Research On Electromagnetism-like Mechanism Algorithm And Its Application In Image Retrieval

Recent years, with the growing science and technology innovation, practical engineering optimization problems have become more and more complex. The study of optimization problems has become an important focus. But traditional deterministic global optimization algorithms have great difficulty in solving high-dimensional, non-differentiable complex issues. Because of its intelligence and efficiency, heuristic global optimization algorithm has become a convenient tool for solving complex optimization problems. Electromagnetism- like mechanism(EM) algorithm is an intelligent global optimization algorithm which introduces the attraction and repulsion mechanism of charged particles in electromagnetic field. EM algorithm has the advantages of using less resource, simple optimization mechanism and strong searching ability. However, when solving high-dimensional complex problems, efficiency and stability of EM algorithm still have to be improved.EM algorithm is studied deeply in this paper. And for the problems of uneven initial population, fixed local search range and poor adaptability in the movement of particles, an improved EM algorithm is proposed. The improved algorithm uses a novel method combined uniform design and good point theory to generate initial population of arbitrary size that distribute uniformly and widely. The efficiency of local search is improved with adding adaptive search parameter to local search formula. The convergence is speeded up by designing an efficient and adaptive move operator. And introduce thoughts of genetic and variance into moving formulas for updating the particl es so that the particles are more likely to move into the other feasible regions. Experimental results show that the algorithm proposed by this paper can converge to global optimums more effectively and accurately.With the rapid development of multimedia and internet technologies, images are more and more widely used in each field, the management and retrieval of image information resources become increasingly important. The way the improved algorithm applied to image retrieval is researched in this paper. In content-based image retrieval, it is difficult to present an image only with color, texture or shape features. As a consequence, it is necessary to construct vectors consisting of multiple features. When features are extracted in this paper, within the geometric segmentation of the image, the invariant moment vectors-average color value of each image block, second color moment and texture eigenvalue are extracted to construct feature vector. And the improved EM algorithm is used to retrieve images. Experimental results show that the retrieval method which is proposed in this paper can retrieve the target image quickly and accurately.In subsequent research, the strategy of EM algorithm improvement will be researched deeply, and the goal is improving its accuracy. O n this basis, the improved EM algorithm will be applied to more practical engineering problems.