�ݺ�ߣ

Projection Computer Graphics
on Moving Surfaces:
A Simple Calibration and Tracking Method
Claudio Pinhanez
MIT Media Laboratory

Frank Nielsen, Kim Binsted
Sony Computer Science Laboratory

Motivations
transforming every surface in
a space into a display
movable displays

enhancing performance
(theater, story-telling, dance)

Structure of the Problem
range

Goal: Correction of the Projection

when the surface moves, the projection is corrected

Problems
1. Calibrate the projector/camera relation

2. Track position/attitude of the surface
3. Correct in real time the projection
4. Compensate for oblique projection
[Raskar, Sketches’99]

Problem 2.a:
Tracking the Position of the Surface
hard for vision-based
systems
(darkness)
“hacked” solution:



4 Infra-red LEDs
5th disambiguating LED

5th marker

Problem 1: Calibration
From stereo vision:
the relationship between points on a planar
surface seen from 2 different cameras is a
homography (3x3 linear matrix) [Faugeras 93]

Geometrically,
pinhole cameras = pinhole projectors

Camera/Projector Homography
camera

projector

c  xc zc , yc zc 

p  x p z p , y p z p 

homography
 xp 
 xc 
 
 
p   y p   Hc  H  yc 
z 
z 
 c
 p

Calibration Process
Only 4 points!
Given: pi  xip , y ip ,1
Read:





i
i
ci  xc , yc ,1



T T T T
P  p1 p2 p3 p4



i  1,2,3,4
i  1,2,3,4



T T T T
C  c1 c2 c3 c4



Since P=HC,



H  PC CC
T



T 1

(manual alignment)

Problem 3:
Correcting the Projection

p=Hc

Projecting Computer Graphics on Moving Surfaces: A Simple Calibration and Tracking Method , Sketch & Applications, Novel Projection Methods track (SIGGRAPH 1999)

Delay and Noise
delay:



position at time t
projected at time t+dt
(100-200ms)

sensor noise
Solution: Kalman filter

Problem 2.b: Estimating the Attitude
4 co-planar points -> unique solution
(4-perspective inversion)

high degree polynomial, very sensitive to physical
parameters -> unstable solution

attitude computed from ratio of
line segments + linear regression
in practice, better and faster
than iterative solution

1/1=1

0.7/1.4=0.5

HyperMask
[Millenium Motel]
white facial mask
with IR markers

camera

mask
projector

(Lip Synching)
sound-based predictive lip synching

neural network: predicts the position of
mouth from the sound stream
Shigeo Morishima (Seikei University)

HyperMask: Performances

Tomorrow, 9 am to 1 pm
Emerging Technologies
Millenium Motel

?
?
?
?

Questions and Answers(?)

?

More info:

http://www.csl.sony.co.jp/
person/nielsen/HYPERMASK/hypermask.html

?

�ݺ�ߣ

Projecting Computer Graphics on Moving Surfaces: A Simple Calibration and Tracking Method , Sketch & Applications, Novel Projection Methods track (SIGGRAPH 1999)

More Related Content

Projecting Computer Graphics on Moving Surfaces: A Simple Calibration and Tracking Method , Sketch & Applications, Novel Projection Methods track (SIGGRAPH 1999)