Research On Key Technologies Of Visual Pose Measurement For Non-cooperative Rotating Spacecraft

In recent decades,human space activities are increasingly frequently,the number of spacecrafts launched increase year by year.A lot of space debris was created at the same time,including debris from collisions of launch vehicles,failed spacecraft and space vehicles.With continuous development of technology,the space-based observation for debrits is attracting great attention.Compared to ground-based observation,the space-based realizes a higher efficiency and better real-time,and with no subject to meteorological conditions and geographical location restrictions.Visual camera has the characteristics of high accuracy,small size and light weight.It can provide a large number of high-resolution images containing spatial target information.It has become the main method of relative pose measurement for spacecraft in orbit service,rendezvous and docking and other space tasks.It can be divided into model-based and non-model-based measurement methods according to whether or not to use the model of the object.The former generally requires the model or prior knowledge of the measured object,and this kind of method is quite common.The non-model-based measurement method does not use any prior information of the measured target,so it is more challenging.Space debris or a failed vehicle often spins around its main inertial axis,although it is in a stable spin state,measuring its motion state is several times more difficult than measuring a stationary object.It is mainly reflected in the rapid extraction and stable tracking of target features,the closed-loop detection and pose optimization of revisit region.In view of these difficulties,this paper mainly studies from the following aspects,the specific content is as follows:Aiming at the reflection optical characteristics of non-cooperative rotating target and the imaging characteristics of visual camera,this paper analyzes the visible light energy of the sun outside the atmosphere,and conducts modeling analysis on the solar incidence,camera observation angle,surface characteristics of target coating materials,etc.We obtain the target radiant luminance characteristics based on specular reflection and diffuse reflection.The imaging signal-to-noise ratio of the camera is calculated by combining the detector and optical system parameters.The key technical parameters of camera imaging measurement are designed,which is of great significance to the camera engineering.For the initialization problem of non-cooperative rotation target pose measurement,we separate two kinds of aspects.In the case of large parallax,if the feature points are located on the same plane,the pose can be estimated by calculating the homography matrix.If the feature points are not on the sam e plane,then the position and pose are estimated by calculating the fundamental matrix.The calculated coordinates of the space points are expressed in Euclidean space.For the case of small parallax,inverse depth parameterization method is adopted to es timate the pose,and the calculated coordinates of spatial points are expressed as inverse depth space.For the nonconvergence problem of the pose measurement of noncooperative rotating target,a measurement method based on position and pose graph optimization is proposed in this paper.We transfer the matching of front and back frames in traditional measurement methods into the matching of current frame and feature database and the optimization of feature points by constructing an optimized target feature database.In order to optimize the position and pose results and ensure the computational efficiency,this paper divides the feature database into local and global types,optimizing the posture in graphical form.The optimized pose information is obtained in Lie Group by computing the residual errors of estimation and measurement in Lie Algebra based on the pose graph theory and exponential and logarithmic operations.Aiming at the problem of posture error accumulation of non-cooperative rotating target measurement,This paper presents a closed-loop detection algorithm based on hierarchical visual bag of word similarity detection and consistency check.The similarity of two images is represented by the 1 norm form of visual bag of word description vector.The 5-point RANSAC method is used to verify the geometric consistency of the current key frame and closed-loop candidate key frame.After the closed-loop similarity transformation matrix is obtained,the joint optimization of the pose transformation matrix of all target space points and all key frames is carried out in the lie algebra space.Finally,we make ues of the pose matrix of the currrent key frame to rectify the parameters of the current frame.