Research On Seabed Terrain Registration Technology Based On Three-dimensional Point Cloud Data

Topography is one of the most important marine environmental parameters.The traditional 3D seabed terrain stitching needs to know the relative position of the carrier,which is supported by high-precision sensors.However,it has high cost and strong dependence on instruments and equipment.With the development of multi beam sonar,the number of bathymetric points is increasing rapidly and the density is increasing.The registration between adjacent lines can be carried out through the correlation of terrain data to realize the stitching of 3D seabed terrain.Supported by the National Natural Science Foundation of China,"key technologies for real-time monitoring of submarine pipeline leakage based on AUV platform",this paper studies the seabed terrain registration technology based on 3D point cloud data,and provides AUV with terrain matching aided navigation means for underwater missions.The main contents are as follows:(1)Terrain overlapping area extraction of adjacent survey lines.The amount of seabed terrain data measured by multi beam sonar is huge.When the terrain of adjacent lines is registered,the point cloud data in non-overlapping area will interfere with terrain registration,and the huge amount of data will increase the processing time and reduce the registration efficiency.Therefore,it is necessary to automatically extract the overlapping area of adjacent lines for subsequent processing through the nearest point search.(2)Detection and elimination of depth outliers of terrain point cloud.Due to the influence of multi beam sonar with the ups and downs of the carrier,noise interference,and errors of auxiliary equipment and other factors,the actual measured terrain point cloud data will produce depth outliers.This paper studies five common methods to detect and eliminate the depth outliers of the simulated terrain data.(3)Feature points extraction of terrain point cloud.By selecting the stable points,rich geometric information and representative points as feature points,this paper introduces several common feature points extraction methods of point cloud,and uses the method based on the distance threshold and the minimum distance to remove the wrong feature matching points,which provides a good position and pose basis for the fine registration of point cloud.(4)Point cloud registration of adjacent survey lines.The feature points extracted from the overlapping area of adjacent survey lines are registered,and the spatial conversion relationship between adjacent survey lines is solved to realize the accurate stitching of seabed terrain.Two classical point cloud registration methods are used to register the simulated terrain,and the application scope of each algorithm is compared and analyzed.(5)Experimental data processing of seabed terrain registration algorithm.This paper deals with the measured terrain data of Songhua Lake in Jilin Province.According to the different characteristics of the terrain in the overlapping area of adjacent survey lines,different registration methods are selected to obtain the relative offset angle of the adjacent survey lines,so as to correct the spatial position information of the corresponding survey lines,and finally obtain the accurate seabed map.