My Reprap-related PhD outline

I am in the process of being admitted to doing a PhD (hopefully starting in March next year) as I thought I should spend a few years working full-time on something that would help Reprap along.

Given my areas of expertise I came up with the below as my Preliminary Outline for an addition to the Reprap ecosystem (I think its ambitious but achievable) and am including it as part of the official paperwork.

Any feedback would be good, and thanks to those who offered suggestions along the way.


The endpoint of this project is a suite of open source software tools wrapped up in a graphical interface, probably including a website, which will then allow the entire world access to the software. This suite of tools are to be the "software toolchain" needed to bring 3D modelling and, when combined with the open source software and hardware for the Reprap project, 3D "photocopying" to the general public.

The Reprap project is a project designed from the ground up to create an affordable 3D printer for the everyday person that can, among other things, print its own parts. See [reprap.org] for more details, pictures, and movies of the machine in action. In its current incarnation it can only produce plastic parts but plans are underway for allowing it to print electronic circuits, and then its own motors and transistors etc.

My project is to complement the Reprap project and will allow the everyday person to create a 3D model of an object by simply taking digital photos of it from different angles. This model can then be used as the first draft of a new object and edited in a 3D editing program such as Blender or Art of Illusion. Alternatively, it could be that this software suite has the capability to automatically send this 3D model to the Reprap software to be printed.

As currently envisioned this suite of tools will have the following parts to it:

1. Create calibration sheets with various symbols and targets on that the rest of the software which will be, to start with, printed on A4 sheets of paper but with the idea that this can be easily expanded to any number of arbitrary sizes.

2. Take photos of an object (which is sitting on/surrounded by calibration sheets) for input and use the calibration sheets in the photo to get camera angles and scale etc. for each photo.

3. Do feature matching and contour matching of the object in the different photos to get a first draft of a spatial cloud of points.

4. Remove the noise and apply any smoothing and other algorithms needed to get a better model.

5. Provide the facility to pass this straight to the Reprap software or convert the point cloud into a typical 3D model file format e.g. STL, which is simply a list of interlinked triangles. It would make sense at this point to have the capability of having conversion plugins between different internal and external representations of the 3D model. This may require some alteration to the Reprap software to make the toolchain flow more efficient.

The ultimate aim is to make it a very small step from this to having a web camera or digital camera mounted on the computer controlled movable toolhead of the printer taking photos from known points and hence having "push button" 3D photocopying available to the average person at low cost.

Hi Reece,

are you in contact with David laserscaner ( [www.rob.cs.tu-bs.de] )?

The combination of this freeware-tool, meshlab ( [meshlab.sourceforge.net] ) and some more informations about 3D-scanning (e.g. [en.wikipedia.org] ) should be a good starting point for optimizing the tool-path ...

Viktor

Wow Reece, that sounds like science fiction!

Good luck and let us know if we can help in anyway (debugging, etc.)

Thanks Viktor, I've bookmarked the site. I had heard of the project and read some papers that referenced the idea but not been able to find much info on it (something to do with not having a lot of time to do any research yet).

Hi Reece,

... glad to help you

If you find or make something like 3D-scanning with a set of two stereo-images or a live-stereo-videostream, it would be great for 3D-micro-scanning!

I've developed simple microscope-optics for web-cams that can capture 3D-stereo-images (e.g. red-cyan-anyglyphs or shutter-3D-sequences), so it would be great to calculate a 3D-representation from a microscopic image - of course, this would run with 'normal' macroscopic scenes too

Viktor

I've seen a digital camera the other day mentioned in very condensed report about the photokina [www.photokina.de]
That camera had a structured light flasher onboard so the images taken contain a very fine projected grid on the photographed surface.
Then the picture is meant be edited with specialized software (just like digiscan2000 [www.rsi.gmbh.de]) in fact is was some other software... but it was a very short report and my brain wasn't fast enough to recognize the "important" information (in other words, I was playing with my cellphone that minute )
well in short terms:
take a picture, run the software and receive a pointcloud for 3d manipulation.

Maybe that is more acceptable for "fielduse" than laserscanning

I'm sorry that I cannot name any brand nor company nor anything else that would be useful to hammer google with...
but maybe you can keep your eyes up on something like that, like I will to let you know more details

'sid

[edit] nothing yet...
BUT: [www.imodeller.com]
check this out, just software and a camera no projectors nor structured light in any way... not cheap, but there's a demo available
oh and guess what it exports stl files

Edited 2 time(s). Last edit at 10/11/2008 08:23PM by sid.

