Multidimensional Data Modeling for Feature Extraction and Mapping

1
Multidimensional Data Modeling for Feature
Extraction and Mapping

• ACSM
• April 19, 2004

E. Lynn Usery usery_at_usgs.gov
http//mcmcweb.er.usgs.gov/carto_research
2
Outline

• Motivation
• Objectives
• Approach
• Theoretical Model
• Implementation
• Scale Dependent Feature Rendering
• Conclusions

3
Motivation

• Conventional GIS model the world in
  two-dimensions with a map model and geographic
  features dependent on geometry for definition
• This map model limits three-dimensional and
  temporal analysis, and multidimensional,
  multi-scale representations
• Cognition studies indicate that humans perceive
  the geographic world as a set of definable
  entities with spatial, thematic, and temporal
  attributes associated

4
Objectives

• Provide a theoretical model based on feature
  orientation
• Develop the model to support unique entities with
  spatial, thematic, and temporal attributes and
  relations for each feature instance
• Implement the model in a feature library and use
  the library for feature extraction to support The
  National Map

5
Approach

• Implement the theoretical feature model in an
  object-oriented library
• Develop feature instances for 20 specific
  features that are relevant to The National Map
• Develop attributes and relationships including
  multiple representations (raster and vector) of
  attributes for each feature instance
• Determine the extraction capability of each
  feature from various image sources

6
Feature Model

• Feature is geographic entity and object
  representation
• One feature, many objects
• Multiple resolutions
• Multiple geometries
• Access from single identity

7
(No Transcript)
8
Definitions

• Feature - A set of phenomena with common
  attributes and relationships. The concept of
  feature encompasses both entity and object.
• Entity - A real-world phenomenon that cannot be
  subdivided into phenomena of the same kind.
• Object - A digital representation of all or a
  part of an entity.
• Attribute - Characteristic of a feature or of an
  attribute value.
• Relationship - Linkage between features or
  objects.
• Feature instance - An occurrence of a feature
  defined by a unique set of attributes and
  relationships.

9
(No Transcript)
10
Databases Supporting Feature Extraction and Map
Generation

• Feature Attributes and Relationships
• Image
• Image Chips
• Spectral Responses
• Digital Number Ranges for Multimodal Images
• Map
• Symbol Specifications
• Symbol Chips
• Inclusion Criteria

11
Feature Library Implementation
12
(No Transcript)
13
Multiple Feature InstanceExample with Actual Data
14
Feature Instance Implementation with Actual Water
Quality Data
15
Relationship Implementation from NHD
16
Time Attribute Implementation
17
National Map Feature Extraction

• Camp Lejeune study site
• 20 features selected
• All attributes and relationships built based on
  DLG-E specifications
• Image chips extracted for storage as attributes
• Spectral responses determined (laboratory and
  from images)

18
Table of the 20 Features
Type Features
Point (5) Helipad, Rock, Tank, Tower, Wreck
Line (6) Bridge, Road, Shoreline, Stream/River, Trail, Transmission Line
Polygon (9) Aircraft Facility, Apron/Taxiway, Building, Lake/Pond, Parking Site, Pier/Breakwater/Jetty, Shrub Land, Swamp/Marsh, Trees
19
(No Transcript)
20
Airport -- DOQ
21
Airport Ikonos Pan
22
Airport Ikonos Pan-sharpened
23
Airport Ikonos MX
24
Airport SPOT Pan
25
Airport CIR Photo
26
Trail DOQ
27
Trail Ikonos Pan
28
Trail Ikonos Pan-sharpened
29
Trail Ikonos MX
30
Trail SPOT
31
Trail CIR Photo
32
Trail Color Photo
33
Airport Map Symbol
34
Trail Map Symbol
35
Geodatabase for the Study Area in ArcCatalog
36
Airport Feature
37
The Study Area Camp Lejeune, NC
38
Scale Dependent Renderer
39
Trees on 130,000-Scale Map
40
Trees on 19,000-Scale Map
41
Buildings Rendered as Polygons 15,000-Scale Map
42
Buildings Rendered as Polygons/Points Based on
the Longest Axis -- 112,000-Scale Map
43
Buildings Rendered as Points on 128,000-Scale Map
44
Buildings Not Displayed on 155,000- Scale Map
45
Conclusions

• A theoretical model of features existing in the
  real world as single geographical entities has
  been developed
• This model shows promise for implementing feature
  extraction methods and scale-dependent rendering
  for The National Map
• Probabilities for extracting specific features
  from multimodal sources can be developed based on
  feature attributes and relationships and
  appearance in various image sources

46
Multidimensional Data Modeling for Feature
Extraction and Mapping

• ACSM
• April 19, 2004

E. Lynn Usery usery_at_usgs.gov
http//mcmcweb.er.usgs.gov/carto_research