1. Spatial Databases: Data Collection
Spring, 2017
Ki-Joune Li

2. Why Data Collection? The cheapest way to build spatial DB
Get existing databases, but
Check if the requirements be satisfied
Metadata: Description of Data
Data Migration vs. Interoperability
Legacy Problem
Building an Information System
No more entirely new system
Integration with Existing systems
Integration with existing DB

3. Integration with Existing Databases
Data Migration
Copy from existing DB
Procedure
Survey on existing DB: Data Clearinghouse
Metadata
Conversion from existing databases
Integration of several databases: Mismatch Problem
Legacy System
New System
Existing System
New System
Existing DB
Copy from Legacy DB
New DB

4. Geo-Data Clearinghouse
Clearinghouse: financial services company that provides clearing and settlement services for financial transactions. (in Wikipedia.com)
Collection of Geospatial Data Servers
Collection of metadata Not data itself.
Providing a single interface to browse metadata of several servers
Z39.50 protocol and client
Example: FGDC in USA (

5. Metadata for Geospatial Data: ISO 19115
Title, and Alternative title, Originator, Abstract, and Data
Frequency of update
Presentation type
Access constraint, Use constraints
Topic category
Bounding coordinates, and extent
Spatial reference system, and resolution
Supply media, and data format
Supplier and Additional information source
Date of update of metadata
Sample of the dataset
Dataset reference date, and language
Vertical extent information
Spatial representation type
Lineage
Online resource
Some items are mandatory and others are optional

6. Generalization: Conversion of spatial data
Conversion from
Large-Scale Spatial DB to Small-Scale Spatial DB
Cartographic Aspect vs. DB Aspect
Cartographic viewpoint: To make maps more visible
DB viewpoint: To reduce the size of data
Six Generalization Operators
Simplification
Elimination
Translation (Cartographic Purpose)
Aggregation
Collapse
Exaggeration (Cartographic Purpose)

7. Simplification Simplification (Line Simplification)
Elimination of internal nodes from a polyline
To minimize the loss of accuracy
Example: which point to remove
Douglas-Peucker Algorithm
A greedy algorithm

8. Elimination Eliminate spatial objects not satisfying given conditions
Example
Eliminate * From Buildings Where area < 100 m2
Propagation of Elimination
Elimination may destroy cardinality condition
Example: What to do in this case ?
1..1
District
District_Office
1..1

9. Aggregation Aggregate a set of spatial objects into a large object
Definition of the boundary of aggregated object ?
Aggregation of Non-Spatial Data
Example: Number of Habitants in the apartment complex
Sum, Max, or Average
Apartment Complex A B
Aggregation
- With Boundary - Without Boundary C

10. Collapse Reduction of Dimensionality Example
From 2-D to 1-D (from surface to line)
From 2-D to 0-D (from surface to point)
Very Rarely from 1-D to 0-D
Example
Computation of collapsed objects
Surface in 1/1,000 Map
Line in 1/50,000 Map

11. Topological Issues in Generalization
Example
How to Correct it
No drop vertex if  Topological Inconsistency
Translation of the object with problem
Left Side
Right Side  Topologically Incorrect

12. Topological Issues in Generalization
Another Example A A
Collapse to point B B
A contains B  A is equal to B at least
B contains A
R(B, A)  RG(B, A)
What is the correct topology after the collapse ?

13. Topological Issues in Generalization
Another Example
Road
Road A A
Buildings
Buildings
Aggregation B B1 B2 B3 B4
R(B, A)= Disjoint
R(B, A)= Overlap
R(B, A)  RG(B, A)
Is it correct ?

14. Mismatch Problems Mismatches
Integration of several databases from different sources
Adjacent Maps
Different Accuracy
Different Dates of Creation
Different Maps
Different Ground Control Points
Etc.

