基于FPN Res-Unet的高分辨率遥感影像建筑物变化检测

针对城市土地资源变化检测工作繁杂、工作量大、自动化程度低等问题，本文提出一种基于深度学习模型的高分辨率遥感影像建筑物变化检测方法，将语义分割的思想引入到遥感变化检测。基于残差结构特征较卷积层提取性能更优和特征金字塔网络多尺度预测的特点，将残差结构和特征金字塔网络融合到Unet模型中，建立FPN Res-Unet模型。该模型以Unet为基础，引入ResNet18的残差结构作为编码路径特征提取层，在每次卷积后使用边界填充，使得输入图像和输出图像尺寸一致；在解码路径每级上采样过程中，拓展支路径将特征金字塔网络融合到模型的网络主干中，将残差结构、Unet及特征金字塔网络的优点相互融合，增强了Unet的特征提取，弥补了语义分割网络对小目标检测的欠缺；在获取深层语义信息的同时关注细节信息，提高建筑物变化检测精度。实验表明，该方法在所用数据集，准确率、召回率、F₁ 3种指标均达到90%以上。     

Global climatic change – the latest scientific understanding

IPCCs statement in its 1995 report (IPCC 1996) that a human influence was discernible in global climate has been widely quoted but often misunderstood. The character of the evidence underpinning this detection statement is explained so that its strengths and weaknesses can be better understood and the subtleties of its message better appreciated. To demonstrate the close linkage between the government-approved summary and the underlying chapters of the IPCC report the detailed evolution of the detection statement from first draft through to the form finally approved by the IPCC is described.     

Numerical study of droplet size dependent chemistry in oceanic,wintertime stratus cloud at southern mid-latitudes

Cloud droplet chemistry is modelled for the first 150 m of rise in a wintertime, mid-latitude, marine stratus cloud using observations made at and near the Cape Grim Baseline Station as a source of input parameters. The emphasis in this work was to study the variation in droplet chemistry as a function of both droplet size and nucleus composition, with a particular focus on the way in which oxidation of dissolved sulfur dioxide varied.At 150 m above the condensation level, solute concentration as a function of droplet size was found to increase by as much as 2 to 3 orders of magnitude for only a factor of 2 increase in droplet radius, primarily as a consequence of the 1/r dependence in the droplet growth equation. This type of size dependence exists at all levels in the model cloud, and has a significant influence on oxidation rate of sulfur dioxide in droplets growing on sulfate nuclei, oxidation by ozone being favoured in the smallest droplets, but oxidation by hydrogen peroxide being favoured in the larger droplets. Oxidation by ozone is favoured at all sizes in droplets formed on sea-salt nuclei as a result of the initially high alkalinity of these droplets, and in the cloud overall is calculated to be the more important oxidation pathway. Although based on a simplified chemical scheme, these results suggest that both size-dependent and nucleus-dependent chemistry of cloud droplets may need to be considered explicitly in cloud modelling work.Volume-weighted mean pH values in the range 5 to 6 were predicted from sensitivity studies in which input variables were varied over reasonable ranges, in agreement with two sets of bulk cloud-water pH data obtained by aircraft near Cape Grim.     

圆形标志中心子像素定位方法的研究与实现

采用回光反射材料制作圆形回光反射标志，经摄影后得到标志的“准二值影像”；对标志的椭圆影像进行亚像素边缘提取，经最小二乘椭圆拟合后得到标志中心的子像素级位置。试验表明，可以达到0．02像素的定位精度。     

中国MODIS卫星影像彩色镶嵌图制作

本文介绍了大范围卫星影像彩色镶嵌图的制作流程．对于大范围卫星影像彩色镶嵌图制作的技术难点（去云、拼接、调色），文章进行了重点介绍并提出技术解决办法．利用MODIS卫星数据制作了两幅全国彩色影像镶嵌图：一幅为红波段（B1）、绿波段（B4）和蓝波段（B3）构成的真彩色合成影像：一幅为红波段（B1）．近红外波段（B2）和蓝波段（B3）构成的假彩色合成影像．两幅影像色彩精美，信息量丰富．     

基于预测的边缘检测方法

提出了一种基于预测的、有一定自适应性的边缘检测方法。介绍了其原理和算法实现过程。通过计算两个相邻像素的灰度值的均值、均方差和梯度来预测下一个像素的灰度值，比较预测值和真实值来判断下一个像素是否是边界点。通过与现有同类算法的比较，证明了该算法的优越性。     

多时相组合分类法在土地利用动态监测中的应用

介绍了土地利用遥感动态监测的基本概念，简述了遥感土地利用变化信息提取等遥感监测方法．重点探讨了多时相组合分类方法的相关技术。对广西2002年和2003年两个时相的MODIS数据．采用多时相直接分类法对土地利用变化状况进行了遥感动态监测。对不同方式波段组合的试验表明。经过差值、比值处理的波段组合具有较差的试验效果(总体精度只有30％～40％)，而经过PCA变换的波段组合则具有相对较好的试验效果(总体精度超过70％)。     

探地雷达在吉林西部地区探测土壤碱化层

吉林西部盐碱地区的探地雷达调查表明,盐碱地与耕地土的雷达回波具有明显差别,根据振幅的衰减特征可以精确区分耕地与盐碱地,根据频谱中高频峰值与土壤有机质含量的相关性可定性评价土壤的盐碱化程度.在雷达瞬时振幅图上,可将盐碱地纵向剖面分为盐分积累带和过渡带.采用探地雷达方法可快速了解盐碱化程度在深度方向上的变化,适宜开展大面积地面调查.     

隧道锚杆锚固质量无损检测技术

采用声波检测仪,应用声频应力波在不同波阻抗面反射的能量和相位的变化原理,对隧道的锚杆锚固质量进行了无损检测试验研究,并对检测技术进行了探讨.结果表明,作为一种工程质量管理辅助手段,采用应力波对锚杆锚固质量进行无损检测,丰富了隧道围岩锚固质量检测方法,为隧道工程建设提供更好的质量保障.