Geospatial Exploitation of Motion Imagery
From MilcordWiki
| Geospatial Exploitation of Motion Imagery
| |
|---|---|
| Mission | To extract and spatiotemporally index moving objects from motion imagery for querying based on object attributes and relationships |
| Sponsor |  Army TEC
|
|
Watch demo video! |
|---|
Overview
GEMI is a geospatially aware and integrated Intelligent Video Surveillance (IVS) software system targeted at real-time and forensic video analytic and mining applications that require low-resolution detection, tracking, and classification of moving objects (people and vehicles) in outdoor, wide-area scenes.
Need
The exponential growth of motion imagery datasets collected by an ever increasing pool of sensors is presenting a daunting challenge for image analysts. Current motion imagery analysis tools offer rather limited analyst support (e.g. simple frame capture and insertion into mosaic imagery). As the current tools fall short of analyst needs, and the analyst workforce remains relatively fixed, most of collected motion imagery datasets either goes through minimal analysis, or is archived for future access.
Approach
GEMI (Geospatial Exploitation of Motion Imagery) indexes the moving object content of video feeds captured from diverse platforms (e.g., surveillance cameras), and enables the query of video imagery for user-specified patterns of movement (e.g. search for video segments containing a vehicle coming to a stop on the side of the highway).
Video data is ingested from ground and aerial sensor platforms. The video motion video is transformed, geo-referenced, indexed, extracted, and stored in a geo-database, enabling pattern discovery and complex queries, such as ‘find a black SUV that stops for 2 minutes within 100 meters of the overpass before checkpoint 541 between 5PM and 6PM and two people exit and return to the vehicle’.
Our technical approach employs an innovative Spatiotemporal Object data structure that provides a compact representation of an object’s attributes, including radiometric (color), geometric (size, shape), and behavioral parameters (velocity, turns, acceleration patterns). When suspicious events are detected either in real-time or from a database search, our geo-referenced view enables users to pinpoint and trace back the activity, and our integration with GIS (Geographic Information Systems) is used to direct response with better situational awareness.
Benefits
GEMI software is a novel approach aiding the analyst to sift through massive video collections (either in real time for incoming or forensics on archived video feeds) to accomplish mission-critical tasks for incorporation in existing motion imagery capabilities of Army topographic terrain units, military intelligence units, and National Geospatial Intelligence Agency (NGA).
Applications
- Reconnaissance and IED Emplacement: Infer suspicious behavioral patterns of humans and vehicles by analyzing motion contained in video surveillance in the IED kill chain
- Critical Infrastructure Protection: Monitor CCTV to alert security personnel for unauthorized or suspicious behavior
- Border Security: Analyze activity near the border to classify intrusions as drug trafficking vs. illegal immigration based on behavior
- Competitive Advantages:
- Unlike current IVS solutions that operate in the image plane, GEMI operates in the geospatial frame of reference
- Enables geospatially referenced queries of video archives
Get Involved
- Want to evaluate GEMI as a beta tester?
- > Please contact us for more information.
- Already a beta user?
- > Please post bugs and feature requests on our Bugtracker.
Documents
- Installation Guide & Tour [ 536.5 KB DOC ] (password required)
- User's Guide [ 1.40 MB DOC ] (password required)
- FAQ [ 1.14 MB PDF ] (password required)
- Overview [ 7.69 MB PPT ] (password required)
Downloads
- GEMI Installer [ 10.09 MB EXE ] (password required)
- Required Apps [ 20.35 MB EXE ]
- Recommended Apps [ 24.97 MB EXE ]
- .NET Framework 2.0 (for MapWindowGIS only) [ 22.42 MB EXE ]
- Sample Project [ 35.68 MB EXE ] (password required)
References
- Windholz, T., Toothaker, M., Stefanidis, A. and Caglayan, A. (2007) “Geospatial Exploitation of Motion Imagery” , Technical Report, DTIC AD Number ADB329555, Milcord LLC, Waltham, MA May 2007.
