CN-121809705-B - Factor graph-based space fragment short arc data association method and device

CN121809705BCN 121809705 BCN121809705 BCN 121809705BCN-121809705-B

Abstract

The invention belongs to the technical field of spaceflight and space situation awareness, and particularly relates to a factor graph-based space fragment short arc data association method and device, wherein the method comprises the steps of acquiring optical angle measurement short arc data from a measuring station and preprocessing; the method comprises the steps of obtaining a track state initial value by adopting an angle observation initial track determining algorithm, constructing a cost matrix, carrying out preliminary screening on the track state initial value corresponding to different short arc observation data, generating a limited number of multiple association hypotheses through disturbance sampling expansion, constructing a nonlinear factor graph under each association hypothesis, solving maximum posterior estimation, selecting an optimal association hypothesis according to residual errors and information criteria, outputting an association cluster, a track state and uncertainty thereof, and carrying out consistency check. According to the invention, under the observation scenes of the foundation and the space-base, the efficient and robust multi-arc segment association and preliminary orbit estimation can be carried out on the short arc data of the optical angle measurement, so that scene self-adaption, instantaneity and high accuracy are realized.

Inventors

LEI XIANGXU
ZHANG SHENGLONG
ZHAO XIANGLEI
LIU ENXU
FAN JUNFU
Lao Zhendi

Assignees

山东理工大学

Dates

Publication Date: 20260505
Application Date: 20260311

Claims (8)

