Research On Detection And Tracking Technology Of Underwater Pipeline In Sonar Image

For decades,in order to make full use of marine resources,human beings have laid a large number of energy transmission pipelines and information and communication cables on the seabed.Due to submarine construction,natural corrosion and other reasons,submarine pipelines are liable to be damaged or even broken,resulting in significant economic and environmental losses.Therefore,it is necessary for underwater vehicles to track and inspect submarine pipelines regularly.The main content of this paper is to collect seabed information by side scan sonar,detect the position and direction of pipelines by image processing algorithm,and guide underwater vehicle to track pipelines.The details are as follows:(1)Studying the side scan sonar image pipeline detection system and image preprocessing.Firstly,the structure of pipeline detection and tracking system is introduced;secondly,the imaging principle of side-scan sonar and the factors affecting sonar image quality are analyzed;secondly,the model of underwater pipeline system is introduced;finally,the methods of removing sonar image noise by mean filter,median filter and Gauss filter are studied.The experimental results show that the effect of Gauss filtering on sonar image filtering is the best.(2)A pipeline detection method based on Cell Average–Constant False Alarm Rate 2D is studied.Firstly,the Cell Average – Constant False Alarm Rate 2D algorithm under the condition of Gaussian noise is introduced.When calculating the average value of the reference unit,it takes a lot of time to extract the gray value of the pixel repeatedly.This paper studies the acceleration of calculation by using integral matrix.Secondly,morphological method is used to smooth the edge of pipeline,and centrifugal index of the connected area of the binary graph is used.Finally,the pipeline edge is detected by Hough transform,and the location and direction of the pipeline are fitted.The experimental results show that this method can effectively detect pipeline targets in sonar images.(3)A two-dimensional average CFAR algorithm for rectangular and cross detection structures is proposed.Firstly,according to the shape characteristics of pipelines,the rectangular detection structure is studied.Compared with the square detection structure,the rectangular detection structure has a lower threshold when detecting pipeline targets,thus effectively extracting pipelines in the image.On the basis of rectangular detection structure,cross detection structure solves the problem of pipeline pixel missing caused by outliers in sonar image by using the ratio of gray mean value of reference units in left and right directions to that in upper and lower directions.The experimental results show that the proposed method can detect more complete pipeline targets.(4)The pipeline tracking method of underwater robot based on PID is studied.Firstly,according to the relative position of the underwater vehicle and pipeline,the strategy of fast approaching pipeline and slow tracking pipeline is formulated;then,Kalman filter is used to solve the problem of pipeline location.The experimental results show that the proposed method can effectively control the underwater vehicle tracking pipeline.