Automatic SAR Image Segmentation Based On Improved Particle Swarm Optimization Algorithm

Synthetic Aperture Radar(SAR)imaging technology has received more attention in both civil and military fields.Unlike traditional optical imaging,SAR images have speckle noise due to their own coherent characteristics.The image processing has brought great difficulties.SAR image segmentation is to mark the target area contained in the original SAR image and divide it into different regions to facilitate the recognition of the object to be studied.However,it is difficult to find a kind of speckle noise and texture feature Segmentation algorithm suitable for all images.The problem of image segmentation can be solved by cluster analysis technique.According to the number of objective functions of image to be segmented,it can be divided into single target clustering SAR image segmentation and multi-objective clustering SAR image segmentation.In this paper,the SAR image segmentation algorithm based on Particle Swarm Optimization(PSO)and the multi-objective particle swarm optimization based on ensemble learning are proposed.The concrete work can be summarized as follows:1.An automatic SAR image segmentation method for single target particle swarm optimization based on graph partition is proposed.The algorithm is based on the particle swarm optimization algorithm,which is based on the traditional single objective clustering algorithm,which is easy to be segmented when the image is segmented.In the case of preprocessing,there are inherent speckle noise in the SAR image,which is smoothed by nonlocal mean filtering,and the edge information of the image can be preserved.In the initial segmentation algorithm,the watershed segmentation algorithm is used to divide the image into non-overlapping regions.The image segmentation method is used to realize the automatic segmentation of SAR images.In the experimental part,the proposed algorithm is compared with the existing particle swarm algorithm,From the segmentation of four texture images and four real SAR images,it can be seen that the algorithm can obtain the segmentation effect of texture image and SAR image more in accord with the real situation.2.A multiobjective particle swarm optimization algorithm based on ensemble learning is proposed for SAR image segmentation.The method of ensemble learning in machine learning is applied to evolutionary multiobjective algorithm.Firstly,based on the particle swarm optimization algorithm proposed in the previous work,the clustering intermediate result and the segmentation result are integrated.In the existing weak classifier,based on the Adaptive Boosting algorithm in the idea of machine learning,which can reduce the variance by averaging multiple decision trees.Although there is a problem of increasing the deviation,the occurrence of overfitting is reduced and the preparation of classification results is improved.It is possible to achieve better segmentation results for artificial SAR images and real SAR satellite images with a large number of categories.