15. Example: Topographic Map and Cadastral Map
Positional mismatches come from several reasons, such as different ground control points, or different scales, especially
different organizations of management of local databases. For example, in our country, we have two types of natioanl
base maps, which are topographic maps and cadastral maps. But they have been maintained by two different governmental
organizations during 90 years, they have significant mismatches like this figures. This problem is very important obstacle in
realizing national gis.
Cadastral Map

16. Example Building DBA Manhole DBB Pole DBC
Find the nearest manhole or pole to Building P
For example, suppose that we have three local spatial databases, and a query is submitted to building DB, which is
to find the nearest manhole or pole to building P.
Building DBA
Manhole DBB
Pole DBC

17. Example Building DB Manhole DB Pole DB
Find the nearest manhole or pole to Building P A B
To process this query, we overlay manhole DB on Building DB.
Building DB
Manhole DB
Pole DB

18. Example Building DB Manhole DB Pole DB result of the query
Find the nearest manhole or pole to Building P A B
Then the nearest manhole to building P seems manhole A, and spatial database system of manhole will give A as the answer.
Building DB
Manhole DB
Pole DB

19. Example Building DB Manhole DB Pole DB result of the query A
adjust the building position B
the correct answer
But, if we consider the mismatches between building DB and manhole DB, the location of building P must
be adjusted like this figure. And we see that the real answer to the query is B instead of A.
This means that we should correct the positional data of local databases to get consistent answer.
must be B.
Building DB
Manhole DB
Pole DB

20. Example Building DB Manhole DB Pole DB result of the query
Find the nearest manhole or pole to Building P
the correct answer
This is a similar example.
Building DB
Manhole DB
Pole DB

21. Elastic Map Transformation: Rubber Sheeting
Elastic transformation of objects in each spatial database.
Building DB
Manhole DB
Pole DB
Transformed Manhole DB
Transformed Pole DB
Reference Spatial DB
or Base Map
Consistent Spatial Databases
Spatial Query
Therefore we need a method to make local spatial databases consistent, and we have proposed a method for that several
years ago. This method is to transform local databases according to a base map so as to make sure that they are consistent
and give correct answers. Like this figure, each local database except base map is to be transformed and we will have
consistent set of spatial databases.

22. Elastic Map Transformation by Delaunay Triangulation
Procedure
1. Select sets of Control Point Pair on MapRef and another MapA, respectively.
2. Delaunay Triangulation with Control Points on MapRef.
3. Triangular Transformation, for each Point p on MapA.
Control Points on the Reference Map
Corresponding Control Points on Other Maps
Corresponding Pair of Triangles
MapRef
MapA
For the detail of the transformation method, I'll give a brief explanation, since this method is an
important starting point of our study.
For the transformation, we select a certain number of control points on the base map, and
select also corresponding control points on other map. For example, we can select a corner of road
as control point.
Then we do delaunay triangulation with the control points on the base map and find corresponding triangle on other map.
The next step is for transforming coordinates of objects on non-base map onto base map. For example, if we want to
transform this point onto base map, we find the enclosing triangle on this map, and the corresponding triangle on base map.
Then we apply triangular transformation method to find the new coordinates on the base map.

23. Triangular Transformation
q1(u1, v2 )
p1(x1,y2 )
q (u, v )
p (x, y )
q3(u3, v3 )
q2(u2, v2 )
p2(x2, y2 )
p3(x3, y3 )
Triangle on MapRef
(u,v) = (a1u1+a2u2+a3u3, a1v1+a2v2+a3v3)
where a1 = b {(y2 – y3)x + (x3 – x2)y + x2 y3 – x3 y2 }
a2 = b {(y3 – y1)x + (x1 – x3)y + x3 y1 – x1 y3 }
a3 = b {(y1 – y2)x + (x2 – x1)y + x1 y2 – x2 y1 }
b = (x1 y2 + x2 y3 + x3 y1 – (x2 y1 + x3 y2 + x1 y3 ))-1
When we are given two corresponding triangle, one is on the base map and another is non-base map,
we can compute the new coordinates of the point on the base map from this point.
This is the equation to compute the new coordinates. It looks a little complex,
but this is a simple linear transformation.

24. Elastic Map Transformation : Example
By this method, every patial object can be transformed onto base map by some control points,
and finally we get consistent spatial databases.
In this figure, we select 7 control points and we get a quite good transformation, as you see.
This method is simple, but has some important drawbacks.

25. Elastic Map Transformation : Example
Control Points

26. Mismatched between Adjacent Maps
Examples
Shift
Discontinuity
Disappearance

27. Edge Matching Rules Priority on More Recent Data
Respect of Pivot Objects
Respect of Predefined Constraints
e.g. Building should be rectangular
Pivot Object