Research On Rough Registration Algorithms For Space Non-cooperative Target Point Clouds

The relative pose measurement of space non-cooperative target is the key technology to realize space autonomous on-orbit service in the future.As non-cooperative targets cannot interact with each other through effective information such as intersatellite link,optical measurement is the main method at present.The position and attitude measurement technology of linear array scanning lidar has the advantages of high measurement accuracy,strong anti-interference ability and little influence by light conditions,which has become an important research direction of leading units in the field of space technology at home and abroad.The relative pose of space non-cooperative target is solved by the point cloud registration method of satellite point cloud data acquired by lidar.Precise registration algorithm is one of the methods widely used in the field of point cloud registration at present.However,due to the sensitivity of precise registration to the initial position of point cloud,its one-way iteration mechanism is prone to fall into the local optimum when the initial position of the target point cloud to be registered is quite different,resulting in the registration failure.In view of the disadvantage of the precision registration algorithm,it is very important to introduce coarse registration to provide a good initial position before the precision registration.Rough registration can avoid algorithm failure,reduce the number of iterations of precise registration,and improve the precision and efficiency of precise registration.Therefore,this paper studies the coarse registration algorithm of spatial non-cooperative target point cloud.The main research work of this paper is as follows:Firstly,aiming at the problem of slow speed and low accuracy of the existing rough registration algorithm in solving the feature information of normal vector estimation,combined with the background task requirements of spatial non cooperative target,a K-Neighborhood threshold selection method based on three-dimensional point cloud data set of spatial non cooperative objects is proposed.On this basis,for the ambiguity of traditional normal vector calculation,an improved algorithm of normal vector estimation based on the point cloud data set centroid point as the field of view point is proposed.Then,combined with the structural characteristics of space non-cooperative targets themselves,a local feature coarse registration algorithm combining SAC-IA(Sampling Consistency Initial Registration)algorithm and FPFH(Fast Point Feature Histograms)feature descriptors is established.Aiming at the selection of threshold parameters required by sac-IA and FPFH local feature algorithms,the concept of bounding box is introduced and an adaptive matching threshold algorithm based on K neighborhood threshold and bounding box is proposed.Finally,the method of obtaining point cloud data of spatial non-cooperative target is proposed,and the feasibility of the proposed algorithm is verified from the perspective of registration error of rough registration algorithm.The experimental results show that the rough registration task for non-cooperative objects in orbit in this paper can provide 2 ° error for precise registration.