Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Check for absolute brightness

Hello,

 

I have a small problem with my low cost vision illumination. The LED-Illumination I use has a cheap power supply and so I have a ~100 Hertz flickering in my pictures. I use a very short exposure time (~400µs) and so I get pictures with varying brightness. I need only ~1 picture per second so I can drop many frames and use only the “good” pictures. Has someone a hint for me, how to reduce a picture to one “brightness” value for thresholding? I can extract of course the histogram but I have no idea to implement a level simple value for – “to dark” – “good” – “to bright” pictures.

 

Best regards,

 

Paul

0 Kudos
Message 1 of 5
(4,970 Views)

Well, one obvious solution would be to take the max pixel value of all image collected in a one second period.

 

1. Snap Picture

2. Snap Another.

3. Compare the two, and build a third image consisting of only the larger pixel values in both images.

4. Return to step 2. until you have enough samples.

5. Start again.

Machine Vision, Robotics, Embedded Systems, Surveillance

www.movimed.com - Custom Imaging Solutions
0 Kudos
Message 2 of 5
(4,962 Views)

Is the flickering even? Can you synchronize the camera frame rate with that? 

0 Kudos
Message 3 of 5
(4,938 Views)

Hi Meisspa,

 

as you suggest, the histogram solution is the probably the most suitable one. It depends on your application, what you define as dark, good and bright. Maybe you have to do it in a trial and error manner. MoviJOHNs suggestions makes sense, if you are trying to build the brightest picture as far as I understood. If this is what you want, then you can do that in that way.

 

Regards Anoj

Anoj Mubarak
National Instruments
0 Kudos
Message 4 of 5
(4,912 Views)

Hi,

 

I have done the job by using the "IMAQ Histogram" Mean-Value as a marker for sorting out dark pictures. Not perfect, but working for my project. I also tried a solution by using "IMAQ Threshold" and calculating the number of pixels "brighter" than the threshold - works well too! Many thanks for your support!

 
Paul

 

0 Kudos
Message 5 of 5
(4,897 Views)