The UK Home Automation Archive

Archive Home
Group Home
Search Archive


Advanced Search

The UKHA-ARCHIVE IS CEASING OPERATIONS 31 DEC 2024


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Ant's homebrew motion detector



On Wed, 15 Nov 2000 19:52:43 +0000, you wrote:

>The technique I used to generate the web page images is extremely
>cumbersome and time consuming, so I'm knocking up a KDE2 app to make
the
>whole process easier.
Sounds rather interesting- if you get past making a nice GUI in
KDevelop and want a Beta tester...

>Yes, pretty much. The model comprises a weighted mean and a weighted
>standard deviation. It's updated every frame from the adaptation
region,
>which is the region of the frame for which no temporal motion has been
>detected in the last n frames. Background motion is excluded from the
>adaptation process to prevent propagation of errors into the model.
Masking out areas of motion sounds good; do you mask areas, or just
AND the inverse of detected motion?

The question with all this processing is CPU cost Vs benefit of
accuracy. Simply comparing adjacent frames happily captures images of
the postman and next door's cat, but at the expense of also saving the
start of rain showers (wet ground starts reflecting sun IR) and
passing car reflections.

Many TV cards also have problems with motion and interlacing leading
to jagged edges of moving objects when alternate scan lines are
combined.

My current solution is to replace the poor mono CCD with a cheap
colour one. This hopefully will reduce the IR smearing, at the
possible expense of low-light sensitivity.

Perhaps your more computational approach is better, particularly when
combined with detection 'zones' on the image like a PIR.

My grand plan is to use small relays to switch between several
cameras, so my ideal detection method must not need a constant signal
and have a low settling time between switching. Storing several
reference background images would seem to be an advantage here.

>At the moment I initialise the background model with 'suitable values',
>but it will be possible to train it with a 'ground truth' image
>sequence, i.e. one in which no motion occurs.
If 'real' world test images are of use, I can generate hours of
640x480 mono or colour.

>I hope to have some more pages up shortly, and maybe a *severely* alpha
>version of the app  and library.
You KDE reference points towards a Penguin rather than a Daemon, but
512x512 is an odd screen resolution. Are you using a dedicated image
processing device?

My kit uses RedHat 6.2 or 7.0 (forced upgrade after a kernel botch).

>This is all taking much longer than I anticipated - I'm meant to be
>finishing my DIY curtain automation! Oh well. ;)
After moving from my last house, I gave up and installed vertical
blinds- much lighter to drive with slow gear motors. Why not try
changing the problem? ;)

TTFN,

James
---
James Derrick    james@xxxxxxx, Cramlington, Near Newcastle, England
Forwarding Service: jderrick@xxxxxxx
Beyond the Horizon of the place we lived when we were young,
In a World of Magnets and Miracles. Pink Floyd.

-------------------------- eGroups Sponsor -------------------------~-~>
eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9699/3/_/2065/_/974324351/
---------------------------------------------------------------------_->





Home | Main Index | Thread Index

Comments to the Webmaster are always welcomed, please use this contact form . Note that as this site is a mailing list archive, the Webmaster has no control over the contents of the messages. Comments about message content should be directed to the relevant mailing list.