Research On Side-scan Sonar Image Processing In Detecting Underwater Targets

Side-scan sonar,as an effective tool for underwater detection,has been widely used in the tasks such as underwater target detection,undersea pipeline's tracking and marine investigation etc.Considering the characteristics like low contrast,heavy noise pollution etc which exist in side-scan sonar images,scholars have proposed a variety of side-scan sonar image preprocessing methods and segmentation methods,in order to meet the requirements of both the segmentation accuracy and speed.But it is usually difficult to achieve the best at the same time.In this paper,we studied a variety of side-scan sonar image preprocessing algorithms and fuzzy clustering algorithms,in order to improve the accuracy of image segmentation,while as far as possible to accelerate the segmentation.In the end,two different kinds of neural network classifier were used in the seabed sediments classification of side-scan sonar image.The main research contents in this paper can be divided into the following three parts:(1)Research on side-scan sonar image denoising and enhancement methods.By means of research on seven denoising algorithms including traditional algorithms,we proposed a kind of BEMD side-scan sonar image denoising algorithm combined with Wiener filter,and then compared and analyzed the results after the denoising experiment.In the aspect of side-scan sonar image enhancement,we studied three commonly used enhancement algorithms and analyzed the pros and cons of each enhancement algorithm by the enhancement experiment.(2)Research on side scan sonar image segmentation method based on fuzzy clustering.Firstly,fuzzy clustering analysis,K-means clustering,fuzzy C-means(FCM)clustering and FCM clustering algorithms bias spatial information were introduced.And then we studied the characteristics of these types of algorithms in depth.Then,In order to improve the anti-noise property and segmentation speed,three improved FCM algorithms were proposed from two aspects including combination with the image texture information and spatial information.Finally,by means of the segmentation experiment,we analyzed the pros and cons of each segmentation algorithm to verify the applicability of the proposed algorithms.(3)Research on seabed classification method.Firstly,four kinds of seabed image were processed by using the methods proposed in(1)and(2).Then,we extracted the seabed image texture features by means of GLCM,and then analyzed its principal components.Finally,seabed classification experiment was proceeded by using BP neural network and self-organizing competitive neural networks respectively.