Multiple Resolution Terrain Features

1
Multiple Resolution Terrain Features

• Dr. Dale D. Miller
• Annette Janett
• Lockheed Martin Information Systems
• ddmiller_at_lads.is.lmco.com,
• ajanett_at_lads.is.lmco.com
• John Nordstrom
• Evans and Sutherland
• jnordstrom_at_es.com
• Dr. Paul A. Birkel
• The MITRE Corporation
• pbirkel_at_mitre.org
• sponsored by
• STRICOM
• as part of the
• SNE Army Science and Technology Objective

2
Operational Navigational Chart - 11,000,000
3
Joint Operations Graphic - 1250,000
4
Landsat TM - 30 m Imagery
5
City Graphics - 125,000
6
Az Zubayr Petroleum Refinery
7
Human Perception of Geospatial Information

• Why should we care?
• Key part of Synthetic Forces (SF) systems are
  models of human behavior
• planning, movement, tactics
• Visual systems for suspension of disbelief
• stimulating human perception
• Human beings reason at varying levels of
  resolution, e.g., a city has representations
• areal feature on a map with some high level
  attributes
• city street map
• experiential

8
Environmental Psychology

• Jean Piaget believed in a map in the head
• adults possess a fully coordinated, complete and
  metrically accurate mental representation of
  their environment
• analogous to ground truth in a simulation
  system
• Contemporary psychologists believe human
  geospatial perception is much more abstract

9
Theory of Geospatial Perception

• Carreiras spatial information is organized
  hierarchically and schematically, at different
  levels of abstraction and in different modes of
  representation so as to maximize the power of
  spatial thought while making most efficient use
  of limited working memory capacity
• Human spatial representations more like an atlas
  than a map
• environment represented on different pages at
  different scales and levels of detail
• the selection of a page of the atlas depends upon
  the application

10
Objective

• Develop an unambiguous reference data model for
  the terrain environment for military simulations
  in which

spatial information is organized hierarchically
and schematically, at different levels of
abstraction and in different modes of
representation so as to maximize the power of
automated geospatial reasoning and the visual
rendering of the environment
See and reason about both the forest and the trees
11
Prior Art Constructive Simulations

• Terrain databases traditionally support a single
  level of detail
• Aggregate simulations JTLS, CBS, Eagle, JANUS
  have low resolution representations with a few
  categories of features
• lines of communication
• features supporting cross country movement
• features acting as barriers to movement
• Platform simulations CCTT, ModSAF, OneSAF
• increased resolution, representing buildings,
  trees, etc.
• OneSAF will require building interiors, windows
  and doors, underground passageways, other high
  resolution features

12
Prior Art Visual Simulations

• Level of detail has long been used to manage
  scene complexity (pixels and polygons)
• render a simple version of the object when its
  far from the eyepoint
• 3-D models
• Terrain
• Aggregate models (basis sets)

13
Prior Art - Classification Schemes

• JSIMS TCDM coverages
• flat, single resolution
• Andersons two-level hierarchy (land-use/land-cove
  r)
• 92 mapping units
• lacks physiographic, obstacle, transportation
  features impoverished cultural features
• Murphys regional landform classification system
• 37 mapping units
• lacks built-up and vegetation features
• Richbourgs observation on classificatory systems
  99F-SIW-108
• depends on the application no single criterion
  exists for classification in any field

14
The Spreadsheets

• Aggregate features and features which comprise
  them
• Building (a column, or feature which may be
  aggregated) needed further specification

• Second spreadsheet contains aggregate features
  vs. building types (BFC_)

• In total, over 30,000 entries evaluated

15
Highest Level Aggregation / Classification
39 mapping units
16
Derived Aggregation Rules

• Feature A contains Feature B if
• 1 Feature B is an innate component of Feature
  A, or
• 2 Feature A has cover type Built-Up Area and
  there are least
• two real-world examples of this
  aggregation, or
• 3 Feature B is fundamental to SF reasoning or
  visual
• perception of Feature A.

• Examples
• Forest feature may be an aggregate of a
  River/Stream feature
• Settlement aggregate feature may contain a
  Windmill feature but not a Quarry feature
• Forest feature should be an aggregation including
  the Road feature

17
Pages from the Atlas

• Mountain - Forest compositional feature hierarchy

18
Pages from the Atlas

• Urban Region (Population 25,000 - 150,000)

Urban Region has a Processing Plant has a Oil /
Gas Facility has a Pumping Station Urban Region
has a Orchard / Plantation has a River / Stream
has a Marsh / Swamp has a Causeway -)
19
Planned Future Work

• From spreadsheets to a relational database model
• Finish and proof the aggregation decisions
• Write a rationale
• Develop an attribution scheme for
  construction/architecture of geotypical cultural
  features
• e.g., residential building in Orlando vs. Az
  Zubayr
• Extensions to higher resolutions to support
  OneSAF and Marine Corps Urban Warrior
• another level of the hierarchy rooms, doors,
  windows, sewers, etc.
• Review by much larger peer group

20
Reference Model - Potential Benefits

• More human-like SF behaviors
• Increased pre-simulation and runtime
  interoperability of heterogeneous simulations
• e.g., WARSIM, CCTT
• simulations would agree a priori on an explicit
  subset of the aggregation reference model
• Decreased cost of terrain database production
• Increased terrain database reuse
• Increase in likelihood that more MS requirements
  will be met in the future data source materials

21
Discussion Topics

• Are the authors deluding themselves - does this
  really have benefit for future simulation systems?

• The model only tells what aggregate features may
  potentially contain, not what any instance of an
  aggregate feature does contain. Following some
  of the has a paths leads to silly combinations.
  Is this really any better than a model in which
  any aggregate feature can contain anything?

• Is this of sufficient interest to the SNE forum
  to establish a working group leading to
  standardization?

• Whats the next step?