Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Computer Vision and Measurements in
Aerospace Applications
Tianshu Liu
Department of Mechanical and Aeronautical Engineering
Western Michigan University
Kalamazoo, MI 49008
Objective
To build a unified theoretical framework for
quantitative image-based measurements of
morphology and motion fields of deformable
bodies like fluids
Geometric Structures:
Points, Curves, Surfaces
Motion Fields:
Points, Curves, Surfaces (Geometric Flow)
Complex Continuous Patterns
Background
• Computer Vision • Photogrammetry
Human Vision Computer Vision
Analogy
Optical Info
Processing
Principles of Physics Mathematical Models
Emission Radiance
in Images
Geometric and Physical-Based Processing
Geometric Features
In Images
Perspective
Transformation
Emission Radiance
in Images
Geometrical, Kinematical
and Physical Representations
in 3D object Space
Mapping of Geometry
and Motion
Mapping of Physical
Quantities
(1) Pressure (2) Temperature (3) Skin Friction
(4) Velocity Field (5) Attitude and Kinematics
(6) Shape
Important Quantities in Aerodynamics
Molecular Sensors!
Combination of Approaches
• Perspective Geometry
• Differential Geometry
• Continuum Kinematics & Dynamics
• Radiometry
Needs for New Methods
Quantitative, Dynamic, Universal, Applicable
to Complex Patterns & Motions of Deformable
Bodies
Relevant Topics
(1) Perspective Projection Transformation
(2) Projective Developable Conical Surface
(4) Perspective Projection of Motion Field
Constrained on Surface
(5) The Correspondence Problem
(6) Composite Image Space and Object Space
(3) Perspective Projection under Surface Constraint
(7) Perspective Invariants of 3D Curve
(8) Reflection and Shape Recovery
(9) Motion Equations of Image Intensity
Perspective Projection Transformation
)P,P,K,K,x,xc,,X,X,X,,,( 21212p
1p
3c
2c
1cΠ
Camera parameters
exterior interior distortion
Formulations of Perspective Transformation
)(
)(cxxx
)(
)(cxxx
T
T
22p
2
T
T
11p
1
c3
c2
c3
c1
XXm
XXm
XXm
XXm
hhh XPx
(1) Collinearity Equations
(2) Homogenous Coordinate Form
(3) W-Vector Form
0)(
0)(
T
T
c2
c1
XXW
XXW
Geometric Image Measurements
Projection 3D 2D
Recover the Lost Dimension
Perspective Developable Conical Surface
0)s()( Dc NXX
0ds/)s(d)( Dc NXX
Conical Surface Equations
Plane parallel to image 3D curve
s)d(s
0
321
tPxPXX 0hcCP
Ray Equation
Reconstruction of 3D Curves and Surfaces
Using Projective Conical Surfaces
3D Space Curve Surface
Reconstruction of 3D Displacement Vectors
Providing a rational
and general method for
Stereoscopic Particle Image
Velocimetry (SPIV) and
Scalar Image Velocimetry
(SIV)
Motion Field of 3D Space Curve
min||t)(|| St XUX
XSt
Variational Problem:
0)]([Gi XU
Physical and Geometric Constraints:
2,1,i
time 1
time 2
Perspective Projection under Surface Constraint
dxdxg|d|dS 2 X
Geometric Structures
Quantities: tangents, normal,
length, angle, area
topology
2
1
1
2
1
dx
dxQ)(
dX
dXSc
T
3 fXm
One-to-One Differential Relation
)X,X(FX 213
Motion Field
)]([U
)]([U
)(
Q
x
x
dt
d
2
1
T2
1
xf
xf
fXmu
S
S
Sc3
Optical Flow
Surface-Constrained Motion Field
Motion Field on Surface
Image
The Point Correspondence Problem
Generalized Longuet-Higgins Relation:
Image 1 Image 2
Epipolar Line
h(1)x
h(2)x
Epipolar Line
0)xx(Q)x(x )1(hh(1))2(hh(2)
Determining Point Correspondence
Four Images or Cameras
Six Longuet-Higgins
Equations
for
Six Unknowns
1
2
3
4
1
2
1(1)m
2(1)m3(1)m
Composite Image Space and Object Space
One-to-One Relation
Provide additional
dimensions
Reconstruct 3D
Displacement Vectors
from Composite Image
Coordinates
Perspective Invariants of 3D Curve
2212,obj
2121,obj
2212,im
2121,im
D
D
d
d
Torsion
)3,'2,2,1(ID
)3,2,'1,1(ID
)3,'2,2,1(id
)3,2,'1,1(id
2211,obj
2122,obj
2211,im
2122,im
Curvatures
2341
4321
2341
4321
DD
DD
dd
dd
Distances
Rectifying plane
Osculating plane
new
new
(Brill, et al.
