– APPENDIX E–
RESEARCH TOPIC NOTES OF WORKING GROUPS
NGA CORE AREAS
Photogrammetry
-
Go to 4-D space-time maps and ability to search and analyze for events and scenarios
-
Use multi-sensor (cameras, sound) and IMUs on people to do internal mapping of buildings in real time (fire fighters; soldiers)
-
Develop situationally aware tools: need to have products and analysis tools suited to the purpose
-
Analytic integration:
-
Using photogrammetry in the aid of social intelligence: (e.g., automated personal identification, crowd estimation, automatic generation of searchable maps)
-
Use of interactive systems, including gaming, needs to be leveraged by the geo-spatial science in a whole different level to support decision science
-
-
Need to move away from four traditional NGA core areas
-
Blending of computer sciences, statistics, electrical and computer engineering, geodesy, geography, bioinformatics
-
Integration of uncertainty and error into sensor models and analysis
-
Characterize multiple sources of uncertainty
-
Sensor errors
-
Confidence in data (subjective sources)
-
Models (empirical vs. physics based)
-
Utilize advanced statistical estimation, numerical methods, optimization
-
-
Adopt new strategies to address complex problems
-
Interdisciplinary
-
Multi-scale and multi-resolution data integration and analysis
-
More effective use of human in the loop
-
-
Leverage consumer photogrammetry and merge metric and non-metric technologies
-
Merge traditional and non-traditional sensing methodologies (kinematic, participatory networks, social media, surveillance networks)
Remote Sensing
-
Exploit hyperspectral imagery
-
Integrate with other data, GIS, etc.
-
Add time (as described in photogrammetry)
-
Exploit other information (culture, context, etc.)
-
-
Adaptive sensing (real-time) based on information value of sensor
-
Link the above with text information to aid classification and event and scenario recognition; link with visual analytics
-
Exploit atmospheric impacts as signals
-
Uses networks of “small satellites” to gain distributed data
-
Adapt products, tools to end user (first responder, soldier, analyst, etc.)
-
Emphasize multi-sensor fusion and information extraction
-
Decrease uncertainty
-
Exploit redundant capabilities
-
-
Greater utilization of state-of the art algorithms
-
Estimation theory – statistics and electrical engineering
-
Robust nonlinear optimization – numerical analysis
-
Statistical sensor measurement models - nonlinear filtering
-
Advanced software – Object oriented C++
-
-
Coordination with other government agencies
-
DARPA, Air Force, Army, Navy
-
-
Exploitation of knowledge sources beyond image data mining; make relevant knowledge sources available; knowledge-based classification
-
Enhance change analysis – beyond the process of measurement and classification to dynamics, behavior, and prediction (issue of sensor control and tasking)
-
Need more than just the inanimate landscape, but also the dynamic, social environment (e.g., the flux of a living city) = GEOINT
-
Metadata and tagging – hey for fusion; relate to other non- GEOINT sources (semantic and tagging interoperability challenge)
-
Augmenting the image analyst –more tools, knowledge, visual analytics, automation, mining given a specific remote sensor
-
Infrastructure implications – data storage, distribution, and throughput to the analyst
-
Remote sensing: We have lots of data (increased availability of commercially collected data). Can we analyze this data?
-
Data collection agency, delivering tools for data analysis (multi-resolution, multi-sensor, multi-platform, multi-temporal, current and future sensor technologies – including new sensors that are not fully understood)
-
Cross-Cutting Issues
-
Data: bring processing closer to the data acquisition system (selective provision of data)
-
How to incorporate third party data and information into NGA processes (reliability, metadata, etc.)
-
Need more comprehensive metadata
-
Processing to support near-real time processing of constant data streams from drones and UAVs
-
Need to blur processing distinctions between satellite, aerial, and terrestrial data acquisition systems
-
Quality of information
-
Reliability and integrity of automatically generated spatial information
-
Scalability
-
More comprehensive use of supporting information (e.g., environmental)
-
-
Quality assurance: system calibration, mission planning for different applications
-
Quality control: verifying the quality of the different products at different levels (sensors, data, information, and knowledge)
-
Develop test sets for different products
-
-
Blending of information:
-
Interface across different information types
-
Information fusion (integration of open source information – quality control of information – evaluating the reliability of this information)
-
-
Information and data presentation
-
How to compress petabytes of data to kilobytes of information for presentation to the end-user
-
Supporting information needs to be more fully utilized
-
-
Automation:
-
Is full automation possible and do we need full automation? (reliability issue)
-
Provide increased human support to carry specific tasks
-
-
-
For example: Tuning the learning models (more of an art that relies on the expertise of the operator ◊ reducing the level of expertise required
-
-
Modeling and data processing:
-
Modeling of non-traditional and emerging sensors (e.g., DSLR, flash LiDAR, range cameras, etc.)
