Research On Image Segmentation Of Side Scan Sonar Based On Combined Features

Side-scan sonar(SSS)measurement and imaging is a common means of detecting underwater targets which is an important task in ocean survey,and SSS image segmentation is the basis of target recognition.Affected by complex marine environment and imaging mechanism,noise pollution and imaging distortion in SSS images are prominent,which seriously restrict the accuracy and efficiency of image segmentation.So the paper studies the image segmentation method based on the characteristics of SSS images,and proposes a joint combination feature and Chan-Vese SSS image segmentation method,which improves the efficiency and accuracy of high-noise image segmentation.The main research work and contributions of the paper are as follows:(1)The composition and imaging mechanism of the SSS system are briefly described.The characteristics of the SSS image and its influence on the segmentation are analyzed,and the potential features favorable for image segmentation are given.(2)Introduced the Non-subsampled Contourlet Transform(NSCT),multifractal,Chan-Vese model and other segmentation theories,studied the target features of the SSS image,and proposed the combined features ?? and ?/? that can be used to enhance the bright region and the shadow edge respectively.(3)Aiming at the shortcomings of the existing segmentation algorithm for the low accuracy of SSS image segmentation with severe noise pollution,a high-noise SSS image segmentation method based on NSCT,combined features and multifractal analysis is proposed,which increases the segmentation accuracy of the traditional segmentation method(Canny operator,Otsu2,maximum entropy,MRF,FCM)from 0.61,0.76,0.78,0.88,0.84 to 0.90,but it is more time-consuming.(4)In order to further improve the segmentation accuracy and efficiency of high-noise SSS images,a joint NSCT,combined feature,edge-enhanced image segmentation threshold automatic calculation and Chan-Vese high-noise SSS image segmentation method are proposed.In the test,the segmentation accuracy reached 0.94,and the segmentation time was 1/8 of the method in(3).The paper studies and solves the problem of target image segmentation of SSS targets with high noise pollution,which lays a foundation for subsequent application research such as target recognition based on side-scan sonar images.