Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Using the same camera

Solved!
Go to solution

Thanks Brad, i thought about taking two images one, with no light so i can see the LEDS and the light on so i can recognise the characters. The cameras i am using are Basler acA1600-20gc. Il look into trying to reconfigure the exposure with the camera. Thanks for the advice.

 

Damien

0 Kudos
Message 31 of 42
(1,569 Views)

Check out the example here:

<LabVIEW folder>\examples\IMAQ\IMAQdx Express Examples.llb\Grab (Express).vi

If you double click on the Express VI, you can select the camera to use as well as the attribute you want to change on the fly (in the last step of the Wizard).

 

Another option that might be more appropiate if you don't need things to be fast is to use a snap and there's an Example of Snap using the Express VI as well.

 

Hope this helps,

Brad

0 Kudos
Message 32 of 42
(1,568 Views)

Thanks for the advice Brad,

 

Iim getting a bit side tracked with other elements of the project. I am looking to take an image every second. For 1 DUT there are nearly 20 tests i perform on it, applying a certain voltage,phase or current and in some cases i have to do character rec/pattern matching and LED recognition.

 

So taking 2 pictures at the moment the sequence is not working, as i have to also take into consideration other hardware devices i need to switch on/off as well as getting info from database. My time of 40 second for the test along with delays for safety reasons it woud be better if i could just take one image.

 

I have reduced the exposure and came up with the following pictures, the first is a different DUT where i just have to pick up the LEDs the second is the one i posted previously with 4 LEDs on and  agreen LED (L) for phase.

 

In the first instance i can leave the light off for the testing which s fine but the second DUT i need to have some light to recognise the characters.

 

Damien

Download All
0 Kudos
Message 33 of 42
(1,563 Views)

OK...sounds good. So do you have a question, or are just updating how things are going? Sounds like you're on the right track. Let me know if you have a problem though.

 

thanks,

Brad

0 Kudos
Message 34 of 42
(1,560 Views)

Yes I have one question its basically the settings to pick up the LEDs i want to pick up and not picking up the other  ones that have a reflective light. I have attempted to use the vision assistant to get four objects from the 4 LEDs but i pick up like 36 because the settings i give it is incorrect. Is it possible to get you to have a look at this and see if you can perhaps get this working so its similar to what you previously posted the number of LEDs is the number of objects detected? With the 2nd picture with the green LED on is it possible for you to have a look at that also. sorry for constantly asking you but you have been a great help

 

Damien

0 Kudos
Message 35 of 42
(1,557 Views)

I'm not sure if I understand your question, but here are two color threshold settings to clearly filter out all the other info and just return the lights that are on for the two images you posted. After this you can use particle filtering to clean up any objects that aren't the expected size (i.e. too small or big), and then particle analysis to find out how many LEDs are on and their location. If this is not what you are asking about, please be more specific/clear.

 

Hope this helps,

Brad

 

Download All
0 Kudos
Message 36 of 42
(1,555 Views)

This is exactly what i am looking for but with the threshold2 i am looking at both the 4LEDs and the phase (L) to be on. With this picture then is it possible to do the character recognition or is it too dark? Is it just a matter of playing around with the settings to pick up the correct LED or is there a guideline as how to adjust the settings?

 

Damien

0 Kudos
Message 37 of 42
(1,548 Views)

The color threshold is really best for isolating just one color. (I got something to kind of work, but it had a bunch of noise..see attached image). You could always call two different color thresholds with the original image and combine the results from the particle analysis on the different images to figure out what's lit up. I don't think doing OCR on the LCD text would work very well in that lighting...too dark for me to read, so I can imagine the algorithm wouldn't like it either. Can you programmatically increase the exposure on the camera so the LCD is easier to read? 

 

Hope this helps,

Brad

0 Kudos
Message 38 of 42
(1,542 Views)

Yeah i understand what you mean its best to isolate the LEDs from the phase LED. I can increase the light slightly so i can get a better reading of the LCD. I have attached a picture with a different light setting this will possibly work if i apply it to the other test steps.

 

Regards,

 

Damien

0 Kudos
Message 39 of 42
(1,539 Views)

Hey Brad,

 

I have a couple of questions for you, I have three template images and i want to compare my live image with these three template images and based on the result, i select to perform testA for imageA testB for imageB etc. What would be the best way to achieve this? I have attached a vi with what i have tried so far but perhaps you would have a better way of achieving it. Thanks again for your time.

 

Damien

Download All
0 Kudos
Message 40 of 42
(1,529 Views)