Tile Image Registration Based on Improved Adaptive Genetic Algorithm

An improved adaptive Genetic Algorithm was proposed, and the method was applied to the optimization process of tile image registration. This paper improved traditional Genetic Algorithm in three aspects. The probability of crossover and mutation was adjusted in a dynamic way according to the change of the fitness of individual during the evolutionary process and in different way when the evolution was in different phase. Random dislocation arithmetic crossover that non-allelic genes of the two father individual did crossover operation was introduced to prevent invalid crossover and to increase searching space and convergence speed so that global optimum was obtained even if some generations fell into premature convergence. Two-population immigrant strategy based on maximum entropy was applied to obtain diversification gene. The method was able to overcome the disadvantage of traditional adaptive Genetic Algorithm which tends to fall into a local optimum answer. The experiment showed the validity and feasibility of this algorithm.