Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

unsteady image acquisition

Hello,

I perform a serie of color snaps of a scene (always the same - static) and compute the 'difference' (sumation of difference of 2 images of the same size divided by number of pixels) between the first snaped image and all the others. The strange fact is that when I plot this 'differences' over time I get an oscilatory signal (nearly aproximates a sinosoidal) with a considerable amplitude. Since my app is very sensitive to this, how could I work this out and what kind of thing may be causing this?

I'm using an IMAQ PCI-1411 board and an ImagingSource 1403/N camera.
I acquire the images directly to the workspace of Matlab 6.5 for further processing.
Camera shutter is set to 1/30000 (later I'll do acquisition of fast pulp crossing acquisition window) and I used a special white lamp (StockerYale 25Khz) for ilumination.
Camera is linked to board using a cable nearly 10 meters long.

Thanks in advance,
Inácio Rodrigues
0 Kudos
Message 1 of 3
(3,170 Views)
Your lamp pulses once for each peak in the sine wave. You will get a pulse during both positive and negative peaks. That means your light is pulsing at 50000 Hz, and your shutter speed is 1/30000 seconds. If you multiply these numbers together, you find out that the light pulses 1 2/3 times during the shutter time. Depending on which part of the pulses you catch, you will have a varying brightness of your image.

One solution is to match the shutter speed to the frequency of the light so that you always catch an integer number of pulses. In your case, a shutter speed of 1/25000 seconds should catch two pulses every time and keep the brightness constant.

Another solution is to use a constant current lamp which provides very steady lighting and doesn't pulse (as far as I know - never checked it with extremely high speeds).

A quick and dirty solution would also be looking at the overall intensity of your images and multiplying by a correction factor to get them to maintain a uniform brightness. This would work as long as a majority of your image remains unchanged.

Bruce
Bruce Ammons
Ammons Engineering
Message 2 of 3
(3,158 Views)
I was afraid of that...
I can´t set the shutter of my camera to 1/25000 and the greater multiple of 25000 I can get is 1/1000.
This is too much time for me and, because of the intensity of the light, the image is totally white.
I think I have to buy a new light as you say...

Thanks,
you were very clear.
0 Kudos
Message 3 of 3
(3,124 Views)