-
Data, information, to knowledge transformation
-
High resolution versus low resolution – local versus global coverage – smart sampling of the landscape
-
Considering the time dimension in geo-spatial data analysis (e.g., pattern of life assessment)
-
-
Fusion
-
Models for determining the optimal sensors and data needed for deriving desired information (requires data repository that have been geometrically, radiometrically, stochastically checked or pre-processed)
-
Information fusion: facial reconstruction, CV
-
Evaluate the results (how it relates to the end goal), understanding the data
-
Cartographic Science
-
Multi-scale to continuous scale maps
-
Beyond tile-based mapping, beyond Mercator projection
-
-
Improve speed of map presentation
-
Multiple scale levels, all scales
-
-
Interactive cartography driven by eye tracking, brain sensing, other body sensors
-
Beyond cartographic scale
-
Investigate semantic aspects of scale
-
Represent human activity at multiple temporal scales
-
-
Need timely access to GEOINT at differing scales based on differing user tasks
-
Incorporation of volunteered geographic information
-
Social “mapping” in space and time in addition to physical objects and terrain… research challenges? Social links and networks have geospatial characteristics, how to deal with them in a spatial sense? … have to interface with other types (agencies) of “INT”
-
Address challenge of how to visually present data and information quality, reliability, and confidence
-
Determine what information is needed by particular users and determine the appropriate evaluation methods
Geodesy
-
Integration of GPS in all aspects of geospatial technology
-
Applications still in infancy
-
-
Increase proficiency in use and interpretation of GPS positioning
-
Provide means of assurance that people using GPS for particular tasks know what they are doing
-
-
Ubiquitous GPS
-
Integration of multiple receivers; phone, navigation
-
-
Expansion of continuously operated reference system (ground based) – CORS
-
Geodesy does not deal with humans directly (classical defn.), but gives information that supports societal and scientific needs… but reference frame is “invisible”
-
Impressive progress in geodetic accuracy… but how to “operationalize” geodesy missions and services? What should NGA do? GPS/GNSS used in many positioning apps… could we cope without it? What about difficult environments where GNSS doesn’t work?
-
Establish a geodetic reference frame at sub-millimeter level; research needed at observational level; drives high performance computing research, etc.
-
Next generation of positioning instrumentation and inertial navigation systems stable to the centimeter level over time
-
Geophysics: collaborative research, could be informative to NGA (in terms of data)
-
Application oriented datum; provide transformation
-
Gravimetry: UAVs; time dependent gravity; GRACE mission
GIS and Geospatial Analysis
-
Continue to pursue temporal dimension
-
True, comprehensive and complete space-time GIS and geospatial analysis does not exist
-
Expand the narrative
-
Geospatial discourse constrains possible tasks
-
Restricted GIS vocabulary to communicate tasks
-
Production of narrative products at multiple levels of explanation
-
-
Incorporation of volunteered geo-information
-
Rating system for accuracy
-
-
Need to understand how to work with the narrative framework
-
Need to achieve timely automated extraction; need OO software approach
-
Automated service and workflow discovery to enable automatic tool application
-
Conceptualize complex information into story line
-
Communication of geo-spatial issues
-
Static and dynamic communication of narratives
-
Visualization of narratives
-
Cross-Cutting Issues
-
Are the core NGA areas “stovepipes?” Are they the right ones?
-
How do people respond, perceive, and trust quality statements, especially for large amounts of data?
-
Need integration of geo information from unstructured sources (text), physical domain, social domain, and knowledge domain with GEOINT
-
Use game based analytics: explore data set in terms of games; analyze game strategy and pattern; use information for interview techniques
-
Cognitive effectiveness of geo-spatial technology
-
Brain scans, MRI, eye/scan patterns, etc.
