Messages

1

General / Re: Photoscan detects only 2 Cams of 12 !

« on: April 22, 2014, 10:16:42 PM »

What are the most important things you need to consider when scanning? I tried a few hours with one object and used 12 cameras for a simple face to scan. But Photoscan detects only 2, sometimes 5 cameras. :-(

Pin sharp focus in all photos - shallow depth of field is your enemy!!!
No movement by subject - your subject should be dead still.
Good even diffuse lighting.
A substantial amount of "overlap" in each image taken (essential)
A subject with good surface texture and features... not just smooth and shiny (smooth & shiny objects are bad).
Cameras placed neither too close together nor too far apart.

Hope this helps.

Cheers

2

General / Re: Octocopter + Agisoft + Bunker = Video

« on: October 31, 2013, 10:29:14 PM »

Nice job!

Cheers,

J

3

Face and Body Scanning / Re: cross-polarization

« on: October 27, 2013, 06:38:39 PM »

Thanks for the advise.

Im thinking to go diffuse reflected strobe off the walls and only place CPL on my cameras to cut a certain portion reflection out.

but im thinking of head scans since its less freeform than body poses, easier to control in small fixed region, and reflection on such closeup is a constant issue - im thinking if using cross polarization is good for this case.

if so, what will be the suggested setup to archive cross polar as well as most even lighting possible? lights direct illuminating from North-west-south-east? what do you mean by 45 degree to camera?

moreover, i tested that if i do cross-polar with camera(or eye viewing) directly pointing to the light. it works by being nearly dark for normal lights, and cut out 70-80% for strobes. so the Cross-polar working is confirmed, but now we need to to cut reflections instead of direct light. I cant get the exact theory understood becasue i imagine even lights coming out fr strobes polarized in one direciton. when reflected, it should be polarized in another direction and hence making the 90 degree orientation of camera CPL irrelevant. and i suspect thats why my cross-polar setup still wont cut all the reflection. see if anyone have an insight on this? is your 45 degree sugguestion relevant to this?

thanks

Hi,

My suggestion of 45 degrees between camera and lights is only an approximation, most likely 30 degrees is more like it... but you get the idea, there must be an angle between the camera and lights.  I've done a lot of testing with cross-polarization over the years and it is always problematic for many reasons, it will never work 100% especially in the context of 3d scanning...again, more lights, more problems due to multiple ray bounces.  Cross-polarization can be helpful under "some" circumstances, but it is far from a magic cure.  I maintain that it is simply best to use large, even, diffuse light sources.  If you still suffer from problems with specular highlights, adjust your lights or remove the specular problems in post production with photoshop.

Cheers,

Joe

4

Face and Body Scanning / Re: cross-polarization

« on: October 27, 2013, 05:09:42 PM »

hi

im experimenting a studio setup of fullbody scan.
i read some web site about cross polarization.
does someone have a detail suggestion on how it can be done properly to give 99% diffuse photos?
1) i try to put the polarizing films on the strobe and camera(with 90 degree rotated direction), however i still got some 40% specular reflection remains. the strobe is directly pointing to the subject. i'd like to check whats done wrong
2) obviously direct pointing the strobe to subject will produce harsh undesirable shadows.  before i come across cross polarization, i used to point the strobe to the white walls and make it a large reflector. can cross polarization works if reflected/diffuse lights like this is used?
 
thanks
viper

Sorry to say that not much can be done.  Lights need to be at about a 45 degree angle from the camera.  The problem is that polarization works because it makes the photons directional.  The more "bounces" the photons have (off walls or other objects) the more scattered they will become and the polarization is vastly reduced. So the more lights you have, the more bounces you have from the rays the less effective the end result.  At the end of the day, large, even, diffuse light sources are most likely your best bet.

Cheers,

Joe

5

General / Re: not as sharp as espected (noisy)

« on: September 28, 2013, 08:31:27 PM »

The problem may be that the object is is quite smooth and lacking "texture" (note the specular highlights). Smooth, shiny, featureless surfaces are non-starters for photogrammetry.  You could try projecting a dense noise pattern onto the object, that might improve your results.

Also, f16 might be a bit high, I don't recommend going past f11 due to the softening of the image from diffusion.

Cheers,

J

6

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 18, 2013, 03:36:43 PM »

Thanks for your kind responses 

Theres not much occlusion on a face i guess, maybe on the ears / nose. But of course you have infinite experience more than me on this

Maybe its better less megapixel, more distance from the subject (so the pixel size/reality is bigger) (less accuracy/resolution of course) > AND more cameras. ?

On Canon 600D Focal length 50 mm and distance 1,5m its 1,3888mm pixel size / real world, maybe the subject in 2 secs doesn?t move 1,3888mm ? I dont know, id like to find these kind of balances so i can adjust budget or to think about a instant capturing method, or semi instant.
What do you think ?

I think as many cameras as you can, as close as possible. Fill the frame with the face. Details, focus and sharpness are key.

Honestly, this is the best advice you will get!  Listen to INFINITE... he knows of what he speaks!

7

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 18, 2013, 03:29:59 PM »

On Canon 600D Focal length 50 mm and distance 1,5m its 1,3888mm pixel size / real world, maybe the subject in 2 secs doesn?t move 1,3888mm ?

Trust me, a subject will move a lot in 2 seconds... I can't stress this enough, human beings are always in movement, the idea that someone can sit still is an illusion.

You need to be able to capture in fractions of a second... in order to stop movement.

