We are a research team at Spatial Sciences Institute, University of Southern California.
We develop computer algorithms and build applications to solve real world problems in spatial sciences.
2. USC SSI Programs
• BS in GeoDesign
• MS and Certificate in GIST
• MS in Spatial Informatics
• PhD in Population, Health and Place
GeoScavenge
3. Spatial Computing: Who we are?
• We are a research team at
Spatial Sciences Institute,
University of Southern
California
• We develop computer
algorithms and build
applications to solve real
world problems in spatial
sciences
4. Spatial Computing @ USC SSI
• Since 2013, we worked with 45 students and 6
postdoctoral researchers
• one local high school student, a number of visiting
international students, and some USC
undergraduate and graduate students
• GeoDisgn, electrical engineering, spatial informatics,
computer science, and data informatics
• A third of the 45 research students are female
students in engineering
5. What is the problem?
5
Large Volumes and Varieties of Heterogeneous
Geographic Data
Manual conversion of large volumes of maps to a usable format for data analysis is time
consuming and does not scale
Problem
6. What are we building?
6
Large Volumes and Varieties of Heterogeneous
Geographic Data
Problem
We build algorithms and tools to bridge the gaps
e.g., Strabo
8. Motivation
• Existing data sources typically contain only
contemporary datasets
• e.g., present place names
• Maps contain detailed geographic
information at various times in the past
• spatiotemporal datasets that cover long
periods of time and large areas
Land reclamation in Hong Kong (http://www.oldhkphoto.com/coast/)
9. Use Case: Identify Contamination Sites in the
Past from Historical Ordinance Survey Maps
11. Use Case: Identify Pollution Sources in the Past
from Historical USGS Maps
Circa
1956
Circa
1921
Railway transportation is a serious
source of pollution but many of the
railroad records no longer exists
12. Exploiting Context in
Cartographic Evolutionary
Documents to Extract and
Build Linked Spatial-
Temporal Datasets
• Editions in map series not
independent
• Change incrementally (updates)
• Overlap in content
• Can be used as training data for
feature extraction!
2012
1964
1950
A Case Study and OutlookMap Processing: Impact & Challenges GeographicContext & Map Processing
US Na-tional Science Foundation award IIS 1564164
and 1563933 to the University of Southern California
and the University of Colorado at Boulder
“Exploiting Context in Cartographic Evolutionary
Documents to Extract and Build Linked Spatial-
temporal Datasets”
13. Information Extraction &
Geographic Context
(1) Building contextual information
• Create generic semantic models:
• Locations, Type & Attributes
• Geometry (e.g., line feature, width)
• Inferring semantic rules ((un)likely situations)
(2) Adaptive graphics sampling
• Collect spatially constrained graphics examples
• “LOCATION” to define sampling areas
• Overlap: map contents & contextual data
(3) Compute feature descriptors: Knowledge base creation
• Shape, color, texture descriptors to be used in matching process
GazetteerAdmin
Records
(x,y)
A Case Study and OutlookMap Processing: Impact & Challenges GeographicContext & Map Processing
17. Building Knowledge Graphs from
Public Data for Predictive Analysis
• A Case Study on Predicting Technology Future in Space and
Time
Ontology-based integration
18. Annotate Other Historical Materials with
Map Content
Murray Burger’s
testimony
The USC Shoah Foundation contains 53,000 audiovisual testimonies of
survivors and witnesses of the Holocaust and other genocides that have been
catalogued and indexed at the Institute
Use map content to enrich the
testimony metadata