This paper addresses the topic of semi-automatic object extraction for GIS data capture. It includes novel algorithms and conceptual strategies, together with performed experiments, encountered problems and adopted solutions. In particular, we present an algorithm and obtained results for semi-automatic extraction of road networks from SPOT imagery using wavelet-transformed images and cost functions expressing local gray value variations and global continuity constraints. For larger scale images and for various object types, we present our technique of least squares template matching for edge extraction, which uses local gray value variations to precisely identify edge locations. Finally, we propose a global approach for semi-automatic object outline detection, whereby least squares matching provides the mathematical foundation, while global continuity is enforced through the introduction of object-type-dependent shape constraints.
This paper discusses a simple strategy for recognising prototypical urban roads from their forms alone using two new dimensions, namely exits and occupancy. Together with average width, these dimensions appear to define the key aspects of roads and enable recognition of both prototypical and untypical cases. The latter are due to the absence of logical links, which cause roads to become combined with neighbouring regions. The recognition strategy therefore provides an additional means for validating topographic data.
There have been many interpolation methods developed over the years, each with their own problems. One of the biggest limitations in many applications is the non-correspondence of the surface with the objects used to support it — usually a set of arbitrarily distributed data points. This is due to the metric methods used to define both the zone of influence of a data point and the set of data points used to estimate surface properties at intermediate locations. Most methods are coordinate system oriented, not object oriented. We describe here an object-oriented approach, in the sense that the data objects themselves form the spatial structure used for interpolation. This has the primary benefit that the surface is described in terms of this structure, permitting easy correspondence between the surface and data values. In addition, the spatial objects used are not restricted to points, but may include more complex objects — currently composed of line segments. The structure used is the Voronoi diagram, which provides easily understood spatial relationships between map objects, as well as the local coordinate systems needed for point/surface correspondence. The methods described here use the Voronoi diagram to generate surfaces which pass precisely through all data objects (points and line segments) and have no inadvertent surface or slope discontinuities. In addition, discontinuities in the slope or the surface itself may be assigned to the line segment objects; and measures of surface consistency may be derived from the values of nearby data objects. This approach permits a consistent method of mapping a fragmented surface with both point and line segment information, and the surface may be guaranteed to be only locally perturbed by the modification of any one data object. While the concepts discussed are here applied to two spatial dimensions plus “elevation”, the Voronoi method may readily be applied to higher dimensions. The Voronoi approach, if used for traditional GIS topological structuring as well as for interpolation, provides a new way to handle an old challenge: to combine object and field data within the same GIS structure.
Global climatological data are often evaluated in grids with perpendicular boundaries in a cylinder projection of the longitude and latitude circles of the earth. Typical datasets contain 106 to 108 values. Each dataset is organized in a different grid, according to the authors's preferences. When calculating the global radiation over the ocean, for example, it is essential that cloud data from at least three different sources can be accessed based on one grid. One approach is to interpolate data from several datasets, each with a different grid, and combine them into a single dataset with an arbitrary grid.To this end, all data sets are managed with two Grid Files each: the data file containing measurements and the description file, containing the information of the area grid partition. This allows for an efficient implementation of point and range queries using the symmetric composite key (longitude, latitude, time, name of quantity) and for a reduction of storage space by factors between 2 and 10 as compared to the equivalent ASCII format; the actual reduction obtained depends on the spatial spectra of the datasets.The interpolation makes use of the description files to calculate the weights which are proportional to the areas resulting from the intersection of the grids.
Graphical representations to illustrate basic concepts of spatial access structures or to explain the behaviour of such data structures are widely used, not only in the relevant literature. Although computer-based tools to create such visualizations can be a valuable help for presentation, analysis or even for debugging of such access structures, to our knowledge no such tool is yet available. As the first approach in this direction, in this paper we present VisTool, a tool for visualizing spatial access structures. Using VisTool, users can visualize such access structures in their static state or observe their evolution during a sequence of operations by choosing some appropriate visualization method. Different spatial access structure implementations can easily be integrated into VisTool, which also serves as an operationally uniform testing environment. The user interface of VisTool has two main components: the control board and one or more Vis Tool windows. The control board allows users to interact with spatial access structures using the set of operations (e.g. create, insert, query, etc.) given by the access structure implementations. Vis Tool windows are X window front-ends which allow visualization, based on a library of visualization methods, of selected aspects of spatial access structures. This paper discusses the basic concepts, the design and implementation and the use of Vis Tool. Since Vis Tool is an ongoing project, we also present further extensions to be made to Vis Tool in the future.
