Research On Pose Estimation Methods Of Non-cooperative Space Objects Towards Space Grapple Applications

Recently, with the constant progress in space technology, the classical on-orbit servicing technology, which is widely used in on-orbit spacecrafts maintenance and life extension, as well as in large debris removal, has gained enormous progress. The target for on-orbit servicing are usually non-cooperative spacecrafts which do not provide any cooperative information. As an critical step in on-orbit servicing, capture of an uncooperative space object has gained much attention. For this purpose, vision-based real time pose estimation and tracking of an uncooperative spacecraft has become a critical research area and has shown its enormous appliance in space activities.In this paper, Pin hole model of the CCD camera is introduced and several mathematical issues concerning geometrical transformation is studied in the very beginning. In the following, thorough solution of real-time pose estimation methods for non-cooperative targets is introduced. A model-based approach combining initial pose estimation algorithm and pose tracking algorithm is introduced in this paper.Secondly, we proposed the through scheme for pose initialization of Non-cooperative targets. In the pose estimation process, we firstly use POSIT pose estimation algorithm to calculate the pose of the target in the reference image. Then concerning the current image, our approach is based on the matching of SIFT features between the current image and the reference image, after the 2D-3D matching is achieved, we uses RANSAC algorithm combining virtual visual servo algorithm to calculate the pose parameters of the target in the current image, the calculated pose is used as the initialization pose in the model-based tracking process.Concerning the problem of pose tracking for uncooperative targets, our approach is based on the classical approach of edge-based Gaussian-Newton nonlinear minimization algorithm, our approach improves the classical one in 2 aspects. Firstly, we integrate Harris key point features with edge features in the minimization process, which accounts for the hybrid tracking algorithm based on edge features as well as key point features. Secondly, we uses M-estimator to improve the robustness of the minimization process by expelling outliers in the non-linear pose minimization process. Accurate pose parameters calculation is achieved with the proposed model-based tracking approach.Finally, we uses monocular CCD camera, 6-DOF industrial robot, infrared pose tracking system, as well as a real non-cooperative model to build the experimental platform for our approach. Several experiments were done using this experimental platform. Experimental results have proved the accuracy and robustness of our approach, and the pose calculation speed is proved to be fairly fast.