image bubble detection in thin fluid layer

[Toby W]

Hi,

i am stuck here with a new topic. My task is to create a bubble detection of a thin fluid layer. The fluid layer will continously be monitored and once a bubble of a certain size occurs, labview should notice.

The basic idea so far is to aquire an image without bubbles and compare this to an image with bubbles. 

My question: Is there an easy way to solve this, maybe via detecting circle-shape objects, or does it make sense to change the image to an array. I don't have NI Vision Acquisition Software.

Thank you so much

Tobi

[MikaelH]

It's all about lighting :-)

Do you have some images to share?

Can you get NI Vision Toolkit?

I wouldn't even try without that, I always use the vision assistant to play with the image to see how to detect the feature the best.

[Toby W]

Hi Mikael,

thanks for your post, I got the  NI Vision Toolkit and an image is attached. How do you suggest to proceed with the  NI Vision Toolkit? 

[ensegre]

Thresholding and flagging blobs larger than a given pixel size should do it, probably without even needing to subtract a background or filter the image in any way.

Check the examples, e.g. "Particle Analysis Report" or "Binary Morphological reconstruction" should give you a headstart.

[MikaelH]

If you can take an image before and after, and subtract them that will help, since you have some background features you’re not interested in.

1. Convert to grayscale (I tried extracting the Red Plane).
2. Apply filters (Smoothing – Median 11x11)
3. Lookup Table Setup (I tried the Square)
4. Threshold (this is the tricky thing to use the right setting to convert it to Binary)
5. Basic Morphology (Close Object Size=21)
6. Advanced Morphology (Fill Holes)
7. Particle Analysis

 

[Toby W]

Trank you two very much, i think i will manage from there on.

[bbean]

The fluid layer will continously be monitored and once a bubble of a certain size occurs, labview should notice.

Just a quick warning....based on the image you attached you have some parallax going on so the measurements of particle size will become more inaccurate (appear larger than actual) as you move away from the center of the image.  You can see it in the ghosting of the droplets in the upper-left corner of the image.   

 

Depending on how accurate you need to be, you may have to go to a telecentric lens or move your camera farther away so that the majority of the image is in the center of the FOV.