Network surveillance cameras
MSc dissertation proposal 2011/2012
Modelling
and Operating Surveillance Cameras
Introduction:
The increasing need of surveillance in public spaces and the recent technological
advances on embedded video compression and communications made camera networks
ubiquitous. Most of nowadays surveillance cameras are however still very
conventional: imaging resolution is usually 640x480 (or less) to keep low
bandwidth usage, pan-tilt-zoom are manually driven, fisheye lenses wider than
180deg are not common due to a harder interpretation by humans, etc.
This project proposal is focused on removing some of the restrictions.
The interfacing standard is network based (i.e. considering IP cameras), and it
is assumed that some processing can be handled in the camera (IP cameras
usually have a computer inside running linux). Some
restrictions to remove involve maintaining scene representations on the camera
and implementing some automatic tracking behaviours.
Note:
this project is proposed in the framework of two projects, a national project
on camera modelling and calibration, and an EU financed project (QREN) on
modelling and using high-definition surveillance cameras.
Objectives:
The main objectives of this project are: (i)
interfacing to novel IP cameras, (ii) calibrating the cameras and the
camera-network, and (iii) detecting and tracking events.
Detailed description:
This project proposal is focused on modelling and operating network (IP)
cameras, and placing memory and processing capabilities on the camera. The IP
camera interfaces are currently receiving much attention in the form of a large
consensus on what the interfacing standard should be. Major surveillance / IP
camera manufacturers have recently introduced the ONVIF standard [ONVIF-www]
which has effectively the capabilities to be a global standard on the video surveillance
market.
Placing memory and processing on the cameras has clear advantages on
keeping low bandwidth constraints. Some of the working directions come from
video encoding technologies. For example saving locally background images
allows sending just incremental changes. Compensating imaging artefacts and tracking
moving objects are additional methodologies allowing bandwidth saving, and
therefore incorporating high-definition cameras within the currently installed
networking hardware.
Exchanging information directly among the cameras is another way to save
network bandwidth. Handing over events among cameras, e.g. informing of moving
objects that are expected to enter the field of view of neighbouring cameras, usually
involves calibrating the camera-network. Given a number of reconstructed
3D-points of the environment and their images we can calibrate the network of
cameras using standard computer vision methodologies [Hartley00]. This is also
one aspect to focus in this project proposal.
In summary, this project proposal involves:
1) Developing a software interface to novel high-definition IP cameras
2) Calibrating the individual cameras and the camera-network
3) Detecting and tracking moving objects; monitoring of tracked events
over the camera network
References:
[Hartley00] R. I. Hartley and
[LoweWWW] David G. Lowe. Demo software: Sift keypoint detector. http://www.cs.ubc.ca/˜lowe/keypoints/.
[Axis] IP camera manufacturer actively involved in the definition of the
ONVIF standard http://www.axis.com/pt/onvif/index.htm
[ONVIF-www] open industry forum for the development of a global standard
for the interface of IP-based physical security products http://www.onvif.org/
Requirements (grades, required courses, etc):
-
Expected results:
At the end of the work, the students will have enriched their experience
in computer vision applied to surveillance camera setups.
Place for conducting the work-proposal:
ISR / IST
More MSc dissertation
proposals on Computer and Robot Vision in:
http://omni.isr.ist.utl.pt/~jag