Available online www.jocpr.

com

Journal of Chemical and Pharmaceutical Research, 2014, 6(3):170-178

ISSN : 0975-7384
Research Article CODEN(USA) : JCPRC5

Application of linear transformation in numerical calculation

Yuhuan Cui*, Jingguo Qu and Guanchen Zhou

Qinggong College, Heibei United University, Tangshan, China

_____________________________________________________________________________________________

ABSTRACT

The linear transformation is the most simple and the most basic transformation. Such as, the linear function is the
most simple and the most basic function. The linear transformation is a main research object of linear algebra. This
text uses a series of related knowledge of linear transformation thought in linear algebra, makes study and analysis
finely to the position and application of the linear transformation thought in mathematics, and sieves several typical
example.

Keywords: Linear transformation, Matrix, Numerical Calculation, The application

_____________________________________________________________________________________________

INTRODUCTION

Linear transformation is one of the core content of linear algebra. It plays an important role in between the whole
structure and linear vector space memory link. Concept of linear transformation is the transformation of coordinates
in analytic geometry, some transformation in mathematical analysis to replace the generalization and abstraction. Its
theory and methods in analytic geometry, differential equations and many other fields, and it has widespread
application.

LINEAR TRANSFORMATION
Linear mapping
Definition 1 Suppose  be a mapping from V to W . If you meet the following conditions, called  is a mapping
from V to W .
(1)For arbitrary    V ,            
(2)For arbitrary a  F ,   V ,   a   a  

Some basic properties of linear mapping:

Condition (1) and condition (2), in the definition 1, is the condition (3) equivalent conditions.
(3)For arbitrary a, b  F and  ,  V
 (a  b )  a ( )  b ( )

:V  W satisfies the condition (1) and (2), for any a, b  F and  ,  V ,

In fact, if the mapping 
 (a  b )   (a )   (b )  a ( )  b ( )

Conversely, suppose (3) was established, made a  b  1 , get condition (1); and b  0 , get the condition (2). In the
condition (2), a  0 get
  0  0

170
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
In other words, the linear mapping will be zero vectors onto the zero vectors.
By (3), as the mathematical induction on n , easy to launch:
  a11   an n   a1  1    an  n 
For arbitrary a1 , an  F and  ,  n  V are established.

Linear transformation
Let V be a vector space of a number field P .

Some basic properties of linear transformation,

1. Suppose A be a linear transformation of V , A(0)  0 and A( )   A( ) .
2. Linear transformation keeps linear combination and linear equation invariant. If  is a linear combination
of 1 ,  2 , L ,  r ,
  k1 1  k 2 2    k r r

Then after the linear transformation A , A(  ) is a linear combination of A(1 ), A( 2 ), , A( r ) :
A(  )  k1 A(1 )  k2 A( 2 )    kr A( r )

If there is a linear relationship between  1 ,  2 ,  ,  r

k1 1  k 2 2    k r r  0

So
k1 A(1 )  k2 A( 2 )    kr A( r )  0

3. Linear transformation can change linear correlation vectors into linear dependent set of vectors.

Example [1]
Case 1 Suppose  be a nonzero vector in geometry space, each vector  to it in the  on the injection of
transformation is a linear transformation, with   expressed, formula is
( ,  )
  ( )  
( ,  )
Here ( ,  ), ( ,  ) denotes the inner product.

Case 2 In the linear space P[x] or P[ x] n , differential quotient is a linear transformation. This transformation
usually use D representative,
D f ( x)  = f (x)

MATRIX OF A LINEAR TRANSFORMATION

The definition of linear transformation matrix [2]

Definition 2 Suppose  1 ,  2 ,  ,  n is a set of basis in n -dimensional linear space V which Belongs to the
number field P . And A is a linear transformation of V .So the basis vectors can be linearly expressed as
 A1  a111  a21 2    an1 n
 A  a   a     a 
 2 12 1 22 2 n2 n

   
 A n  a1n1  a2 n 2    ann n

With the matrix representation is

171
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________

A(1 ,  2 , ,  n )   A(1 ), A( 2 ), , A( n )   (1 ,  2 , ,  n ) A (1)

Among them
 a11 a12  a1n 
 
a a22  a2 n 
A   21
   
 
 an1 an 2  ann 

Matrix A is called the matrix of linear transformation A on the base of  1 ,  2 ,  ,  n .

