[ILUG] Webcam for John "Kate" Looney

Conor Daly conor.daly at met.ie
Tue Jun 27 09:57:34 IST 2000


-----Original Message-----
From: Caolan McNamara <cmc at stardivision.de>
To: ilug at linux.ie <ilug at linux.ie>
>
>Turned it into security camera when some lowlife walked off with my
>cdburner from openplan postgrad area. Could carefully review 10 hours
>of security guards walking around and lights coming on and off. Would
>be neat to have some software intelligent nough to review a camera feed
>searching for possible persons, or at least able to flag a significantly
>large enough delta between images as a possible interesting feature
>worthy of human review.
>
>C.
>


Shouldn't be too difficult a task (at least to get a delta setup).
Something like the following;

1.  Read a histogram of each image
2.  Normalise the histogram
        Pick something like the upper and lower 1% point as the bounds and
stretch the middle section over the span of the brightness range.
3.  Do a compare between a successive pair of histograms
4.  If there's a sufficient difference, flag the image for review.

The normalisation should take care of lights turning on and off and the
brightening with the dawn (I know most of the list only see's the dawn from
the wrong end :-).  A human in the mid or far field probably wouldn't make
that much difference to the histogram but someone in the near field would
fill considerably more of the image and should alter the histogram
sufficiently to cross a threshold.

I might be able to supply code to do the normalisation but that's quite
trivial anyhow.  The comparison task would be a bit more complex.  Maybe
something like fitting a curve to the histogram and setting thresholds on
the curve parameters to trigger a flag.  I did a bit of work on image
comparison for my thesis which used a bi-variate histogram of satellite
images to identify cloud types.  That algorithm could be adapted to general
image comparisons also.  With two identical images, it produces a straight
line while differences will result in a broadening of the line.  Could be a
very nice method!

Have you a couple of like and different images from your cam, I'd be
interested to see how they compare.

---
Conor Daly
-------------------------------------
General Forecast Division
Met Eireann
Glasnevin Hill
Dublin 9.

ph +353 1 8064255
fax +353 1 8064275

conor.daly at met.ie
-------------------------------------






More information about the ILUG mailing list