Lecture 3D Geometry Processing
Scanning and Fabrication
Prof. Dr. David Bommes
Computer Graphics Group

3D Content Creation Pipeline

3D Content Creation Pipeline

3D Content Creation Pipeline

Digital Michelangelo Project

  • Scan large statues (~5m) at high resolution (1/4 mm)
    • Capture chisel marks
    • Record color/reflectance for re-lighting

Digital Michelangelo Project

  • Several challenges
    • Fragile, precious sculptures
    • Bulky equipment
    • Visibility

Digital Michelangelo Project

1G sample points
8M triangles
4G sample points
8M triangles

  • Another major challenge: huge data sets!

Stanford University 1997 - 2000

Aachen Dome

RWTH Aachen 2015

3D Scanning Techniques

Contact Scanners

  • Probe object by physical touch
    • Can be highly accurate
    • Even works on transparent objects
    • Often used in manufacturing control
    • Slow scanning, sparse set of samples

Contact Scanners

  • Modeling approach (historic) at Pixar
    • Hand-held contact scanner
    • Allows direct placement of mesh vertices
    • Imprecise and lots of manual work

concept sketch
physical model
manual scanning
digital model

Contact Scanners

  • Invasive: unsuitable for fragile objects!

3D Scanning Techniques

Time-of-Flight Scanners

  • Probe object by laser or infrared light
    • Emit pulse of light and measure the time until the reflection from the surface is “seen” by a detector
    • Speed of light & measured round-trip time ==> distance to surface

Time-of-Flight Scanners

  • Capturing a full surface
    • Sweep beam over the surface, or
    • Use time-of-flight camera
  • High speed of light makes accurate measurement difficult

SwissRanger by Mesa
Kinect One

Time-of-Flight Scanners

  • Laser LIDAR
    • Light Detection and Ranging
    • Good for long distance scans
    • 6mm accuracy at 50m distance

Forecast 3D Laser System

3D Scanning Techniques

Triangulation Principle

  • Two “cameras” identify the same 3D point
    • Compute depth from angles and baseline

Triangulation Principle

  • Two “cameras” identify the same 3D point
    • Compute depth from angles and baseline

  • Passive Stereo Matching
    • Find and match features in both images

Passive Stereo Matching

  • With high resolution cameras, skin has a lot of details!

[Beeler et al., SIGGRAPH 2010]

Passive Stereo Matching

  • Passive stereo capturing is fast (just one shot)

[Beeler et al., SIGGRAPH 2010]

Ten24 Face Scanning

Ten24 Face Scanning

  • More than two cameras!
  • Same triangulation principle: one camera pair at a time.

Multi-view Stereo

3D Scanning Techniques

Structured Light Scanner

  • 1 camera and 1 projector
    • Project special patterns to identify pixels

Structured Light Scanner

  • 1 camera and 1 projector
    • Project special patterns to identify pixels

Structured Light Scanner

  • Consumer-level scanner: e.g. Microsoft Kinect
    • Project infrared pattern, but just one frame
    • Fast (30fps), but noisy data

infrared pattern

[Denis John]

Structured Light Scanner

  • Comparison to custom-built scanner (>$10k)

LGG Lab Scanner (2009)

Kinect (2011)

Aside: Face Tracking

LGG Research Prototype 2013

Aside: Face Tracking

LGG Spin-Off Company: Faceshift 2015

Aside: Face Tracking

Apple’s Animojis

FaceID

Apple’s FaceID

3D Scanning Techniques

Laser Scanning

  • 1 camera and 1 laser beam
    • Sweep laser, identify laser in camera image

Laser Scanning

Digital Michelangelo Project

  • Laser scanning!

3D Scanning Techniques

3D Scanner Data

  • Common issues with scanned data:
    • Noise (especially bad with structured light scanning)
      Kinect (2011)

3D Scanner Data

  • Common issues with scanned data:
    • Noise (especially bad with structured light scanning)
    • Holes

3D Scanner Data

  • Common issues with scanned data:
    • Noise (especially bad with structured light scanning)
    • Holes
    • Insufficient resolution

The next steps…

  • Scanning is only the beginning!
  • Transform acquired samples into triangle mesh representation
    (surface reconstruction)

physical model
acquired point cloud
reconstructed model

3D Content Creation Pipeline

Questions on Scanning?

Digital fabrication: bringing digital geometry to the physical world

  • Fabrication technologies
    • Subtractive
      • Laser cutting
      • CNC (computer numerical control) turning/milling
    • Additive
      • 3D printing
  • Designing parts for 3d printing

Laser cutting

  • Cut/engrave flat sheets with a laser
  • Supports many materials (wood, metal, paper, acrylic, …)

Laser cutting

Laser cutting: how it works

  • Mirrors direct laser beam to precise (x, y) location
  • Laser intensity/exposure time tuned for the material and desired effect (cutting vs engraving)

Laser cutting

Laser cutting: from file to object

  • Design is just an image; mode chosen by color:
    vector cutting, vector engraving, raster engraving

Laser cutting

  • Great for fabricating 2D or 2.5D designs
  • Can also produce 3D shapes:

Laser cutting: designing sliced parts

Rhino/Grasshopper

Laser cutting: 3D shapes by folding

Pepakura Designer

Laser cutting: curved surfaces

  • Sheet metal/paper only bend into developable surfaces
    (Gaussian curvature K = 0)
  • Inserting cuts enables stretching, permitting nonzero K
    Beyond Developable: Konakovic et al. 2016

CNC turning (lathes)

  • Rapidly spin material, insert cutting tool
  • Great for surfaces of revolution, screw threads, etc.

