Research On Visual Relative Pose Measurement Technology For Non-cooperative Spacecraft

In recent years,with the vigorous development of the aerospace industry,the Earth’s orbit has been occupied by increasingly uncontrollable space debris and disabled spacecraft,which seriously affects the normal work about space missions.The on-orbit service for the purpose of cleaning space has been increasingly concerned by astronauts.As a key point in the service,the visually relative pose measurement for non-cooperative targets has naturally become a hot issue in current research.For non-cooperative targets,there may exist some problems,such as uncontrolled target motion,less or completely unknown prior information,large changes in target pose and scale,disturbance about spatial illumination and so on.Undoubtedly,these problems present a significant challenge to visually relative pose measurement for non-cooperative targets.Work on the above key issues has been carried out in this paper as follows:(1)With respect to the problem of sequential pose measurement for target,based on the ICP algorithm,combined with the increase on the amount of sequence frame data,the method about adaptive down-sampling by means of uniform grid is designed to effectively suppress the dense points,which ensures the stable time when ICP algorithm is conducted.At the same time,in the case of the rapidly sequential frame changes,the greatly changes in pose and scale on target,the axisymmetric features about target model and so forth.The sequential pose measurement algorithm about fusion of two-dimensional depth and three-dimensional point cloud is proposed.Taking advantage of roll angle deviation calculated from the adjacent frame in two-dimensional depth map to assist the registration of the 3D point cloud,which improves the accuracy of the sequential pose measurement algorithm.(2)In case of the completely unknown information about target,an algorithm about pose measurement which uses the original collected data to construct the target coordinate system for completely non-cooperative target is proposed.Firstly,the common typical features about target need to be detected on the 2D depth map and mapped into the 3D space.Then,the coordinate system of target can be constructed by means of the center of the arc and the border of the line.Last but not at least,the pose relationship with the fixed camera coordinate system need to be calculated in order to achieve the pose measurement of a completely non-cooperative target.(3)In order to carry out digital simulation experiments,an algorithm about simulation for point cloud data based on laser imaging radar is designed.First,the 3D CAD model of the spacecraft need to be downloaded on the NASA official website,then,in terms of the fixed grid size,the 3D CAD model is divided into the finite elements which are jointed together into model point cloud.According to the operating principle of the laser imaging radar,the measurement point cloud data with the information about position and attitude can be simulated,which provides the convenient for diversification of simulation experiments.(4)In order to verify the algorithm proposed in this paper,the ground test of the target which operates with free rolling in space was designed on the platform equipped with robotic arms.The experimental results show that the proposed algorithm can perform real-time position estimation on non-cooperative targets reliably at the state of tumbling.