Approach To Fast Relative Position And Attitude Determination Based On Mono-vision For Non-cooperative Spacecrafts

Non-cooperative visual measurement is an important research topic in the field of space-orbit service and offense and defense services. Especially, the methods used in pose calculating and image processing of monocular vision are key technologies. This thesis takes the background of a sub-project of 863 :"Research on fast capturing pose of××××". Based on the systematical conclusion of current research statue in this field, a deep research develops monocular vision algorithm theory and implementation in the demand of rapid and robust performance of non-cooperation spacecrafts.This study focused on two main aspects: one is to find and verify a quick and feasible visual image corner detection operator; the other hand, monocular vision pose determination algorithm would be proposed.First of all, this paper conducts a systematical description and research on the corner detection technology. For practical applications, key points are set on the exploration of SUSAN operator and Harris operator from the point of fast, accurate and robust corner detection. This kind of method which is directly on the foundation of image template could complete the quick and accurate corner detection.Second, for the measurement of monocular vision pose, three algorithms are given. Similar iterative algorithm is much in common with the two-stage method, both of which gain the final results by solving the absolute position value and the depth of field and in two iteration steps. But they also have a common drawback, that is, the algorithms are depending on the initial value. If the initial value selected is not appropriate, it will lead to divergent results of the final operation. Based on RANSAC, monocular vision relative pose navigation parameters determination algorithm can effectively eliminate matching errors of value and model. The final estimate is calculated by POSIT algorithm, which enhances the fastness of the algorithm and does not have dependence on the initial operation with more precise operation.Finally, monocular vision pose estimation algorithm based on randomized RANSAC is proposed. Through simulation, this algorithm could eliminate gross errors and matching errors effectively and robustness and fastness of the algorithm have been proved strongly. Single image obtained by monocular vision in real time when applied by this algorithm can be used to gain sufficiently accurate non-cooperative target spacecraft relative pose. When affected by gross errors, the relative position accuracy is smaller than 2% and relative attitude determination accuracy is smaller than 0.01°. The requirement of hardware and real-time calculation is not harsh for this algorithm which is an advantage to satisfy the robustness and fastness expectation in the application of monocular vision navigation. Therefore, this algorithm can be used as an effective method for non-cooperative target satellite in quick pose determination...