1. The factor graph-based space debris short arc data association method is characterized by comprising the following steps of: S1, acquiring optical angle measurement short arc data from a measuring station, preprocessing including time unification, coordinate conversion and scene correction, and outputting standardized short arc observation data and observation noise covariance, wherein the measuring station is a foundation or a space-based observation platform; S2, aiming at each short arc observation data, adopting an angle observation initial orbit determination algorithm to obtain an initial orbit state value; S3, constructing a cost matrix based on a Markov distance, performing preliminary screening on initial values of track states corresponding to different short arc observation data by utilizing an optimal matching strategy based on a Hungary algorithm, and generating a limited number of multiple association hypotheses through disturbance sampling expansion, wherein each association hypothesis corresponds to one candidate cluster; s4, under each association assumption, constructing a nonlinear factor graph comprising a track state variable, an observation factor, a dynamic factor, an association factor and a priori factor; S5, solving maximum posterior estimation for a nonlinear factor graph by adopting a Gaussian-Newton method, a Levenberg-Marquardt algorithm or a iSAM algorithm, and selecting an optimal association hypothesis according to residual errors and an information criterion; S6, outputting the association cluster, the track state and the uncertainty thereof, and performing consistency check; the specific implementation mode of the step S3 is as follows: s3.1, mapping the initial value of the track state to an observation angle domain, wherein the initial value of the track state corresponding to different short arc observation data after mapping is the short arc to be associated, and executing the following operations on all the short arcs to be associated: S3.1.1, combining the short arcs to be associated in pairs to form all possible short arc pair candidate association assumptions, and obtaining each pair of candidate short arcs; S3.1.2, aiming at each pair of candidate short arcs, calculating the mahalanobis distance between the short arcs and a cluster center in an observation angle domain based on the initial value of the track state and the corresponding observation noise covariance matrix, wherein the cluster is a group of short arcs belonging to the same space target, and the cluster center is the statistical or logic center of the initial value of the track state of the group of short arcs in the observation angle domain; s3.1.3, organizing the mahalanobis distance of each pair of candidate short arcs according to the dimension of 'short arc number multiplied by short arc number' to obtain a cost matrix, wherein the matrix element value is the association cost of every two short arcs; S3.2, inputting the cost matrix into a Hungary algorithm, and screening out a preliminary optimal association hypothesis set; S3.3, aiming at screening results, combining noise distribution and geometric constraint conditions, and generating a limited number of multi-association hypotheses by using disturbance sampling, wherein each association hypothesis corresponds to a candidate cluster, and the candidate clusters are potential short arc sets belonging to the same space target; in S4, the nonlinear factor graph specifically includes: A track state variable comprising a track state variable for each candidate cluster; the observation factor, the top center projection model is adopted when the measuring station is a foundation observation platform, and the platform posture-camera geometric model is adopted when the measuring station is a space-based observation platform; The dynamic factor, adopt the dynamic model of two body orbit, the orbit state of the last moment propagates to the next observation moment, is used for constructing the consistency constraint that the state changes with time, the process noise includes J2 perturbation influence; The correlation factor applies constraint to geometric consistency and time sparsity, and the robustness is enhanced by adopting a Huber or Cauchy kernel function; a priori factors, relaxed constraints are applied to the track states to avoid unconstrained drift.
2. The method for correlating space debris short arc data based on factor graph according to claim 1, wherein in the step S1, the non-cooperative target observed by the station is a low orbit space target, the number of maneuvers of the low orbit space target is assumed to be no more than one in an observation time period, the observation arc is short arc, and the observation data is angle-only observation data.
3. The factor graph-based space debris short arc data association method according to claim 1, wherein in S1, the preprocessing process is as follows: S1.1, unifying time, namely unifying time stamps of all optical angle measurement short arc data to UTC (universal coordinated time); s1.2, coordinate conversion, unifying the observation coordinates of the measuring station to a geodetic inertial coordinate system, wherein the standard form is a J2000.0 coordinate system; s1.3, scene correction, wherein the foundation observation platform is a foundation scene, and the correction process comprises the following steps: Atmospheric refraction correction, namely compensating elevation angle observed quantity by using an atmospheric refraction model; The earth rotation and polar motion correction is carried out, the earth rotation angular velocity parameter and polar motion correction quantity are introduced, and the station measuring posture at the observation moment is dynamically adjusted; The method comprises the steps of compensating a station position and meteorological parameters, combining accurate geodetic coordinates of the station and real-time meteorological data during observation, correcting the influence of meteorological conditions on an atmospheric refractive index, and simultaneously compensating the interference of station position errors on coordinate conversion, wherein the accurate geodetic coordinates comprise longitude and latitude of the station and elevation, and the real-time meteorological data during observation comprise temperature, air pressure and humidity; The space-based observation platform is a space-based scene, and the correction process comprises the following steps: Platform attitude compensation, namely correcting satellite attitude at the observation moment by utilizing real-time attitude measurement data of a star sensor and a gyroscope in the space-based observation platform; correcting exposure delay and rolling shutter effect, correcting the time stamp and angle quantity of each piece of observation data through camera parameters calibrated in advance, and eliminating imaging time sequence errors; Correcting errors of a camera coordinate system, obtaining distortion parameters through camera calibration, and carrying out distortion correction on original angle observation data; Short-time geometric correction under J2 perturbation, introducing a J2 perturbation simplified model, and compensating and correcting the geometric position of a target orbit in a short arc observation period aiming at short-time orbit perturbation of a low earth orbit target due to the earth equator doming effect; s1.4, outputting standardized short arc observation data and observation noise covariance which unify time and a coordinate system after preprocessing.
4. The method for associating space debris short arc data based on factor graph according to claim 1, wherein in the step S2, for each short arc observation data, a Gauss three-point method or Gooding angle observation method is adopted to obtain an initial value of a track state under a geocentric inertial coordinate system when the measuring station is a ground observation platform, and an expansion Gooding algorithm or a relative track approximation based on Clohessy-Wiltshire equation is adopted to obtain the initial value of the track state when the measuring station is a ground observation platform; After the initial value of the orbit state is obtained, whether abnormal observation exists is identified by calculating an observation residual error, and when the abnormal observation exists, outliers are subjected to weight reduction or rejection by a robust estimation method.
5. The method for correlating space debris short arc data based on factor graph according to claim 1, wherein in S5, a gaussian-newton method, a Levenberg-Marquardt algorithm or a iSAM algorithm based on incremental smoothing mapping is adopted to perform nonlinear least square optimization on each factor in the nonlinear factor graph, solve maximum posterior estimation, and score and select multiple hypotheses with normalized residual error, posterior negative log likelihood and information criteria to obtain optimal correlation hypotheses, wherein the information criteria adopts red pool information amount criteria or bayesian information criteria.
6. The method for associating space debris short arc data based on factor graph according to claim 1, wherein in S6, the association cluster is a candidate cluster corresponding to an optimal association hypothesis, the track state is a maximum posterior estimate after solving the nonlinear factor graph, and the uncertainty is a covariance matrix representation of the track state.
7. The method for correlating space debris short arc data based on factor graph according to claim 6, wherein in S6, consistency check is performed by propagation re-projection and multi-platform cross-validation, if the correlation and estimation result are considered valid, otherwise, step S3-S5 is re-performed; Propagation reprojection verification is that based on the track state of the association cluster, the theoretical track state at the observation time is calculated through a track propagation model, then the observation angle domain mapping back calculation theoretical observation value is utilized, the theoretical observation value is compared with the actual observation value of the standardized short arc observation data, the normalized residual error is calculated, if the residual error is smaller than the threshold value corresponding to the observed noise covariance, the consistency of the track state of the association cluster in the observation angle domain is verified, and the verification is passed; The multi-platform is a foundation-foundation, a space-base or a foundation-space-base mixed observation platform, the multi-platform cross verification is that for a multi-platform observation scene, the space-time references of all the platforms are unified firstly, then the track states of different platforms for the same association cluster are extracted, the position and speed deviation of the platforms are calculated, and if the deviation is smaller than a set association threshold, the consistency of the track states of the association cluster among the multiple platforms is verified, and the verification is passed.
8. A factor graph-based space debris short arc data association device is characterized by comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the method of any one of claims 1-7 is realized by executing the program by the processor.

