Project Number:
09-05
Research Project:
Thousand Eyes on California's Streets, Roads, and Infrastructures
P.I. Name & Address:
Martin Krieger
University of Southern California
School of Policy, Planning and Development
Los Angeles, CA 90089-0626
Email: krieger@usc.edu
Website: http://www.usc.edu/schools/sppd/faculty/detail.php?id=18
Phone: (213) 740-3957
Fax: (213) 740-1801
Co-P.I.
Ramesh Govindan
University of Southern California
Department of Computer Science
Los Angeles, CA 90089-2905
Email: ramesh@usc.edu
Website: http://cs.usc.edu/~ramesh/
Phone: (213) 740-4501
Fax: (213) 740-7285
Project Objective:
This project will explore the development of technologies that will dramatically increase the level of video surveillance of our transportation network and its infrastructure. This technology leverages two key ideas. First, the ubiquity of cell phones permits citizen sensing, in which ordinary citizens can report accurate and timely information using these net-connected mobile devices. Second, the emergence of video capability al.
This project strives to design and develop a real-time information system that allows for dense and widespread surveillance, but centralized evaluation and comparison. Essentially, this allows citizens to take videos of ongoing traffic and of the transportation infrastructure, and automatically, and almost instantaneously, transmit them to a centralized server. Thus, truckers might report on commercial goods movement problems, and ordinary citizens might capture problems of safety and security. These videos are annotated with position information (through GPS), and they are also time-stamped.
At the centralized server, and at any Internet-accessible site, these reports can be analyzed quickly by transportation and security experts, who can then appropriately respond to the incident at hand. Compared to existing methods, this approach requires little investment, and can increase, by an order of magnitude, the number of perspectives into the transportation system.
Task Descriptions
Video clips are captured on smartphones operated by users, then automatically transmitted to and stored at one or more servers located at different points on the Internet. These records are automatically indexed to enable easy search and navigation of the corpus. Finally, users examine these records using novel displays that help them analyze the submitted reports.
An important component of this system is capture and transmission. This component enables the collection of the videos clips by relatively unsophisticated users (truck drivers, citizens, law enforcement officers). These reports are immediately transmitted back to the server. Our capture and transmission software, called vcaps, has a simple and intuitive user interface on the phone (essentially, a shutter button) and in the display.
The collected video records are tagged with various pieces of information: the time of capture, the location, as well as other technical pieces of information (such as the kinds of encoding standards used for video and audio, or the ambient light levels, camera exposure settings, microphone gain levels, and so forth). These records can be robustly transmitted by vcaps software in the cellphone, which discovers available network connectivity options and attempts to upload the records with minimal user intervention. In a fully developed system, we envision the need for multiple servers distributed across the Internet, for reasons of scale (if successful, the system could have a large number of participants) and robustness (replication of records across servers would increase the availability of the video clips).
At the servers, back-end software will automatically collect and store retransmitted records, replicate them if necessary to ensure availability, and then index the records to enable various forms of search and navigation of the corpus. For example, it will be possible to simultaneously view all videos taken at "nearby" locations, at a given time. Similarly, a user may wish to view videos taken at different times at a given location.
To enable this, our software will present multiple videos in a tiled fashion (similar to the views available in commercial surveillance cameras, for example). Such displays can present multiple aspects of a complex event (e.g., a traffic incident), and enable viewers to build sophisticated internal models of the event, resulting in more accurate analyses. Moreover, it is possible for mobile phone users with Web access to view the videos, so that these reports can be available to the public at large.
Milestones, Dates:
August 15, 2008 – September 14, 2009
Total Budget:
$79,650
Student Involvement:
One research assistant
Relationship to Other Research Projects:
Related to 08-12; part of the safety, security, and vulnerability focus area
Technology Transfer Activities:
Project report will be posted soon
Potential Benefits of the Project:
This project strives to make the transportation system safer, improve travel times for commuters as well as commercial goods, and provide real-time traveler information.
TRB Keywords:
Infrastructure, citizen sensing, goods movement, security
