A pose estimation method based on RGB-D system in the process of attaching asteroids

Based on RGB-D system, a pose estimation algorithm framework integrating point cloud normal vector information, point cloud information and two-dimensional gray information is designed to solve the relative pose between spacecraft and non-cooperative target interest surface (NCTIS) during attaching asteroids. First, the framework fuses the organized point cloud obtained by the TOF (time of flight) camera with the color image to obtain the organized RGB-D point cloud data. Then, the normal vector feature of the point cloud is extracted and fused with the gray image information to perform variable scale template matching to track the NCTIS. The matching algorithm considering the normal vector feature improves the robustness and accuracy of the matching results. Finally, the relative pose between NCTIS and the spacecraft is estimated in real-time by the MRPs kinematics model based on the normal vector of NCTIS and the Kalman filter. The effectiveness of the proposed algorithm framework is verified by ground experiments, and the accuracy and real-time performance of pose estimation meet the mission requirements.