Research On Improved Cuckoo Search Algorithm And Its Application

Cuckoo search(CS)is a meta-heuristic optimization method inspired by the breeding behavior of the cuckoos.Because of its simple structure,fast operation speed,and strong global search ability,cuckoo search has stood out from various bionic optimization algorithms since it was proposed,and has quickly become a hot topic for scholars at home and abroad.Based on the previous research work,this paper proposes two new variants to improve the convergence speed and the optimization accuracy of CS,and applied them to multi-threshold image segmentation.The main contributions of the paper can be summarized as follows.In this paper,a hybrid algorithm based on cuckoo search and differential evolution(CSDE)is proposed to solve the premature convergence problem of CS.During the process of global search,the improved algorithm first judges whether the individuals are clustered or not,and then applies differential evolution(DE)operation to the individuals in stagnation to increase the diversity of the population.Besides,it combines the based random walk strategy and the differential evolution strategy to improve local search capability.Simulation experiments,carried on 28 benchmark functions from CEC 2013,demonstrate that CSDE algorithm has greater competitiveness than the standard CS and other similar meta-heuristic algorithms.For the lack of information exchange mechanism in population and the useful information of those updated individuals are not used in time,an improved CS with dual-subpopulation and information-sharing strategy(DSIS)is proposed in the paper.In DSIS strategy,the population is divided into two subpopulations which are assigned different update tasks,and make the latter subpopulation can effectively use the information in the convergence direction applied by the previous updated subpopulation,to accelerate the convergence speed.Extensive experiments suggest that the DSIS strategy can help both CS and its variants to achieve better optimization performance.Finally,this paper applies CSDE algorithm and DSIS?CS algorithm to multithreshold image segmentation.Based on the Otsu method,the objective function of the algorithm optimization is determined,and the practicability of the two segmentation methods is verified through multiple test images and conditions at multiple different thresholds.