International Core Journal of Engineering 2020-26 | Page 144

F ( x i , w j ) Y a fit 1  Y b fit 2
detection operator is converted into a two-dimensional
matrix of 0,1, in which white Representing a binary one,
black represents a binary zero, and finally converts the
binary image into a one-dimensional watermark sequence.
Since the white pixel contains the main information of the
copyright image, the white pixel is used to control the
embedding of the watermark.
There Y a 0.4 , Y b 0.6 , avg a : average value of data
blocks before watermark embedding; avg b :average value of
data block after watermark embedding; V a 2 : variance of data
block before watermark embedding; V b 2 :variance of data
block after watermark embedding; | after ( i , j )  before ( i , j ) | :
Indicates the amount of data change after embedding the
watermark, the abscissa i is the id value, and the ordinate j
represents the attribute column.
The database performs block preprocessing, connects to
the database, counts the attribute names of the database,
randomly selects a list of attributes, and counts the number of
the columns, and compares the number of the numbers with
the set threshold. When the number of the numbers is greater
than the threshold Count all the ids of the tuple in which the
mode is located as a block into the table, and set an index for
each block to facilitate traversal.
The initial population is sorted by embedding the initial
population into the database, and the individuals with better
fitness are obtained with smaller sequence values. Then, the
sorted individuals are given a serial number R i by using the
following formula: R i N  i ; i 1, 2,..., N ; Next, the
selection probability p i of each individual is calculated by
the serial number and the population number:
R
p i
; i 1, 2,..., N ;
N
B. Watermark embedding algorithm
Definition 1. In the m * n relational database, the
relationship is represented by R ( p , A 1 , A 2 ,..., A n ) , where P is
the primary key, and A i (1 d i d n ) 0 is the attribute in the
relational database R, then r j (1 d j d m ) is a tuple of the
database, and ( r j , A i ) represents the value of the i-th
attribute of the j-th row tuple.
(3) From the population X i 0
Each data block is assigned a random 8-bit binary
watermark sequence, and the one-dimensional array
generated by the image is traversed. When the pixel is 0, no
operation is performed. When the pixel is 1, the following
operations are performed:
^
`
U i g  1
^ u
i 1
g  1
`
, u i 2 g  1 , u i 3 g  1 ,..., u iD g  1 ; i 1, 2,3,..., N
(5) According to the selection formula, variation of
individuals to generate new population after crossover
operation u i g  1 compared with x i g , if the candidate
`
1 m m
¦¦ | after ( i , j )  before ( i , j ) | 2
m * n i 1 j 1
`
(4) The parent x j g and the v i g  1 generated after mutation
were partially crossed according to the crossover formula to
obtain the experimental vector :
(2) The fitness function is calculated for the population
by the following formula, and the fitness function is set to
two parts:
fit 2
, x i 2 0 , x i 3 0 ,..., x iD 0
is an evolutionary algebra.
(1) The scaling factor F and the crossover probability CR
are initialized, and F 0.5 , CR 0.3 . N individuals are
generated from the database partition by means of a
uniformly distributed random function. x i 0 represents the i-
th individual of the 0th generation, and each individual is
composed of the values of D coordinate points, that is, D-
x i 1 0 , x i 2 0 , x i 3 0 ,..., x iD 0 ; i 1, 2,3,..., N .
dimensional: X i 0
(| avg a  avg b |  | V a 2  V b 2 |) * N
D
0
i 1
generated, and then a random number a is generated from
0~1, and individuals x r 2 and x r 3 are randomly selected. If
the probability values of the selected individuals are greater
than a and mutual If it is not the same, the selection is
successful, otherwise it is re-selected, and it is mutated by the
above variation formula to generate the intermediate
individual
V i g  1 v i 1 g  1 , v i 2 g  1 , v i 3 g  1 ,..., v iD g  1 ; i 1, 2,3,..., N ; where g
Specific steps:
fit 1
^ x
randomly selected An individual x r 1 different from x i is
Definition 2. The least significant bit LSB (Least
Significant Byte), which represents the smallest unit in a
binary number, also refers to the smallest weighted byte in a
multi-byte sequence. For the low redundancy of the
numerical database, the watermark is embedded by changing
the least significant bit.
^
(10)
individuals u i g  1 , the fitness function is better than x i g
chose u i g  1 as a child, otherwise choose x i g as a child,
generate
the
final
population:
X i g  1
x i 1 g  1 , x i 2 g  1 ,..., x iD g  1 ; i 1, 2,3,..., N The final
^
(8)
`
population is the optimal watermark embedding position.
(6) Iterate through the random watermark sequence
allocated for the data block. When the watermark parameter
is 0, the position to be embedded is traversed, and the least
significant bit of the position data is subtracted. When the
watermark parameter is 1, the least significant bit of the
position data is added.
(9)
Add weights to two formulas:
(7) Determine whether all the data blocks are traversed, if
not, return (1).
122