抗视差的宽基线弱纹理影像自动拼接算法

针对目前带有视差突变的宽基线弱纹理影像拼接效果差及需要人工干预的问题,本文从影像匹配和影像配准两方面进行改进,提出了面向宽基线弱纹理影像的抗视差全自动拼接算法。首先,采用融合了影像视角几何纠正的局部特征变换模型,由粗到精地实现弱纹理特征的准密集对应;然后,基于匹配点和深度神经网络,泛化学习宽基线影像间的可靠透视变换,以获取全局配准视差,局部视差则通过薄板样条函数来精确拟合;接着,将影像拼接结果的多边形边界进行规则化处理,通过全卷积网络将其训练为规则化矩形,在有效剔除空白区域的同时,最大限度地保留影像拼接内容;最后,选取4组无人机和地面近景宽基线弱纹理立体像对进行测试,并将本文算法的影像匹配及配准各阶段结果分别与现有代表性算法结果进行对比。试验结果表明,本文算法在匹配点数目、匹配精度及影像拼接质量等方面具有显著优势,并能够在影像弱纹理区域及视差突变场景表现出较好的稳定性。

智慧城市综合感知与智能决策的进展及应用

城市感知与决策是实现城市智慧化的首要前提。针对当前的最新研究进展,分别从基于空天地集成化传感网的城市综合感知、城市时空大数据的管理与分析、面向多层次用户的智能决策等3方面进行了分析论述,对比分析了当前城市时空信息平台及应用,在此基础上,提出了未来智慧城市感知决策面临的“三高”挑战,即高度融合的物理感知与社会感知,高度智能化的城市管理分析能力,以及高置信度的城市信物融合系统决策。

遥感影像智能解译样本库现状与研究

我国遥感对地观测等项目顺利实施,获取了大量时效性强、覆盖范围广、信息量丰富的遥感数据。但遥感影像智能化自动处理技术发展仍相对滞后,无法满足区域/全球大范围地物信息快速提取的需求。近年来,人们利用深度学习技术显著提高了影像特征提取成效,但由于所使用的深度学习样本数量和类型有限,对于多源遥感影像的自动解译能力仍然不足。本文面向大范围多源遥感影像地物信息智能解译需求,在分析现有样本集现状及问题的基础上,研究提出遥感影像智能解译样本库设计方案,并在此基础上设计了基于互联网的样本协同采集与共享服务框架。本文将为多源遥感影像样本库建设提供参考,为大范围遥感影像智能解译提供支持。

智能遥感深度学习框架与模型设计

近年来,随着遥感技术的快速发展,遥感对地观测数据获取量与日俱增。在对海量遥感数据的特征提取与表征上,基于深度学习的智能遥感影像解译技术展现出了显著优势。然而,遥感影像智能处理框架和信息服务能力还相对滞后,开源的深度学习框架与模型尚不能满足遥感智能处理的需求。在分析现有深度学习框架和模型的基础上,针对遥感影像幅面大、尺度变化大、数据通道多等问题,本文设计了嵌入遥感特性的专用深度学习框架,并重点讨论了其构建方法,以及地物分类任务的初步试验结果等。本文提出的智能遥感解译框架架构将为构建具备多维时空谱遥感特性的深度学习框架与模型提供有力支撑。

影像解译中的深度学习可解释性分析方法

深度学习的迅速发展极大地推进了多种图像解译任务的精度提升,然而深度学习网络模型的“黑箱”性质让使用者难以理解其决策机理,这不仅不利于模型结构优化和安全增强等,还会极大地增加训练调参成本。对此,本文围绕影像智能解译任务,对深度学习可解释性国内外研究进展进行了综合评述与对比分析。首先,将当前可解释性分析方法分为激活值最大化分析法、代理模型分析方法、归因分析法、扰动分析法、类激活图分析法及样例分析法等6类方法,并对方法的原理、侧重点及优缺点进行了回顾。其次,对8种衡量各类分析方法所提供解释的可靠性的评估指标进行了回顾,并梳理了当前公开可用的可解释性开源算法库。在当前开源算法库的基础上,以遥感影像智能解译任务中的可解释性分析为例,验证了当前深度学习可解释性方法对遥感影像的适用性;试验结果表明当前可解释性方法在遥感解译中还存在一定的局限性。最后,总结了现有基于自然影像的可解释性算法在遥感影像解译分析中存在的问题,并展望了设计面向遥感影像特性的可解释性分析方法的发展前景,旨在为相关研究者提供参考,推动面向遥感影像解译的可解释性方法研究,从而为深度学习技术在遥感影像解译任务中的应用提供可靠的理论支持与算法设计指导。

