基于约束满足神经网络的整体影像匹配

将影像匹配看作一个约束满足问题,并用约束满足社会网络来实现整体影像匹配,根据新松驰标号法对网络的结构和迭代方式进行了蛤 能够处理复杂地形条件下影像匹配中存在的“零匹配”和“多匹配”问题。实验表明,该匹配算法可快速,有效地处理复杂地形条件下的影像匹配问题。     

基于空间约束的神经网络影像匹配

提出了一种可充分利用影像灰度信息和特征信息的基于视差变权空间连续性约束的整体影像匹配方法,并将其映射到 Ｈｏｐｆｉｅｌｄ网络,利用神经网络所固有的大规模并行处理能力,可以高速、可靠、高精度地完成影像匹配,并大大提高了复杂地形条件下影像匹配的可靠性。实验证明该算法具有一定的实用性。     

基于空间约束的神经网络影像匹配

提出了一种可充分利用影像灰度信息和特征信息的基于视差变权空间连续性约束的整体影像匹配方法,并将其映射到Ｈｏｐｆｉｅｌｄ网络,利用神经网络所固有的大规模并行处理能力,可以高速、可靠、高精度地完成影像匹配,并大大提高了复杂地形条件下影像匹配的可靠性。实验证明该算法具有一定的实用性。     

半全局约束下的多基线立体影像MVLL匹配方法

多视铅垂线轨迹法（multi-view vertical line locus,MVLL）能够以物方地面铅垂线为几何约束,匹配获取物方空间点的最佳高程,是一种实用的多基线立体影像匹配方法。针对MVLL匹配方法中的物方空间点独立匹配、缺少整体性约束问题,将物方空间局部光滑特性应用至匹配过程,提出了一种半全局约束下的多基线立体影像MVLL匹配方法。首先,将物方空间点沿地面铅垂线上的准确高程搜索,等效至像方空间沿核线方向上的准确视差搜索;其次,采用MVLL匹配方法计算物方空间点在多张影像上的等效像方匹配测度;然后,采用半全局匹配法（semi-global matching, SGM）对匹配测度进行多路径聚合分析,得到物方空间局部光滑约束下的等效视差图;最后,将等效视差图转化为物方空间点匹配高程,并对多基线立体影像进行多分辨率匹配处理,实现整体最优条件下的MVLL匹配。采用多类型地形特征与局部区域影像进行匹配实验与对比分析,实验结果表明,所提方法能够对多类型地形特征的物方空间匹配测度进行优化,获取更加可靠的匹配结果,具有更高的影像匹配性能。     

运动平滑与核线约束的特征匹配提纯算法

针对基于最近邻距离比率约束和几何条件约束的影像特征匹配方法,在处理重复纹理影像以及视角和尺度变化大的宽基线影像时,存在匹配正确率低和匹配点对数量少的问题,设计了结合运动平滑约束与核线约束的AKAZE特征匹配提纯算法。①为提高影像特征的不变性、显著性和时效性,采用AKAZE算子提取影像特征,经暴力匹配快速构建初始匹配集;②基于网格运动统计区分真假匹配,剔除不能满足运动平滑约束的匹配点对;③为消除局部相似特征引起的误匹配,采用核线约束提高匹配纯度。实验结果表明:所提算法实现了重复纹理影像、倾斜影像和宽基线影像同名点对的高精度匹配,增加了匹配点对数量,拓展了特征匹配的适用范围。     

多重约束条件下的影像直线匹配算法研究

针对直线匹配中“一配多”“多配多”难点问题,文章提出多重约束条件下的影像直线匹配算法.该方法利用直线邻域范围内的同名点计算每一个待匹配直线的单应性矩阵,然后利用单应矩阵将目标影像上待匹配直线映射到搜索影像,根据距离约束确定候选直线,依据直线段倾角相似度等约束再在候选直线中进一步确定可能的同名直线;并利用核线约束条件和灰度相似性约束确定最终的同名直线.最后分别采用无人机影像和UCX数字航空影像进行实验,验证了本文方法的有效性.     

