Transcript Representation of geographic concepts

Representation of geographic
concepts
Geog 495: GIS database design
November 7, 2005
Outlines
• [D] Review of the article: Egenhofer et al 1999
Discussion will be focused on
–
–
–
–
Relational DB and GIS
Cognitive approach to GIS
GIS and DBMS
GISystems and GIServices
• [L] Representation of spatial entity
Lecture is organized into three parts
– Spatial data model
– Spatial data structure
– Design consideration
DB implementation of geographic concepts
is divided into three subfields in GIScience
• Spatial data model/spatial query language
• Spatial data type/operators
• Spatial access method
Q. What are they about?
Q. Indicate a level of abstraction for each
Q. What are different focuses on these?
Mathematics and GIS
• It is said that database implementation of
geographic concepts heavily borrows from
mathematics. So which subfields of mathematics
are influential in the followings?
• Representation of geographic entities
– e.g. Spatial primitives of vector data (point, line,
polygon)
• Representation of spatial relationships
– e.g. Connectivity in road network
Database and GIS
• Since most of GIS is implemented using RDB,
let’s focus on RDB.
• Examples: building Country database, how can
you store geometric features of Country in table?
Is multi-valued attribute allowed? Then break it
down to 4NF. And then… What about query such
as displaying boundary of France? What about
query such as returning adjacent countries of
France?
• Now generalizing your point, go to the next slide
Inadequacy of RDB to GIS
• What is the inadequacy of employing relational
database model for representing geographic
concepts?
• What is the legacy of storing spatial information
in relational database when creating GIS
database?
• How can we accommodate this problem? Think
from different perspectives. If you are DB
developer, If you are network administrator who
makes decision on GIS/DBMS purchase, if you
work in ISO committee, if you are a researcher...
Cognitive approach to GIS
• Do you agree with authors’ view on cognitive approach?
• Where do you think cognitive approach will be useful for
making a better GIS (that’s the goal of GIScience)?
Remember the current GIS is a just snapshot. You could
be a person who can create a better GIS.
• Query based on natural language (for visually impaired?)
• Design of better user interface design
• Development of spatial data model (beyond
vector/raster?)
GIS and DB systems
• Four different approaches to integrating GIS with
DB systems
– Extension of GIS with DB functionalities
• e.g. Arc/Info, Arcview support ODBC
– Coupling of a GIS with commercial DBMS
• e.g. spatial data is stored in relation DB so that SQL can be
called upon the spatial data
– Extension of DBMS with spatial functionalities
• e.g. Illustra’s DataBlades, Oracle’s Cartridges
– Open toolbox approaches where GIS provides
specialized services
• e.g. you work in your computer where you can download
data or processing tools from server
GISystems and GIServices
• Would a gradual shift toward GIServices
replace GISystems?
• What would be the likely impact of
GIServices? What about the impact on the
integration of GIS with DB systems?
Representation of spatial entity
Data hierarchy
• Data model: how the real-world is viewed
– e.g. Relational DB, Object-Oriented DB Model
– e.g. Object view, Field view
• Data structure: how data is stored in the computer
– e.g. Vector, Raster, TIN
• Data (file) format: how data storage is specified in s/wspecific way, you need particular viewer to display the
data
– e.g. VPF, shapefile, MrSID
Spatial data model:
