Cartographic Generalization
 Process of abstraction which react on change
of map purpose or scale
 Process is influenced by symbolics and
character of displayed area
 Process encompasses
 Reduction of complexity
 Emphasis of essential and suppress unimportant
 Maintenance of relationship among map objects
 Preservation of aesthetic quality
Definitions
(ICA'73–McMaster'92–Ruas'95)
 Selection and simplified representation of
detail appropriate scale and/or purpose of map
 Process of deriving from data source through
the application of transformations, where
derivation objectives are to reduce data in
scope of amount, type and cartographic
portrayal with maintenance of consistency and
clarity of presentation
 Process of thematic and geometric resolution
reduction
Automated Generalization
 Algorithmisation of transformations
 Three main types
 Computer aided generalization – cartographic
generalization provided by humans with computer
based procedures
 Multiple representation – to design adaptable
cartographic visualization of geodata
 Real-time generalization – automated map
derivation from geodata
Roles of Generalization
 Map production
 Electronic map design
 Web cartography
 Mobile cartography
 Geodata integration
 GIS
Map Production
 Classical task (purpose oriented universal
maps)
 Off-line processing
 Computer aided generalization
 Reduce time of map compilation
 Reduce map inconsistency
 Ability of creation of new special purpose
maps
Electronic Map Design
 Adaptive zoom – smooth change of map face
and legend according to scale
 Adaptive visualization - change map face and
legend according to purpose and/or context
 Audience
 Position
 Time
 Conditions
 Multiple representation/real-time
 Less importance of annotations
Web & Mobile Cartography
 Reduction of amount of transferred data
 To generate more individual maps
 Often opposition to map production – as less
data as is possible
 Client/server distribution of tasks
 Multi-scale map-face
 Homogenize geo-data from various sources
Data Integration
 To update geodatabase from various
resources
 To find identical features in different scales
 Reverse engineering
 Harmonize geodata creation
 Homogenize existing geodata
 To support geoprocessing based on various
resources
GIS
 To save geoprocessing time
 Iterative geoprocessing
 Identification of spatial patterns
 To eliminate random errors
 Support for cartographic analysis of geodata
 Visual interface support for GIS
 To avoid redundancy and save storage space
Models of generalization
 Framework how to generalize
 From holistic process to procedures sequence
 From map compilation to geodatabase
processing
 From empiricism to analytics
 Separation of generalization into two parts
 Model generalization
 Visual generalization
Ratajski model '67
 Quantitative processes
 Reduction of map content according scale
 Qualitative processes
 Transform form representation to more abstract
forms
 Generalization point – scale which map
capacity is decreased to the level where
qualitative change is necessary
Bertin model '83
 Similar to Ratajski
 Conceptual generalization
 New conceptualization of phenomena
 Structural generalization
 Simplify phenomenon reperesentation (shape,
symbol or distribution)
 Structural = Quantitative, Conceptual
extending qualitative
Brassel-Weibel Model '88
Generalization
controls
Source DB
Structure of
original data
Process
library
Target DB
Target map
Process recognition
Structure recognition
Process execution
Data displayProcess modeling
Process types
& parameters
Brassel-Weibel Model '88
 Structure recognition
– through cartometric
analysis
 Identification object or
aggregates
 Their spatial and
semantic relations
 Their importance
 Generalization
controls
 Purpose
 Scale difference
between source and
target
 Source quality
 Symbols
specifications
 Process recognition
 Data modification
 Parameters setup
 Process modeling
 Compile rules and
procedures
McMaster-Shea model '92
 Why to generalize
 Theoretical issues
 Reducing complexity
 Maintaining accuracy, consistency, logical hierarchy
 Maintaining aesthetic quality
 Application issues
 Scale
 Audience
 Clarity
 Computational issues
 Algorithm cost
 Maximum data reduction
McMaster-Shea model '92
 When to generalise
 Geometric conditions
 Congestion
 Coalescence
 Complication
 Inconsistency
 Imperceptibility
 Controls
 Operators
 Algorithms
 Parameters
 Measures
 Density
 Distribution
 Length and sinusoity
 Shape
 Distance
 Gestalt
McMaster-Shea model '92
 How to generalize
 Simplification
 Smoothing
 Aggregation
 Amalgamation
 Merging
 Collapse
 Refinement
 Exaggeration
 Enhancement
 Displacement
 Classification
 Symbolisation
 Typification (added
later )
Operations (source ESRI)
Kilpelainen(Sarjakoski) model
'94
 Model generalization separation
 Incremental generalization
 Generalization task is divided into modules which
represent separable part of processed
geodatabase
 Generalization operations are divided into
 increasing conflict (displacement, exaggeration) and
non increasing conflict (rest)
 Content (selection, aggregation,amalgamation and
reclassification) and graphic (rest) ones
Contemporary approaches
 rule based
 constraint based
 agent based
 combinatorial optimalisation based
 continuous optimalisation based
 amplified intelligence
Rule based
 expert systems legacy
 If something then action
 blind end ?
 generalization is too complex
 not easy to identify rules
 rigidity
 not exactly !
 generalized modus ponens
 reverse engineering
 non-inventory objects
Constraint based approach
 set of conditions
 for example : size, min. distance...
 is impossible to satisfy all
 optimization
 synthesis of conditions
 flexibility of method choice
 results oriented
Agents
 communicating objects
 have plans how to reach better status
 perceive environment
 collaborate
 combine local and global view
 explicitly define map objects and its relations
 follows incremental model
 but too complex
"hill climbing” (source Weibel '05)
Activate Characterize &
evaluate
collect plans
select & trigger best
plan
characterization &
decision
worstbetterperfect
Pasive
no more plans
backtrack
backtrackno plan left at node
Combinatorial Optimization
 generate trial solutions
 test against criterion (a)
 trials are combination of methods
 cost of constraint violation
 discrete search space
 complicated tuning of parameters
 sometimes leads to deformations
Continuous Optimization
 To find min. or max. of function defined in
continuous space
 Good for continuous
and partial deformations
Define
constraints
Formulate
constraints
like functions of
point locations
Measure
Optimize by
standard
math. method
Generalize according
to point locations
Least Square
Finite Elements
Elastic Beams ...
Amplified Intelligence
 Improvement of computer aided generalization
 Computer offer possible solutions
 WYSIWYG conflicts display
 Parameters are transformed in “human
readable” form
 Holistic
 Strongly depend on user
Implementations
 more computer aided generalization
 Intergraph first
 MGE MapGeneralizer
 DynaGen
 Geomedia extension (???)
 McMaster Model
 various simple functions
 WYSIWYG parameters tuning
Implementations
 ESRl in last years
 simplification support
 big plans in future
 Laser - Scan is most developed
 related to AGENT and MAGNET
 CLARITY environment
 Various NMA's products
 CartACom - IGN
 Change – roads and buidings complex
generalisation, Univ. of Hannover
Implementations
 Open Source world
 JTS is heavy used
 Plug-ins
 web services
 ICA working group for algorithms sharing