3D scanning

3D scanning is the process of making a 3D file from a statue, model, or building.

Structured-light scanning

Structured-light scanning is making a 3D file of an object just using a camera or a camcorder with either 1) a projected grid from a video projector or 2) a projected line(s) from low power laser light source (generally, a laser-pointer or similar). It should be born in mind that 3D scanning produces clouds of discrete points, not contiguous surfaces. A scanned primitive shape like a cube or a sphere is not recognised by CAD programs, and point cloud processing software is often the most expensive component of commercial systems.

DIY

SplineScan is free and open source software. It has reached a high level of polish and sophistication. The website and documentation does not yet reflect this. See also the RepRap local documentation stub SplineScan.

David Laserscanner is closed source but there is a *free* basic scanning software version, and a commercial edition that costs €199, for which you get additional tools for assembling and combining multiple scans of the same object into a watertight model, and for applying textures. David requires a 3-plane "corner" with calibration targets, a laser line generator, red or green, and a web cam or a video recording of the line being swept over the target in the "corner". There's also a "for sale" starter kit. The kit costs €400 and you need a Windows computer. You get a suitable camera and laser line, and the panels to build and calibrate the "corner". They've spent a lot of time on their website, and there's good feedback and advice from users and the vendors through good forums, and many real-world user examples. If you already have a laser line generator, and a web or video camera, and three 600mm x 600mm sheets of MDF or similar, you can build a David scanner and be operating it within a couple of hours. Scan output is .obj, and if you're using the free version, Meshlab can assemble and stitch multiple scans into one object.

Commercial

There are probably a number of expensive solutions for doing this, based on good software, a camera, and a laser pointer. And good marketing.

Photogrammetry scanning

This is making a 3D file of an object just using a camera, or a camcorder, even without a laser-pointer-like light source.

Technically this is called 'photogrammetry' or more precisely 'object reconstruction'. The following is a list of free programs, but the full list is available from wikipedia: http://en.wikipedia.org/wiki/Photogrammetry

	Platform	Standalone / Plugin	Automatic modeling	Scalability	Data source	Inception	Vendor / creator	Software License	Description
123D Catch (Beta)	Windows/OSX/iOS, Web-based	Standalone part of the 123D suite	Yes	No, limit of 70 images (40 on iphone app)	Images/Video	2011	Autodesk	Freemium	General purpose
Ames Stereo Pipeline	Linux/OS X	Standalone	Yes	No, 2 images only	Images	Unknown	NASA	Open Source	map mars landscape with only 2 cameras
ARC3D	Web-based	Standalone	No, use MeshLab	Yes, multiple images	Images	2005	KU Leuven	Proprietary	Upload pictures, download point-cloud
CMVS	?	Standalone	?	Yes, multiple images	Images (from the internet)	Completed?	University of Washington	GPL	“build” Rome in a day from pictures on the internet
Correlator3D	Windows	Standalone	Yes	Yes, multiple images	Images	2003	SimActive Inc.	Proprietary & Patented	3D mapping from aerial and satellite images
Enwaii	Windows/Linux/OS X	Standalone/Plugin (Maya)	No	Yes, multiple images	Images/Video/LIDAR	2008	Banzai Pipeline Ltd	Proprietary	High-end 3d scans of sets for Hollywood CG.
Harvester	TBD	Standalone	TBD	Yes, multiple images	Images	Incomplete	User:Midge	GPL	DIY 3D scanning
Hypr3D (now [1] Cubify Capture)	Web-based	Standalone	Yes	Yes, multiple images	Images/Video	2010	Viztu Technologies	Proprietary	upload pictures, download .stl
Insight 3D	Windows/Linux	Standalone	Yes?	Yes, multiple images	Images	Incomplete	Lukas Mach	AGPL 3	generate point-clouds of scenes (e.g., buildings)
libmv	Matchmoving backend for Blender	Blender backend	Yes	Yes	Images/Video	Incomplete	community effort	MIT License	primarily matchmoving, but also reconstructing 3d models
Mementify	iOS	Standalone	Yes	Yes, multiple images	Images	2012	Tretja dimenzija, XLAB	proprietary	App-based photo upload and model download
My3DScanner	Web-based	Standalone	No	Yes, multiple images	Images	2010?	My3DScanner	Proprietary	upload pictures, download point-cloud
PHOV	Web-based (free trial)	Standalone	Yes	Yes, multiple images	Images	2010	XLAB	Proprietary	upload pictures, download point-cloud
Python Photogrammetry Toolbox	Windows/Linux/OS X	Standalone	Yes	Yes, multiple images	Images	Unknown	Arc-Team	GPL?	Runs CMVS on any platform without installing Cygwin
StereoScan	Windows/Linux/OS X	Standalone	Yes	No, 2 images only	Images	2010	Agisoft	Proprietary	Generate 3d meshes from 2 pictures
VisualSFM	Windows/Linux/OS X	Standalone	Yes	Yes, multiple images	Images	Unknown	Changchang Wu	Apache 2.0 (Open Source, but non-commercial)	GUI for CMVS
WebDLT	Web-based	Standalone	No	Yes, multiple images	Images	2012	B. Molnar, BME FMT	Proprietary?	Upload pictures, type physical measurements, and download DXF

Touch-probe scanning

Main page: CMM

• "Robot Arm Type": here a human guides a 'finger' of a device that looks like a robot arm over a model or sculpture. This deals well with overhangs and precise geometries like boreholes or outer diameters.

• "Generic Touch Probe": we can mount a touch probe toolhead on a 2+1 axis machine like a CNC router, mill, or RepRap. This is convenient if we don't have a robot arm handy.

Further reading

See also David_Project.

• "Beagle Board Challenge: Beagle Scan" uses structured-light scanning.