Our structured light scanning rig

[maxem]

Hi Gunter,
The setup above does not work well because I used regular mirrors that i had. To get it to work you need to get optical mirrors or "First Surface Mirrors" because of the 2nd reflection there is a ghosting effect

you can see the "ghosting" of the lines laying into the black of the circles

did you experience this with your own mirror setups to scan around a person with laser?
I did not know how expensive first surface mirrors are about $120-200 for the 4 of them! forget it!
Anyway it was just an idea of how to get two angles with the equipment i had laying around. It only cost me my weekend haha

Yes you are right about the post processing but what is great about adobe lightroom is that is saves all these settings and you can just "sync" the rest of the photos which applies the same changes to all of them. Of course different settings for Left and Right side crops. The only thing i would add is that when shooting calibration shots i rotate the calibration plane towards the mirror side i am calibrating so that i get the proper "V" line that david likes to see, otherwise i get an upside down "V". So two sets of calibration shots with the calibration corner rotated towards respective sides so that the camera appears to be looking directly at the calib corner and the projector appears to be above the camera.

[MagWeb]

hi,

yes, I know the refraction problem of using normal second surface mirrors.
I used those mirrors in dimensions from a human bust and bigger. With a laser, I found that the ghost image easily can be filtered by the camsettings. That may also work for the gray code patterns (patterns 06-11) which own a clear BW contrast, but the fringe, sin-wave patterns (patterns 02-05) may be problematic.
There are some cheaper first surface products (e.g "Dibond" mirror) or foil mirrors as used in theatres. Maybe some gives the needed quality. I already tried Dibond - successfully - if you´re lucky and Mr Postman didn´t ruin it (by deforming its surface). Years ago I also worked with those theatre foil mirrors. These are of a high quality if you apply the foil accurately...
Simon also announced (he´s a bit busy now) that in future releases the pattern sequence may be configurable by the user . So one could use only gray code patterns ( means more patterns to get an equal res).

So I wouldn´t drop that idea

I´d add : there is a possibility to get a third aspect. If there would be a gab between the first mirrors in front of the cam, you could capture the object also directly.

Gunter

[maxem]

hmm i didn't really think about other first surface mirror materials because i think they must be really perfect to be distortion free, and i have enough trouble keeping my calibration panel completely flat! haha
The $120 was for acrylic backed and $200 for glass backed but i didn't really shop around and these are for 12"x12" which i know are way oversized especially on the first two 45deg mirrors right in front of the lens.

I thought about the 3rd perspective but in this config you can see that the projector is directly in line with the cam. See the perfectly vertical white lines in the middle between the left and right images? This is actually a very small gap between the 1st two mirrors and you are seeing between them straight at the subject, this is the same angle the projector sees so no 3D depth info.

I did this because if this setup worked flawlessly I was going to machine out aluminum and make it light weight so it could be tripod mounted, I wanted the cam and proj in middle for nice center of gravity.

The image above also illustrates what i meant by turning the panel for calib, this photo is of left view calib setup.

If I could change the conversation back to the nervous line issue? I am putting the mirrors aside and going back to the standard structured light setup that was shown by moe at the beginning of this thread in order to get my structured light scanning down good. This is really my first time posting on any forum ever so feel free to tell me if this is not the place or if I write something in the wrong place or to many photos! haha I have been reading here a lot and see that you are quite the guru and was glad you saw my setup because i knew you had mirror experience!

to answer your hints from before about nervous line:
1- I made my own patterns to fit 1280x720 and have been displaying them with a fullscreen picture viewer (computer res set to 1280x720). They are correct cropped not scaled, here they are:

1280x720Patterns.zip

(35.41 KiB) Downloaded 198 times

2- I have read the entire thread you posted several times and I see what you mean by high contrast from the example given. What i noticed was that every pattern is only black and white except the sin-wave patterns (02-05) they contain black, white and 5 gray tones so i extracted the 7 part grayscale and made a projector grayscale calibration image with these exact tones.

This photo is the result of calibrating my projector (brightness,contrast) to the white panels, then calibrating camera (fstop, shutter, iso, contrast) to the projected image. (my projector sucks i know its a 5 yr old lcd that is almost dead I had to reset the internal hours clock three times to keep it going over the years!) I then use these settings to shoot everything. but this gives me a very weak line, so i boost the contrast after having taken the photos in order to look very contrasty almost all black and white in the sin-wave patterns, like examples given by others. I get a strong line but it still jumps around here and there sometimes worse with added contrast.

3- In the 1st picture i posted "black calib dot.jpg" the intersection angle appears low because i have dual monitors and it squashes this screen vertically (the images are scaled down to 2304x1536 which fit width wise on my dual monitors but not height wise) the pic i posted is cropped to show the multiple lines.

The only solution i have now to combat this is to do the "panel-less" advanced setting and record a short section of the "reference" line just enough to get it to say "rotating laser motion detected". This is about 1/3 down the screen so i get the bottom 2/3 of the scan distortion free...

I have been playing around with the "laser line detection" numbers and this gives me so much more data! If i just get rid of the jumping/split personality line problem I would be golden!

Any help is greatly appreciated.

[MagWeb]

Hi,

concerning the third aspect through a gab, you are absolutly right - sorry I didn´t pay attention on the needed intersection..

