Professional Documents
Culture Documents
SS21
meetings:
Lesson every Monday 8:00-9:30 and
10:00 – 11:30 online in BBB
Exam:
written exam at end of semester about 120 min
Chapter 1 – Introduction
1. Development of Robotic, Robotic today (Industry robotic, Service
robotic, Self driving cars, Medical robots, Aeronautic and military
robots)
2. Autonomous robotics: Definition Robot
3. Artificial Intelligence – AI: Neural Networks, classification flow –
superviced learning, over and under fitting,
4. Ethic Aspects
5. Vision based Robotics: Main components of vision based robots,
Classification of a camera as a sensor, Why vision in robotics?,
Challenges of robot vision, Structure of vision sensing
https://www.pinterest.de/pin/409616528588699674/ https://www.pinterest.de/pin/311381761708387505/
https://www.zdnet.com/pictures/15-vintage-robots-that-definitely-show-their-age/10/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
Care-O-bot
by
Fraunhofer
1. Introduction
IPA
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://automationline.rbsonline.de/multiturn-sensor-fuer-smarten-serviceroboter/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://www.wired.com/story/waymo-launches-self-driving-minivans-fiat-chrysler/
1. Introduction
Waymo
https://www.buyherepayhereusa.com/blo
g/safety-tips-learn-from-self-driving-cars/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://intelligente-welt.de/medizinroboter/
https://intelligente-welt.de/medizinroboter/
1. Introduction
http://www.agriexpo.online/prod/bosch-deepfield-robotics/product-168586-1199.html
1. Introduction
https://www.bostondynamics.com/wildcat
https://www.pinterest.de/pin/409616528588699674/ http://de.moviepedia.wikia.com/wiki/Terminator_Genisys/Wikia-Kritik
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
•
1. Introduction
neural networks
• support vector machines
• decision trees
• fuzzy logic
• deep learning
• ….
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
e
Optimum
1. Introduction
training data
validation data
over fitting
under fitting
Optimum
under fitting over fitting
N
? ? ?
Ethic Aspects
Will the AS replace Who ultimately bears
humans? the responsibility?
Feature Landmark
Extraction Recognition
passive
1. Introduction
aktive
sensors
proprioceptive exteroceptive
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
high cost!
• Reduce cost
• Do not emit light or waves
• Maintaining of high degree of intelligence,
flexibility and robustness
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
intensive
https://www.abb-kundenmagazin.de/fertigungsautomatisierung/trommelwirbel/
P
PQ Q
P
Q
origin x axis
y axis
yR
P2
P1
{R}
ξR
P
xR
origin x axis
P yR
RP
OP
Oξ
{R}
R xR
{O} x axis
yC
P CP
xC
yR
RP
{C}
OP Rξ
C
Oξ
{R}
R xR
{O} x axis
Algebraic rules
Ay
By
P
AP
y
cos ϕ
BP Bx
y
BP sin ϕ
x ϕ
Emission
spectra for
sun, tungsten
lamp and
response of
human eye
Transmittance
of sun light in
5 m water
illuminant
illuminance luminance
E(λ) L(λ)
reflectance
R(λ)
Spectra
response of
3. Physics of Light and color
three
different cone
cells of
human eye
(normalized)
https://wisotop.
de/lab-
3. Physics of Light and color
farbmodell.php
illuminant response
M(λ)
illuminance luminance
E(λ) L(λ)
reflectance
R(λ)
Spectra of
light from
3. Physics of Light and color
source till
human eye
Red: λr = 700nm
Green: λg = 546.1nm
Blue: λb = 435.8nm
Color
3. Physics of Light and color
matching
functions
for
standard
observer
r-g-chro-
3. Physics of Light and color
maticity
plane
Color
matching
3. Physics of Light and color
functions
for standard
observer
based on
imaginary
primaries
X,Y,Z
x,y-chromaticity plane
3. Physics of Light and color
[https://en.wikipedia.org/wiki/Impossible_c
olor]
https://en.wi
kipedia.org/w
3. Physics of Light and color
iki/HSL_and_
HSV#/media/
File:HSV_colo
r_solid_cylind
er_saturation
_gray.png
https://sensing.konicaminolta.
