Team: Parker Coleman | Keith Alphonso | Yancy Matherne
Support: Vance Lowe | Marc Cenac
Concept: During the course of the investigation of the Boston Marathon Bombings, authorities had the benefit of unprecedented levels of surveillance, which was augmented by crowd sourced imagery and video from the many people that were recording the event. According to media reports, this proved to be useful during the course of the investigation.
The Problem: Authorities requested all imagery from the scene of the attack to be sent to an email address. Additional information was requested — such as time and location of photo so authorities would not have to manually extract that information from Exchangeable Image File Format (EXIF) data tagged in the photo.
This request suggests Boston Emergency Operations Center (EOC) Authorities have neither automated nor efficient means to extract EXIF data or use that data to quickly create feature collections using EXIF data already captured in the photo taken by modern smartphones and cameras.
It took 5 days between the bombing and the public release of the suspects’ images. This timeline could have been compressed if a more automated means of georeferencing, filtering and sorting large quantities of crowd sourced media was available for analysis.
Design: The application monitors an email address. Any mail sent to that address will be checked for pictures containing geolocation data in the EXIF metadata. If any such pictures are found, this geolocation data will be extracted, and a link to the image, along with the text contents of the email (which might include valuable information) are inserted into a PostGIS database table. This data is made available on any Open Geospatial Consortium (OGC) compliant map view by means of GeoServer.
The Dash: Devise a solution that allows the public to email a photo, parse the EXIF data and known email address, add that photo to a file server, perform a database insert into a PostGIS table, and view those images on a map.
EOC authorities should eventually be able to filter these images by location, time range and media or device type and view those results on the map.
First line analysts should also be able to flag images for further analysis or dismiss as irrelevant.
Retrospective: We did what we set out to do. Users can email a photo and in a very short amount of time, it will show up as a feature on a map.
We learned from previous mad dash attempts. As such, we decided before the start of the dash on an initial design and hosting strategy.
Using cloud services significantly increased our ability to stand up and make this service available.
There are a number of enhancements that will need to be completed for this to be useful with crowd sourced data collection for EOCs.
This should integrate with existing COP solutions such as OpenCOP. The Mapmypic tool will ultimately be impractical for emergency operations without integration.
Follow On Work:
- Add analysis workflows
- Make tool ‘event’ driven
- Add support for video and audio
- Implement image verification system
- Allow for multiple methods to ‘Map My Pic’ beyond simple email
- Consider scraping publicly available media for data. For instance, public images posted to twitter, facebook or flickr exposes image metadata. Cross referencing an Area of Interest (AOI) and date/time for a defined event can ingest a larger dataset for analysis.