Sei sulla pagina 1di 5

Assignment 2

Benjamin Reich

May 3, 2011

Part I

Rotations

Figure 1: Image Rotation

The image rotation in Figure 1 occurs due a rotation occuring in the xz-plane. Say we took the photo and put it on an axis in
R3 , then we picked an arbitrary point located within the image which we will use to track the rotations we perform on the
image.

 
−1
The image on a 3 dimensional axis is show in Figure 2 and we'll pick the point A =  2  which we have also shown in
0
o
Figure 2. If we want to rotate the image 90 around the y-axis, and hence the point A we would have to map the point A to a
 
0
new point B which is represented by  2  (deduced from inspection).
−1

Figure 2: Image placed on Axis with arbitrary point

1
Ben Reich Assignment 2

From our knowledge of rotational matrices we can at least draw upon the matrix for rotations around the y-axis and then seek
out to proove that it works. So our rotational matrix about the y-axis is

 
cos(θ) 0 sin(θ)
Ry (θ) =  0 1 0 
− sin(θ) 0 cos(θ)
π
Now if our previous statement holds true, then solving Ry (θ) · A = B should yield θ= 2.

    
cos(θ) 0 sin(θ) −1 0
 0 1 0  2  = 2 
− sin(θ) 0 cos(θ) 0 −1
   
− cos(θ) 0
 2  = 2 
sin(θ) −1
This leaves us with 3 equations, obviously the middle one is trivial but the other two should each yield a general solution with
a common intersection occuring.

cos(θ) =0
π
θ =πn − , n∈Z
2

sin(θ) = − 1
π
θ =2πn − , n∈Z
2
Now by simple inspection you can see that these two solutions for θ have an intersection when n = 0. This gives us the answer
of θ = − π2 , which is basically saying that the point underwent a rotation around the y-axis in an anti-clockwise direction of 90o .
So going back to the original question, when the image was displayed facing a mirror and the image appeared as if it had been
ipped horizontally, what actually happened was each point in the image (which we will represent as matrices) underwent a
simple rotation of π. Finally to get the rotational matrix that performs this ip we merely have to solve Ry (π). So without
further ado

 
cos(π) 0 sin(π)
Ry (π) =  0 1 0 
− sin(π) 0 cos(π)
 
−1 0 0
= 0 1 0 
0 0 −1
This can be quite easily checked using our previous point A. Rotating A by π will merely negate the x-ordinate as the
y-ordinate will always stay the same (an eigenvector) and because it is a half rotation the z-ordinate doesn't change either. So
applying this transformation matrix on A we get

  
−1 0 0 −1
Ry (π) · A =  0 1 0   2 
0 0 −1 0
 
1
= 2 
0
Just as we suspected.

2
Ben Reich Assignment 2

Part II

Eigenvectors and Eigenvalues


The matrix equations to show that v1 and v2 are eigenvectors of A is simply

A · v1 =λ1 · v1 (1)

A · v2 =λ2 · v2 (2)

The transpose of equation 2 is merely

T T
(A · v2 ) = (λ1 · v2 )
v2T · AT =v2T · λT2

But since A is symmetrical and λ2 is a number, there transpose is equal to their original value.

v2T · A =v2T · λ2

Multiplying this equation from the right by v1 leads to

v2T Av1 =v2T λ2 v1

Rearranging equation 1 for v1

v1 =A−1 λ1 v1
T
Subbing that back into the equation (v2 Av1 = v2T λ2 v1 ) yields

v2T AA−1 λ1 v1 =v2T λ2 A−1 λ1 v1


v2T AA−1 λ1 v1 =v2T λ2 v1
v2T Iλ1 v1 − v2T λ2 v1 =0
v2T v1 (Iλ1 − λ2 ) =0

Now if we go ahead and solve that equation it gives us the following two equations with which to ddle with.

v2T v1 =0 (3)

Iλ1 − λ2 =0 (4)

Now since anything multiplied by an identity matrix just leaves it the same, the second equation cannot be true as we have
been given the information that λ1 6= λ2 =⇒ λ1 − λ2 6= 0 therefore allowing us to deduce that v2T v1 = 0 must be true.

