Pose Measurement and Tracking of Non-cooperative Satellite Based on Stereo Vision

Abstract

To realize non-cooperative satellites captured by a space robot, a method for autonomous measurement and tracking of space non-cooperative satellites based on stereo vision is proposed. The pose measurement takes the general structure of satellites, i.e. satellite frame and satellite-rocket docking ring as recognition features, which facilitates edge determination by the ellipse center. Then, the 3D coordinates of corner points are obtained through binocular stereo vision by which the pose of the satellite can be subsequently computed. On this basis, kernelized correlation filter (KCF) is used to track the identified corner points. The measurement and tracking results are verified through the visual simulation system. The results show that the methods can realize relative position and attitude measurement for non-cooperative spacecraft with high accuracy and the recognition algorithm can retain effective when features are partially occluded.