1992)
Geometrical Flow
Problem
Shape and Reflection
)]aa(p[Ec
Ec)(I
LNsdlssys
aasys
VLVNLN
x
ss
Motion Equations of Image Intensity
)t(Lc)t(I g,a,X,x,
Image Intensity and Radiance
TN21 )p,,p,(p pPhysical parameters:
TM21 )q,,q,(q qGeometric parameters:
Radiance: L(X, p, q)
Generic Motion Equations of Image Intensity
L
dt
dL
dt
dL
t
LcI
t
IqpXsysx
qpUu
TN21 )p,,p,(p pPhysical parameters:
TM21 )q,,q,(q qGeometric parameters:
Radiance: L(X, p, q)
Optical Flow Velocity Field
Governing Equation
Emitting Passive Scalar Transport
Luminescence
2XD
tdt
d
U )t,(c)t,(L XX
Perspective Projection onto Image Plane
Radiance
Motion Equation of Image Intensity
for Emitting Passive Scalar Transport
xx
Ih
x
IhD
x
Iu
t
I 2
)]([U
)]([U
)(x
x
dt
d
2
1
2
1
xf
xf
fXm
Gu
S
S
Sc3
Optical Flow and Velocity Field on Surface
Governing Equation
Light Transmitting Scalar Transport
2XD
tdt
d
U
)dsexp(LLs
0ext0
Perspective Projection onto Image Plane
x
I
x
II
xx
ID
x
Iu
t
I 12
2
Ix
Iu 1212
U
Optical Flow and Path-Averaged Velocity
Motion Equation of Image Intensity
for Light Transmitting Scalar Transport
2
1
2
1
3
3
12
12
Xd
Xd
U
Uwhere
Motion Equations of Image Intensity
Schlieren Image of Density-Varying Flows
0]dX)II([dX)II(t
2
20
2
20
X
X
2K1212
X
X
2K
U
2
1
3
2schl
K
K dXX
CI
II
Image Intensity and Density Gradient
2
1
2
1
3
3
12
12
Xd
Xd
U
UT21
12 )X/,X/( where
0)]II([)II(t
T2
121212T2
12
U
Motion Equations of Image Intensity
Shadowgraph Image of Density-Varying Flows
Image Intensity and Second-Order Density Derivative
2
1
3212shad
T
T dXCI
II
)II( T2
12
T212 II
solution
where
0)]II([)II(t
T1212T
U
Motion Equations of Image Intensity
Transmittance Image of Density-Varying Flows
Image Intensity and Density
2
1
3trans
T
T dXCI
II
Typical Applications
• Aerodynamic measurements: pressure and
temperature sensitive paints, videogrammetric
attitude measurement, stereoscopic PIV, laser-
tagging technique, schlieren, shadow and
transmittance imaging, oil-film/liquid-crystal skin
friction measurements.
• Metrology and kinematics of large inflatable space
structure.
Unification of Measurement Systems
Conventional Techniques Image-Based Techniques
Force Balances Accelerometers
Pressure Taps & Probes
Hot-Wires & Films
Temperature & Heat
Transfer Gauges
PSP & TSP Videogrammetry
Global Velocimetries
(PIV, DGV)
Schlieren/Shadowgraph
Oil & LC -Film Skin
Friction Meters
Data Fusion and Understanding
Measurements Data Base & Models
Integrated data & data
at discrete locations
Distributions on surfaces
Fields in 3D space
Aerodynamics
data base
Theoretical models
CFD
“SuperAerodynamicist”:
Intelligent Expert System!?
Conclusions We will see unified image-based instrumentation providing
non-contact, global measurements of important physical,
geometric and dynamical quantities in wind tunnel testing.
Unified Measurement
Techniques
Wind Tunnel Testing
CFD
Theories
Aerospace
System
Design
Ideal Aerodynamics Lab