The related theorem [3]

Theorem 1 Suppose  1 ,  2 ,  ,  n is a set of basis in linear space V ,  1 ,  2 ,,  n is n vector arbitrary in V .
There is a linear transformation of A
A  i   i , i  1,2,  , n

Theorem 2 Suppose  1 ,  2 ,  ,  n is a set of cardinality in n -dimensional linear space V which Belongs to the
number field P . In this base, every linear transformation according to the formula (1) corresponding to a n  n
matrix. This corresponds with the following properties.
(1) Linear transformation sum corresponding to the matrix sum.
(2) The product of linear transformation corresponding to the matrix product.
(3) The number product of linear transformation corresponding to the matrix number product.
(4) Invertible Linear Transformation corresponding to invertible matrix.

Theorem 3 Suppose matrix A is the matrix of linear transformation A on the base of  1 ,  2 ,  ,  n . On the base
of  1 ,  2 ,  ,  n , vector  coordinate is ( x1 , x 2 ,  , x n ) .So the A  coordinate, on the base of  1 ,  2 ,  ,  n ,
is ( y1 , y 2 ,  , y n ) .And

 y1   x1 
   
 y2   x
A 2 
    
   
 yn   xn 

Theorem 4 Suppose matrix of linear transformation of A, which in linear space V , under the two bases of (2) and (3)
are respectively A and B .From the base (2) to (3) the transition matrix is X ,and B  X 1 AX .
 1 ,  2 ,,  n (2)
1 ,2 , ,n (3)

Example[3]

Case 1 Suppose  1 ,  2 ,,  m is a set of basis in subspace W which belongs to n -dimensional linear
space V (n  m) .It expanded a set of basis in V as  1 ,  2 ,  ,  n .Specifies a linear transform A meet
 A i   i , i  1, 2, , m

 A i  0, i  m  1, , n

So the linear transformation A called a projection in subspace W .It is easy to find that

172
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
2
A =A

Matrix of projection A under the base of  1 ,  2 ,  ,  n is

1 
 
 1 
  
 
 1 
 0 
 
  
 0 

By determining a set of basis, established a mapping from linear transformation in n -dimensional linear
space V which Belongs to the number field P to the n  n matrix.

Case 2 Suppose the matrix of linear transformation A, which in three dimensional linear spaces V , under the base of
1, 2 , 3 is

 a11 a12 a13 

A   a21 a22 a23 
 a31 a32 a33 
(1) Seeking the matrix of A under the base of  3 ,  2 ,  1 .
(2) Seeking the matrix of A under the base of  1 ,  2 ,  3 , among them   P and   0 .

Solving:
A  1  a11 1  a 21 2  a31 3
A  2  a12  1  a 22  2  a32  3
A  3  a13 1  a 23 2  a33 3

So
A3  a33 3  a 23 2  a13 3
A  2  a 32  3  a 22  2  a12  1
A  1  a 31 3  a 21 2  a11 1

The matrix of A under the base of   3 ,  2 , 1  is

 a33 a32 a31 

a a 22 a 21 
 23
 a13 a12 a11 
In the same way,
 21
A 1  a11 1  ( ) 2  a31 3

A  2  a12  1  a 22 2  a 32  3

A  3  a13 1  ( 23 ) 2  a 33 3


173
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
The matrix of A under the base of  1 ,  2 ,  3 is
 a11 a12 a13 
 21  23 
 a 22 
   
 a31 a32 a33 

Case 3 Suppose V is a Two-dimensional linear space in the number field P , and  1 ,  2 is a set of basis. The
matrix of linear transformation A under the base of  1 ,  2 is
 2 1
  1 0
 

Calculating the matrix of A under another base of 1 , 2 ,among them

 1 1
1 ,2    1 ,  2   
 1 2 

So, the matrix of A under the base of 1 , 2 is that

1
 1 1  2 1   1 1  2 1  2 1   1 1 3 2   1 1 1 1
 1 2   1 0   1 2   1 1  1 0   1 2   1 1   1 2   0 1
             