how spatial things are viewed
• Object view
–
–
–
–
See that the world is composed of discrete entity
Interested in precise location of geographic features
Space is measured, attribute is controlled
Human beings prefer discrete perception of things rather than fuzzy
perception
– e.g. parcel, lake, city
• Field view
–
–
–
–
–
See that the world is composed of continuous field
Interested in smooth variation in attribute over space
Space is controlled, attribute is measured
Measuring attribute for all locations are infeasible due to its continuity
e.g. elevation, temperature
Q. So which view is implicitly supported by DB model such as OO or relational?
Spatial data structure:
how spatial things are stored in the machine
Way to represent spatial things
• Vector: point, line, polygon, node, vertices
• Raster: a set of grid cell
Way to represent spatial relationships
• TIN (Triangulated Irregular Network): a set of
triangulated line between neighboring points
• Matrix: stores attributes of relationships between
spatial objects
Note on spatial data structure
• Spatial data structure appears to correspond to
spatial data model (i.e. nature of phenomenon:
continuous/discrete)
• However, data structure is largely determined by
the manner of data input regardless of the
nature of geographic phenomenon
– Vectorization of paper map (i.e. digitizing)
– Satellite image, Air photo
• You should choose data structure suitable for
the nature of phenomenon given users’ need
More on vector
• Higher-level vector
– Overlapping polygon (e.g. flycircle)
– Multi-part polygon (e.g. Hawaiian island)
– Hole in polygon (e.g. lake in island)
• Spaghetti model vs. topological model
– Cartographic purpose: spaghetti will do
– What is topology? Why do we need topology?
– e.g. Network model commonly uses link-node
representation
• Planar vs. non-planar
– How to represent overpass in contrast to intersection?
How topology is stored
Q. What kind of topology is measured? {containment, connectivity,
contiguity}
How to build topology
• Do some exercise in the lab
• 1) Digitize features over backdrop (say
airphoto) say at arcview
• 2) Save the file and import this file into
arc/info coverage (use SHAPEARC)
• 3) Use CLEAN command in Arc/Info
workstation
• 4) Read attributes at feature attribute table,
say AAT and PAT
Planar vs. non-planar
More on Raster
• Same values are often repeated in the cell
• Efficient storage method (or compression
method) is needed as it uses space rather
inefficiently
• These are methods for storing
– Run-length code: linear representation of cell
values
– Quadtree: hierarchical organization of raster
structure (e.g. SPAN)
Run-length code, Quadtree
Run-length code
3A1B2A2B2A2B1A3B
Quadtree is also known for spatial access method
• See how, runlength code can
save space, and
save time in
searching the
location of Well
Triangulated Irregular Network (TIN)
• Point is not
sampled in a
regular manner
• Sample points
are connected by
lines to form
triangles
• Compact
representation of
surface (compare
this to pointbased DEM)
Matrix
• Used to represent attributes of spatial
interaction (e.g. commuting flow, traffic
flow, commodity flow, migration, and so on)
• Conversion between matrix and relational
table often necessary as matrix is not well
supported in GIS software
Same phenomenon can be
represented differently (data structure)
Q. How topography is stored in GIS?
• In Raster
• In TIN
• In Vector
Spatial file format
• This is system-specific, so you need a specific software
to view them
– TIGER/Line, DXF, SHP, TIFF
– SDTS (Spatial Data Transfer Standard)
– MrSID
• Need for metadata to exchange info
– FGDC standard, XML
• Need for interoperability
– Converting file formats are tedious and sometime it loses
information
– OpenGIS
Special aspect of spatial data
• Spatial data, unlike attribute data, are designed
for display in addition to query
• Therefore, some technical aspects of
cartographic representation should be taken into
account when you design database for GIS
– e.g. some cartographic symbol can be exaggerated to
enhance readability without regard to its positional
accuracy
Which model to choose will be
influenced by …
• Users need/application
– e.g. routing application requires link-node model, road
maintenance application requires polygon
representation of road as width of road can be
important
• Geographic scale
– Line generalization (e.g. Douglas-Peuker algorithm)
– Multiple representation
– Annotation
Terms
• Interoperability: the ability of two or more
systems or components to exchange information
and to use the information that has been
exchanged
• Open GIS Consortium: Non-profit organization
that works for standardizing the implementation
specification of geospatial information; designed
to form the architecture for interoperability;
promotes COM technology
• Middleware: software that glues two separate
applications; e.g. link a database system to
application; balances traffic between the two
Terms
• MBR (Minimum Bounding Rectangle): spatial access
method uses filter-refine step. Where filter select
candidates using minimum bounding rectangle.
• R-trees: spatial access method that stores overlapping
MBR into tree-structure
• Quadtree: spatial access method that stores nonoverlapping multi-scale grid into tree-structure, also
known as the method for compressing raster data
structure
• Which version of SQL has the spatial extensions?
1) SQL1 2) SQL2
3) SQL3