-
-
-
logical
-
physical
-
-
Broader cross training of students in geo-spatial workforce … computer science, behavior, …
-
-
understanding … geophysics, geodesy, …
-
facilitating interdisciplinary training and research
-
CROSS-CUTTING THEMES
Forecasting
-
Challenges
-
Predicting human behavior - relating social factors to physical factors. Geospatial elements need to become part of social network theory
-
-
-
No grand unified social science theory. There are multiple theories from many different parts of the social sciences
-
Low-hanging fruit - gross human behavior may be predictable to some level
-
More study required on foundational framework of social science integration with geospatial data
-
Rare events - perhaps some focus on predicting the unpredictable
-
-
Spatial data analysis methods need to be incorporated to get better predictions that put in spatial relationships and meaning
-
Need to tie together of spatial data and temporal forecasting
-
What are the validation methods? Need to develop general validation approaches. Need sensitivity analysis
-
Should we distinguish between prediction and forecasting? Be sure that all aspects of these areas are being covered
-
The forefront of modeling. More complex models that are combinations of very different models for actionable results
-
Technosocial predictive analytics
-
Interesting interplay between social networks and physical infrastructure
-
Needs systematic work on defining priors
-
-
black swans are a challenge since not enough data on extreme events
-
-
Needs visual analytics as a visual tool to gain better insights
-
Links possible with geocollaboration
-
-
collaboration over time, space, expertise
-
-
Beginnings of applying computing to sociology and anthropology exciting!
-
-
Modeling of human behavior – more interactive and real-time forecasting tools where problem domain is constantly changing. Use of normality modeling and anomaly detection as alternative to deductive based forecasting
-
Computational modeling, prediction, and analysis are important research topics for the future
-
Potential to guide data collection and assimilation
-
Participatory Sensing
-
Uneven distribution of sensed data
-
Privacy issues
-
Crowd sourced data aggregation methods need to be developed
-
Understanding when crowd sourcing is useful
-
What about foreign countries or areas where you can’t apply your structure?
-
A very powerful way of collecting GIS data “unstructured collection.” Need on-the-fly planning
-
Use the GIS as a framework. May already have some 3D models, images, etc.
-
Building shared spatial knowledge bases with participatory input and sharing. Active knowledge bases
-
Directed planning; opportunistic planning. Situationally aware models. Need to get actionable results. Spatio-temporal models of social, political dynamics
-
Trust and confidence; how to account for biases and keep this information with collected data. How to do quality control in a messy data environment. This needs new ideas
-
Add reference data (reference models?) as points of validation with data of uncertain accuracy and provenance. This could be a general approach
-
Embed social networking in spatial-temporal. Insert the idea of locality and spatial structure in social network analyses
-
Quality control
-
Need methods to aggregate measures of quality
-
Timeliness is an important dimension
-
Measures of trust, reliability, provenance: don’t trust; verify
-
Spot-checking with high-quality, calibrated sensors to improve trust and quality
-
-
Judicious use and context of information collected
-
For example, owner-defined property lines and conditions valuable in non-legal contexts
-
-
Systematic approaches to integrate information from multiple sources:
-
Domain knowledge and expertise such as local context (cultural)
-
Participatory data analysis – Wikipedia over GIS
-
Counterpoint to the deep and intensive thinking of the analyst
-
-
How to engage all relevant sub-groups (age, gender, socio-economic) in participatory data collection?
-
Develop the wider model against which participatory data can be tested. Use of prior knowledge for improved registration and classification
-
Understanding of the quality compromises and strengths of having mixed use of authoritative and public participatory data – requires broader development of the models of use
-
Understanding the relationship of culture and social factors to policy and practice of collection and use of public participatory data. Research into security issues of participatory data
-
Participatory sensing: Integration is important!
-
How to influence social media to generate data that is needed
-
Research to calibrate and judge quality of sensor in participatory sensing to allow decision making
-
Data fusion from this data with serious geo information?
-
Visual Analytics
-
Specific interfaces for specific users? Emphasize the generalization. What are the underlying fundamentals?
-
How to get from visualization to underlying methods? Need to understand the domain areas. Can general principles be extracted?