Case 4 Suppose  1 ,  2 ,  3 ,  4 is a set of basis in four dimensional linear spaces V , known the matrix of A under
the base of  1 ,  2 ,  3 ,  4 is
1 0 2 1 
 1 2 1 3 

1 2 5 5
 
 2 2 1  2

Seeking the matrix of A under the base of 1   1  2 2   4 ， 2  3 2   3   4 ，  3   3   4 ，

 4  2 4 .
Solving:
1 0 0 0
 2 3 0 0
(1 , 2 , 3 , 4 )  ( 1 ,  2 ,  3 ,  4 ) 
 0 1 1 0
 
 1 1 1 2

So
1
1 0 0 0 1 0 2 1  1 0 0 0
 2 3 0 0  1 2 1 3   2 3 0 0
  
 0 1 1 0 1 2 5 5  0 1 1 0
     
 1 1 1 2  2 2 1  2  1 1 1 2

174
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________

2 3 3 2
2 4 10 10 
 
 
3 3 3 3
8 16 40 40 
  
3 3 3 3
 0 1 7 8 

Case 5 Two bases of P 3 of given [4]

 1  (1,0,1) ，1  (1,2,1) ，
 2  (2,1,0) ，  2  (2,2,1) ，
 3  (1,1,1) ， 3  (2,1,1) 。

The definition of linear transformation A:

Ai   i ， i  1,2,3 。
(1) Write out the transition matrix from  1 ,  2 ,  3 to 1 , 2 , 3 .
(2) Write out the matrix of A under the base of  1 ,  2 ,  3 .
(3) Write out the matrix of A under the base of1 , 2 , 3 .
Solving:
 1 2 2  1 2 1
 2 2 1  0 1 1 X
   
 1 1 1 1 0 1
 3 3 
1  2  2 2 
1 2 1 1 2 2  
3 3 
X  0 1 1  2 2 1   1
   2 2 
1 0 1  1 1 1 
1 1 5 

 2 2 
Because of

Ai  i   i A

So
(1 , 2 , 3 )  ( 1 ,  2 ,  3 ) A
 3 3 
 2  2 2 
 
3 3 
A X  1
 2 2 
 1 5
1  
 2 2 

And

175
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
1
 3 3   3 3   3 3 
 2 
2 2   2 
2 2   2  2 2 
 3 3   3 3   3 3 
1  1  1 
 2 2   2 2   2 2 
1 1 5  1 5 1 5
 1   1  
 2 2   2 2   2 2 
 3 3 
 2 
2 2 
 
3 3 
1 X
 2 2 
 1 5 
1 
 2 2 

APPLICATION OF LINEAR TRANSFORMATION IN DEFORMATION MEASUREMENT

Digital camera belongs to the non-metric camera. It has no fiducial mark and orientation equipment; when it goes on
field photography can not make camera horizontally, therefore the initial inner and outer orientation elements of
value can not be obtained when the camera at the moment of photography. So the past commonly used method is no
longer applicable, we must adopt new data processing method [5].

In this study using the direct linear transformation (DLT) method, the method doesn’t need fiducial mark and the
initial inner and outer orientation elements of value. Therefore this method is suitable for industrial and civil
Photogrametry. The principle is
x  ( L1 X  L2Y  L3 Z  L4 ) / ( L9 X  L10Y  L11Z  1)  0
(4)
z  ( L5 X  L6Y  L7 Z  L8 ) / ( L9 X  L10Y  L11Z  1)  0

Considering the effects of camera lens distortion measurement of non quantity, in the formula (4) is introduced
corresponding correction terms x, z .

x  ( x  x0 )r 2 K1
z  ( z  z0 )r 2 K1
Obtained
x  ( x  x0 )r 2 K1  ( L1 X  L2Y  L3 Z  L4 ) / ( L9 X  L10Y  L11Z  1)  0
(5)
z  ( z  z0 )r 2 K1  ( L5 X  L6Y  L7 Z  L8 ) / ( L9 X  L10Y  L11Z  1)  0

Among them K1 is the coefficient symmetry lens distortion, x0 , z0 is the main points of the image coordinatograph
coordinates. The image diameters is
1
r  ( x  x0 )  ( z  z0 ) 
2 2 2

