A Robust Estimation Using Quaternions For Relative Pose Of Space Object

Relative pose estimation of space objects (including cooperative object and non-cooperative object) is a key technology of space rendezvous and docking. Because of the importance for national security, this technology has also become a leading research in the filed of international spatial technology. Based on the deep analysis of the close quarters space relative motion of objects, this dissertation,which extends the classical theory and methods of photogrammetry, uses quaternion theory as a mathematical tools for sptial rotation transformation, and also studies the theory and methods of the quaternions estimation and robust estimation for relative pose of space object. The main achievements and innovations are summarized as follows:1. Theoretical system of quaternion and its application in the spatial rotation transformation are discussed systemically, and the direct solution of spatial similarity transformation based on the unit quaternion is improved.2. Considering the randomicity of the goal spacecraft orbit, the relative motion equation, which is suit for common ellipse motion, between the chaser and the target based on the T-H equation is established. The analytic expression of relative motion equation is also given at the same time.3. Combining two restrictive conditions of relative distance, orientation and analytic solution of T-H equation, the flight orbits of tracing spacecraft are designed based on the movement mode of the chaser to the target along the orbit.4.The principle and method of space resection using orientation matrix is proposed in this dissertation. It successfully avoids frequently transformation between Euler angles and orientation matrix in relative pose estimation.5. An optical measurement algorithm of seIf-calibration in orbit for relative pose estimation is presented. CCD camera interior parameters can be self-calibrated in this algorithm when relative pose parameters are gained.6. A new model of P3P direct solution based on the theory of pyramid is improved. In this improved model, matrix expression of unit quaternion is used to replace the tranditional rotation matrix, and the algorithm efficiency is improved.7. A new estimation algorithm of space resection based on quaternion is proposed. This algorithm uses quaternion to describle the rotation matrix, linearizes collinear equation which is describled by quaternion, and solves the linear equations iteratively by the least square principle.8. A globally convergent algorithm of relative pose estimation is designed and improved. In this algorithm, the space resection problem of 3D space to 2D space is translated into absolute orientation problem of 3D space to 3D space iteratively.9. Considering that there are some coordinate outliers in observed value, the theory of M-estimation and RANSAC-estimation are used into relative pose estimation of space object.10. In order to enhance the performance of RANSAC algorithm, an improved RANSAC- estimation method (R-RANSAC) is designed. Proactive testing process is introduced in this algorithm before relative pose parameters are gained, so that, the algorithm efficiency is enhanced.