Integrating Data Layers to Support The National Map of the United States

1
Integrating Data Layers to Support The National
Map of the United States
E. Lynn Usery Michael P. Finn Michael Starbuck
usery_at_usgs.gov, mfinn_at_usgs.gov mstarbuck_at_usgs.gov
http//carto-research.er.usgs.gov/
U.S. Department of the Interior U.S. Geological
Survey
2
Project Team

• Current
• Austin Hartman
• Mark Barnes
• George Timson
• Mark Coletti
• Past
• Bryan Weaver
• Gregory Jaromack
• Ryan Stelzleni

3
Outline

• Goals and Objectives
• Approach and Data
• Test Sites and Methods
• Preliminary Results
• Conclusions

4
Goals and Objectives

• The National Map will consist of integrated
  datasets
• Current USGS digital products are single layer
  and not vertically-integrated
• The objective is to develop procedures for
  automated data integration based on metadata
• Framework for layer integration based on metadata
• Framework for feature integration
• Example results for Atlanta and St. Louis

5
Approach

• Integrating disparate networks
• Federated database design via schema mapping
• Physical integration processes -gt vertical
  horizontal
• Layer-based (vertical)
• Use existing seamless datasets
• Determine feasibility based on resolution and
  accuracy
• Feature-based
• Implement integration on feature by feature basis
  using developed feature library

6
DataFocus on 5 layers best available concept

• Orthoimages from 133 priority cities of the
  Homeland Security Infrastructure Program
• National Hydrography Dataset (NHD)
• National Elevation Dataset (NED)
• Transportation (DLG, TIGER, State DOT, others)
• National Land Cover Dataset (NLCD, others)

7
Data Sources, Resolution, and Accuracy
8
Test Sites

• St. Louis, Missouri
• Initially the Manchester and Kirkwood quadrangles
• Atlanta, Georgia
• Initially the Chamblee and Norcross quadrangles

9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
Challenges Facing The National Map

• Institutional (Masser and Campbell, 1995)
• Variation in participant priorities
• Variation in GIS experience among participants
• Differences in spatial data handling
• Technical
• Most datasets are outdated and inaccurate
• Vertical horizontal data integration

13
Technical Factors Complicating Integration

• Total length of coincident participant boundaries
  and network feature density at these boundaries
• Complexity (attribute precision) of the global
  schema

14
MethodsLayer Integration

• Determine compatible resolutions and accuracies
• Use metadata to automatically combine appropriate
  datasets
• Determine transformations possible that integrate
  datasets of incompatible resolutions and
  accuracies
• Determine limits of integration based on
  resolution and accuracy

15
Cartographic Transformations from Keates

• Sphere to plane coordinates projection
• Mathematical, deterministic, correctable
• Three-dimensional to two-dimensional surface -
  planimetric
• Mathematical, deterministic, correctable
• Generalization
• Non-mathematical, scale dependent, humanistic,
  not correctable

16
Scale and Resolution Matching (Mathematical
Transformations)

• Working postulate If data meet NMAS (or NSSDA),
  then integration can be automated based on the
  scale ratios
• If linear ratios of scale denominators are gt ½ ,
  then integration is possible through mathematical
  transformations (12 / 24 K 0.5)
• For ratios lt ½ , generalization results in
  incompatible differences (12 / 48 K 0.25)

17
Generalization Issues

• Selection common features may not appear on
  data layers to be integrated (Topfers Radical
  Law)
• Simplification lines may contain reduced
  numbers of points and have different shapes
• Symbolization for map sources, symbolization
  may result in areas shown as lines or points
• Induction features may have been interpolated
  and appear differently on different sources

18
NHD on Ortho
19
Transportation Overlay on Orthoimages
MODOT
Census TIGER
20
Hydrography Overlay on Orthoimages
St. Louis County Hydro
USGS NHD
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
GA DOT on Ortho (12K)
28
USGS DLG on Ortho (12K)
29
(No Transcript)
30
Sample Results of Visual Interpretation of
Integration
31
Raster to Raster Integration
32
Vector-to-Vector Integration
33
Feature Integration

• Metadata exists on a feature basis
• Accuracy, resolution, source are documented by
  feature
• Use Feature Library with an integration
  application

34
(No Transcript)
35
(No Transcript)
36
Integrating Vector Data with Orthoimagery

• USGS grant partially funding work of Ching-Chien
  Chen, Cyrus Shahabi, Craig A. Knoblock
• University of Southern California
• Department of Computer Science Information
  Sciences Institute
• Approach to identifying road intersections from
  orthoimagery
• Classify pixels as on-road/ off-road
• Compare to road network nodes (intersections)
• Filter algorithm to eliminate inaccurate pairs

37
Technique To Automatically Identify Road
Intersections
38
NGTOC III Algorithm to Emulate USCs Method of
Automated Road Integration

• 1 - Locate nodes (intersections) in vector data
• 2 - Create buffer around nodes and create within
  buffer a image template of road segments
  (geometrically accurate of attribute width)
• 3 - Drop buffer template into the original raster
  imagery
• 4 - Perform pattern matching to identify the best
  match to the template
• 5 - Repeat steps 3 and 4 for all nodes in the
  vector data
• 6 - Filter poorly identified intersections
• 7 - Perform rubber sheeting transformation to
  correct the vector roads

39
Example of Localized Template Matching
40
Vector Intersections (circles) Corresponding
Imagery Intersections (rectangles)
41
Preliminary Results
42
Manually Edited the goal
43
Manually edited -- enlarged
44
MO-DOT and Orthoimagery IntegrationUSC
Algorithm(red MO-DOT yellow processed roads)
Showing improvement
45
MO-DOT and Orthoimagery IntegrationUSC Algorithm
Showing some degradation
46
Preliminary Evaluation of USC Integration Results
Completeness the percentage of the real roads
in images for which conflated roads were
generated Correctness the percentage of
correctly conflated roads with respect to total
conflated roads Displacement - the average
distance between every portion of the conflated
road network to the nearest roadsides of real
road network
47
NGTOC III Algorithm
48
NGTOC III Algorithm
49
NGTOC III Algorithm
50
NGTOC III Algorithm
51
Comparison of NGTOC III USC Algorithms
52
Comparison of NGTOC III USC Algorithms
53
Conclusions

• Geospatial data integration of layers for The
  National Map can only be accomplished with
  datasets that are compatible in resolution and
  accuracy
• Mathematical transformation can automate data
  integration with limited ranges of scales, but
  cannot correct generalization differences
• The National Map will leverage partners data ,
  but technical and institutional integration
  present many challenges
• This research illustrates a design for an
  integration approach (vector transportation data
  with orthoimagery) for geospatial datasets
• Design should support a variety of data sources

54
Integrating Data Layers to Support The National
Map of the United States
E. Lynn Usery Michael P. Finn Michael Starbuck
usery_at_usgs.gov, mfinn_at_usgs.gov mstarbuck_at_usgs.gov
http//carto-research.er.usgs.gov/
U.S. Department of the Interior U.S. Geological
Survey