Where f is the number of the clusters in a chromosome, and Value (Ccej)is the value of the three bands of cluster center Ccej. It is the sum of the obtained weights multiplied by 255.
Value (pixelir) are the values of the three bands of the pixel on the left side of the cluster center Ccej in the chromosome(. . . pixel i;1 pixeli;2pixeli;3 Ccej pixeli+1;1 pixeli+1;2 pixeli+1;3 Ccej . . .
), and y is the number of pixels in each cluster. The fitness F(Chromosome) of each chromosome xi in the current population P(0) is the objective function, and according to the fitness value, each chromosome will be given a number of chances to be selected using the roulette wheel selection process 46, where the lower in cost is given more possibilities formating. Two chromosomes are selected randomly, and they are mated (by replacing cluster centers from one parent with the other parent and vice versa) to create two new children. Mutation operation works on individual chromosomes by changing a cluster value “intensity” with another cluster value that is selected randomly from the available clusters in the image. The probability of crossover can vary between 40% and 70%, while mutation can vary between 10% and 20%. These reproduced children replace their parents, and their fitness values are calculated again. The reproduction process is tailed by the Hill-Climbing process in order to slow the fast convergence of GA toward a local optimal solution. In the first experiment, SPOT5 is used where three bands are enhanced with respect to spatial resolution using the panchromatic image.
The size of the SPOT image is 360 X 360 pixels and the resolution is 5 meters see Figure 2a. SPOT remote-sensing satellites program was created by France in partnership with other European countries such as Belgium.The deployment of SPOT satellites in orbit facilitated the mission of observing basically the entire planet in 1 day. Field work is carried out to verify the classified image where 110 different samples are collected from three classes (1-Urban settlements(light brown), 2-Bare soil (light green), and 3-Vegetation (dark green)).
The results are verified based on a collection of samples combined with the confusion matrix 47. The matrix covers information about real and projected classifications done by a specific system. Performance of such systems is commonly evaluated using the data in the matrix.
The confusion matrix (Table 1) demonstrates that the accuracy value of the HyGA method is 82%.