Hello,
About 10 years ago I did a bit motion analysis for a tyre manufacturer. They
wanted to see the bouncing of tyres at different speeds between 50km/h up to
more than 100km/h during a few seconds at an out-door test range. As in your
case the distance between camera and car was some metres. We used a digital
black and white camera with 256 x 256 pixel at 500 frames per second and 256
x128 pixel at about 1000 frames per second. Well, the oscillating was close
up to subpixel movement and we used special motion tracking software for our
offline(!) studies. The system was to rent for $750.- a day. It was sold for
about $30'000.-.
For details see the comments to your questions.
Regards,
Walter
-----Ursprüngliche Nachricht-----
Von: "tdtran tran" <tdtran_84@xxxxxxxxxxx>
An: "List for Photo/Imaging Educators - Professionals - Students"
<photoforum@xxxxxxxxxxxxxxxxxxxxxx>
Gesendet: Donnerstag, 27. April 2006 15:54
Betreff: Re: High Speed Camera Application for new user
> The intended application for the camera is to capture the movement of the
> vehicle as it moves on the road. I would like to capture it's bouncing
> motion due to the road surface.
>
> 1) My budget would be anywhere from $100 to $10000. Although I would
like
> to know what I
> achieve if money was not an issue.
>
(Well, with $100.- a good standard lens would be your own... ;-) Real
digital high-speed cameras with megapixel at 500 - 1000 frames per second
and more cost about $30'000,- to 60'000,-. (Open end, of course.)
Renting is about $1'500,- a day. Lower resolution and frame rate would
decrease costs to some thousand $.
> 2) I would like to track about 2-7 seconds of movement.
Inexpensive industrial ("machine vision" = image processing) FireWire
cameras and so on can directly write to the HDD with some MByte/second
and resolutions from let's say VGA (640 x 480) to some megapixel with frame
rates of some 100 fps depending on resolution.
> 3) How many images I take will depend on the accuracy I can obtain from
the
> pixel dimensions.
> That is I don't want to capture repeated images (fast frame rates), while
at
> the same time I want to be able to track the movement of the smallest
> feature without too much 'jump' from frame to frame.
That are contradictory demands, indeed. For movement you need et least to
images of one object at different times. And if the images are separated by
too
much time it is hard to find and calculate the movement. Refer to 4)
> 4) 80-100 km/hr across the camera. I.e. the camera will be placed on the
> side of the road.
That sounds like a lot of movement! 2 seconds at 100km/h make 56m.With
perpendicular movement and a closest distance of 5m to10m between fixed
camera and car you'll need an extraordinary angle of view demanding a
fish-eye lens with massive distortion, what speaks against accuracy named
in 5).
>
> 5) I just need to track the smallest feature, so anywhere between 2-5
pixels
> square would be adequate.
Therefore you have to select a greater distance, but then resolution demands
are increasing, refer to 6)
> 6) It is important to track the vertical bounce of the vehicle. Not sure
> what I need the pixel of the scene to be for this.
On a flat (and good) road the vertical movement is within some cm (= 0.01m).
To see interesting things you'll need at least about 1 pixel/cm. OK, then
some extended line scan camera with 5600 x 5 pixel?
And just to mention: If the car doesn't drive really perpendicular, but in
an angle to or from the camera, the measurement results must be corrected,
if possible at all. Because coming closer is seen as movement in up
direction and vice versa.
Consider if it wouldn't be better to work with more than one camera or a
moving or turning camera as it was mentioned in an earlier reply. Or what's
about mounting the camera inside the car?
Or positioning an intensive light source (searchlight, spotlight, flasher,
stroboscope; caution, a laser pointer may show a too small light beam)
inside the car and try to find it with a standard video camera. Following
the "light tail" could to be handle in real-time without too expensive
equipment. Perhaps a long-time exposure would provide some sinusoidal waves.
> 7) This experiment will be conducted during the day, so the lighting
> available would be just the normal sunlight.
That should be sufficient at first sight. There are open-air crash tests at
1000 frames per second and military tests with even higher frame rates
without additional illumination.
> I hoped that clarified what I need.
> thx,
> TriDat
>
[snip]