Description

Factor graph-based space fragment short arc data association method and device Technical Field The invention belongs to the technical field of spaceflight and space situation awareness, and particularly relates to a factor graph-based space fragment short arc data association method and device. Background With the rapid development of geospatial activity and constellation deployment, the number of LEO (low earth orbit) space fragments continues to increase, constituting a collision risk for on-orbit spacecraft. Optical goniometry is widely used for debris monitoring due to its passivity, broad coverage and low cost. However, due to factors such as observation window, target brightness, weather and platform geometry, most of the practical results are short angle arcs (e.g. right ascension RA, declination DEC or azimuth Az, elevation El) with very short time spans (typically less than 1 minute). Short arc information is insufficient, the noise ratio is high, and the geometric constraint is weak, so that the precision and the efficiency of the traditional association and IOD (initial orbit determination) method in a complex scene are limited. In the prior art, methods such as multi-hypothesis tracking (MHT), data Joint Probability (JPDA), kalman filtering, least squares, RANSAC, etc. are used for correlation and estimation. But the existing method has the problems of insufficient robustness and expandability in face of uncertainty under the condition of short arcs, association hypothesis combined explosion, scene differences (such as the atmospheric refraction and earth rotation of a foundation, the relative motion and posture influence of a space-based platform) and real-time requirements. In the prior art, as CN111457916B focuses on random finite set modeling, US7105791B1 optimizes from the illumination observation angle, but the problem of LEO short arc data association is not solved in a system under a unified factor graph framework, and especially a general modeling and efficient optimizing means for simultaneously considering the foundation/space-base and the mixed scene is lacking. Researchers have performed a lot of work and made several key advances regarding short arc correlation, angle observation only initial trajectory determination, TLE/SGP4 propagation, and factor graph optimization. Cai and the like provide an improved tracklet correlation method for short arc optical measurement, and the accuracy and the robustness of short arc correlation are remarkably improved by improving a loss function and an initial value searching strategy under the short arc condition. Vallado and the like systematically sort and revise SGP4/SDP4 realization specifications, demonstrate feasibility and limitation of SGP4 taking NORAD/TLE as input in engineering application and situation awareness comparison, and lay an engineering foundation for observation-cataloging matching based on cataloging comparison. Gooding the Lambert problem-based angle only initial orbit solver is widely adopted and expanded and improved in subsequent work to adapt to multiple observation points and short arc conditions, and it is verified that the initial orbit solution with engineering significance can be obtained under the angle only observation conditions. Dellaert and other systems illustrate and popularize the application of factor graphs and incremental solvers (such as iSAM/GTSAM) in large-scale sparse Bayesian estimation, and indicate that the factor graph framework has good modeling expression capacity and calculation efficiency in the joint data association and orbit/state estimation problem, so that a mature methodology and a tool chain are provided for introducing the factor graphs into short-arc data association. In conclusion, the prior literature has systematic research results in the aspects of short arc observation association, angle initial orbit determination, TLE/SGP4 orbit propagation modeling, factor graph sparse optimization and the like. However, under the condition that the observation information is limited (short arc and angle are only measured), how to uniformly incorporate the multi-platform (foundation and space-based) heterogeneous observation data into a factor graph frame after time synchronization and coordinate conversion, so that the arc segment association and orbit joint estimation with high consistency and high confidence degree is realized, and the challenges in engineering realization and algorithm coordination still exist. Disclosure of Invention According to the defects in the prior art, the invention aims to provide a method and a device for correlating space fragment short arc data based on factor graphs, which can efficiently and robustly correlate and preliminarily estimate multiple arc segments of optical angle measurement short arc data under the condition of foundation and space-based observation, and realize scene self-adaption, real-time performance and high accuracy. In order to achieve