Research On Pipeline Detection And Location Algorithm Based On Underwater Three-dimensional Point Cloud

Funded by a Major National Science and Technology Project,this article uses the AUV for automatic inspection and tracking of subsea pipelines as a background,and uses a multi-beam sonar as a means of three-dimensional reconstruction of underwater scenes,and studies an underwater pipeline detection and positioning technology based on point cloud data.The research contents of this paper include the following five aspects:(1)In this paper,multi-beam sonar is used for underwater scanning to obtain underwater images,and the sonar image is gray-scale transformed by dynamic brightness adjustment to improve the contrast between target and background.Each underwater image represents the cross-section information of the pipeline.Through the edge detection of the image,the edge contour of the pipeline section can be obtained.Subsequently,each image is stitched in order,and the target pixels are converted into the form of three-dimensional coordinate points.underwater three-dimensional point cloud data is obtained,and the underwater three-dimensional scene reconstruction is completed.(2)Due to the large amount of initial point cloud data,the point cloud is reduced using random downsampling;According to the distribution characteristics of underwater target point clouds and noisy point clouds,two common point cloud denoising methods are introduced: point cloud Density denoising,Statistical filtering denoising.Through algorithm simulation,the denoising ability and algorithm stability of the two algorithms are analyzed and compared.Finally,the statistical filtering is selected to complete the denoising of threedimensional point cloud.(3)The segmentation algorithm based on the Difference of Normal(Do N)and a three-dimensional cylinder detection method based on Random Sample Consensus algorithm(RANSAC)are used to extract the point cloud of the pipeline.By introducing the Do N segmentation algorithm,it is possible to effectively remove the flat area in the underwater scene while retaining the point cloud data of the pipeline,so as to ensure the success of subsequent RANSAC extraction.Subsequently,a cylindrical model is established for the point cloud of the pipeline,and the RANSAC algorithm is used to complete extraction of pipeline point clouds.(4)By performing point cloud registration on the point cloud of the pipeline collected at different time periods,the relative position offset between different pipelines can be obtained.In this paper,the Principal Component Analysis is used for rough pipeline registration,and then uses fast Iterative Closest Point algorithm improved by K-D Tree to complete accurate registration of pipeline point clouds,so as to obtain the spatial position relationship between pipelines.(5)Finally,this paper uses multi-beam sonar to conduct experiments in the silencing pool,and gives the experimental equipment and layout environment.The experimental data collected are processed according to the algorithm flow given in this paper,and the experimental results are analyzed to verify the feasibility of the proposed scheme.