Study On Relative Navigation And Experiment Of Spacecrafts Based On Natural Features

Relative navigation is a key technology of space rendezvous and docking andthe in-orbit spacecraft approaching service operation. According to thecharacteristics of relative navigation tasks and the spacecraft’s structure, thisdissertation proposes a relative navigation technology based on monocular visionwith points and lines and binocular vision with non-coplanar double circles. Thensemi-physical and mathematic simulations validate the feasibility of thesealgorithms and schemes. The main achievements obtained in this dissertation are asfollows:Firstly, according to structural characteristics of the spacecraft and the relativenavigation of cooperative target spacecrafts using feature points consisting ofspecial marks, this dissertation presents a scheme with lines or a combination oflines and points when the special marks in cooperative spacecrafts are in failure orthere is no marks in the no-cooperative spacecrafts. Using the unified descriptionmethod of the collinear error of points and the coplanar error of lines, we solveparameters of position and attitude with a rapid and global convergence methodcalled orthogonal iteration and the algorithm is validated by simulation.Secondly, the problem of the3D information of circle recovered from the2Dimage and the relative navigation are studied since most spacecrafts are designedwith circles and a circle can provided many properties such as its ellipse perspectiveprojection in any arbitrary orientation and its high image-location accuracy. In orderto solve the duality problem in pose estimation of a single circle in monocularvision and the problem of location estimation without the radius of the circle, abinocular scheme is presented in this dissertation to identify the unique pose andlocation result. After the3-D location estimation, we use non-coplanar doublecircles to estimation the relative position and attitude parameters and mathematicsimulation validates the feasibility of the scheme and the high-accuracy of the circlerecovery. Finally, this dissertation uses Hough transform to extract lines and circles weconcerned on the target structure since the simple thresholding cannot dual with itthen proposes a scheme to recover the problem of edge roughness; vision camera iscalibrated with the approach of Zhang according to the experimental needs. Thenorthogonal iteration algorithm is validated by this semi-physical simulation. Thesimulation results compared with mathematical simulation indicate that theserelative algorithms and scheme are efficient.