-
Developing a repeatable body of knowledge within visual analytics; for example, generic rules applying to the interpretation of data. Develop evaluation criteria
-
Integrated tools. Integrated, iterative, interactive—this is the new thing that visual
-
Interactive part of visual analytics is a key aspect of its contribution here analytics can bring, even using existing analysis tools (no toolkits)
-
New ideas derived by looking at aggregates. Individual locations to aggregations that make sense for groups. Functional and meaningful scales and multi-resolution methods. Attach meaning to aggregations. Space and time aggregations
-
Visual is not the only sense as you only reach a small part of the population (19%)
-
Interactive analytics is may be the right term
-
Metaphors for interaction with models and animations need to be developed
-
Integrated spatial and temporal analytics
-
Understanding the use of animation
-
Modeling, simulation, and high performance computing
-
Proper depiction of data quality and error uncertainties
-
Games
-
Social interactivity
-
Importance of design and art as an additional skill to be embraced
-
Workflow
-
Domain-driven integration of information from multiple sources
-
Take advantage of human cognitive abilities
-
-
Need to address how techniques work across scales
-
Agent-based approaches, links, etc.
-
-
Need new advances in interaction for visual analytics
-
Further strengthen bridges between visual analytics and other areas
-
Visual narratives
-
Causality
-
-
Quality of the visualization
-
Develop techniques to measure quality of the presentations
-
Minimizing unintended artifacts, illusions, confounds, etc.
-
-
Visualizing and communicating uncertainty. Development of interactive visualization tools - dynamic feedback with analyst through eye-tracking and other sensors
-
Collaborative two-way participatory augmented reality
-
Achieving the correct balance between full automation and visual analytics assisted decision making – how to decide which to use in specific situations?
-
Computational modeling and/or visual analytics
-
How to enable human reasoning with large amounts of heterogeneous geospatial data?
-
Data fusion
-
Deal with users
-
-
Science of interaction: Need to develop adaptive visual analytical methods to support geospatial users
Beyond Fusion
-
Data fusion
-
Relate to geo-space:
-
represent spatial and non-spatial dimensions
-
incorporate spatial structure: spatial variation or spatial correlation
-
couple spatial and non-spatial algorithms
-
-
-
-
-
time dimension?
-
-
Vector space and graph space; opportunities to integrate or couple? Cross-correlate outcomes? How to represent and handle uncertainty?
-
-
Different forms of spatial data
-
High-resolution, attributes cross space
-
Location (point or area), boundary, space of different scales
-
Models
-
Best way to combine GIS data layers, coding, incorporating uncertainty
-
-
Non-spatial data fusion (as in Haesun Park’s talk): cognitive domain
-
Cognitive aspects of knowledge fusion
-
-
Fusion challenges
-
Scale
-
Semantic interoperability
-
Different resolutions
-
Fusion at different levels (data, information, and knowledge)
-
-
Heterogeneous data of different fields and kinds of knowledge, disparate terms and understanding
-
GPS positioning
-
Data on positioning and gravity are uncorrelated, nicely separated
-
2-, 3-, 4-D geodesy, not much to gain from data fusion
-
Essentially, it’s about data understanding
-
-
Fusion has a lot to achieve, let alone beyond fusion
-
Is it the same as merging? Conflation is part of fusion
-
Need for clarification, vocabulary, a scientific language
-
-
Can disparate data, information, and knowledge be put together? Redcross, trusted feedback, outdated geospatial data together
-
Would techniques presented take care of these?
-
Overarching issue of uncertainty labeling for broad NGA data set needs to be addressed; what is uncertainty of high-dimensional data?
-
-
A set of techniques for understanding relations in high-dimensional data
-
See also manifolds, etc.
-
Applications to GI data not shown
-
Loss of visibility of space and time at “preferred” scales
-
-
Powerful, but evaluation methods need to be developed
-
Do not stand alone—insight needs to be developed alongside
-
Analyst interaction important
-
-
Also need methods to understand large disparate data bases
-
Interrelationships possibly not understood
-
-
Both broader understanding and uncertainty reduction will likely require complementary, non-GI data
-
Comparison of fusion algorithms from visual analytics with existing fusion algorithms
-
Early fusion, mid fusion, late fusion
-
Bayesian fusion algorithms
-
Hard-soft fusion using hard sensor data and text, human generated, web derived information
-
Compare and evaluate the accuracy and applicability of these two types of fusion algorithms
-
Need scalable algorithms to handle large volumes of data in real-time and interactive mode
-
Would approximate, but faster algorithms be desirable?
-
Need to develop systematic approaches to matching computationally driven interfaces to user work practice
-
Need to investigate existing standards such as the Predictive Model Markup Language (PMML) to use the same data for different classification algorithms.