The dot product of two vectors can also be represented as matrix multiplication of the transpose of the rst vector (in matrix
form) and the second vector (in matrix form). i.e.

a · b =aT b
T
So using this, an interesting property that pops up from the previous answer (v2 v1 = 0) is that the vectors v2 and v1 are
perpendicular.

3
Ben Reich Assignment 2

Part III

Integration by Parts
To begin with I split the integral up into two parts so that I could perform integration by parts on it. Such that

ˆ
I(n) = sinn (x).dx
ˆ
= sin(x) · sinn−1 (x).dx

Performing some integration by parts on that integral gives us the following

ˆ ˆ
sinn (x).dx = − sinn−1 (x) · cos(x) + (n − 1) cos(x) sinn−2 (x) cos(x).dx
ˆ
n−1
= − sin (x) · cos(x) + (n − 1) cos2 (x) · sinn−2 (x).dx

Using the trigonometric identity of sin2 (x) + cos2 (x) = 1 we can get that cos2 (x) out of there and get some cancelling going on
n−2
with that sin (x).

ˆ ˆ
n n−1
sin (x).dx = − sin (1 − sin2 (x)) · sinn−2 (x).dx
(x) · cos(x) + (n − 1)
ˆ ˆ
= − sinn−1 (x) · cos(x) + (n − 1) sinn−2 (x).dx − (n − 1) sinn (x).dx

Now that is left is to tidy the integral up and get it in terms of I(n − 2)

ˆ ˆ ˆ
sinn (x).dx = − sinn−1 (x) · cos(x) + (n − 1) sinn−2 (x).dx − (n − 1) sinn (x).dx
ˆ ˆ
n sinn (x).dx = − sinn−1 (x) · cos(x) + (n − 1) sinn−2 (x).dx
ˆ ˆ
n−1 1
sinn (x).dx = sinn−2 (x).dx − sinn−1 (x) · cos(x)
n n
 ˆ 
1
= (n − 1) sinn−2 (x).dx − sinn−1 (x) · cos(x)
n

Now performing our substitution to get it into the form we need, gives us the following

1
(n − 1)I(n − 2) − sinn−1 (x) · cos(x) + c

I(n) =
n
Now to use this newly constructed integral to show that

ˆ π/2
dx 4
=
π/4 sin4 (x) 3
1
I(−6) + sinn−1 (x) · cos(x) ,
 
This integral can also be represented as I(−4) and hence leads us to the equation I(−4) = − −4 at
rst glance it appears to be a behemoth of an equation taking us in circles but on further inspection you can actually use this
to your advantage. If we take the integral of I(0) and then rearrange we can quite easily determine the integral of I(−2) and
that will easily set us on the path to solving I(−4).

4
Ben Reich Assignment 2

n · I(n) =(n − 1) · I(n − 2) − sinn−1 (x) · cos(x)


(0) · I(0) =(0 − 1) · I(0 − 2) − sin0−1 (x) · cos(x)
I(−2) = − sin−1 (x) · cos(x)
= − cot(x) + c

Now that we have the value of I(−2) we can set up an integral to solve for what we are chasing I(−4). That integral is

1
−3 · I(−4) − sin−3 (x) · cos(x)

I(−2) = −
2
As you can see we can just rearrange, substitute what we know and it gives us what we need.

−2 · I(−2) = − 3 · I(−4) − sin−3 (x) · cos(x)


3 · I(−4) =2 · I(−2) − sin−3 (x) · cos(x)
1
I(−4) = 2 · I(−2) − sin−3 (x) · cos(x)

3
1
2 · cot(x) + sin−3 (x) · cos(x) + c

=−
3
4
Going back to the original question we can show that the denite integral equals
3.

ˆ π/2
dx 1 π/2
4 =− 2 · cot(x) + sin−3 (x) · cos(x) π/4
π/4 sin (x) 3
1
= − [(2 · (0) + (0)) − (2 · (1) + 2)]
3
1
= − (−4)
3
4
= 
3

Potrebbero piacerti anche