A Policy Based Media Controller in Project ReSCUE

1
A Policy Based Media Controller in Project ReSCUE

• Jooyoung Park Mentor Prof. Sharad
  Mehrotra
• Monitoring media-instrumented spaces via vision
  processing on cameras

Introduction My project goal is to perform
image processing on cameras at the multi-media
sensing environment within Rescues Responsphere
infrastructure. Privacy Preserving Workspace
Within such a system, infrastructure that
supports privacy-aware constraints can be
employed to preserve privacy of monitored
individuals. Efficient Monitoring System
An automated triggering mechanism can detect
high-level events and dispatch necessary alarms
as fast as one still image per 1 millisecond.
2. Rescue Biohazard Drill Video Processing

CAMASVM Project and Its Progress 1. The Coffee
Cam Project
Coffee Cam project aims to detect
violations of coffee room policies using images
acquired from Linksys WV54G cameras with audio.
The coffee room has two cameras, one for
capturing mug shots from the coffee machine and
the other for capturing a broader area of the
coffee room for the purpose of detecting
violators. With images received from these
cameras, the CAMASVM coffee cam system detects
high-level coffee room alerts for events
including the brewing of a fresh pot, the
emptying of the pot, an accidental leaving of the
burner on, and coffee spill. When any of these
violations occur, the CAMASVM gets a shot of
possible violators. My job in this project was
to integrate the work of graduate students to
make the system operable. For detecting persons
in the images, I used JAI (Java Advanced
imaging).
My job is to process videos from a biohazard
drill conducted by the ReSCUE team and detect
events including counting the number of people in
the camera scope, tracking humans, detecting
firefighters and rescue teams by the color of
their t-shirts ( bright blue for rescue team and
red for firefighters), and performing face
recognition. To execute this job, I used openCV
(open source computer vision for real time
computer vision).
Further development Little Sister
Recognition need to detect small mobile robots
with attached cameras. Event Log create a
log of events with corresponding timestamp.
real-time processing Real-time processing using
CAMASVM software which produces jpeg files from
streaming video may be done using Java.
Future development Development of web-based
system that integrates multi-camera
infrastructure along with privacy preservation
policies. Face recognition using openCV or
similar library will be deployed for person/face
recognition of coffee-room violators. Voice
Alarm System When violation occurs, an
automated voice system recorded in .wav file will
be played in the coffee room to alert the
violator that he/she has made a violation.
Acknowledgements Dr. Weigi Yan Cal IT2 Daniel
Massaguer, Hojjat Jafapour, Shengyue Ji grad
students
S ummer U ndergraduate 2 R esearch
0 F ellowship in 0 I
nformation 6 T echnology
jooyounp_at_uci.edu www.research.calit2.net/stude
nts/surf-it2006 www.calit2.net