CS计算机代考程序代写 Multiple View Geometry: Solution Sheet 3

Multiple View Geometry: Solution Sheet 3
Prof. Dr. Florian Bernard, Florian Hofherr, Tarun Yenamandra
Computer Vision Group, TU Munich
Link Zoom Room , Password: 307238

Exercise: May 12th, 2021

Part I: Theory

1. (a) M =
(
I T
0 1

)
=




1 0 0 tx
0 1 0 ty
0 0 1 tz
0 0 0 1




(b) M =
(
R 0
0 1

)
=



r11 r12 r13 0
r21 r22 r23 0
r31 r32 r33 0
0 0 0 1




(c) M =
(
I T
0 1

)(
R 0
0 1

)
=

(
R T
0 1

)
=



r11 r12 r13 tx
r21 r22 r23 ty
r31 r32 r33 tz
0 0 0 1




(d) M =
(
R 0
0 1

)(
I T
0 1

)
=

(
R RT
0 1

)
=



r11 r12 r13 r1T
r21 r22 r23 r2T
r31 r32 r33 r3T
0 0 0 1


,

where r1, r2, r3 are the row vectors of R: R =


— r1 —— r2 —

— r3 —


.

2. Let M := (M1 −M2) =:


m11 m12 m13m21 m22 m23
m31 m32 m33


.

”⇒”:
We show that M is skew-symmetric by distinguishing diagonal and off-diagonal elements of
M :

(a) ∀i : 0 = e>i Mei = mii where ei = i-th unit vector
(b) ∀i 6= j : 0 = (ei + ej)>M(ei + ej) where ej = j-th unit vector

= mii +mjj +mij +mji ⇒ mij = −mji

hence, mii = 0 and mij = −mji, i.e. M is skew-symmetric.

1

https://tum-conf.zoom.us/s/62772800235?pwd=SUpZN2QrV0JpeXJyR2R1TWx5cHEwdz09

”⇐”:
using M = −M>, we directly calculate

∀x : x>Mx = (x>Mx)> = x>M>x = −(x>Mx)
⇒ x>Mx = 0

Alternative for ”⇐”:

∀x : x>Mx = x>(M̌ × x) = 0

Because M is skew-symmetric, Mx can be interpreted as a cross product. The result of any
cross product with x is orthogonal to x.

3. We know: ω = (ω1 ω2 ω3)> with ||ω|| = 1 and ω̂ =


 0 −ω3 ω2ω3 0 −ω1
−ω2 ω1 0


.

(a)

ω̂2 =


 −(ω22 + ω23) ω1ω2 ω1ω3ω1ω2 −(ω21 + ω23) ω2ω3

ω1ω3 ω2ω3 −(ω21 + ω
2
2)


=




ω21 − (ω
2
1 + ω

2
2 + ω

2
3)︸ ︷︷ ︸

1

ω1ω2 ω1ω3

ω1ω2 ω
2
2 − (ω

2
2 + ω

2
1 + ω

2
3)︸ ︷︷ ︸

1

ω2ω3

ω1ω3 ω2ω3 ω
2
3 − (ω

2
3 + ω

2
1 + ω

2
2)︸ ︷︷ ︸

1




=


 ω21 ω1ω2 ω1ω3ω1ω2 ω22 ω2ω3

ω1ω3 ω2ω3 ω
2
3


−


 1 0 00 1 0

0 0 1


= ωω> − I

ω̂3 = ω̂ ω̂2

= ω̂ (ωω> − I)
= ω̂ ω (ω>)− ω̂I
= (ω × ω) ω> − ω̂
= −ω̂ (as ω × ω = 0)

2

Alternative solution for ω̂3:

ω̂3 =


 −(ω22 + ω23) ω1ω2 ω1ω3ω1ω2 −(ω21 + ω22) ω2ω3

ω1ω3 ω2ω3 −(ω21 + ω
2
2)


 ·


 0 −ω3 ω2ω3 0 −ω1
−ω2 ω1 0


=




0 ω3 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

−ω2 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

−ω3 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

0 ω1 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

ω2 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

−ω1 · (ω21 + ω
2
2 + ω

2
3)︸ ︷︷ ︸

1

0




= −ω̂

(b) The formulas for n even and odd can be found by writing down the solutions for n =
1, . . . , 6:

ω̂

ω̂2

ω̂3 = −ω̂
ω̂4 = −ω̂2 as: ω̂4 = ω̂3ω̂ = −ω̂ω̂ = −ω̂2

ω̂5 = ω̂ as: ω̂5 = ω̂4ω̂ = −ω̂2ω̂ = −ω̂3 = −(−ω̂) = ω̂
ω̂6 = ω̂2 as: ω̂6 = ω̂5ω̂ = ω̂ω̂ = ω̂2

For even numbers: ω̂2

ω̂4 = −ω̂2

ω̂6 = ω̂2

For odd numbers: ω̂

ω̂3 = −ω̂
ω̂5 = ω̂

even: ω̂2n = (−1)n+1 ω̂2 for n ≥ 1
odd: ω̂2n+1 = (−1)n ω̂ for n ≥ 0

Proof via complete induction:
i. For even numbers 2n where n ≥ 1:

– n = 1 : ω̂2 = (−1)2ω̂2

– Induction step n→ n+ 1 :

ω̂2(n+1) = ω̂2n · ω̂2

= (−1)n+1 · ω̂2 · ω̂2 (assumption)
= (−1)n+1 · ω̂3 · ω̂
(a)
= (−1)(n+1)+1 · ω̂2

3

ii. For odd numbers 2n+ 1 where n ≥ 0:
– n = 0 : ω̂1 = (−1)0ω̂
– Induction step n→ n+ 1 :

ω̂2(n+1)+1 = ω̂2n+1 · ω̂2

= (−1)n · ω̂ · ω̂2 (assumption)
= (−1)n · ω̂3
(a)
= (−1)n+1 · ω̂

(c) We know: ω ∈ R3. Let ν = ω‖ω‖ and t = ‖ω‖. Hence, w = νt, ω̂ = ν̂t.

eω̂ = eν̂t

=
∞∑
n=0

(ν̂t)n

n!

= I +

∞∑
n=1

t2n

(2n)!
ν̂2n +

∞∑
n=0

t2n+1

(2n+ 1)!
ν̂2n+1

(b)
= I +

∞∑
n=1

(−1)n+1
t2n

(2n)!︸ ︷︷ ︸
1−cos(t)

ν̂2 +
∞∑
n=0

(−1)n
t2n+1

(2n+ 1)!︸ ︷︷ ︸
sin(t)

ν̂

(def.)
= I +

ω̂2

‖ω‖2
(1− cos(‖ω‖)) +

ω̂

‖ω‖
sin(‖ω‖)

4