I also looked at your patterns - seem to be o.k

so i boost the contrast after having taken the photos in order to look very contrasty almost all black and white in the sin-wave patterns, like examples given by others

Maybe this is the problem? If you postprocess your images, you should apply the same filtering (e.g. contrast) on all images (and I don´t know whether that is a good idea anyhow). The first two images (total projections of black and white) show DAVID what can be seen in the scene with and without light. Processing the other images these values are subtracted from the pixelvalues (if I got that right). Maybe your postprocessing disturbs this process?


Gunter

[Moe]

Here's my two cent on the postprocessing topic - Short on time as always...

I've had good success with applying Photoshop's auto contrast filter to all images EXCEPT the first image - saved several scans which came through very bad before postprocessing.
Gotta check if i can improve scan quality by auto contrasting gray code images only.

Cheers,
Moe

BTE: Amazing setup you built - you could try checking laser suppliers for some front coated mirrors (sadly not really cheap) to get rid of the double-reflections and aberrations.
EDIT: sorry - just read you are already past that discussion

[Kirch]

I am following this. I want to scan people as well. Thus the need for speed.
It looks like you have already figured the following questions out. IF you would have a moment for a quick response it would be greatly appreciated.

what projector are you using? Moe, Gunter,and Maxim-

I have a canon 21mp-hi res textures- shoots 8 fps. If I have to take 12 shots I should be able to get a shot in 1.5 secnds...right

How is the camera synched to snap in unison with the projector's displayed image

Regards,
Karl

[Moe]

Hey Karl,

i'm using a rather old DLP-Projector (1024x768 native resolution) - gonna head down to the lab later and get the exact name (although i doubt that it's available on the market anymore - it was acquired way back and paid in Deutsche Marks).

As far as synching up camera/p projector goes I'm using a short piece of Processing to project the patrerns and an arduino hooked up to the projecting computer and the DSLR (camera) to synch everything nicely. Started working on dual video capture - although processing's built-in video libraries seem kinda slowish to me. Gonna put it all online as soon as the thesis is done. Promised

Cheers,
Moe

EDIT:
Hey Maxem - sorry I just overread some lines. By Cranking up sensitivity I actually meant decerasing the value for MinLineIntegral. I set it to 2 and it magically saved some of my scans. As far as the multiple lines in your scanline-images go my theory is that they tend to appear when the subject is moving (jitter). Just a guess - have to verify that... Cheers

[Kirch]

Thx Moe,
I looked at Gunter's setp on another post here and I see what he is using. This gives me an idea of the parts to use. Once I get my hands on things, I usually get it working Just want to make sure I don't end up with too many extra parts!
Best,

[MagWeb]

Hi,

I would recommend my 3M MPro110 only for scanning small objects. Here the mini size of the pocket projector plays his card: his small size and its possibility to get it focused on a least diagonal of about 12cm. But this small size also brings problems as the limited light power.
There are new little bigger LED projectors available with a better res of 800x600, more light power and a better contrast ratio. I would consider to use one of these if the object is bigger than,let´s say 20 cm.

Gunter

[Moe]

I'd also HIGHLY recommend a Projector with optical zoom (which mine unfortunately doesn't have) - saves a lot of time and pain moving everything around. The zoom-Lens gives you the possibility to make the Patterns as small or large as you need them. Optical distortion correction would be an additional plus (but not as badly needed as a zoom

Cheers,
Moe

[Kirch]

Thx Gents-
That is very helpful-
I'll see what I can find.
In the meantime I will be following your progress!
Best,
Karl

[maxem]

Hi Karl, Moe, Gunter,

Yes I would say get a DLP projector
I currently have LCD projector Panasonic pt-ae500u 1280x720, when it was brand new the specs were 850ansi Lumens, contrast 1300:1 (probably half that now its 5yrs old which is the problem with my extra lines, i explain later)
I'm looking at getting a new one

On another post someone pointed out that most regular size projectors don't focus till about 1m (41") creating a large pattern. I have what is called "close up filters" from photography lenses its like a poor mans macro lens adapter. They come in different powers (+1,+2,+4)

It basically shortens the focus on camera lens, but i put it on my projecter! same effect here are the results, the red line marks the closest sharp focus with denoted filter:

EDIT: Rough conversion 41in=104cm, 20.5in=52cm, 12.75in=32cm, 9in=23cm

I originally got them as a 4 set (+1,+2,+4,+10macro) the whole set cost $20. I think a +1 is all you need really, because the drawback of these is that it also shortens the distance which is in sharp focus so you can't have deep object. But usually small object is not deep.
You can see below I am projecting pattern #06 the first of the Grayscale I chose it because it has sharp Black and White lines to see sharp focus on the mouse but then just behind it the white background is out of focus. This is with +4 close up filter, projected pattern is not cropped thats the full projected 1280x720 pattern!

@ Moe- Thanks for the info i have been playing with MinLineIntegral # it has helped a lot!
@ Gunter- I figured out what the multiple or nervous line was from, it's my faulty projector! i'll post later, just a few more tests!

[maxem]

Results with +2 close up filter:
Interpolate 20
Smooth Average 0
Smooth Median 1
no texture applied, 1 scan no merges
fingerprint ridges visible on thumb and in lower palm...

i attached the last photo to show size of fringe pattern on hand

[Moe]

Hey Maxem,

awesome results! indredible detail. I'm impressed. About that multiple stripes in your first scans - did you figure out the exact causes for them? Still trying to improve accuracy and ease of use on our system and coming to a point where I need some ideas

Cheers,
Moe

[Khalid Khattak]

great work.. can you show us more about structured light scannining technique..what is the price of projector you are scanning with?