Research On Measurement Of Relative Pose Between Two Non-cooperative Spacecrafts In Short Range Based On Computer Vision

With the development of space technology, proximity operations have become more and moreimportant in many space-related applications, in which the measurements of relative position andorientation are the key points, such as Rendezvous and Docking Satellite Servicing Satellite Capture,etc.For relative pose measurement between two non-cooperative spacecrafts in short range, a newmethod based on combination of binocular stereo vision and three-dimensional laser scanner isproposed. When the distance between the two satellites reduces to100m, the target satellite isdetected by the stereoscopic system. The shape of the target satellite is extracted and tracked by thestereo vision for relative orientation determination of the service satellite. Simultaneously the rangeinformation is obtained with the help of three-dimensional laser scanner. When the distance decreasesto20m, the relative orientation is determined by the analysis of the geometric structure, and highprecision range distance is measured by the combination of binocular stereo vision andthree-dimensional laser scanner.Because of the influence of space environment and variation in illumination, almost all the targetimages degrade in a measure, such as affine transformation and image blurring. A combination ofaffine moment invariants and blurring invariants is proposed for target recognition of degradedimages according to the mathematical principle of invariants. Tests performed on simulated imagesand actual imageries have shown the excellent robustness and accuracy of the developed invariants.The average recognition rate has reached92%, which outclasses that of affine moment invariants.During the image feature extraction, Harris algorithm is researched, and in order to eliminate theimpact of rotation and illumination variation in wide baseline stereo matching, SIFT local descriptoris also researched. For the sake of feature line extraction, Hough transforms algorithm is studied andthen a novel line detection algorithm is proposed which can describe the boundary of the object.Finally, according to the principle of stereo depth calculation, an algorithm for measurement ofrelative pose based on stereo vision is proposed. Furthermore, a novel model between stereo depthand stereo disparity is proposed through analyzing the error of relative pose measurement. Theexperimental results indicate that the develop model can meet the requirement.