Research On Underwater Acoustic Image Segmentation Based On The Genetic Algorithm

With the rapid development of computer science and technology, sonar technology playsmore and more important role. Among all kinds of information material which human getfrom underwater environment, underwater sonar image material account for a large part. Tohuman, how to effectively obtain valid information from underwater sonar image matrial is asignificant subject. Study sonar automatic identification technology is one of possible ways toaccomplish this topic. In this process, a very important processing step is underwater acousticimage segmentation, so the study on underwater object segmentation technology of the sonarimag has been carried on.Firstly, this paper discusses the background and signification of image segmentationtechnology, the current research situation at home and abroad, the basic theoretical knowledgeof image processing and three classifications of image segmentation algorithm. The threeclassifications include region-based segmentation, edge detection-based segmentation and thesegmentation algorithm based on a particular theory.Secondly, the paper introduces the concept of image segmentation briefly, classicalthreshold segmentation algorithm and summed up their advantages and disadvantages, whichfocus on Otsu algorithm. Combined with the characteristics of the underwater acoustic image,this paper researched two improved methods, one is to obtain the target of images, and theother one is to obtain the shadow of images.Thirdly, the paper studies the Genetic Algorihtm which belongs to intelligentoptimization algorithms. For the shortcomings of the standard Genetic Algorihtm applied insonar image segmentation is the phenomenon of premature and unstable, this paperresearched the improved Genetic Algorihtm that based on dual populations. The simulationresults show that the improved Genetic Algorihtm overcomed the premature and unstableshortcomings.Finally, the improved Genetic Algorihtm and improved Otsu used in underwater acousticimage segmentation are conbinated. Simulation results show that this method with goodsegmentation efficiency, good retaining of the object outline and reduceing the timecomplexity and very stably.