Application of improved artificial immune algorithm in parameter optimization for image registration

The paper proposes the tabu search artificial immune algorithm (TS-aiNet) based on the aiNet model and the tabu search algorithm. It introduces a tabu list that tabooes the cells whose affinity do not increase any more in the network. In some phrase the tabooed excellent cells are released according to the aspiration criteria. For saving mature memory cells a memory table is added to the network. Moreover, it redefines the expression of the Gauss mutation for diversity seeking, and uses the Markov chain to prove the global convergence. The performance optimization analysis of the proposed algorithm was carried out with typical system experiments, and it was compared with the CLONALG and the opt aiNet algorithm. Finally the TS-aiNet algorithm was applied to the image registration for visible and infrared images, and the matching accuracy of 0.5 pixels was achieved. Both the theoretical analysis and the simulation results show that the presented approach has preferable global convergence ability in multi modal search space, and it can avoid prematurity effectively. It has better performance in improving accuracy and speed of image registration, and is an efficient global optimization algorithm.