Research And Implementation Of Honey Bee Mating Optimization Algorithm For Image Segmentation

With the increasing demand for information,computer applications can be seen everywhere,and the acquisition of information includes not only text,sound,video,but also images.In order to make the information more intuitive,people must preprocess the images before they get the information contained in the images.As an indispensable step in image processing,image segmentation has been a hot and difficult research topic.Therefore,image segmentation is regarded as the main research object in this paper,aiming at improving the segmentation accuracy.At present,there are many methods of image segmentation,such as clustering based,region based,edge detection and so on.This paper mainly uses the threshold-based image segmentation technology.Whether the threshold can be accurately searched is the key of this method,so this paper improves the accuracy of the threshold by improving the traditional honey bee mating algorithm.Honey bee mating algorithm is one of the swarm intelligence algorithms,which mainly mimic the breeding behavior of bees in nature.This algorithm searches threshold by the swarming behavior of bees.It has good global search ability and fast convergence ability.At present,there are many kinds of improvements to traditional honey bee mating algorithm,but there are still some defects in these improved methods.Such as not considering the influence of propagation probability and mutation probability on the algorithm,the non-additive information of the image itself is ignored in the process of segmentation,and so on,these defects greatly affect the image segmentation accuracy.Basing on the existing defects at this stage,a Tsallis Entropy Threshold Image Segmentation Algorithm Based on Improved Honey Bee Mating Algorithm is proposed.In this paper,an adaptive definition method for propagation probability and mutation probability is proposed,which effectively avoids the influence of traditional methods,and then improves the global convergence ability of the algorithm.Furthermore,the algorithm is combined with the elite retention strategy,which shortens the optimization time of the algorithm.In addition,Tsallis entropy is introduced as an adaptive function to avoid the non-additive information being ignored and the segmentation accuracy is improved by the nonextensibility of Tsallis entropy.Finally,the proposed algorithm is simulated and compared with other improved algorithms.The comparison includes segmentation effect,algorithm running time,peak signal-to-noise ratio and so on.Through the combination of theory and experiment,the superiority of the proposed algorithm is proved more comprehensively.