us/blog/understanding-the-
cie-lch-color-space/
Dichromatic
reflection on a
3. Physics of Light and color
surface
a) Molecular
view
b) Theoretical
view
https://www.researchgate.ne
t/figure/The-dichromatic-
reflection-model-of-surface-
reflectance-in-
inhomogeneous-materials-
a_fig1_293121272
http://time.com/5234787/di
rty-dozen-pesticides/
3. Physics of Light and color
a)
b)
c)
https://www.elmar-
baumann.de/fotogra
fie/lexikon/gammak
orrektur.html
Perspective Projection of an object by pin hole and by lense [Corke, Robotics – Vision
and Control]
https://www.arthipo.com/artblog/en
/drawing-techniques/basic-
4. Image formation
principles-of-perspective-drawing-1-
diminution.html
P = (X, Y, Z)
4. Image formation
z
x
P‘ = (X‘, Y‘ ,Z‘)
xC
zC {C}
Camera origin
y yC
https://de.wikipedia.org/wiki/CCD-Sensor#/media/File:Delta-
https://de.wikipedia.org/wiki/Active_Pixel_Sensor#/media/File:Ac Doped_Charged_Coupled_Devices_(CCD)_for_Ultra-
4. Image formation
tive_pixel_sensor_prototype.jpg Violet_and_Visible_Detection.jpg
P = (X, Y, Z)
4. Image formation
z
x
P‘ = (X‘, Y‘ ,Z‘)
H
xC
zC {C}
Camera origin
y W yC
https://de.wikipedia.org/wiki/Chromatische_Aberration
https://en.wikipedia.org/wiki/Spherical_aberration
http://hyperphysics.phy-astr.gsu.edu/hbase/geoopt/coma.html
https://de.mathworks.com/help/vision/ug/camera-calibration.html
Without distortion (left), Pincushion distortion (mid) and Barrel distortion (right)
http://vrguy.blogspot.com/2013/07/what-is-geometric-distortion-and-why.html
Barrel distortion
http://www.scandig.eu/Kameraverzerrung.html
u u ( k1 r 2
+ k 2 r 4
+ k 3 r 6
+ ... )
+
2 p1uv + p2 ( r 2
+ 2u 2
)
=
v v ( k1r 2 + k2 r 4 + k3r 6 + ...) p1 ( r 2 + 2v 2 ) + 2 p2uv
radial tangential
https://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=i
mages&cd=&cad=rja&uact=8&ved=2ahUKEwiy5qyJqPLgAhUGJ
1AKHaxVC44Qjhx6BAgBEAI&url=https%3A%2F%2Fwww2.infor
matik.hu-
berlin.de%2F~stuebing%2Fpublic%2FStudienarbeit_Sascha_Stu
ebing_2010.pdf&psig=AOvVaw1bTNeojvyH8AGSAGz_Ywkg&us
t=1552126514887415
http://www.signkwik.com/security
-camera-calibration.html
http://www.vision.caltech
.edu/bouguetj/calib_doc/
htmls/example.html
https://upload.wikimedia.
org/wikipedia/commons/
c/c3/The_Squirrels_0048.
jpg
P = (X, Y, Z)
4. Image formation
z R x
Θ Φ
u
P‘ = (u,v)
{C}
Camera origin r
Φ
v
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Fisheye lens camera 103
Stereographic r = 2 f tan ( )
Equidistant (Equiangular) r = f
Equisolid
r = 2 f sin
2
orthographic r = f sin ( )
http://www.fisheyelens.de/html/def.html
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Fisheye lens camera 105
4. Image formation
http://www.fisheyelens.de/html/def.html
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Catadioptric camera 106
4. Image formation
https://commons.wikimedia.org/wiki/File:Phoe
nix_500mm_f8_mirror.jpg
https://www.math.drexel.edu/~
ahicks/design/what-are.html
Imaging L m
Perspective 0 f
Stereographic 1 f
Fisheye >1 f
Catadioptic (elliptical, 0<ε<1) 2 2 ( 2 p − 1)
1+ 2 1+ 2
Chroma-keying:
5. Image processing
Iu = Ku * I Iv = Kv * I
5. Image processing
Derivate of
I u = K * G ( ) * I
Gaussian:
5. Image processing
I ( u, v ) I ( u, v )
2 2
1 2
( I ( u, v ) − I ) ( I ( u, v ) − I )
2 2
1 1 1 1
f(·)=min(·)
Dilatation imdilate O=IꚚS
f(·)=max(·)
Open imopen O = I ○ S = (I Ө S) Ꚛ S
Erosion then dilatation
Close imclose O = I ● S = (I Ꚛ S) Ө S
Dilatation then erosion
https://de.mathworks.com/help/images/ref/bwskel.html
classification
representation
features
0, if I u , v t
C u , v =
6. Image feature extraction
1, if I u , v t
t..threshold, C..Class, I…intensity
t u , v = (W ) − k (W )
mpq = u v I u , v
p q
( u ,v )I
where p,q are order of moment
0. moment:
m00 = I u, v
( u ,v )I
Center of mass:
m10 m01
uc = vc =
m00 m00
