Research On Segmentation Algorithms Of Multi-objective Sonar Image

With China becoming more powerful,it is very urgent to protect the legitimate Marine rights and interests of our country.As one of the important techniques in the detection of underwater,sonar technology has an irreplaceable role in underwater navigation,underwater object tracking and underwater communications etc.While sonar images carry the majority of the information sonar equipment detected,the research on it has become one of the important branches in the field of digital image processing.Sonar image segmentation is an important foundation part of sonar image processing which is similar to the segmentation of optical images.Sonar image has the characteristics of low resolution,serious noise.Along with the wider application of sonar image,People are increasingly demanding requirements for sonar image segmentation.Under the background of related subject,multi-objective image segmentation algorithms are studied in this paper.Sonar image segmentation is mainly used in underwater target detection field,tracking underwater targets field and so on.Generally image segmentation methods mean separating the target from the background,and do not feature different targets for research division.Due to the difficulty of obtaining sonar images and sonar equipment range of applications is very small,and the noise of sonar image is serious,the goal was not obvious,blurred edges,uneven distribution of gray value targets and other characteristics,sonar image segmentation research is complex and difficult than the optical image segmentation research.With the promotion and upgrading of sonar equipment,sonar image resolution is also increasing,which makes it possible for the study of sonar image segmentation problem under more complex cases.As underwater activities of countries become more frequent,sonar image content people get more and more varied and complex,and therefore,Research is necessary for multi-objective sonar image processing methods,this paper mainly studies the division of multi-objective sonar image.According to the characteristics of sonar image,the paper first introduces the research status of image segmentation at home and abroad,and then researches the related sonar image denoising algorithms;and proposes an improved algorithm based on median filtering,through the pixel stretching highlights the target regions and shadow regions.Experiments show that the proposed algorithm can be filtered on the sonar image at the same time,enhance the target and background contrast,compared to traditional filtering methods,which greatly reduces the difficulty of sonar image segmentation goals.Then the paper introduces some commonly used algorithms and theories of multi-objective image segmentation,including the method of support vector machine,the method of geometric snake model,the level set method,theclustering method and the method based on region segmentation,and the advantages and disadvantages of each method are discussed.Then this paper focuses on the research of the multi-objective in sonar image segmentation method based on fuzzy clustering.In view of the shortcomings of fuzzy clustering,the paper reconstructs the sonar image by using the method of empirical decomposition.Experimental results show that this method can reduce the influence of noise in sonar image segmentation.At last,the paper studies the multi-objective sonar image segmentation approach based on hierarchical gray histogram.The multi-objective segmentation of sonar image is transformed into the single-objective segmentation of sonar image by histogram stratified,and the experiment of image segmentation by improved Otsu method and particle swarm optimization algorithm.Experimental results demonstrate the correctness and feasibility of this method.