Cheers,

Joe

I know i need less than a fraction of a sec to stop movement, however, by 2 seconds, i meant the time lapse between the first photo, and the last one, being each photo's shutter speed 1/125.
If the subject, from the first photo, till the last one, moves less than 1,38mm > Then, what i?ve understood (which may be perfectly wrong), missalignment shouldnt be noticeable IF -> the distance from the lens to the subject is 1,5m & using a canon 600d with 50mm lens.

I understand that human beings are always in movement, as little as it can be, however if the accuracy is below that, you won?t notice those movements. (heart pumps, micro muscle movements...)

This is what i meant, im not sure if thats what you meant also

Look, many of us have travelled down this road before and our advice to you is that you simply are going to need more cameras and those cameras must fire in sync with each other with exposure times measured in a fraction of a second... that is if you want good results.   Honestly, we would all have been happy to use a lot less cameras if it were possible... but honestly,  it's not.

So my advice at least is the same as before, build your rig up little by little over time, eventually you will get to where you want/need to be.

Keep at it!

Cheers,

Joe

8

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 18, 2013, 03:01:11 PM »

On Canon 600D Focal length 50 mm and distance 1,5m its 1,3888mm pixel size / real world, maybe the subject in 2 secs doesn?t move 1,3888mm ?

Trust me, a subject will move a lot in 2 seconds... I can't stress this enough, human beings are always in movement, the idea that someone can sit still is an illusion.

You need to be able to capture in fractions of a second... in order to stop movement.

Cheers,

Joe

9

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 17, 2013, 07:13:26 PM »

Okay i understand, im trying to reduce costs, maybe vertical 8 cameras mounted on a curved motorized rail. But if the error results to be to big i?ll consider on buying more cameras, (the motor would take maybe 20 secs from the start till the end.  I?d love to make a system balanced between accuracy and budget (i know its really hard )

What number of cameras do you suggest for a decent multicam face scan ? just the face, not the complete head / hair...

Thank you!

It all depends on your requirements.

Honestly, there is no such thing as doing this inexpensively.... sorry to say, but true.

My best advice to you is start with 8 cameras and start to build your set-up, little by little as you can afford to do so.

You will quickly find out that it is not just the cameras that are expensive, but the lenses, the lights, the cables, the stands...etc... the list goes on and on.

Scanning costs a lot less than it used to, but to do quality work... it still isn't inexpensive.

Slow and steady wins the race!

Cheers,

Joe

10

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 17, 2013, 06:30:00 PM »

- Objects on a turntable move relative to the camera, why theres no pixel deviation here? you mean that the error is on the movement of our facial muscles, so the geometry doesnt match ?

Human beings are always moving... no matter how still you try to be there will always be movement, unless you are dead.  A turntable can be helpful, but will not solve the problem as a person simply can't stay totally still over a period of time... even a short period of time.  So yes, this will cause the geometry not to match exactly and this will cause errors.

Remember when you are talking about accuracy that is measured in microns... even the smallest movement might as well be a mile ;-)

Cheers,

Joe

11

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 17, 2013, 04:04:08 PM »

1m distance and you work on approx 50micron precision/deviation per pixel ( 18Mpix image ) so even slightes movement is problematic as Meshmaster wrote even about the multicam setup.....

Yes, I should have also said a multi camera setup using bright enough lights (either from strobe or continuous sources) to provide for reduced lens aperture and high enough shutter speed to eliminate movement from photos. Rock solid camera mounts and stands are also most helpful in reducing any wobble and shake :-)

Cheers,

Joe

12

Face and Body Scanning / Re: First Face Scan, need tips

« on: September 17, 2013, 02:00:39 PM »

Hello 


I'd be really gratefull if anyone could answer these questions:
Taking account that im using one camera for tests purposes, and different views :

- Does it matter if the subject moves slightly from the first photo, till the last photo ? (this is important)

Yes, it matters.... it matters a lot.  Movement between photos is NOT desirable and will lead to errors, you want to avoid this at all costs.  This is why multiple cameras are really required for capturing human subjects.  Humans are always in motion... unless they are dead ;-)  So, if you aim to scan people and want the highest quality scans then you are going to need multiple cameras to do the job correctly.

Cheers,

Joe

13

General / Re: Strobe-based Image Projection

« on: September 11, 2013, 08:22:36 PM »

I think it has real advantages.  Firstly, depending on how you proceed, and if you decide to build a custom system you can potentially use VERY bright light sources... far brighter than anything that a projector will produce.  Second, you can sync the output to your cameras as well, or to a trigger delay... that opens up many possibilities in respect to scanning.  On the downside, it takes a good amount of experimentation to land on a solution that will work for you.  That said, I believe that a strobe driven speckle projector has a lot of real advantages.  At the very least, it's worth having a go!  Up till now, it's sort of been one of the "secret" weapons in my toolbox.  Like I said, I've not specifically tried this unit, but works on exactly the same method as what I cooked up here.

Cheers,

J

14

General / Re: Strobe-based Image Projection

« on: September 11, 2013, 07:43:32 PM »

Ok, I'll let the cat out of the bag.

I came up with a similar idea back in 2005.  I've built several strobe based speckle projection systems since.. initially I tore apart traditional slide projectors and replaced the light source with a strobe, since then I've come up with several very custom solutions. The upshot is that it works :-)

I've not tried this thing... but I see no reason why it wouldn't work.

Cheers,

Joe

15

General / Re: Strobe-based Image Projection

« on: September 11, 2013, 07:32:29 PM »

Yup, I've done this myself!

Cheers,

J