Application Of Meta-Heuristic Algorithm In Image Processing

With the continuous advancement of computer science,optimization methods are successfully applied to various fields of daily life.As a kind of optimization method,the meta-heuristic algorithm shows a good performance on solving nonlinear optimization problems with complex constraints.As an important medium for information transmission,the application of image processing technology is widely applied to industry,military industry,aerospace,scientific research and so on.However,traditional image processing methods often encounter problems such as large data processing and nonlinear processing functions.This paper focuses on the improvement of meta-heuristic algorithm and its application on image processing,main tasks as follows:(1)Firstly,a discrete artificial bee colony algorithm and genetic algorithm are proposed to solvethe discrete variable optimization problem.The algorithm combines the artificial bee colonyframework with the genetic algorithm framework,which not only retains the advantages ofthe strong global search ability of the artificial bee colony algorithm but also saves the finelocal search ability of the genetic algorithm.The algorithm employs the concept ofcomplement to design a specific update,crossover,and mutation strategies to complete theupdation of individuals.Finally,the improved algorithm is applied to the extraction ofhyperspectral endmembers.The experimental results demonstrate the effectiveness of the algorithm.(2)In view of the problem that the traditional artificial bee colony algorithm's weak local search ability and slow convergence speed,the search strategy of the employed bee and onlook bee are improved to speed up the algorithm's search ability and convergence speed.Furthermore, in order to avoid the algorithm falling into local optimum,the step size of learning to global optimal solution adopts an adaptive method,and the disturbance factor is added to the designed step size element to ensure the global search ability of the algorithm.Finally,the improved algorithm is applied to the image segmentation problem.The experimental results verify the effectiveness of the algorithm.(3)The artificial bee colony algorithm has strong global search ability,but it is easy to receive the constraint of dimensionality disaster,especially on the function of inseparable variables. Then its convergence speed is very slow on such functions.For conquering this shortcoming, this paper proposes an improved differential grouping method.Firstly,by analyzing the functional properties of separable and non-separable variables,we propose an improved differential grouping strategy into the artificial bee colony algorithm,and then introduces different update strategies to improve the performance of the artificial bee colony algorithm based on the grouping results.This paper totally proposes three update strategies to assist individuals to complete the update:(a)to balance the local search ability and global search ability of the artificial bee colony algorithm,a method which consider the historical solution of the population is put forward.(b)In order to reduce the number of function evaluations,the scout bee's new initialization strategy is propsed.(c)The multidimensional update method is used for the non-separable variable among the individuals who with the worst performance during each iteration.Finally,the improved algorithm verifies the effectiveness of the algorithm in the basic test function,CEC2010 test function,CEC2103 test function and CEC2014 test function value.