Research On Key Technologies For Pose Estimation Of Non-cooperative Target In Space Based On Point Cloud

Non-cooperative targets including failed spacecraft,rocket bodies at the end of rockets,abandoned fuel tanks,and post-collision debris,occupy a large number of orbital resources and has become the biggest threat to other spacecraft in orbit.High-precision scale measurement,accurate attitude estimation,and acquisition of motion parameters are crucial for capturing non-cooperative target at close range.Especially under the unfavorable factors of lack of cooperation signs,unknown model and attitude information,higher requirements are put forward for the measurement system and processing algorithm.This paper conducts research on the key technologies of spatial non-cooperative target pose measurement.and designs a multi-sensor fusion measurement system.Based on the theory of conformal geometric algebra,a new point cloud contour extraction algorithm and cross-source point cloud fusion algorithm are proposed.These algorithm provides a new technical idea for the on-orbit measurement of space non-cooperative targets.The main research work and innovative results and of this paper are described as follows:Taking space debris in space as research objects,this paper designs a point cloud-based noncooperative target close-range measurement system,provides a data acquisition platform for multi-sensor fusion measurement,and produces a point cloud dataset of satellite models.Conformal geometric algebra,and the conversion relationship between Euclidean space and conformal space are studied in depth.Through the spherical inversion principle of conformal mapping,the points in Euclidean space are mapped to five-dimensional conformal space and straight lines are converted into circles,validating that the point transformation in three dimensional space can be realized in conformal space by other geometric objects.During the research,it was found that the Grassmann inner product structure in the five-dimensional space can provide a unified description of the positional relationship between the geometric objects in the point cloud processing,and the attitude description of the rigid body spiral motion in the generalized homogeneous coordinate system transform linearization.Therefore,the introduction of conformal geometric algebra can reduce the complexity of point cloud processing related algorithms.Aiming at the characteristics of space debris without a priori model and irregular geometric shape,a method to extract the outline of space debris with a small volume at close range is proposed.The objective is to obtain the structural size of space debris and to capture space debris in an active removal technology.A strip contour point extraction algorithm based on conformal geometric algebra is proposed.Boundary sphere is constructed to guide the search direction of the contour point.The outer product symbol of the geometric object in the conformal space is used to determine the positional relationship between the point and the straight line and the candidate is obtained by calculating the probability distribution of the points in the minimum connected domain contour points.Subsequently,multi-view registration is performed on the point cloud contour sequence and the redundant points are removed using the judgment properties of bump points in the conformal space to obtain the final contour model.Experiments show that the proposed algorithm for contour point extraction is superior to other similar algorithms in three aspects: accuracy,speed and robustness,and it retains the details of short contours.An approach for pose and motion parameter estimation of non-cooperative targets based on cross-source cloud fusion algorithm is proposed,which solves the problem that the limitation of self-imaging mechanism,a single sensor can not obtain higher precision position attitude and model information in space non-cooperative target measurement..This method utilizes the different geometric meanings of the inner and outer areas in conformal geometric algebra to construct rotation-invariant feature descriptors.The best transformation matrix between two point clouds is obtained by using the translation and rotation description in conformal geometric algebra.For the problem of scattered distribution of point cloud after fusion,a point cloud plane clustering algorithm is studied.Finally,using the representation of the torque in the helical motion,combined with the C-W equation,the attitude estimation of 6degrees of freedom and the parameter calculation of velocity and angular velocity are performed for the non-cooperative spacecraft.The proposed algorithm is verified by numerical simulation and semi-physical experiments,and compared with other advanced algorithms.The results show that the problem of large cloud density difference and low overlap rate at different source points is solved,and the pose estimation results meet the accuracy requirements.In addition,owing to the real-time requirements of spatial non-cooperative target pose measurement,the method of FPGA hardware design is used for realizing the three key modules as the farthest point sampling with long delay,the construction of the rotation invariant,and the corresponding point search.The feature point detection accuracy in the point cloud registration algorithm influences the matching of corresponding points and the solution of the rotation transformation relationship.Therefore,through the analysis of the input point cloud coordinate quantization bit width,and the truncation bit width setting of the main parameters in the algorithm,accuracy of fixed point quantization and registration error are consistent with software simulation.Finally,the Xilinx Vivado validation platform was used to perform logic simulation and timing synthesis of hardware design.Under a clock frequency of 100 MHz,the total delay of key modules reaches 101 μs,and the total delay of the algorithm is 0.76 s.The algorithm velocity has been greatly improved.In summary,this paper takes data acquisition,algorithm design,experimental verification,and FPGA hardware design as the research route.Taking conformal geometric algebra as the theoretical basis,tasks such as measurement system design,point cloud contour extraction,cross-source point cloud fusion,contour model reconstruction,pose estimation,and FPGA hardware verification of key modules have been completed.The new algorithm proposed in this paper has important application value for the measurement of non-cooperative targets.