Schlieren Images for the masses

[pezman]

Steve,

I'll try this at work today and see what comes out. However, you might want to use a random background. A regular pattern raises a risk of aliasing.

Also, the more bits that are present in the flow-field, the beter the results (i.e., best to zoom in).

One thing is that you can put little "fiduciary marks" on the background (e.g. a little red cross or something). If these are outside of the flowfield, then it makes it a litle easier to mach up the photos for processing.

[thecheat]

I'm sorry for being such an idiot but what IS this supposed to be?

PS: anders, interesting you mentioned greyscale, that's something I know about, and greyscale is usually used on calulators, which, I'm assuming is where you got that from correct?

[pezman]

Steve,

I ran a synthetic schlieren on the pictures. The results are attached. Not that interesting -- there are two features that might be shocks (see density.bmp).

The gradient plots show some stuff, but I think that the are just measuring ripples in the paper -- which brings up the interesting point that youcan actually extract 3-d features from objects by projecting patterns on them, taking shots from two slightly different perspectives and using Digiflow to extract the parallax information.

At any rate, here are some tips for improved results:
- You definitely want a textured background -- just random dots on paper will do, or might be able to find something that has a textured surface.
- Shoot the pictures in a lossless format
- Crop the pictures to a consistent size. For example, a 1024x780 picture that is zoomed in on the flow field would be good.
- Make sure that nothing moves between shots because shifts of the background pattern will be treated as a change in refractive index. Note that the algorithm is pretty good at compensating for small shifts, but the less movement the better.
- If you use a fluid with a different density than ambient air (e.g. propane, or heated air), you'll get much better results, since the shifts will be more pronounced.

[steve]

Steve,

I ran a synthetic schlieren on the pictures. The results are attached. Not that interesting -- there are two features that might be shocks (see density.bmp).

that is probably turbulence resulting from the flow seperation at the tip of the nozzle.


I'll try to get some better images tonight.

[Rossco]

Righto Pez.
Its about time you give us a quick rundown on how you did that in "DigiFlow"....Pleeeease?

I had a play with Steve's pics after i decided that it was useless with my camera and couldnt get anything out!

Just give us a hint or two to catch up can you?

Rossco

[pezman]

Ok, here's what I did.

I cropped Steve's pictures to just look at the business end of the process (1308 x 982 pixels sliced form the lower left of the image that Steve provided).

I opened DigiFlow

I selected "Analyze", "Synthetic Schlieren", "Pattern Match" from the menu

I loaded the "running" shot as the experiment
I loaded the "static" shot as the background

I set the Max x and Max y displacements to 7. Basically, this is how large a region the pattern matching algorithm will search in order to find a match. Make it as small as you can. Too large and you'll get "aliasing". Too small and the algorithm can't match regions.

I set the "medium" to "displacement" and the coordinate system to "default". The medium and camera geometry stuff only matter if you are doing quantitative schlieren.

Then, just run the program. You will get density, xgradient and ygradient pictures. Click on any of these pictures and then select "Adjust scaling and other aspects of image appearance" from the toolbar. By tweaking the gamma, saturation and black-level, you can emphasize certain features of the image.

The help file that is bundled with the software is pretty good. It even explains the theory behind the process pretty well.

Best bet -- do your first experiments with a candle or the exhaust plume of a pulse-jet -- something that really generates some optical turbulence. Also, you can amplify the displacement by moving the subject away from the background a bit. Another trick is to project a pattern through the flow-field and against a white background -- it will perturb the image twice as it passes through the field (both times in the same direction).

One other trick -- make sure that you use a strobe for any high-speed flow. Even then, xenon strobes are about 10us, so the flow field will actually move several 10'ths of an inch as you photograph it.

[steve]

here are pics from my latest test. the originals were almost 7mb so they had to be scaled down- i'm sure that probably screwed everything up, but I'd figure I'd post them anyway.

[steve]

I tried running digiflow but cant seem to get it to "run the program" no pictures show up under output.

[Rossco]

Pez, thanx.
This is what i was doing... nearly. The crop helps!
Im learning more and more, although i still cant quite match the clarity of your files? and, from steves original shots, i still cant get a hint of those...whatever they are... "turbulances"?
What format are you saving the crops in? My DigiFlow wont recognise Tiff?

Steve, ill have a play with the new ones.
Do it just like Pez said. there is no "run" as such tho. Just select the experiment file as the one in action, and the background the one thats not.
Then change to the output tab and hit file. Type in the file name that you want them to come out as for the three. X, Y, and Displacement. Then hit OK. It will run the sequence and youll end up with just the output files up when its done.

Rossco

[Rossco]

Ive had a play, just as they are.
i think that im getting some shift in the images tho, creating a speckling of changes throughout the image.
The bmp. files are better as far as i can see.

The sand paper is a GREAT idea!
Maybe try it further back from the subject tho. This should give a greater deflection.

Rossco

[pezman]

Well, I'm not so sure that those little glints on the first set of photos are anything real -- they might be little artifacts that are caused by something else.

The sandpaper is pure genius!

I'd recommend practicing on a simpler problem first -- e.g. the heat plume from a candle, butane coming out of a lighter etc. Once you have confidence that you're doing it right, then do your experiment.

DigiFlow does not like TIFF files, but you can convert them by reading them into Paint and saving as BMP.

[Rossco]

I just cant get it to work!!!!!!!!!!!!!!!!!!!!!*&%$

what can you do with these? Just as a ref. for me.

Rossco

[pezman]

Rossco,

I'll play w/ it tomorrow at work.

I'd suggest using a smaller aperature, though -- might give better results.

Also, might need a background that has a little more white in it.

[steve]

I got my candle to work, but I can probably get better results then this if I work at it. its a start anyway.

[Anders Troberg]

anders, interesting you mentioned greyscale, that's something I know about, and greyscale is usually used on calulators, which, I'm assuming is where you got that from correct?

I choose grey scale for two reasons:

* 8 bits instead of 24 in the example saved me some work.

* As an example, grey scale is easier to understand for people not into computers. It's just a simple value describing the brightness of a pixel.