Research On Coarse Registration Algorithm For Spatial Instability Target Point Clouds Of Multi-Geometric Feature

In recent years,with the rapid development of the aerospace industry in various countries,space frequency orbit resources are in short supply,and space on-orbit services have become an urgent need.Space instability targets cannot effectively interact and obtain pose information,which makes their in-orbit service a difficult problem.Therefore,the pose measurement of space instability targets has become a key technology to break through in space in-orbit services.The data acquisition of space instability targets includes visible light and lidar,because of its advantages such as strong anti-interference ability and high accuracy of lidar,lidar is currently used extensively for data acquisition.Registration of points collected by lidar is the key to 3D reconstruction.Point cloud registration is divided into fine registration and coarse registration.The fine registration error is small and the accuracy is high.Because of the high requirements for the initial position,when the pose difference is large,it is easy to fall into the local optimum,resulting in registration failure.To solve this problem,use the coarse registration of multiple geometric features and then transform to provide a good initial pose,and finally the fine registration can effectively improve the accuracy and efficiency.In this paper,the research on the coarse registration of multi-geometric features of the spatial instability target point cloud is as follows:First,it discusses the preprocessing of the coarse registration of spatial instability targets.Start the analysis and processing from the required experimental point cloud data,introduce the source of the point cloud data,judge the experimental point cloud data as a scattered point cloud by the relationship between the points,and classify the experimental point cloud data by noise,which is convenient subsequent processing retains multiple geometric features;At the same time,it analyzes the processing effects of three point cloud filters on different noises,explains the rigid body motion of the experimental data,explains a variety of rigid body solution methods,and selects the optimal solution method.Secondly,it discusses that under the existing spatial instability target data,there are too many similar features with different parts and uniform distribution,which leads to the timeconsuming and poor accuracy in the coarse registration process,an improved through filtering SAC-IA coarse registration algorithm is proposed.First,carry out coordinate transformation by analyzing the principal component direction of the point cloud data;Then the straight-through filtering removes the single plane feature of the solar panel outside the limited range,and retains the rich part of multiple geometric features such as curved surfaces;Finally,the feature calculation registration is performed to improve the registration efficiency and accuracy.Then it discusses the large number of spatial instability targets and the uneven distribution of plane features.A mathematical registration method based on index filtering for point cloud segmentation normal distribution is proposed.Point cloud segmentation is performed through index filtering,planar features are removed,and parts with multiple geometric features such as circles,surfaces,and polygons are retained.Finally,mathematical normal distribution algorithms are used for registration to accurately and quickly calculate the pose transformation.Finally,a method of combining the two registration methods in this paper is proposed,using VS2015,QT5.9 and PCL1.8.1 to compile spatial instability target registration software to meet the needs of different experimenters.