Robust Monocular Vision Navigation For Spacecraft On-orbit Servicing

With the development of space technology,many countries are actively researching spacecraft on-orbit servicing technology,and the relative navigation technology is the prerequisite for rendezvous and docking of on-orbit servicing.The traditional monocular vision navigation is often applied in approaching cooperative target spacecraft at final approaching stage,but it has certain limitations due to the on-orbit servicing spacecraft are mostly damaged or broken.As a result,this dissertation concentrateds on robust vision navigation method for non-cooperative spacecraft on-orbit servicing,and the detailed research contents are as follows :Firstly,for the system model uncertainty and kinematic state jump of non-cooperative spacecraft,two kinds of relative vision navigation algor ithms based on strong tracking filter are designed.The traditional strong tracking filter needs the nonsingular measure matrix and relies on prior i knowledge,which has been improved.Simulation results indicate that algor ithms are both robust to system model uncertainty and kinematic state jump.Secondly,the solution of projection point extraction fault is studied.Case 1: The position of the fault projection point is reconstructed based on inverse projection principle,which can ensure the solution of the relative pose.Case 2: The traditional depth iterative algor ithm is improved and the depth iteration of the fault projection point is completed based on the feature point position relation,besides the iterative speed is optimized.Case 3: Based on Unscented Kalman F ilter,the quadratic exponential smoothing method is used to predict the new value of fault measurement value and the proportion matrix is introduced to adjust covariance matrix and filter gain,which can reduce the correction of state estimation of fault measurement value.Thirdly,on the basis of the above researches and considering large feature extraction error when observing the target spacecraft stureture,the relative vis ion navigation strategy of non-cooperative spacecraft is studied.First the “selfie stick” camera installation structure and camera on-orbit calibration scheme are designed,then to select different observation target according to the field of view,next the optimization model is built based on inverse projection principle,at last the ant colony search algorithm and the absolute position algorithm are respectively applied to solve the depth and the relative pose quickly.Finally,in order to verify the effectiveness of the algorithm proposed in this dissertation,a hardware-in-the-loop simulation system is designed based on monocular camera,which has the advantages of simple calibration,low cost,good performance,etc.And relative pose measurement software based on Lab VIEW is developed to validate the robust fast depth iterative algor ithm proposed in this dissertation.The demonstration experiment of the relative pose measurement based on three-axis Air Bearing Table is completed.