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Central Moments of an image 128
pq = ( u − uc ) ( v − vc ) I u, v
p q
6. Image feature extraction
( u ,v )I
Symmetric matrix:
20 11
J =
11 02
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Central Moments of an image 129
Sides of ellipse: 2 1
a=2 b=2
6. Image feature extraction
m00 m00
λ..eigenvalue of J
Aspect ration: a
r=
b
Angle of orientation:
x22
= tan −1
x21
x2… last column of eigenvector
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Other Moments 130
ij
normalized central moments: ij = i+ j
1+
2
6. Image feature extraction
00
Hu-moments:
H1 = 20 + 02 H 2 = (20 −02 ) + ( 211 )
2 2
H5 = (30 − 312 )(30 + 12 ) (30 + 12 ) − 3 (21 + 03 ) + ( 321 −03 )(21 + 03 ) 3 (30 + 12 ) − (21 + 03 )
2 2 2 2
H 6 = (20 −02 ) (30 + 12 ) − (21 + 03 ) + ( 411 )(30 +12 )(21 +03 )
2 2
H 7 = ( 321 −03 )(30 −12 ) (30 + 12 ) − 3 (21 + 03 ) − (30 − 312 )(21 + 03 ) 3 (30 + 12 ) − (21 + 03 )
2 2 2 2
Circularity:
6. Image feature extraction
4 m00
= 2
p
p…perimeter length
m00 …0.order moment (area for binary images)
v = −u arctan ( ) +
sin ( )
6. Image feature extraction
θ η u
η θ
ρ
η
θ
v
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Detecting points 134
6. Image feature extraction
interest point
interest point
any point on line (corner point)
( I u + + i , v + v + j − I u + i , v + j )
6. Image feature extraction
s ( u , v, u , v ) =
2
u
( i , j )W
C ( u, v ) = min s ( u, v, u , v )
u ,v
s ( u , v, u , v ) = W ( i , j ) ( I u + u + i , v + v + j − I u + i , v + j )
2
( i , j )W
6. Image feature extraction
with
I u + u + i , v + v + j I u + i , v + j + I u u + i , v + j u + I v u + i , v + j v
results
s ( u , v, u , v ) = W ( i, j ) ( I u u + i, v + j u + I v u + i , v + j v )
2
( i , j )W
s ( u, v, u , v ) = u2 W ( i, j ) I u 2 u + i, v + j + v2 W ( i, j ) I v 2 u + i, v + j + u v W (i, j ) I u + i, v + j I u + i, v + j
u v
( i , j )W ( i , j )W ( i , j )W
u
s ( u , v, u , v ) = ( u v ) A
v
W ( i, j ) I u 2 u + i, v + j W ( i, j ) I u u + i, v + j I v u + i , v + j
( i , j )W ( i , j )W
A=
W ( i, j ) I u u + i, v + j I v u + i, v + j W ( i , j ) I v 2 u + i , v + j
(i , j )W ( i , j )W
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Harris corner detection (2) 137
can convert to
6. Image feature extraction
G ( I I u2 ) G ( I I u I v )
A=
G ( I I u I v ) G ( I I v2 )
CH ( u , v ) = det ( A ) − k tr ( A )
with
n
det ( A) = 12 = sgn ( ) ai , (i )
i =1
tr ( A ) = 1 + 2 = ( a jj )
n
j =1
Gaussian sequence
6. Image feature extraction
- - -
https://de.mathworks.com/help/vision/object-detection-using-features.html?s_tid=CRUX_lftnav
P = (X, Y, Z)
7. Multi Vision and VBC
z
x
P1‘ = (X1‘, Y1‘ ,Z1‘)
zC {C1}
y W xC
yC
{C2}
yC
SS 2021 Robotic Vision – Prof. Dr.-Ing. Maria Schweigel
Corresponding Points 144
7. Multi Vision and VBC
Describe points
Descriptor vector
Find similar points
T
P2 ' F P1 ' = 0
7. Multi Vision and VBC
F K 2 −T t X R K1−1
l1 F T P2 '
l2 F P1 '
https://www.uio.no/studier/emner/matnat/its/nedlagte-emner/UNIK4690/v16/forelesninger/lecture_6_3_stereo_processing.pdf
https://www.uio.no/studier/emner/matnat/its/nedlagte-emner/UNIK4690/v16/forelesninger/lecture_6_3_stereo_processing.pdf
bx bx
Z= f X = xL
bx
Y = yL
d d d
https://de.mathworks.com/matlabcentral/fileexchange/46148-image-mosaicing
7. Multi Vision and VBC
http://leblogdex.overblog.com/image-mosaicing
Camera motion
https://de.mathworks.com/help/vision/ug/monocular-visual-odometry.html
P = J P ( P, K , C )
P = J P ( P ', Z ) ( )
+
J P , Z
P 1 1
=
( P '* − P')
vx
fˆ
−
u Z
0
u uv
Z fˆ
−
u 2
fˆ + v
fˆ
vy
vz
(
J P , Z
P n n )
= w
v fˆ v ˆ v 2 x
−u
uv
0 − Z Z f + fˆ −
fˆ
wy
w
z
fˆ
− 0 v vx
P = Z vy
7. Multi Vision and VBC
fˆ
0 − −u wz
Z
https://www.servomagazine.com/magazine/article/a-look-at-holonomic-
locomotion