Human beings use hierarchies extensively to simplify their conceptual models of reality and to perform reasoning more efficiently. Hierarchical structures are conceptually imposed on space and allow better performance of complex tasks in very large contexts. To understand how spatial hierarchies are formed and used is one of the most important questions in spatial reasoning research. In this project, wayfinding in large road networks is studied as a particular case. Humans can find fastest paths even in very large street networks quickly, applying a hierarchical strategy. Standard, non-hierarchical algorithms show performance that degrades rapidly with increasing network size. A hierarchical structure can be found as an abstraction from the hierarchy of street classes (expressway, highway, local road). This reduces the number of nodes involved in a search process, and allows to perform the search process in subnetworks more efficiently. We propose an algorithm which searches for an optimal path in the subgraph of the highest possible level. This leads to an efficient wayfinding algorithm, even where standard simple graph search algorithms for the shortest path become inadequate.
The topic of the paper is information processing using representations of spatial relations, called symbolic spatial indexes. The paper describes how symbolic spatial indexes can be used in several problems including information retrieval, composition of spatial relations, route planning and update operations.
We describe the XYZ GeoServer, an autonomous server for geometric computation. The server supports the storage of temporary data, incremental operations and the definition of macro operations. The data model of the GeoServer is composed of a geometry independent and of a geometry specific data model. This distinction makes it possible to easily modify the geometric data model. The server is based on the XYZ GeoBench, a loosely coupled collection of carefully crafted software packages for geometric computation. The GeoServer is a first example of a server supporting as communication interface the COSIMA-standard (Cooperative Spatial Information Management). COSIMA defines a protocol allowing different services to be linked via a communication network.
The major goal of spatial access methods in query optimization is to deliver the exact or a minimal superset of the result set and to perform this task at minimal cost. We present a clustering spatial access method that directly delivers exact result sets. Minimal cost is guaranteed through a cost-based adaptation strategy that dynamically determines and realizes storage clusters best suited for a set of spatial range queries. We introduce a tessalation of the data space which allows irregular, arbitrary small and large patches. Such patches can be adapted to query ranges in order to answer queries at minimal cost. The adaptation cost is kept small by performing only local repartitioning. Thus only a small number of neighbouring patches are merged or split during an adaptation step. The directory part has to be simple to perform range queries at minimal cost and to allow frequent adaptation updates at moderate cost. An implementation and evaluation in a database prototype system environment is under developement.
Spatial Data Management and Spatial Data Processing often requires technology of many different computer science disciplines like Data Management, Computational Geometry, Statistics, and Data Visualisation. Todays GIS try to integrate this technology into one single, large, software package. We propose a new flexible solution for keeping the specialised software packages independent and autonomous and linking them together via loose communication links. This open network of highly specialised services is called the Cooperative Spatial Information Management Architecture (COSIMA). It allows the integration of different types of services in a distributed computing environment. We describe a basic cooperation protocol that allows to dynamically link the independent components for any specific complex task. We define the minimum protocol elements all partners have to agree on and give examples for extensions of this basic protocol.
The purpose of the project is to develop a software system for decision support in case of accidents with environmental impact. The program we develop is based on a Geographical Information System (GIS) because of the large amount of geographical data, needed in the decision making process. Information on the characteristics of the polluting substance, predictions of the “behavior” of the pollution and expert knowledge for recommending actions are also of great importance. We combine standard GIS functionality with relational data bases, mathematical models of environmental processes and knowledge processing in an integrated, easy to use software system. We develop a special architecture to ensure successful integration and easy development of the system. Knowledge bases and knowledge processing modules play a key role—they are used for control of the whole system, for management of the user interface and for processing domain dependant knowledge (for example, estimation of the hazard to humans or recommending substances for neutralization and cleaning). The paper describes a specific software project. The main ideas and the architecture of the system will be useful in developing decision support systems, employing geographical information.
The implementation and proficient use of GISs has become so complicated that it is growing towards a domain in itself. The "Virtual GIS" (VGIS) aims at the use of the "knowledge base" that is inherent in a GIS for the automatic interpretation of remotely sensed images. VGIS allows the user to deal with abstracted data, hiding technical information about the database and allowing concentration on the user's particular tasks. The key to its independence from the actual GIS used is its focus on the user's tasks. Each task can be subdivided into elemental building blocks that act like atoms in a molecule. As a first step, an interface is introduced which frees image processing specialists (as well as other inexperienced GIS users) from learning the internals of GISs. This paper presents the VGIS rationale, its concept, its structural design, and first steps toward implementation.<>
Since the beginning of 1993 our group is a member of the SFB “Interactions of Continental Substance-Systems and their Modelling” 1. In this interdisciplinary project about 20 groups of the University of Bonn, mainly geoscientists, are participating. To provide the SFB with database technology as soon as possible, we decided to develop GEOSTORE, a first prototype system which is restricted on the management of geologically defined geometries such as 3D-points and triangulated 3D-surfaces of geological layers and faults. The functionality of GEOSTORE includes the management of three dimensional geological surfaces, the checking of spatial integrity constraints, the 2D-visualization of the examinated area in the Lower Rhine Embayment and a 2D-visualization of horizontal and vertical intersections with 3D-faces. GEOSTORE has first been realized on top of a relational DBMS and secondly on top of an ODBMS. For the 3D-visualization of the geological layers and faults a coupling with GEOCON, a 3D-construction and modelling tool for geological surfaces, is realized via remote procedure call.