Inverse calculating the principal point coordinates of image

x0  ( L1 L9  L2 L10  L3 L11 ) / ( L2 9  L210  L211)
(6)
z0  ( L5 L9  L6 L10  L7 L11 ) / ( L2 9  L210  L211)

Inverse calculating the principal distance of photograph

Fx2   x0 2  ( L12  L2 2  L32 ) / ( L9 2  L10 2  L112 )
(7)
Fz2   z0 2  ( L5 2  L6 2  L7 2 ) / ( L9 2  L10 2  L112 )

176
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
Among them
F   Fx  Fz  2

Direct linear transformation formula is a set of nonlinear equations; it is calculated with the method of least squares
iterative method. The algorithm consists of two steps,
( x  Vx )  ( L9 X  L10Y  L11Z  1)  ( L1 X  L2Y  L3 Z  L4 )  0
(8)
( z  Vz )  ( L9 X  L10Y  L11Z  1)  ( L5 X  L6Y  L7 Z  L8 )  0

The error equation of positive operator

Vx 
Vz  
 
 L1 
L  (9)
 xZ  A( x  x0 )r 2    1  X 
2
1  X Y Z 1 0 0 0 0  xX  xY
*  *     * 
A  0 0 0 0 X Y Z 1  zX  zY  zZ  A( z  z0 )r 2    A  Z 
 L11 
 K1 

By (9) can be calculated for 11 L coefficients and a distortion coefficient K1 , the above formula written as
V  B  L W
The Method equation is
BT  B  L  BT  W  0
The solution is
L   L1 , L2 , , L11 , K1    BT  B   BT  W
T
(10)

This operation step is an iterative process; the difference of two adjacent operations is less than 0.01mm as a limit
cycle iterative solution method. Coefficient of each A is calculated by L value. And once again put the
coefficient of A into the error equation, calculation of L coefficient, so the cycle until less than setting limits
0.01mm [6].

When was solved, we can obtain the coordinatograph coordinates,

x  x  x  ( x  x0 )  r 2  K1
(11)
z  z  z  ( z  z0 )  r 2  K1

Because each photograph has respective coefficient L , the process of positive operator should be carried out with
piecewise.
X 
Vx  1  L1  xL9 L2  xL10 L3  xL11    1  x  L4 
Vz   A *  L  zL L6  zL10
 Y  
L7  zL11    A  z  L8 
（12）
   5 9
 Z 

For each photograph of each image point can have type (12), In order to obtain the X , Y , Z , at least two photograph.
The formula is
V  N S Q

The Method equation is

NT  N  S  NT Q  0

177
Yuhuan Cui et al J. Chem. Pharm. Res., 2014, 6(3):170-178
______________________________________________________________________________
The solution is
S   X Y  Z    N T  N   N T  Q
T 1
(13)

The above calculations take the point by point manner, to accelerate the computation speed, reducing the occupied
memory capacity [7].

CONCLUSION

This paper introduces the concept, basic knowledge and correlation theory of linear and linear transformation matrix,
and given many numerical examples, embodies the importance of linear transformation thought in numerical
calculation. Finally introduces linear transform in the application of deformation measurement.

REFERENCES

[1] Peking University Department of Mathematics algebraic group in geometry and algebra Research Laboratory,
higher algebra[M], Higher Education Press,2001
[2]Li Xin, Matrix theory and its application[M], Chongqing University press,2005
[3]Chen Wendeng, Huang Xiankai, Mathematics Review Guide[M], World Book Inc,2004
[4]Yu Chengxin, Zhang Xiangdong, Mou Yuzhi, Song Chuanzeng, Quan Jin, Ding Ning, Li Xianli, The application
study of the direct linear transform in deformation observation[J],Journal of Shandong institute of ARCH and
ENG,2002:23-28
[5] Lipschutz S., Theory and Problems of Linear Algebra[M], McGraw-Hill，1991
[6] Griffel D H., Linear Algebra and its Applications[M], Marcei Dekker, 1998
[7] Jacobson N., Lectures in Abstract Algebra II. Linear Algebra[M], Graduate Texts in Math.31, 1953

178