结合离散化描述与同名点约束的线特征匹配

由于受到多重因素的影响,影像中提取的线特征经常出现断裂、变形、遮挡等问题。为克服上述问题对线特征匹配的影响,本文提出一种结合离散化描述与同名点约束的航空影像线特征匹配算法。算法首先利用单应性约束结合核线约束精简候选线特征数量;然后采用“以点代线”的方式,通过统计线上同名点的分布情况确定线特征初匹配结果;最后根据线特征与其邻域内同名点间的距离关系对匹配结果进行核验,以保证匹配的可靠性。选取AMC580相机系统所获得的航空影像进行线特征匹配试验,结果表明,本文算法匹配正确率高,匹配速度相对较快,并可实现曲线匹配及断裂线特征的多对多匹配。与已有匹配算法进行比较,本文算法在影像间存在有尺度、亮度、旋转变换等复杂条件下的匹配正确率均高于90%,具有较好的匹配稳定性。     

SURF与归一化互相关测度相结合的SAR影像匹配方法研究

合成孔径雷达(SAR)匹配是SAR影像处理的重要环节,但由于SAR影像的特殊成像机理,导致SAR影像在匹配时会出现精度、成功率都较低等问题。特别是山地区域,SAR影像富含纹理信息稀少,影像间几何畸变和辐射变形严重,导致特征点提取困难,匹配精度低。鉴于此,本文提出了采用SURF算法和归一化互相关测度相结合的方法,该方法能够快速获得大量匹配点对,实现SAR影像高效的自动化匹配。首先,应用SURF算法获取影像特征点约束匹配搜索范围;然后,采用归一化互相关和测度筛选提取到的匹配点对;最后,利用双向一致性约束剔除错误匹配点对。本文选取地形复杂区域SAR影像数据进行匹配实验,结果表明,该方法能够获得高精度、特征稳定且分布均匀的匹配点对,可以有效地避免几何变形对SAR影像匹配的影响。     

一种基于三角网约束的立体影像线特征多级匹配方法

针对直线匹配可靠性问题,提出了一种基于三角网约束的立体影像线特征多级匹配方法。首先采用SURF算法匹配一部分可靠的种子点,利用这些种子点约束其邻域内的直线匹配;然后,将这些种子点构建三角网,利用三角网约束直线匹配的搜索范围,进行三角网约束下的线-线匹配;再次,在三角网的约束下,进行线-面匹配。为了提高直线匹配相似性测度的可区分性,提出了基于移动窗口的自适应直线相关方法,不仅在表面非连续区域能取得可靠匹配,在纹理缺乏区域也能取得可靠匹配结果。利用具有典型纹理特征的近景影像和航空影像进行了试验分析,结果表明,本方法能获取可靠的直线匹配结果。     

面向无人机倾斜影像的高效SfM重建方案

针对无人机影像分辨率高、数据量大导致稀疏重建效率低的问题，提出了一种减少影像匹配对数量、提高外点剔除效率的算法。首先利用无人机飞控数据、相机安装角计算影像的粗略POS（positioning andorientation system）信息；然后基于拓扑连接分析设计了最大生成树算法（maximum spanning tree expansion，MST-Expansion），最大程度地降低影像配对数；考虑到初始匹配的高外点率，设计了分层运动一致性约束（hierarchical motion consistency constraint，HMCC）算法，提高几何验证算法的效率。4组不同倾斜设备采集的无人机影像的实验结果说明，该方案能够在保证重建精度的前提下，实现无人机倾斜影像高效和稳健的稀疏重建。     

What about people in pedestrian navigation?

Pedestrian navigation has become an important theoretical and practical research topic in many disciplines such as cartography, geographical information science, global and indoor positioning, spatial behavior, psychology, sociology, and neuroscience. Many research studies view pedestrian navigation using process-oriented and goal-directed approaches. However, this paper revisits people’s needs in pedestrian navigation and classifies their needs as three layers: physical sense layer, physiological safety layer, and mental satisfaction layer according to Maslow’s theory. This paper introduces a people-centric framework for pedestrian navigation theory based on these three layers and discusses theoretical challenges for meeting each layer of people’s needs. These challenging theories may represent promising and valuable research and promote usage of pedestrian navigation systems or devices in the future.     

高分卫星自然资源调查需求满足度评估

当前高分对地观测数据应用来源众多,且其应用方式各异,如何对需求其进行规范化和结构化的需求综合及聚类,进而开展卫星观测能力对于用户需求的需求满足度进行评估,是当前高分对地观测系统建设及应用的关键.本研究基于层次分析法开展自然资源遥感调查的需求建模与评价研究,以可见光近红外类型的自然资源遥感调查需求为主要研究对象,提取自然...     

