Geospatial Data Production NSF Workshop on GeoSpatial and GeoTemporal Informatics Jan 8-9, 2009 J. Chris McGlone SAIC, Inc 14668 Lee Rd Chantilly, VA 20151.

Geospatial Data Production

NSF Workshop on GeoSpatial and GeoTemporal Informatics

Jan 8-9, 2009

J. Chris McGlone

SAIC, Inc

14668 Lee Rd

Chantilly, VA 20151

[email protected]

703-676-9228

Energy | Environment | National Security | Health | Critical Infrastructure

What is solved (90%)?

• Photogrammetry: determination of image<->world relationship – Bundle adjustment, direct georeferencing, camera self-calibration

• Synthesis of photogrammetry, computer vision, computer graphics– Geometric and metric properties– Stochastic characteristics– Feature extraction

• Photogrammetry for the non-technical user: photogrammetry (e.g., Photosynth)

• Semi-automated construction of 3d urban models

Energy | Environment | National Security | Health | Critical Infrastructure

Texture-mapped lidar building models, w/ road centerlines

Energy | Environment | National Security | Health | Critical Infrastructure

Visual simulation database examples

Energy | Environment | National Security | Health | Critical Infrastructure

What is almost solved?

• Lidar exploitation (40%)

• Data fusion (based on geometry) (70%)

• Image sequence registration (50%)

• Semi-automatic feature extraction (40%)

• Building interior generation from plans or scans (50%)

• Interaction with 3D worlds (30%)– Navigation– Query– 3D – Immersive worlds (e.g, Second Life)

Energy | Environment | National Security | Health | Critical Infrastructure

What has failed (so far)?

• Automatic cartographic feature extraction– Still scene/image specific – Low productivity due to training, parameter fiddling, or editing requirements

Energy | Environment | National Security | Health | Critical Infrastructure

What is missing?

• Database maintenance models – Updates: scheduled, need based – Reasoning on status

• Integration of crowd sourcing into standard production flows– Verification – Characterization: accuracy, completeness, timeliness

• Simultaneous interaction with multiple datasources, SDIs – Understanding differences (timeliness, accuracy, coverage, design goals,

etc) – Automatically select best source for current purpose

• Non-geometric feature attribution (e.g., function, materials, description)

• Data fusion based on semantics

• Consumer understanding of data quality issues

• New business models for data acquisition, processing, distribution

Energy | Environment | National Security | Health | Critical Infrastructure

What is next?

• Improved realism and interactions for 3D models s

• Crowd-sourced photogrammetry (distributed acquisition)

• 3D modeling as consumer video camera application

• Video game boxes as photogrammetry/modeling/geospatial data platforms

• Mapping from image sequences

• Continuous mapping instead of discrete updates

• Better sensors, new sensing modalities (e.g., lidar + video)