-
How to retrieve geospatial documents and extract geospatial information from text is still a challenge
-
How to use existing geospatial ontologies to inform the information extraction process
-
How to enable human computer interaction when complex modeling is involved
-
Develop methodology to create heterogeneous benchmark data sets for research
-
Formulation of standards for methodology and data structures
Human Terrain
-
Human landscape is a better term – human condition, biophysical conditions
-
Economy, sociology, transportation, anthropological, ethnic, religious, cultural, historical
-
-
Geospatial, social, cultural data integration and analysis
-
More systematic approaches in collection, coding, displaying, understanding
-
Categorizing trivial and non-trivial data
-
Voluntary and non-voluntary contributors may not be aware of the consequence of making data available
-
-
Data uncertainty, quality, consistency, reliability, disparity, fuzzy
-
Tools to filter and clean up data
-
Identify what data is necessary for a given task
-
-
Collaborative tools for crowd-source data
-
Interactive tools
-
-
Proper analysts with specialized knowledge
-
Human intervention to double check the quality (human in the loop)
-
-
Human terrain
-
Relate analytical outcome based on the significance of consequences of prediction errors; should we weight the outcomes accordingly?
-
Assess possibility or level of confidence on data and analytical outcomes
-
Interoperability: customized system vs. open system; closed sourced black box? scalability? Need to consider modularized system and develop API to couple with other systems
-
Need a stronger geospatial component in social network analysis; dynamic relationships over space and time,
-
Social networks in virtual space vs. in physical world
-
-
Cross-cutting
-
Complexity of analysis: ability to interpret the results
-
-
-
Absolute single result vs. multiple possible outcomes; means to assess and communicate uncertainty in decision support
-
Develop an architecture for supervisory level model analysis that combines outcomes from multiple models to mediate meaningful and coherent advice
-
historical studies: run models against historical or past data
-
compare outcomes from multiple models
-
-
Differential uses of words or dialects in different places
-
how to understand how people use language in the context of place (place-dependent use of words or phrases)
-
identify clues used in a language, relate the outcome in an analytical manner back to the spatial context (to know where the communication took place)
-
-
-
Methods to enable analysis in native language
-
Human terrain-based dynamic network analysis seems to serve well as one basis for structuring a broad range of social phenomenology in space-time
-
Representation and visualization in GI space an issue
-
-
Quality assertion, quantification an issue
-
Highly disparate underlying data quality levels; need agreed ontology
-
NGA to develop technical and ethical best practices for collection?
-
Interplay between space-time accuracy and relational accuracies
-
Deception possible, not easy to detect
-
-
A form of “narrative?” Perhaps useful to assess commonalities, distinctions in these methods
-
A larger issue lurking here? Methodological synthesis to deal with the space-time dynamic
Cross-Cutting Issues
-
Computation (cloud computing, mobile computing, analytical servers)
-
Distribution of data and data storage
-
Customization of products—making dynamic products for end users to dissect, modify, traceability of evidence and logic (case files FBI or doctors)
-
Validation, data quality, spatial uncertainty
-
Populist information: privacy, uncertainty, NGA’s role? Use to validate directly gathered data
-
-
Multiple levels of uncertainty (data, model)
-
Move to knowledge, wisdom, insight
-
New paradigm of uncertainty (based on analytical needs at hand)
-
-
Advancement of sensors
-
Sensor calibration
-
Smart sensors, miniaturization, on-board computing
-
Infrared, radar (better sensors)
-
Don’t lose focus; don’t forget the sensors
-
Don’t forget the core areas
-
Scenario modeling to deploy appropriate sensor for task (weather, geography, etc.)
-
-
Temporal analytics
-
-
Partner with NSF and other government entities, and with other international science entities
-
How do Advances in the Cross-Cutting Themes Shape the 5 Core Areas?
-
Can’t lose track of the 5 core areas
-
Cross-cutting themes support the 5 core areas, but can’t ignore or replace them
-
Cross-cutting themes need to show value to the core areas, not a substitute
-
Mathematics, visual analytics can directly benefit NGA and its missions
-
-
Adding time to space
-
Rich extension
-
How to do this? Visual analytics, 4D GIS. Time is difficult to represent; temporal analytics?
-
-
No stove-piping in 5 core areas
-
Also applies to cross-cutting areas
-
These areas blend together (look for and/or promote innovation at the intersection of these areas)
-
-
Science development needs to be plugged into international science community