Well the imodell looks similar to what you want but they the models it make are fairly basic. It cant do concaities. Because they are only doing it to get 3d image they just use textures for areas that it cant figure out.

So I think you are defiantly going to have to use more than just two images to get a good print able model.

Also I forget where it was but there is some software I found once that was meant to be used contractor measureing houses and stuff. Basically you printed out a black square that was exactly 6in X 6in ( I think you could use any size square really but not sure) and placed in front of the house. Then you took a picture. Then in the software if figured out the perspectives and you could get fairly accurate measurements of the house.
I realize you are not dealing with a house but to help with scale and perspectives in the pictures you might try having something similar like a 1inX1in square flat on a table next to the object. If not used for measuring the perspective it might be useful for "linking" photos together so calculating camera locations/ photon paths might be a little easier.

Well found the software i was talking about...kind of seems they have moved on to using anything in the photo as a reference point you just need its dimension to plug in. Before it was called iphotomeasure but for what ever reason they changed it. [www.uphotomeasure.com]

Edited 1 time(s). Last edit at 10/12/2008 01:08AM by ronanwarrior.

My original PhD Outline has had to be re-written as it was 'too much development, not enough research' and it had no links to previous research etc.

So, the new outline that has been signed off on by one of the two heads of department is:

Is it practical for the everyday man to build and use his own 3D fax/photocopy machine?

The ability to do 3D faxing/photocopying relies on two technologies that have until now been separate; 3D scanning and 3D printing. In the commercial world, the hardware and software for 3D printing is available (for example [1],[2]) and slowly coming down in price so that it is becoming practical for small business use, see [3], but is still out of the range of all but the most wealthy and motivated hobbyist. Commercial 3D scanning is also available but can involve elaborate controlled conditions and/or expensive hardware such as lasers [4], [5].

Recently however, there have been a number of projects started to produce homemade 3D printers and chief among these is the Reprap project [6], a project designed from the ground up to create an affordable 3D printer for the everyday person that can, among other things, print its own parts, (so that eventually the cost can go down to the cost of the raw materials for the parts). As the Reprap project is an open source hardware and software project, anyone can contribute and add whatever functionality they like, including incorporating the ability to do 3D scanning.

With the advent of cheap digital cameras, web cams, software and algorithms to process images and movies of an object to create an approximate 3D model [7], [8], [9], which in turn have lead to online plans for do it yourself 3D scanners [10], [11], [12], [13] and 3D scanning web service sites [14], [15] the possibility has opened up that the everyday person may be able to create a 3D computer model of an object, and maybe even build their own 3D scanner.

Therefore this research is to be in three stages: firstly, to see if open source software can be created to take as input the aforementioned images from a digital camera and/or movies from a web cam and turn them into a 3D model using the method that requires the least controlled environment and least cost and number of additional tools like calibration sheets, such as shown in [16].

Secondly, to then extend the software and hardware of the Reprap project and incorporate this and additional functionality. The software could be extended so that the 3D model could be emailed to an online Reprap which could then print it out i.e. 3D faxing. The hardware could also be modified to create a 3D photocopier by the addition of a digital camera or web cam mounted on a tool-head which can be controlled by the software. As the calibration sheets are primarily used to deduce the camera position, and the exact position of the camera is known in this scenario, they may not be needed.

And lastly, to see if this can be put together by the

Hi Reece,

... excellent! - now we have a bunch of 3D-scanning-related links together ...

Viktor

Hi Reece,

I found the camera I talked about earlier, it seems that what I've seen was some sort of flashback (at least I think so)
nevermind:
It was the "minolta 3d 1500"

in corporation with metacreation.com (now viewpoint.com afaik)

but as far as I can see that was eight years ago, but I've seen it THIS year, on that review... weird...

'sid

[edit]
canestavision may be another good looking approach
[www.canesta.com]

Edited 1 time(s). Last edit at 11/21/2008 07:58AM by sid.

Dear Reece,

About 3 years ago, I had a very similar idea. I took it to the university where I teach (Penn), which has a center that helps students and others bring ideas to fruition (the Weiss Tech House).

I was wondering how you're idea has developed?

Mark

I just published my software as a 0.8 beta.

See the post [forums.reprap.org]