Study On Particle Swarm Optimization And Its Application In Image Retrieval

Particle Swarm Optimization (PSO) algorithm is put forward according to the simulation of the bird flock in their food-searching. Because of not much parameter adjusting, simple operation and easy application, nowadays PSO algorithm is paid more and more attention in the optimizing application. But PSO Algorithm has backwards of easy falling into local optimization and low convergence precision. For these reasons, some improvements of the algorithm were proposed. Though gained some achievements and progress, these improved algorithms still have disadvantages of high complexity computation and low convergence speed etc. So proposing efficient improvement of PSO is still a hot topic among the researchers now.Particle swarm optimization (PSO) algorithm has the disadvantage that once it gets into the local optimization it is very hard to jump out from the local optimization. For that reason, a novel improved particle swarm optimization algorithm is proposed in this paper. The algorithm can use statistical laws of particle fitness to classify the particles, and take different evolution model for different kinds of particles. And for the particle evolved in full model, learning factor is adjusted dynamic, which can enhance the evolution efficiency and precision greatly. By the experiments and analysis, the optimization variation rule which evolved with the learning factor is achieved, and the function expressions of learning factor are given in this paper. The simulation results showed that, compared with other PSO algorithms proposed before, it is improved virtually on both optimization precision and optimization efficiency by using the improved PSO algorithm to optimize 4 typical benchmarks.Whereas the parallel and memory quality of PSO algorithm the CBIR based on interactive PSO algorithm is proposed in this paper. In this method, Firstly, each image in the collection is segmented into a constant number of sub-images, and the content in each sub-image is computed to make up the feature vector of the image. Like this, "the image database" was transformed into "the feature space", where every particle represents a feature vector in the feature space. We can Regard object image as the solution to question, and the process of retrieval image may be regarded as the process of searching optimal solution using PSO in the feature space . In the retrieving, this method evaluate fitness of particle (image) using the man-machine interaction. This method can not only solve the difficulty construct of fitness function, but also ensure the objectivity of fitness evaluation.Because of the combination parallel and memory with interactive, this method can automatically search optimal solution, at the same time, its every step will be adjusted by manual participation. As a result, it assures that the image obtained can consistent with our attention. Finally, through the experiment simulation, the availability of the improve method was further confirmed...