Intelligent Optimization Algorithms And Their Applications

Intelligent optimization algorithm, which is a global, parallel optimization of high performance, robustness, versatility, no problem specific information, etc., has been widely used in computer science, optimization of scheduling, transportation problem, combinatorial optimization, optimization design and other fields. Domestic and foreign scholars are attracted wide attention to it. More and more people pay attention to typical intelligent optimization algorithms which are based on particle swarm optimization algorithm and clonal selection algorithm. Research focus on algorithm is especially to improve these algorithms and their application in many fields. Performance requirements are also increasingly high. How to design a robust, good, fast, high precision and stable performance of the intelligent optimization algorithm is still the focus of the study question.In this paper, some research is done in the intelligent optimization algorithms and its applications in the registration area. The main intelligent optimization algorithms in this paper are QPSO and improved clonal selection algorithm, combined with feature point extraction based on Hausdorff distance to regist the different sensors images. A large number of experiments show that the algorithm performance is effectiveness. Details are as follows:(1) Image registration is an important step in image fusion. In this paper, a new automatic image registration method is proposed. First, two images are transformed by gradient transformation, and feature points are extracted using fuzzy clustering segmentation methods. Then, Hausdorff distance is used as objective function, and these feature points are matched and the affine transformation between the two images is obtained through a matching technique based on the QPSO. Simulation results indicate that using feature points to search for the affine transformation make the computation as load of Hausdorff distance decreased significantly. At the same time, image registration through QPSO based on Hausdorff distance does not require to establish an explicit point correspondence between images, so it can tolerate errors introduced by the presence of outlier points (noises) as well as the absence of some missing points. This registration method can be applied to images with large misalignment and is a fast method of image registration.(2) Aiming at the slow convergence speed of the traditional immune clonal selection algorithm (ICA), adaptive immune clonal selection algorithm without memory(AICA)and adaptive immune clonal selection algorithm with memory(AICAM)are proposed respectively based on the combination of adaptive algorithm of clonal probability, immune probability , and group disaster algorithm. The two proposed algorithms have been applied to TSP problem. The application of group disaster algorithm can break away from the situation of slow-convergence and hold the diversity of antibody to some extent. The adaptive algorithm has strong global search ability and weak local search ability at early evolution. Global search ability is weakened and local search ability is enhanced with the process of evolution in order to find global optimal point. Simulation results indicate that compared with the traditional immune clonal selection algorithm(ICA), the proposed algorithms can enhance viration of population, overcome the problem of early-maturing, and can accelerate convergence speed in some extent.(3) Improved clonal selection algorithm has global search ability and can accelerate convergence speed in some extent. This paper propose the algorithm using feature points to search for the affine transformation and Hausdorff distance as objective function based on improved clonal selection algorithm to regist multi-sentor images. This registration method and is a fast method of image registration. If images are quite different and similarity of feature points, it can still obtain the registration parameters.