Supporting Multiple Representations

in Spatio-Temporal Databases

Stefano SpaccapietraDatabase Laboratory Ecole Polytechnique Fédérale Lausanne (EPFL)

on behalf of the

MurMur Consortium

http://lbdwww.epfl.ch

2

SpatioTemporal

Evolution of Data Modeling in Databases

Expressive power

Data Models

Codasyl

Relational

OOER

Extended ER

UMLODMG

Multi-representation

3

Multi-scale Map Production

Cartographic Generalization

The Database Output Maps

4

Multi-resolution Databases

Cartographic Generalization is costly:

-> store the result for reuse

How do we express the linksbetween different representations ?-> update propagation

5

Resolution Level 1 Resolution Level 2

Multiple Geometries for the Same Object

One possible solution : stamping spatial attributes with the spatial resolution

Spatial integrity constraints : Sinuosity (River.geometry[2]) = Sinuosity (River.geometry[1]) Length (River.geometry[2]) = Length (River.geometry[1])

River described as an area or as a

line

Rivermr geo

M

6

Multiple Abstraction Levels: aggregation of objects

Grouping of objects, e.g., according to their spatial relationships Example: a set of buildings close to each other is

replaced with a built-up area

7

Aggregation Link Grouping of objects according to

their semantic and spatial relationships e.g., a set of buildings and adjacent

fields belonging to the same farmer grouped into a single object Farm

Derivation of attribute(s): Farm.geometry= Spatial Union

(Field.geometry,Building.geometry)

Aggregation constraint: the fields and the buildings composing the

same farm must belong to the same farmer and the fields must be adjacent.

Farm

Field Building

Composed

Composed

1,n1,n

8

« Typification » of objects

Generalization operation (typification): no 1-1 or n-1 mapping between the ground and cartographic buildings

n-m relationship

5 ground buildings (1,2,3,4,5)

represented by 3 cartographic

buildings (a,b,c)

A ground building can participate into 0 or 1  typify  relationship

GroundBuildin

g

Cartographic

Buildingtypify

t = ( {1,2,3,4,5} , {a,b,c} )

9

Topological Relationships

Level 1

Level 2

At resolution level 1, the road is adjacent to the enbankment.

At resolution level 2, the embankment is no longer represented. The road is seen as adjacent to the building.

Embankment

Road

Near

M

M

M

10

Classification hierarchy and hierarchical value domains

Describe the same characteristic at different abstraction levels Hierarchical value domains for attributes Classification hierarchy for objects

cultivated area

rose iriscarnation

flower cereal oleaginous

corn barley rape sunflower

11

Multi-representation

Car

Vintage Car

Collectible

Transport Mean

Vehicle

Land vehicle

Ford

Imported GoodMovie Accessory

12

Multidimensional Representation Space

Classification

Space granularity

Viewpoint

Time granularity

two representations of the sameobject in the same viewpoint at two different resolutionlevels

13

A Mono-resolution Database

Classification

Space granularity

Viewpoint

14

A Map

Classification

Space granularity

Viewpoint

15

Classification Dimension

students

facultiespersons

technicians

secretaries

• Current Status: refinement hierarchies

Person

Faculty TechnicianSecretary

StudentEmployee

faculties

technicians

secretaries

Is-a

16

Limitation: Roles

car-owners companiespersons

Person Car-owner Company

Person-with-car Company-with-carintersectionclasses

partition constraint

Car-owner = Person-with-carCompany-with-car Person-with-car Company-with-car = Ø

17

A More Direct Representation

Car-owner

OR IS-ACar-owner CompanyPerson

MAY-BE-A MAY-BE-A

+ partition constraint

Intersection link

18

Viewpoint Dimension

Relational DBMS support (mostly non-updatable) views, but semantics is poor

Object-oriented DBMS have rich semantics but poor view mechanisms

Object-relational DBMS: ?

Object-oriented expressiveness augmented with intersection links, roles and revised inheritance rules will provide the best solution

19

Multidimensional Representation Space

Classification

Space granularity

Viewpoint

Time granularity

How is the representation space - presented to users?- implemented in Ddatabases?

20

Possible architectures

One single multi-resolution, multi-viewpoint schema

One schema per viewpoint and/or per resolution range

One schema per resolution range and per viewpoint

with an intrinsic schema

21

A single schema

owner

landuseParcel

Building M

Cartographicbuilding

owner

landuse Parcel/use

agr/use

Parcel/ownerPlot

Castlecomposed

Typify

Road M along

near

on/under

Bridge

agr/owner

22

A schema per viewpoint and resolution

Local schema : mono resolution and/or mono viewpoint schema

Local schemaLocal schema

23

A multi-resolution schema per viewpoint

Building M

Cartographicbuilding

agr/owner

owner

landuse

landuseParcel/use

agr/use

Parcel/owner

Building MPlot Castlecomposed

Typify

Road M along

near

Parcel

Bridge

on/under

owner

Viewpoint 1

Viewpoint 2

24

agr/use

A schema per resolution and viewpoint

Cartographicbuilding

on

Building

Road

Building

Castle

Plot

Bridge

Road

On / under

Parcel/use

near

Parcel Parcel/owneragr/owner

composed

25

An intrinsic schema

Intrinsic Schema : description of real world entities independently of any viewpoint

Intrinsic schema

Local schemaLocal schema

26

Murmur IST Project (2000-2002)

A conceptual data model supporting space, time, and multirepresentation (extension of MADS)

A corresponding query language (multirepresentation algebra)

Two application cases (cartographic, risk assessment)

A schema editor for visual data definition (DDL)

A query editor for visual data manipulation (DML), including intelligent zooming and temporal travelling

Implementation on a GIS (from Star Informatic)

27

The MURMUR Layer

VisualSchemaEditor

VisualQuery

Language

VisualBrowser

Query bySketch

Tool

Murmur Kernel

-> OR or OO -> RDBMS -> Open GIS -> GIS

Open GISWrapper

Translation kernel

OO/ORDBMS RDBMS

GIS

STAR Info

GISGIS

28

Murmur : the Consortium

STAR Informatic (GIS provider)

IGN (geodata provider and map producer)

Cemagref (public research center)

Free University Brussels

University of Lausanne

Swiss Federal Institute of Technology Lausanne (EPFL)

29

Contact

http://lbdwww.epfl.ch