The Target Image Recognition Based On A Hybrid Optimization Algorithm

With the fast development of computer technology, the target recognition is a very meaningful and difficult task in the study and application of computer vision and artificial intelligence.At present there have been a lot of optimization methods to solute the problem of target recognition, but most of these methods have some limitation. As a novel simulated evolutionary algorithm, Ant Colony Algorithm (ACO) has many merits as parallel essence, positive feedback and coordination, so it can solve the problem of target recognition well. The thesis study the target recognition based on the hybrid ant colony algorithm.In order to get over the disadvantages of the slow convergence speed and stagnation behavior, combined with Simulate annealing, the thesis proposed converse ant colony algorithm based on simulated annealing. Inducting converse ants into the ant colony and the number of converse ants is adjusted by simulated annealing. The ability of searching for global optimal solution has improved. The algorithm can solve the combinatorial optimization problem -Traveling Salesman Problem. The simulated results show that the algorithm can solve combinatorial optimization problem well. The ability of optimization and the convergence speed have improved a lot.To solve the function optimization problem, a hybrid optimization algorithm based on ant colony algorithms and chaos optimization algorithm is proposed. The algorithm utilizes the pheromone positive feedback effect to guide chaos search. The working ant colony searches different searching space according to the concentration of the pheromone to reduce the searching blindly. The simulation results prove that the efficiency of the algorithm was higher than that of common ones, the regularity of the algorithm has been improved.Lastly, this thesis has mainly discussed the application of target recognition based on mixed ant algorithm. Preprocessing the collected image, and extract the feature point. Convert the feather points matching into a combinatorial optimization problem, and solve it with the improved ant colony algorithm in the paper. So the task of target recognition was accomplished. The results proved that the new method has high rate of recognition, and can be applied in automatic recognition system.