DATA MODEL AND VISUALIZATION OF 3D CITY LANDSCAPE BASED ON INTEGRATED DATABASES

Data acquisition and modeling are the two important, difficult and costful aspects in a Cybercity project. 2D-GIS is mature and can manage a lot of spatial data. Thus 3D-GIS should make the best of data and technology of 2D-GIS. Construction of a useful synthetic environment requires integration of multiple types of information like DEM, texture images and 3D representation of objects such as buildings. In this paper, the method for 3D city landscape data model and visualization based on integrated databases is presented. Since the data volume of raster are very huge, special strategies(for example, pyramid gridded method) must be adopted in order to manage raster data efficiently. Three different methods of data acquisition, the proper data structure and a simple modeling method are presented as well. At last, a pilot project of Shanghai Cybercity is illustrated.

A DISTRIBUTED HYPERMAP MODEL FOR INTERNET GIS

The rapid development of Internet technology makes it possible to integrate GIS with the Internet,forming Internet GIS.Internet GIS is based on a distributed client/server architecture and TCP/IP & IIOP.When constructing and designing Internet GIS,we face the problem of how to express information units of Internet GIS.In order to solve this problem,this paper presents a distributed hypermap model for Internet GIS.This model provides a solution to organize and manage Internet GIS information units.It also illustrates relations between two information units and in an internal information unit both on clients and servers.On the basis of this model,the paper contributes to the expressions of hypermap relations and hypermap operations.The usage of this model is shown in the implementation of a prototype system.

DESIGN AND IMPLEMENTATION OF AN INTERNET GIS

The development of Internet provides convenient environment for information searching and browsing.It also offers a new platform for geographic information processing and analysis.This paper discusses organization and processing approach to Internet geographic information,and provides a new method to cross_platform and distributed geographic information and its software implementation.The example is used to show the practicality and superiority of this method.

AN INTEGRATED DATA MODEL IN THREE DIMENSIONAL GIS

The current GIS can only deal with 2-D or 2.5-D information on the earth surface. A new 3-D data structure and data model need to be designed for the 3-D GIS. This paper analyzes diverse 3-D spatial phenomena from mine to geology and their complicated relations, and proposes several new kinds of spatial objects including cross-section, column body and digital surface model to represent some special spatial phenomena like tunnels and irregular surfaces of an ore body. An integrated data structure including vector, raster and object-oriented data models is used to represent various 3-D spatial objects and their relations. The integrated data structure and object-oriented data model can be used as bases to design and realize a 3-D geographic information system.

AN ALGORITHM FOR AUTOMATIC CARTOGRAPHIC SOUNDING SELECTION

Soundings on nautical charts provide information about the shape of ocean bottom between chart depth curves. A single chart may have thousands of soundings posted on it. This paper describes a new algorithm for selecting soundings automatically. At first,the problem encountered in automatic cartographic sounding selection is presented. Then the authors give the principle of the new algorithm,and the implementation is illuminated in detail. At last the experiments prove the rationality and efficiency of the new algorithm.

Computer-Assisted DEM-Based Aero Photography Design

In aerial photography, the primary factor is terrain undulation. However, most of the external aerial photography software used for aerial photography design do not take terrain undulation influence into consideration. Therefore, the design result has comparative randomicity and ＂gaps＂ are expected. An aerial photography design system is developed by analyzing the terrain undulation influence to the design result with DEM data so that the forward overlap and side overlap can be justified according to the block terrain undulation to meet specifications or standards. The data designed by this system is compared with the real flying data. The results show that making use of DEM to assist in aerial photography design can ensure that the designed result fits the real terrain better.

MULTI-RESOLUTION SEAMLESS IMAGE DATABASE

This paper presents the basic concepts and principles,data structure and high efficient spatial index for multi_resolution image database.The database is characterized by arrangement of multi_resource image data and seamless mosaic,distribution_based storage and management,integration with other spatial database software such as GeoStar and GeoGrid developed by Wuhan Technical University of Surveying and Mapping.

Geospatial Semantic Interoperability Based on Ontology

In GIS field, great varieties of information from different domains are involved in order to solve actual problems. But usually spatial information is stored in diverse spatial databases, manipulated by different GIS platforms. Semantic heterogeneity is caused due to the distinctions of conception explanations among various GIS implements. It will result in the information obtaining and understanding gaps for spatial data sharing and usage. An ontology-based model for spatial information semantic interoperability is put forward after the comprehensive review of progress in ontology theory, methodology and application research in GIS domain.