Glacern Machine Tools - YouTube

CNC milling

  • Rotating cutter, complex insertion paths (5-axis)
  • Produce much more general 3D shapes

CNC milling: advantages

  • Great accuracy
  • Large parts
  • Great mechanical properties:
    part is solid material, not built from slices

CNC milling: limitations

  • Wasteful: material milled away is lost

  • Geometry limitations:
    cutting tool must reach all surfaces

  • Planning tool paths/selecting cutting tools can be challenging

    • Higher fixed-costs
    • Slower turnaround time
  • Usually large, expensive and dangerous

Infill Optimization: Wu et al. 2017

Desktop CNC milling

Pocket NC Version 2 - YouTube

3D printing

  • Slice 3D object into many layers
  • Fabricate layers one at a time, from bottom to top

3D printing: advantages

  • Pay only for the material you use
    • Material isn’t wasted like in CNC milling
    • Cost and fabrication time independent of complexity
  • Objects with highly complex topology can be fabricated
  • “Just click print”
    • Less complicated planning process than CNC
    • Shorter turn-around time

Infill Optimization: Wu et al. 2017

Applications: industrial design

Rensselaer Polytechnic Institute

Applications: customized products

Makerbot
Makerbot

Applications: aerospace

Boeing 787 - Wikipedia
Titanium Part for 787 - Norsk Titanium

Boeing is saving $3 million on each 787 jet using 3D printing
(by avoiding material waste).

Applications: medical/dental

Robohand
Amos Dudley

Applications: new materials

James Zhou

Create new materials by printing microstructure

Applications: shoes

Adidas and Carbon3D

Spatially varying properties for optimal sports performance

Applications: mechanical clock!

Christoph Laimer - YouTube

3D printing: technologies

  • Fused Deposition Modeling (FDM)
  • Stereolithography (SLA)
  • Selective Laser Sintering/Melting (SLS/SLM)
  • Material Jetting (MJP)

FDM technology

3D Hubs

Extrude melted plastic through a nozzle to form each layer

FDM limitations

  • Lower resolution; limited by
    • nozzle diameter
    • spatial precision of extruder head
    • flow of melted plastic
  • Poor bonding between layers (anisotropic properties)

3D Hubs

FDM free-form use: wireprint

FDM free-form use: weaving / flowers

Vessel - David Lobser
drooloop flowers - Mark Peeters

SLA Technology

3D Hubs

Cure liquid resin into a solid by shining UV light

3D printing: Laser SLA (Form 2)

Formlabs Form 2

3D printing: Laser SLA (Form 2)

Form 2 in action

3D printing: DLP SLA (B9 Creator)

Project and cure entire layer geometry simultaneously

SLA pros and cons

  • Advantages
    • Excellent resolution (25um layer height)
    • Layers almost perfectly bind; isotropic properties
    • Wide range of resins to choose from
  • Disadvantages
    • Requires cleaning and curing postprocess (place in UV oven)
    • Color and mechanical properties shift over time as curing continues
    • Only one material at a time

SLS/SLM Technology

3D Hubs

Laser shines on a bed of plastic or metal powder, fusing or melting the particles into a solid part.

Material Jetting Technology

3D Hubs

3D printing: what can go wrong?

3D printing: support structure

The defect on the overhanging feature is caused by insufficient support; support structure required.

SLA can print overhangs, but not local minima (e.g., bear hand).

3D printing: support structure removal

Removing support can be difficult or completely intractable.

3D printing: support structure for SLS

3D Hubs

SLS doesn’t need support structure!

3D printing: support structure for Multijet

3D Hubs

Multijet printers can print with dissolvable support!

3D printing: stringing and gaps (FDM)

Stringing can be fixed by lowering temperature/speed or asking printer to retract filament.

Gaps can be fixed by increasing the extrusion rate.

3D printing: warping (FDM and SLA)

First few layers can warp/peel from build platform.

Consider printing on “raft.”

3D printing: design rules

  • Minimum thickness constraints

3D printing: design rules

  • Minimum thickness constraints
  • Support requirements

3D printing: design rules

  • Minimum thickness constraints
  • Support requirements
  • Clearance requirements

3D printing: design rules

  • Minimum thickness constraints
  • Support requirements
  • Clearance requirements
  • No enclosed voids (SLA/SLS)

Optimizing for 3D printing

What exactly can we optimize?

  • Printing parameters
  • Support structure
  • The design itself

Optimizing printing parameters

  • Adjust orientation/slicing to
    • minimize print error
    • keep prints fast: use adaptive slicing!

Optimizing support structure

Branching Support Structures for 3D Printing: Ryan Schmidt and Nobuyuki Umetani

These branching supports use 75% less material than the manufacture’s default support structure–reduce print time/cost.

(Note: part orientation matters here too!)

Minimize printing costs/maximize robustness

  • Optimize printed design itself
    • minimize printed volume without breaking
    • maximize resilience for given print volume
  • Cost-effective Printing of 3D Objects with Skin-Frame Structures
    [Wang et al. 2013]

Fabrication services

Pushing 3D printing to the extremes

Large scale: concrete printers

Andrey Rudenko - Youtube

Pushing 3D printing to the extremes

Small scale: nanophotonics

Nanoscribe
Two photo lithography - nffa.eu

Microstructure with nanoscribe

Emerging Digital Fabrication Technologies

Automatic Machine Knitting
Wire Bending
Zippables