|
|
|
All papers and posters are to be published in the conference proceedings of extended abstracts, which will follow a common in-house format. It will enormously help the preparation and editing of this book if authors would submit material as far as possible according to the following guidelines.
All revised abstracts must be submitted by Monday 19th February 2001 in order to meet our publication deadline. Authors who fail to meet this deadline risk their papers being omitted from the proceedings of extended abstracts. Alternatively, the original submission may be included instead (if submitted in the specified format). This is at the discretion of the editors, depending upon the quality of the paper and the amount of editorial work required.
Please re-submit modified abstracts to gisruk@glam.ac.uk (with your abstract reference number in the subject header):
As a Microsoft Word file labelled “GISRUKxxx.doc”, where “xxx” refers to your allocated reference number; 1 inch (2.54 cm) margins (top, bottom, left, right); no headers or footers; and all text set in Times New Roman (or similar) typeface, 10 point font. The title and primary headings should be bold capitals, subheadings should be bold title case. Please do not use more than two levels of subheading. Each biography of the principal author(s) should not exceed 5 lines, and the abstract, including references and figures should not exceed 1500 words or equivalent. For references please follow the style used in the example below. Figures should be included in the Word document wherever possible (but disable the "Float over Text" option!).
Alternatively, the paper or poster can be submitted as a Notepad or Wordpad ASCII text file labelled “GISRUKxxx.txt”, where “xxx” refers to your allocated reference number; any figures included as separate attachments “GISRUKxxx_1.gif” (or .jpg etc. with consecutive figure numbers) and place the figure captions in the body of the text where you want it to appear.
(NB This example is illustrative only - its precise appearance will vary according to your web browser).
David B. Kidner1, Christopher Eynon1 & Derek H. Smith2
University of Glamorgan
1GIS Research Centre
School of Computing
2Division of Mathematics
Pontypridd, Rhondda Cynon Taff
WALES, U.K., CF37 1DL
e-mail: dbkidner@glam.ac.uk
ABSTRACT
Digital terrain modelling addresses the problem of characterising the Earth's surface using a finite set of terrain measurements. Ayeni (1982) defines a digital terrain model (DTM ) as "the numerical and mathematical representation of a terrain by making use of adequate elevation and planimetric measurements, compatible in number and distribution with that terrain". Ideally, the resolution of the terrain model is best suited to the specific application of the user. Naturally, the accuracy of a terrain model will increase with resolution, but so does the cost of storing, manipulating, analysing, and visualising the model (De Floriani et al., 2000). However, most DTMs utilise a fixed-resolution data structure and can be considered ill-suited for the requirements imposed by multiscale or real-time applications. Furthermore, the dissemination of terrain data, particularly across the Internet, is hampered by the constraints of existing data models and the strategies of national mapping agencies in partitioning geographic data.
Access to spatial data at variable scales can be achieved by storing multiple representations of the data at predetermined scales; storing a single large-scale version from which smaller scales are derived using generalisation algorithms; or using a multiresolution data structure specifically adapted to retrieving data at varying levels of detail (Kidner et al., 2000). Storage of multiple representations results in significant storage overheads owing to data duplication between different versions. Retrieval from a single version can incur major processing overheads when deriving a representation of much smaller scale. Multiresolution or multiscale data structures represent a compromise between these approaches.
This paper will describe alternative models to the traditional regular grid digital elevation model (DEM) and triangulated irregular network (TIN) for representing multiscale terrain surfaces. A new data model is then presented which hierarchically encodes very large (i.e. national) digital elevation models using an implicit quadtree pyramid. This is illustrated for the complete representation of Great Britain using the 812 tiles of the 20x20 km Ordnance Survey, 1:50,000 scale DEMs at a sampling resolution of 50 m (Figure 1).
Figure 1 - The 812 tiles of the O.S. 1:50,000 scale DEMs at 50m resolution. The small squares are shaded with respect to the standard deviation of elevations within each 20x20 km tile. The larger squares define the 100x100 km National Grid tiles for GB. (© Crown Copyright Ordnance Survey. An EDINA Digimap / JISC supplied service).
The data are differentially encoded at each level of the hierarchy and then statistically modelled for better storage and transmission. Users can download the data or retrieve the DEM at any prescribed resolution down to its original 50m grid spacing (Figure 2).
Figure 2 - Illustration of the quadtree pyramid for the higher levels of the GB DEM (this derivation uses all 131 million elevations of the O.S. 1:50,000 scale DEMs). (All figures: © Crown Copyright Ordnance Survey. An EDINA Digimap / JISC supplied service).
Our premise is that the original data should be transmitted error-free (i.e. losslessly encoded) at the highest resolution (e.g. 50m), but with the flexibility to retrieve data at lower resolutions if required (e.g. 100, 200, 400, 800m, etc.). An analogy can be drawn with progressive image compression techniques, in which images are transmitted over a communications line and are decompressed and viewed in real time. The entire image can be viewed in a low-quality format, which is then improved as more and more of the image is received (Saloman, 1998).
The paper will provide an overview of techniques such as elevation pyramids and surface patch quadtrees, before outlining our proposed method and presenting its performance statistics.
METHODOLOGY
Blah, Blah, Blah, ...,
(This abstract will no doubt be updated in due course to include all the referees' comments).
REFERENCES
AYENI, O.O., 1982, Optimum Sampling for Digital Terrain
Models: A Trend Towards Automation. Photogrammetric Engineering & Remote
Sensing, 48(11), 1687-1694.
DE FLORIANI, L., MAGILLO, P. and PUPPO, E., 2000, Variant: A System
for Terrain Modeling at Variable Resolution. Geoinformatica, 4(3), 287-315.
KIDNER, D.B., WARE, J.M., SPARKES, A.J., and JONES, C.B., 2000,
Multiscale Terrain and Topographic Modelling with the Implicit TIN. Transactions
in GIS, 4(4), 379-408.
SALOMAN, D., 1998, Data Compression: The Complete Reference. (New York:
Springer-Verlag).
BIOGRAPHY
Dave Kidner is a Senior Lecturer in the GIS Research Centre of the University of Glamorgan. He holds a degree in Mathematics & Computing (1987), and a PhD, entitled "Digital Terrain Models for Radio Path Loss Calculations" (1991). He subsequently became a post-doctoral research fellow investigating Deductive MultiScale Databases. His research interests include digital terrain modelling, intervisibility analysis and visual impact assessment, spatial data compression, and multiscale databases.
Last modified by Dave Kidner on 1st February 2001