Sonar Image Segmentation Algorithm Based On Markov Random Field

With the continuous development of science and technology,in order to explore new resources to meet people's increasing living requirements,people have never stopped exploring new fields.People are also exploring some areas where environmental conditions are more demanding.Therefore,in such a region where conditions are relatively severe,acoustic signal detection based on radar system will have a greater advantage.Sonar image is a signal imaged by acoustic technology based on radar system.Segmentation of sonar image is a very important step for acoustic technology based on radar system.Because the sonar image is seriously affected by the coherent speckle noise,the sonar image segmentation method based on Markov random field not only considers the single pixel gray value,but also considers the structural characteristics of the sonar image region.Therefore,we usually choose the sonar image segmentation method based on Markov random field to achieve better segmentation effect.However,in the image segmentation algorithm of Markov random field,the influence parameters of all pixels in the sonar image are usually set to a fixed empirical value.For all the pixels in the sonar image,the influence parameter is fixed.The segmentation effect of the region disturbed by a large number of coherent speckle noises will be seriously affected.In view of this limitation,the correlation based on Manhattan distance is introduced into the influence parameter of the potential function.At this point,each pixel in the image has its corresponding influence parameter value according to its neighborhood characteristics.Through the analysis of the simulation results,it can be found that the improved algorithm in this paper reduces the influence caused by coherent speckle noise,and also greatly improves the accuracy of sonar image segmentation.In the segmentation algorithm of Markov random field,there is a problem that the detail information cannot be retained well after the segmentation of the image region with many details.To solve this problem,the weight function is introduced into the energy model of Markov random field.Then change the proportion of energy component of image pixel information field and segmentation marker field in energy model to explore its influence on image segmentation effect.After the exploration of the simulation experiments of different image details,this paper introduces the weight function based on the concept of sonar image neighborhood variance into the energy model of Markov random field.Through the comparison and analysis of the simulation results,we find that the improved Markov random field energy function method in this paper can effectively retain the image information in the areas with many details.And after the improvement of the algorithm,the details of the edge part of the image are more complete and comprehensive.After the experiment,we can find that the improved algorithm can effectively improve the anti-noise characteristics in the segmentation process and effectively retain the details of the image.