"OM" said:
Erm... OK... this is a weird question. : )
My friend has had their car damaged in the exact same place 15 times in
the last few months.
The car is parked outside of their house.
I wanted to know how good a webcam is for spying at 10 meters?
I've never tried this before... and thought I'd ask before buying a
webcam for my friend!
The idea is that they would put the camera discretely in the front room
window.
Or should we get serious and get more sophisticated hardware?
If so... what?
Any recommendations would be useful.
Thanks.
OM
I've tried focussing a 640x480 to cover outside my house,
and the head of a bystander is tiny in the picture. The
lense in that case was a fixed focus and fairly wide
angle.
Some possible approaches:
1) 640x480 fixed focus. You get to see the height, and
general features of the perps.
2) 640x480 adjustable zoom. The scene is framed better
this way, but there is still a compromise in terms of
field of view and ability to identify the individual.
3) Camera with large number of pixels:
This device has a Firewire interface, and dumps 5 frames
per second at 2208 x 3000, in monochrome. 5 frames per
second is sufficient to collect a number of samples
when the vandals are at work. Note the price:
http://www.pixelink.com/products_info.asp?id=39
4) Two cameras. First camera takes in the whole scene,
as in (1). Second camera has adjustable pan, tilt,
and zoom. A computer controls the second camera,
and uses the first camera to detect scene changes.
The second camera tracks the detected disturbance in
the picture. I expect the software for this would
cost a small fortune.
5) Place four cameras inside the car (compass points).
The subjects are now closer to the camera, so a
640x480 resolution is not as much of a problem.
(To disguise the cameras, they might have to be the
pinhole type, as you don't want expensive lenses poking
every which way.) There are wireless transmitters that
can remote the signals, for recording or analysis.
http://www.supercircuits.com/index.asp?PageAction=VIEWCATS&Category=311
Now, the other element to a solution like this, is
illumination. For the 5 frame per second camera, the
target wouldn't be too impressed, if an electronic
flash went off 5 times a second. The solution is a steady
source of infrared illumination. Silicon based cameras
can see into the infrared, and as long as the illumination
source is strong, and slightly outside human response
wavelengths, you can pour a lot of light on the subject.
This one is an example of a simple filtered light source.
http://www.supercircuits.com/index.asp?PageAction=VIEWPROD&ProdID=3903
There are other, perhaps weaker illuminators, that use arrays of
LEDs. They are invisible, as the wavelength selected would
be slightly longer than the above example. The tradeoff is,
as the light becomes more invisible to humans, it is
also becoming more invisible to the camera. So, there is
a limit to how long the wavelength can get and still be
effective. Do not underestimate the need for illumination.
The two CCD cameras I own, need direct halogen illumination
to tame the video noise in the CCD chip. Frame averaging
works wonders for these cameras - averaging just two adjacent
frames is enough to eliminate most of the noise. In that
experiment, I was dealing with static images. With movement,
you could always do averaging as a post-processing step, like
when the subject stops moving for a few frames, snip out those
frames and average them.
When the target is illuminated with an infrared source like
the one above, there likely won't be a lot of color information.
That is why a monochrome camera would be as good as anything. You
can still use a color camera if you want, assuming there
is plenty of light on the illuminator. I think the color
camera still responds to infrared, just like the monochrome
camera. (The color camera has the benefit of giving a color
image during the day, but with the IR illuminator at night,
I wouldn't expect to see much color information in the image.
Color cameras generally require more lux than a monochrome
camera.)
Hope that gives you a few ideas. To do this right, will
cost a few bucks. You can simulate the camera coverage, by
using a digital camera, and downsampling the collected
pictures to 640x480. That will give you some idea, for
a given field of view, distance from subject, and resolution,
exactly what an object the size of a human would look like.
One of the fun tasks, is trying to figure out what kind
of lense to use, what kind of lense mount, to get the
same effect from your surveillance camera.
I've given up on my cameras and put them back in the box.
So far, the only thing they were good for, was interior
photography under ideal lighting conditions. And even then
they needed frame averaging to get a usable picture.
If there was some way to hook up an ordinary digital camera,
such that it would take a picture, and immediately the picture
would be downloaded, I bet that would give better results
than the kind of cameras I've got. (Just trying to do it
cheaper than the Pixelink camera.)
Paul