阿克苏河-塔里木河水系水质污染遥感监测与评价

 采用Landsat TM数据分析了阿克苏河—塔里木河断面水质污染状况,通过波段的DN值和常规监测数据建立能反映水质状况的 污染物监测模型。结果发现,将2000年常规监测数据代入模型后,与遥感数据的结果基本吻合| 重建阿克苏河—塔里木河的连续水体 污染变化曲线,得出污染物浓度随着远离上游而增加。     

Monitoring drought dynamics in the Aravalli region (India) using different indices based on ground and remote sensing data

The hard-rock hilly Aravalli terrain of Rajasthan province of India suffers with frequent drought due to poor and delayed monsoon, abnormally high summer-temperature and insufficient water resources. In the present study, detailed analysis of meteorological and hydrological data of the Aravalli region has been carried out for the years 1984–2003. Standardised Precipitation Index (SPI) has been used to quantify the precipitation deficit. Standardised Water-Level Index (SWI) has been developed to assess ground-water recharge-deficit. Vegetative drought indices like Vegetation Condition Index (VCI) and Temperature Condition Index (TCI) and Vegetation Health Index (VHI) have been computed using NDVI values obtained from Global Vegetation Index (GVI) and thermal channel data of NOAA AVHRR satellite. Detailed analyses of spatial and temporal drought dynamics during monsoon and non-monsoon seasons have been carried out through drought index maps generated in Geographic Information Systems (GIS) environment. Analysis and interpretation of these maps reveal that negative SPI anomalies not always correspond to drought. In the Aravalli region, aquifer-stress shifts its position time to time, and in certain pockets it is more frequent. In comparison to hydrological stress, vegetative stress in the Aravalli region is found to be slower to begin but quicker to withdraw.     

Landslide risk analysis between Giri and Tons Rivers in Himachal Himalaya (India)

The area around Sataun in the Sirmur district of Himachal Pradesh, India (falling between the rivers Giri and Tons; both tributaries of the Yamuna River) was studied for landslide vulnerability on behalf of the inhabitants. The study was made using extensive remote sensing data (satellite and airborne). It is well supported by field evidence, demographic and infrastructural details and aided by Geographic Information System (GIS) based techniques. Field observations testify that slope, aspect, geology, tectonic planes, drainage, and land use all influence landslides in the region. These parameters were taken into consideration using the statistical approach of landslide hazard zonation. Using the census data of 1991, vulnerability of the populace to the landslide hazard was accessed. As most of the infrastructure in the region is concentrated around population centres, population data alone was used for vulnerability studies.     

Gravity Changes Before Large Earthquakes in China: 1998-2005

Although it is well known that coseismic gravity changes take place during an earthquake, previous research has not yielded convincing evidence demonstrating that significant gravity changes occur before large earthquakes. Furthermore, even if we suspect that gravity changes occur before large earthquakes, we have yet to demonstrate how to consistently observe these changes for useful earthquake forecast that would bring benefits to society. We analyzed ground gravity survey data obtained in 1998, 2000, 200...     

GNSS系统及其技术的发展研究

阐述了国际上卫星导航系统的总体发展,对美、俄、欧、中的四大全球系统,以及相关的区域系统和增强系统进行了简单的介绍,研究分析了GNSS系统及其技术的发展趋势,同时分析了我国Compass全球系统面临的机遇和挑战,并提出了我国未来工作的建议.     

“4D”技术及其应用

介绍“４Ｄ”（即数字地面模型、数字正射影像、数字栅格地图和数字线划地图）系列产品的生产流程、关键技术、及其在耕地保护、防灾减灾救灾和城市规划等方面的应用方法。     

Virtual Australia: its key elements – know,think, communicate

Abstract

Virtual Australia is not a well-defined or agreed concept. This discussion paper conceptualises Virtual Australia as a ‘virtual [digital] model containing and representing all non-trivial objects and their contextual environment – from blue sky to bedrock – in real world Australia’. It describes a scenario for Virtual Australia, one or two decades from now, in which the locations and conditions of non-trivial objects and their environment are updated automatically through a combination of remote sensing and wireless communication technologies in support of a ‘Supranet’. It then examines the concept of the ‘Supranet’ – a pervasive information network based largely on wireless technology linking the physical world to a virtual model in real time – and develops and discusses three principal characteristics of the Supranet: the ability to have or collect specific information (know); the ability to process information (think); and the ability to communicate that information. If, in the near future, any or all non-trivial devices are to some extent able to know, think and communicate, the potential for object autonomy will be realised.