula sp. NJ298 were investigated at -3 ℃, 0 ℃ and 8 ℃. The results showed that plasma membrane protein content was greater at -3 ℃ than that at 8 ℃, and a unique membrane polypeptide composition with an apparent molecular mass of 94.7 kDa was newly synthesized with SDS-PAGE analysis; GC analysis showed that the main changes of fatty acids were the percentage of unsaturated fatty acids (C18∶ 1 and C18∶ 2) and shorter chain saturated fatty acid (C10∶ 0) increased along with the decrease of the culture temperature from 8 ℃ to -3 ℃; HPLC analysis indicated that astaxanthin was the major functional carotenoids of the plasma membrane, percentage of which increased from 54.6±1.5% at 8 ℃ to 81.9±2.1% at -3 ℃. However the fluidity of plasma membrane which was determined by measuring fluorescence anisotropy was similar at -3 ℃, 0 ℃ and 8 ℃. Hence these changes in plasma membrane's characteristics were involved in the cellular cold-adaptation by which NJ298 could maintain normal plasma membrane fluidity at near-freezing temperature.     

Information for authors

正1 Aims and Scope Advances in Polar Science(APS)is an international,peer-reviewed journal jointly sponsored by the Polar Research Institute of China and the Chinese Arctic and Antarctic Administration.APS is a comprehensive academic journal dedicated to presentation of multi-disciplinary achievements in Arctic and Antarctic expeditions and research.Its primary purpose     

JOURNAL OF OCEAN UNIVERSITY OF CHINA

<正>Copyright Submfnion of a roanusatpt inks:that the work described has not been published before(exoept In the form of an abstrW or as part of a published lecture,review,or thesis);that it is not under consideration for publication elsewhere that its publication has been approved by ail     

Information for Authors

<正>The Journal of Ocean University of China(Oceanic and Coastal Sea Research)is a comprehensive academic quarterly sponsored by the Ocean University of China,published one volume per year.This journal is devoted to the publication of the theoretical and applied research results on oceanography and marine fisheries.The fieldcovered include the physical,chemical,biological,geo-     

Hydraulics and Hydrology in Mountain Areas

<正>The 35th IAHR World Congress was held in Chengdu,China from September 8 to September 132013.The title of the conference was:"The Wise Finds Pleasure in Water:Meandering through Water Science and Engineering".During the congress,1239 papers were presented in 139sessions for oral presentations and 7 sessions for poster presentations.After the Congress,a Committee was appointed to select some of the best papers presented at the Congress for publication,after a     

Sea ice thickness measurement in spring season in Bothnian Bay using an electromagnetic induction instrument

As an important component of the cryosphere,sea ice is very sensitive to the climate change.The study of the sea ice physics needs accurate sea ice thickness.This paper presents an electromagnetic-induction(EM) technique which can be used to measure the sea ice thickness distribution efficiently,and the successful application in Bothnian Bay.Based on the electromagnetic field theory and the electrical properties of sea ice and seawater,EM technique can detect the distance between the instrument and the ice/water interface accurately,than the sea ice thickness is obtained.Contrastive analysis of the apparent conductivity data obtained by EM and the value of drill-hole at same positions allows a construction of a transformable formula of the apparent conductivity to sea ice thickness.The verification of the sea ice thickness calculated by this formula indicates that EM technique is able to get reliable sea ice thickness with average relative error of only 12%.The statistic of all ice thickness profiles shows that the level ice distribution in Bothnian Bay was 0.4-0.6 m.     

Thermohaline structure inhomogeneity associated with polynia at the northern margin of Emery Ice-shelf

Based on the hydrographic data in austral summer during the 22nd Antarctic Expedition of China(2005/2006),some features can be found about the northern margin of Emery ice shelf as follows.The heat content in the surface layer(0-50 m) at the eastern end and the western end of the ice-shelf margin is much higher than that at the middle.The upper mixing-layer depth and the seasonal thermocline depth at the middle of the ice-shelf northern margin are much shallower than those at the both ends.However there is much less difference between the middle and the ends in the bottom layer.The remote sensing photos show that the inhomogeneity in the surface-layer water is closely related to the spatial distribution of the floes and polynia in the area.     

Phylogenetic analysis of cultivable bacteria isolated from Arctic sea-ice

Phylogenetic analysis based on 16S rDNA of 8 strains of cultivable bacteria isolated from Arctic sea-ice was studied.The results showed that strain BJ1 belonged to genus Planococcus,which was a genus of low mole percent G C gram-positive bacteria;strain BJ6 belonged to genus Burkholderia of β-proteobacteria and the rest 6 strain all belonged to γ-proteobacteria,of which strain BJ8 was a species of Pseudoalteromonas,strain BJ2-BJ5 and BJ7 were members of genus Psychrobacter.Phylogenetic analysis also indicated that bacteria of genus Psychrobacter of the isolates formed a relatively independent phylogenetic cluster in comparison with other bacteria belonged to genus Psychrobacter.     

Foreword

正Rapid changes of Arctic sea ice cover have been in the focus of the international climate research community in recent years.Quite a few of nations have completed a large number of related surveys and research projects in the Arctic Ocean.Up to now,China has performed six research cruises to the Arctic Ocean resulting in a significant volume of research output.Improved knowledge on the atmospheree-sea ice-ocean interactions in the Arctic is a