Microsoft Kinect - Cheap 3D scanner?

Hey all! I just caught this video and thought to myself "This is the perfect cheap 3D scanning option! As you can see in the link below, people have managed to connect to the device and reconstruct what it sees as 3D objects. By meshing multiple views, we can recreate full 3d objects. Just thought i'd share

[gizmodo.com]

Cheers!

... the accuracy is something poor with around 10mm best resolution

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

If you take multiple views i think that it would be higher. You would have to sweep an object from specific angles and then use some 3d mesh repair software to combine them all.
I believe MeshLab does this.

Resurrecting an old topic, but I've just done this: http://eclecti.cc/hardware/facecube-copy-real-life-with-a-kinect-and-3d-printer

This is only from a single view, so the utility is somewhat limited. It wouldn't be too difficult to form a complete mesh of an object on a turn table though.

Just making an effort to resurrect this thread and hopefully get some QuickStart details from people who have gotten their kinect to work as a 3d scanner.
@nrp: I tried your Python script but no matter what i do i get errors. First was that Python couldnt locate the Freenect module. Then it could not find some functions in the module. Could you help in getting python properly setup for Kinect use?

Also as a challenge to the community, anyone have a suggestion on the simplest way to go from scan to finish? I have scoured the net and have not found a simple approach to get the data from Kinect into any application for 3d printing.

Thanks!

Ive managed to take the Cocoa OSX Kinect application by fernLightning and add PLY export ability. I have tried many many different options available on the internet and this was the quickest way to get the Kinect connected with a computer. It took some figuring out to get the data to format right for Point Cloud file format, but anyway...its done now. I wanted to share in case there was interest in the community.

Original application and source code found here:
fernLightning - OSX Kinect

Since i have not been able to get ahold of anyone from that site, i dont feel comfortable sharing my modified code.

How to save a PLY file: Simply connect and preview your pose in the window. When you are ready to save the pose, click "Save PLY" and a snapshot is stored. Next enter in a filename and location and click save. Only PLY ASCII file format is supported. Next open your file in your favorite point cloud software and complete the transaction.

I have not been able to find a good work flow after this. I am now stuck with point clouds, but no easy way to combine multiple angles or create solids. If anyone wants to help with this i would be very grateful.

Enjoy!

Thanks for the info.

Was waiting for this, KinectFusion has been posted online:
Kinect Fusion open source
Coolest thing about this is that it no only lets you move the kinect about an object to generate a point cloud, but it refines the data as you move it about, generating more and more accurate depictions. There's a neat vid where they're able to capture the embossed logo on the back of a monitor, which is only approx. 1mm deep.

Has anyone tried downloading and compiling the source code for KinectFusion? Those videos are amazing!

That's pretty awesome. And with Microsoft just announcing a 'Near mode' [microsoft-news.com] in the upcoming revamp of the hardware, maybe the ability to scan smaller objects accurately is not too far away.

so Im making a light duty CNC/3d printer and wanted to add a 3d scanner so I could reproduce items, out of a stroke of luck I found this guy who is making software that uses a kinect to make a real time 3d model of anything, not only is it super fast but what amazed me was that it refines the model the more it sees the object. you really have to see it to believe it. you dont even have to place it on a turn table or in a special set up corner like dave3d, in another of his videos he walks around with the kinect and it maps his whole desk in ever increasing detail.

[reconstructme.net]

[www.youtube.com]

the guy just sent me an a message and now they are taking beta testers... you can go to his site above or his blog to find out more but I think his demo videos kinda speak for themselves on how awesome this is.

its exactly what we all have been waiting for and it outputs a standard STL file that can be imported into any CAD or 3d modeler software...

Reconstructme is great! I don't have a printer up and running yet but I have scans of friends and family which have been saved directly from the scan software to stl. With he newer versions you can use multiple scanners to create 1 scan. I am now working on a system to increase the accuracy and decrease interference from the infa-red grid from other kinects.

Hi,

I posted a video on youtube that shows how to scan and print a person with a Kinect scanner: [www.youtube.com]

The software is Artec Studio 9.0, but you can use Freeware tools like reconstructme as well.

Enjoy.

Best,
Bernhard

see what happens, if you connect your Kinect with a professional 3DScanner Software [youtu.be]

www.youtube.com/virtumake

The Kinect Fusion is bundled with the 1.7 SDK now. You should be able to Scan any Live object in 3D. The next step however is to find a Library to take that row Depth Data and convert it into and STL file. So it can be sliced and printed. I am familiar with the SDK and Depth Data, but now nothing about drafting file formats. Dos anyone know of such a Library, hopefully Open Source? or books of generating such files?

One of the output formats of the scans within the new Fusion SDK is .stl. I don't think there is any work to be done, or did I not understand your question fully?