[plug] open source video (burglar) surveillance
Denis Brown
dsbrown at cyllene.uwa.edu.au
Wed Sep 24 22:48:59 WST 2003
Dear PLUG list members,
I'm wondering if anyone has investigated or used open source software for
the purposes of video surveillance for theft. The context is my son's
preschool where several pilferings have occurred. I'm musing about the
possibility of using (something like) the Wizard mini PC, the USB ports
interfacing a video camera or two and the processing being done on a
frame-difference basis mitigated by time of day, since we'd be interested
in out-of-hours activities, region of interest and by the change magnitude
- to eliminate false triggerings by dogs, cats, birds, ....
The school is considering getting broadband so that would make keeping an
eye on things (no pun!) a bit easier. This should be a modern spin on
something we (industry) used to use many years ago in the slow-scan TV
line. A frame capture device would compare successive frames of video
and, in the presence of a significant change in a defined
region-of-interest, would raise an alarm. It was pretty crude by today's
standards but it did the job :-)
I've just done a bit of Googling and had a hunt on Sourceforge and
Freshmeat without too much in the way of direct hits. There are several
commercial approaches, including a turn-key solution from iomojo, several
offerings from Axis (manufacturer of video and other servers), etc. I
don't doubt there is something out there but neither do i fancy ruffling
through Mr Google's nearly 8000 hits on "surveillance, video, linux, open
source" and so on keywords. Freshmeat turned up net-cam but documenttion
is sparse and it seems to talk of Axis-2000-something cameras, probably
related to the other Axis reference.
In summary: one or two (or more?) webcams connected (USB?) to some cheap
Linux-based hardware. Within a defined time of day, and region of
interest in the camera(s) field(s) of view, a snapshot of the DIFFERENCES
is taken. If the quantity of different pixels is sufficient, this
information is written to storage, along with the next to-be-decided (5,
10??) seconds of video. A flag is set, ready for interrogation at some
future time. A remote PC could then poll the surveillance unit from time
to time and retrieve any captured "footage" for analysis / action.
I suppose that one would have to be lucky to capture a useful image of the
villain (focus issues, movement artifact, etc) but it might be better than
nothing. There might also be interesting legal issues about admissible
evidence, privacy, ... Urk!! What say you all?
Cheers and TIA,
Denis
PS. While riffling through likely web sites I came across this reference
http://www.cl.cam.ac.uk/users/rja14/
which relates to a Cambridge Professor's site (Ross Anderson) talking
about the European Union's proposed(?) intellectual property laws.
_______________________________________________
plug mailing list
plug at plug.linux.org.au
http://mail.plug.linux.org.au/cgi-bin/mailman/listinfo/plug
More information about the plug
mailing list