Research On Preprocessing And Matching Of Side-scan Sonar Image

With the continuous decrease of land natural resources,the speed of exploration and development of marine resources is accelerated.Because of the bad marine environment and the limited range of visible light in the sea water,the commonly used means of ocean exploration is mainly sonar.We usually transform sonar data into sonar image,display it by waterfall display or match the sonar image into a large seabed map for analysis.There are geometric distortion,gray distortion and noise interference in side scan sonar image due to sonar data acquisition model,bad marine environment,attenuation of sonar signal energy,attitude and speed change of underwater vehicle or towed fish,and there are great differences in sonar image obtained by sonar scanning the same area of seabed in different directions,all of which are for side scan sonar Image matching and stitching brings great difficulties.In order to solve the above problems,this paper analyzes the actual sonar data and proposes a complete set of solutions,including side scan sonar image filtering,slant correction,speed correction,gray-scale correction and other preprocessing,and studies the corresponding matching algorithm for the side scan sonar image under different conditions.The image matching method of side scan sonar studied in this paper can not only be used for the image splicing of side scan sonar,but also for the positioning of submarine.The main research work is as follows:There is a lot of noise in the complex marine environment,which leads to a lot of noise interference in the original side scan sonar image converted from the data collected by sonar.Because the side scan sonar is imaged by the echo model at oblique range and the speed of the sonar carrier is unstable,the original side scan sonar image will have geometric distortion.In this paper,the side scan sonar image will be filtered and processed,and the sea scan sonar image will be processed In the correction of slant distance of side scan sonar image,the velocity correction method of side scan sonar image is proposed by analyzing the change of carrier velocity,and the geometric correction of side scan sonar image is completed.In the process of sonar signal transmission,there will be energy attenuation,acoustic scattering and other situations,which lead to the gray distortion of the original side scan sonar image.In this paper,the current gray-scale correction method is studied,and a side scan sonar image gray-scale correction method based on retinex theory is proposed.The experimental results show that the image correction effect of the method proposed in this paper is better and faster.After preprocessing,the side scan sonar image can display the seafloor scene normally,and then can be matched and stitched.In order to match and splice the side scan sonar image,this paper studies the commonly used matching methods such as sift,surf,orb and the newly proposed GMS method,and applies them to the side scan sonar image matching.It is found that these mainstream matching methods are only suitable for the side scan sonar image with obvious partial features,small noise interference and small deformation of seafloor scene in the same area However,most of the side scan sonar images contain a lot of noise,the features are not obvious,and the seafloor terrain features in the same area have a large deformation.For this kind of side scan sonar image,this paper proposes a side scan sonar image matching method which combines the semantic segmentation network and template matching method,that is,first,the semantic segmentation network in depth learning is used for side scan The sonar image is segmented,and the template matching is used to match the side scan sonar segmented image to obtain the relative position information of the two side scan sonar images.Finally,the Poisson fusion method is used to fuse and splice the two side scan sonar images,and the experiment verifies the validity and applicability of this method.