Improved Aquila Optimizer And Its Application In Multi-Threshold Image Segmentation

Image segmentation is the classical problem in image processing,and for many image segmentation methods,threshold segmentation is the most commonly used and effective.In recent years,various intelligent optimization algorithms have been developed and applied to the image segmentation.In this paper,the Aquila Optimizer is improved and applied to the maximum Kapur entropy multi-threshold segmentation.The specific work is as follows:Analyze the defects of the Aquila Optimizer,and propose an improved Aquila Optimizer that integrates Gaussian walk and somersault strategy.In order to solve the problem of insufficient local search capacity and poor algorithm accuracy,Gaussian walk is proposed instead of Levy flight.Gaussian walk adaptively generates step size according to the number of iterations can solve the problem caused by the unstable step size generated by Levy flight,so that the global search and local search capability can be enhanced.To deal with the late diversity reduction and tend to fall into local optimum problems,the somersault strategy is introduced into the Aquila Optimizer,all individuals perform somersaults.If the individual’s fitness value becomes better after somersaulting,the original individual is replaced.The effectiveness of the strategies of the improved algorithm is demonstrated using experiments on the solution of 12 public test functions.Apply the improved Aquila Optimizer to multi-threshold segmentation,and Kapur entropy is chosen as the image segmentation criterion,i.e.,it is used as the fitness function to find the optimal set of thresholds to make the image segmentation better and time loss less.The simulation experimental results of segmentation with different threshold numbers for different types of images show that the improved AO proposed in this paper ranks first in comparison with the standard AO and another improved AO,and has better values for each image segmentation evaluation index.The time loss is similar to AO and much lower than another improved AO,the threshold set that makes the segmentation effect better and more efficient can be found.In comparison with other intelligent optimization algorithms,the threshold set that makes the optimal image segmentation metrics can also be found under a specific number of thresholds.It shows that the improved algorithm proposed has better applicability in image multi-threshold segmentation.