JW 
V4 
Zhaobao Zheng 
  
random from 0~15. Then the elements at the crossover points in the P, directly exchange with the same elements from 
P» and inherits two children .The detailed algorithm of the crossover mechanism is shown in Fig.3. 
Pi 
E oF | b» veo D24 
Crossover poit ; i 
point 1 Crossover point 2 . . . Crossover point 24 
  
  
  
(b) 
Fig.3 An example of the crossover 
(a) Two parent chromosomes P, , P; 
(b) Two child chromosomes P' ; and P' » generated by crossover operation between P, and P» 
2.2.4 Mutation Operator. Mutation is a secondary search operator which increases the variability of the population 
and the ability of exploring the optimal solution. The mutation operator creates new individuals by changing one or 
more of the gene values in the chromosome with a probability equal to the mutation rate P,,,55, . The operator entails 
the two decision phases . The first is to randomly select a chromosome from the population . The second is to randomly 
select a bit from the chromosome as the mutation point and take its reverse value as its value, e.g. O 1,1 0. 
2.2.5 The Procedure of Deciding MRF Optimal Parameters Based on Genetic Algorithm. The procedure of 
deciding MRF optimal parameters based on genetic algorithms is given below ,where P(t) is a population of candidate 
parameters at generation t. 
I0; 
initialize P(t) 
evaluate P(t) 
while not (termination condition) 
begin 
t=t+1; 
reproduce P(t) from P(t-1) 
recombine P(t) by crossover and mutation operator 
evaluate P(t) 
end; 
end while 
3 TEXTURE CLASSIFICATION BASED ON MRF PARAMETERS 
Traditional texture classification methods based on MRF parameters regard the value of MRF parameters as the main 
features for classification. But during our study, we discover that the signs of parameters of different textures are not all 
the same. In other words, the signs of parameters of the same textures must be the same. If the signs of parameters of 
two textures are different, they must not be the same texture. So we presents following procedure of texture 
classification: 
(1) According to texture samples, standard MRF parameters of different texture classes are acquired by genetic 
algorithms. 
(2) To solve MRF parameters of a unknown texture image by GA. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 1051