RESEARCH ON THE SECURITY OF NATIONAL GEOSPATIAL DATA CLEARINGHOUSE BASED ON ASP

On the basis of the authors’ experiences of setting up an NGDC Web site,this paper attempts to present some significant aspects about the security of NGDC based on ASP.They include data storing,database maintenance,new technical support and so on.Firstly,this paper discusses how to provide the security of data which is saved in the hosts of NGDC.The security model of "Networks_DB Server_DB_DB Object" is also presented.In Windows NT Server,Internet Information Server (i.e.,IIS) is in charge of transferring message and the management of Web sites.ASP is also based on IIS.The advantages of virtual directory technique provided by IIS are emphasized. An NGDC Web site,at the Research Center of GIS in Wuhan Technical University of Surveying and Mapping is also mentioned in this paper.Because it is only an analoge used for case study,the transmission of digital spatial products is not included in the functions in this NGDC Web site.However,the management of spatial metadata is more important and some functions of metadata query are implemented in it.It is illustrated clearly in the functional diagram of the NGDC Web site.

A DESIGN OF THREE-DIMENSIONAL SPATIAL DATA MODEL AND ITS DATA STRUCTURE IN GEOLOGICAL EXPLORATION ENGINEERING

The key to develop 3-D GISs is the study on 3-D data model and data structure. Some of the data models and data structures have been presented by scholars. Because of the complexity of 3-D spatial phenomenon, there are no perfect data structures that can describe all spatial entities. Every data structure has its own advantages and disadvantages. It is difficult to design a single data structure to meet different needs. The important subject in the3-D data models is developing a data model that has integrated vector and raster data structures. A special 3-D spatial data model based on distributing features of spatial entities should be designed. We took the geological exploration engineering as the research background and designed an integrated data model whose data structures integrats vector and raster data byadopting object-oriented technique. Research achievements are presented in this paper.

RESEARCH ON DATA CONSISTENCY IN SPATIAL DATABASE SYSTEM

In this paper,two cases of data consistency problems are discussed in Multi_user Geographical Information System.When a user updates some spatial data that have been accessed by other users in database,or executes an UNDO/REDO operation,data consistency will probably be destroyed.To solve the problem caused by the first case,a “Notification_Reread Method" is introduced.As for the second case,there are many problems to be studied.

THE INTEGRATED SPATIAL DATABASES OF GEOSTAR

GeoStar is the registered trademark of GIS software made by WTUSM in China.By means of the GeoStar,multi_scale images,DEMs,graphics and attributes integrated in very large seamless databases can be created,and the multi_dimensional dynamic visualization and information extraction are also available.This paper describes the fundamental characteristics of such huge integrated databases,for instance,the data models,database structures and the spatial index strategies.At last,the typical applications of GeoStar for a few pilot projects like the Shanghai CyberCity and the Guangdong provincial spatial data infrastructure (SDI) are illustrated and several concluding remarks are stressed.

THE CNSDTF TOPOLOGICAL VECTOR MODEL AND STRUCTURE

Geo-Spatial Data Transfer Standard is an important part of 'National Spatial Data Infrastructure(NSDI)' ,as well as a necessary means for data sharing. 'Chinese National Geo-Spatial Data Transfer Format (CNSDTF)' was approved by National Quality Technology Supervise Bureau in 1999 with the standard serial number of 17798-1999. It is designed to support vector and raster spatial data. This paper describes the vector part of CNSDTF, including design ideas, main characters, conceptual model, definition of spatial object, and file structure.

OBJECT-ORIENTED SPATIAL DATA ORGANIZATION AND MANAGEMENT IN GEOSTAR

Spatial data, including geometrical data, attribute data, image data and DEM data, are huge in volume and relations among them are complex. How to effectively organize and manage those data is an important problem in GIS. Several problems about space data organization and management in GeoStar which is a basic GIS software made in China are discussed in this paper. The paper emphasizes on object model of spatial vector, data organization, data management and how to realize the goal, and the like.

AN EFFICIENT SOLUTION FOR BUILDING POLYGONAL OBJECTS

Polygonal object is a fundamental type of geometric data in vector GIS. The key step of cleaning topological relationship after data collection of polygonal layer is to build polygonal objects from digital arcs. The raw digital arcs may intersect with each other. The algorithm for building polygonal objects after the raw arcs have been split at all intersections is presented. The build-up of polygonal objects in this paper is designed to be implemented by two steps. The first step is to extract all the polygons needed for build-up of polygonal objects from arcs. The second step is to organize polygonal objects from these polygons. For the first step,a tracing algorithm is proposed. The algorithm merely extracts the polygons needed for the build-up of polygonal objects, which is a subset of all the possible polygons that can be induced from the arcs. For the second step, an algorithm based on a specially designed order of polygons is advanced. All the topological relationships among the polygons are shown in a single scan. Experiments show that the two algorithms together offer a robust and efficient solution for building polygonal objects from intersected arcs.