The Improvement Of Grey Wolf Optimization Algorithm And Its Application In The Image Segmentation

Grey wolf optimizer(GWO)is a novel meta-heuristic intelligent optimization algorithm.This algorithm mimics the leadership hierarchy and hunting mechanism of grey wolves in nature.Because of the simple structure,less number of parameters and fast convergence speed,the GWO algorithm has been widely applied in the practical engineering.However,as it has been proposed for only a short time,the research on the theoretical basis and application is not perfect.It also has many shortcomings,such as low precision and premature convergence in dealing with complex optimization problems.In this paper,two improved algorithms are proposed based on the theory and evolutionary pattern of GWO and applied to the multi-threshold image segmentation problem.The main contents are summarized as follows:(1)We describe the source,theory and procedure of GWO in detailed,discuss its pros and cons,summarize the various improved algorithms conclude the fields of application.(2)As social hierarchy influence the grouping hunting behavior of grey wolves,an improved grey wolf optimization algorithm is proposed to strengthen the social hierarchy of grey wolves.For this algorithm,the grey wolf has two kinds of hunting patterns,ie.directive hunting-mode and independent exploration hunting-mode.This two hunting patterns can reflect the leading role of the higher level grey wolf on the lower level grey wolf and exert the autonomy and activity of individual on the basis of fully exploring the position information of swarm.They also can improve the diversity of the population and avoid trapping in the local extreme.The simulation results demonstrate that the algorithm has stronger global exploration ability and higher accuracy.(3)On the basis of the analysis of both the advantages and disadvantages of the grey wolf optimization algorithm and the differential evolution algorithm,a hybrid algorithm is proposed to take full advantage of both GWO and DE algorithm.This hybrid optimization algorithm obtains both local and global search ability and can be used to solve the complex high-dimensional function optimization problem.The experimental results show that the hybrid algorithm has better convergence speed and optimization performance,and it is more suitable for solving various functions Optimization.(4)A new multi-threshold image segmentation is proposed to improve the accuracy of the threshold selection and the speed of segmentation when apply the above hybrid algorithm to the maximum entropy-based multi-threshold segmentation.The experimental results proved that the algorithm can find the optimal thresholds of the image segmentation effectively and efficiently.