Point Cloud Data Registration And Pose Optimization Of Spatial Non Cooperative Target

The relative position and attitude pose measurement of space non cooperative target is the core of space on orbit service technology,and its position and attitude pose information can provide necessary navigation information for on orbit service satellite to approaching operation.Due to the coupling environment of multi physical fields in space,the traditional vision measurement methods are greatly limited.Lidar has become the mainstream measurement load in the field of space measurement with its strong anti-illumination and electromagnetic interference ability,wide area measurement,high frame rate,high directivity and high coherence.Therefore,the data matching and pose optimization of lidar visible point cloud are important means to solve the relative pose of space non cooperative target.Due to the complexity and non-cooperative characteristics of the target motion state,the traditional registration algorithm is not ideal and easy to fall into the local optimal solution,resulting in the failure of pose solution.Therefore,on the basis of the traditional registration algorithm,the front-end point cloud clustering segmentation processing is introduced,the traditional registration algorithm is optimized by clustering point cloud,and the back-end is supplemented by optimization link to optimize the relative pose obtained from the registration solution,so as to complete the research of point cloud data matching technology.The main work of this paper is as follows.Firstly,on the premise that the clustering parameters are suitable for all frames of the spatial non cooperative target visual point cloud data set,aiming at the clustering pollution problem of general features and salient features,this paper proposes a clustering segmentation algorithm of RGB image optimized region growth.The algorithm uses the mapping relationship between the image color information and the depth information of the point cloud to reduce the dimension of the three-dimensional point cloud,uses the boundary extraction algorithm to separate the salient features from the general feature background,then uses the boundary inversion and Boolean erasure to jointly optimize the clustering results,and finally uses the inversion criteria to restore and update the data,so as to achieve the goal of thinning clustering and extracting the salient feature point cloud.It provides effective information for identifying the load type carried by the target.Then,using the advantages of small-scale clustering point cloud and significant features,it establishes a positive feedback connection with the iterative nearest point registration algorithm,reduces the interference of general features on the registration process,and uses the reverse registration method to ensure that the corresponding point relationship between the clustering point cloud and all the visible point cloud can be established stably,which improves the efficiency and accuracy of the registration algorithm,and provides the reference for the on orbit satellite Quasi algorithm,real-time detection of non-cooperative targets to provide theoretical and technical support.Finally,according to the registration results,the real-time positioning,map building technology and graph theory optimization are combined to optimize the position and attitude of the non-cooperative target.The relative scanning path of the probe star is optimized,and the optimization results are reacted to the registration results.Experiments show that the total angle of nutation angle is more closer from 334.31909 ° to 352.67061 ° before optimization.It is close to the true value of 360 °.