This paper studies efficient path computation methods that can be applied on digital elevation models represented by a triangulation (called polyhedral terrains). The quality of the path is measured in the following aspects: Euclidean length, lying below a given height Z, avoiding large slope regions, and total height difference. Known techniques that can be used are reviewed, and it is shown how to find paths that satisfy one, or a combination, of the given quality measures.
While recent hardware and software developments made Geographic Information Systems (GIS) systems widely available there are still many problems to be solved. Partially in contrast to the mainstream in GIS research, the most important technical issues of real-world projects point to processes which support an efficient, economic but long living ‘data life cycle’, namely: 1. Data capturing; 2. Continuous map revision, database update and quality checking; 3. Map generalization; 4. Spatial data integration and transfer. All four processes are discussed and ways are shown to overcome those practical problems which contribute to the majority of the resources in realworld GIS projects. Most of the techniques mentioned have not been investigated very far or are still in an early stage though there are some pragmatic starting points available. Therefore the author proposes under the terms of ‘Sustained Spatial Data Management’ (S2DM) a pragmatic research focus and provides some examples and ideas to investigate further.
Efficient management of spatial data is becoming more and more important and for very large sets of 2-dimensional data, secondary memory data representations are required. An important class of queries for spatial data are those that extract a subset of the data: they are called window queries (also region or range queries). In this paper we propose and analyze the hybrid linear quadtree for the efficient secondary memory processing of three kinds of window queries, namely the exist, the report and the select query. In particular we show that it is possible to answer to all the above queries for multiple non-overlapping features with a number of accesses to secondary memory never greater than the number of pixels inside the window. More precisely, we prove that for a window of size n×n in a feature space (e.g., an image) of size T×T (e.g., pixel elements) using the hybrid linear quadtree the exist and report query can be answered with O(nlog r T) accesses to secondary storage, while the select query can be answered with O(nlog r T+n2/r) accesses to secondary storage. This is an improvement in worstcase time complexity over previous results [Nar93] and shows that multiple nonoverlapping features (i.e., coloured images) can be treated with the same I/O complexity as single features (i.e., black and white images).
The report describes software developed at the Institute for Nuclear Safety (Moscow) for the analysis of off-site consequences connected with nuclear technology. A number of computer environments developed and widely used are presented. A general description of the Central Data Bank on Chernobyl accident is presented. Attributive information has been accumulated for about 10,000 populated areas belonging to 15 oblasts (regions) of Russia. The cartographic database includes electronic vector and raster maps of different scales. Special software was developed for creating the applied information system for contaminated from Chernobyl accident regions of Russia. The main goal of this system is investigation of long term consequences of Chernobyl accident.
In this paper all major issues related to definition, design and implementation of a prototype of the geographical information system CARTECH are presented and discussed. The prototype proposes solutions for the most important requirements for geographical information systems (GIS). Namely, the double nature (spatial and descriptive) of geographical data, the fact that querying geographical information usually involves the specification of conditions which regards both the descriptive component of the data and the spatial one, the need of efficient spatial access methods to provide efficient and reliable management of large quantities of data. Moreover it also considers the important issue that user interaction with systems dealing with complex application realities, like geography, is typically incremental and iterative and proceed trough a process of view definition, updating and refinement. System highlights are: a system integration methodology to reuse existing modules so to deal in an efficient way with specialized aspects of the data and to build an overall system without re-inventing the wheel; a unifying logical data model to provide the end user with a uniform view of geographical entities, and to the designer a formal tool for the correct specification of data schemes and operations; support for the maintenance of consistency during the view building operations which are the result of the user-system interaction activities.
Traditional approaches to spatial and geometrical data management are based on mappings of spatial objects to individual disk pages. However, modern operating-system and hardware technologies allow variable-sized and large I/O operations to be handled almost as efficiently as fixed-sized individual block operations, and, based on this technology, our DASDBS Storage Manager implements a nested-relational storage model supporting arbitrary-sized objects. This paper describes our experiments in exploiting such larger storage clusters for the management of spatial data and its indexing. Our experiments show the impact of the new storage manager technology on the I/O behaviour of spatial data structures. We show that performance even of very simple spatial data managers can be significantly increased. Our results lead to a new I/O cost model for storage managers, and this can be used to provide new design criteria for spatial data structures.
One challenge in spatial data modelling is related to multiple representation of geographic objects. While often associated with cartographic change of scale and generalization, it applies in fact to a wider range of data analysis situations. We discuss multiple representation issues from a data modelling point of view, with emphasis on hierarchical structures inherent to multiple resolution levels. Then, we consider specific features related to multiple representation from a database modelling and querying point of view. Some guidelines for extending the object-oriented query language O2SQL are finally given. Although motivated by spatial applications, this extension is useful for a variety of other applications such as cooperative query answering [CCL91]. In particular it allows querying without exact data knowledge.
