A method for constructing a space target behavior feature library

Abstract

The application discloses a kind of space target behavior characteristic library construction methods, comprising the following steps: determining space target type and space target behavior characteristic type;For each behavior characteristic type, respectively determine characteristic data information;According to the value of each kind of space target each characteristic data information, construct initial characteristic database;The behavior characteristics of each kind of space target in initial characteristic database are modeled, and behavior characteristic model library is constructed;The value of the characteristic data information of the space target to be evaluated is obtained, and each type of space target behavior model in behavior characteristic model library is matched, and the target type of the space target to be evaluated is determined according to the matching result;The target type and its characteristic data information are added to initial characteristic database, and initial characteristic database is updated to form space target behavior characteristic library.The application realizes the automatic classification analysis of space target action behavior characteristics.

Description

TECHNICAL FIELD

[0001] The application relates to a space target behavior feature library construction method and belongs to the technical field of space target physical feature research. BACKGROUND

[0002] Space targets are various, and their features are different, according to different features of the space targets, the space targets can be classified, and the space target types are determined, such as remote sensing space targets, investigation space targets, communication space targets and combat space targets. The space target action behavior features corresponding to different space target types are obviously different, by constructing the space target behavior feature library, after the space target acquires part of the action behavior features of the space target, the space target can be matched and classified according to the physical characteristics of the space target. The traditional space target database generally includes a physical database and a logical database, and is limited to the physical feature level, and there is no public space target behavior feature library establishment related material. Therefore, it is necessary to establish a set of space target behavior feature library construction and data update management method which can comprehensively summarize the physical characteristics of different types of space targets. SUMMARY

[0003] The technical problem solved by the application is to overcome the shortcomings of the prior art, provide a space target behavior feature library construction method, integrate and summarize the common characteristics of the same type of space target, construct a general common type feature set, and finally realize intelligent classification and perfect the database according to the target action behavior features.

[0004] The purpose of the application is achieved by the following technical scheme:

[0005] The application discloses a space target behavior feature library construction method, which comprises the following steps:

[0006] Determine the space target type;

[0007] Determine the space target behavior feature type;

[0008] Determine the feature data information for each behavior feature type;

[0009] According to the value of each type of space target and each type of feature data information, an initial feature database is constructed;

[0010] Modeling the behavior features of various space targets in the initial feature database, a behavior feature model library is constructed;

[0011] Obtaining the value of the feature data information of the space target to be evaluated, matching the behavior model of each type of space target in the behavior feature model library, and determining the target type of the space target to be evaluated according to the matching result;

[0012] The target type of the space target to be evaluated and characteristic data information thereof are added to the initial characteristic database, and the initial characteristic database is updated to form a space target behavior characteristic library.

[0013] In the space target behavior characteristic library construction method, the space target type is remote sensing, communication, navigation, reconnaissance or attack and defense.

[0014] In the space target behavior characteristic library construction method, the behavior characteristic type is orbit anomaly, relative motion, rotation motion or load action.

[0015] In the space target behavior characteristic library construction method, the characteristic data information is determined for each behavior characteristic type, and the specific method is as follows:

[0016] When the behavior characteristic type is orbit anomaly, the characteristic data information is a time sequence including an anomaly flag and an anomaly speed increment: Oaq={{Flagyd(t1), Vex(t1), Vey(t1), Vez(t1)}, {Flagyd(t2), Vex(t2), Vey(t2), Vez(t2)}, …}, wherein Flagyd is the anomaly flag, and Vex, Vey and Vez are three-axis anomaly speed components;

[0017] When the behavior characteristic type is relative motion, the characteristic data information is a time sequence of relative motion information: Maq={{x(t1), y(t1), z(t1), dx(t1), dy(t1), dz(t1)}, {x(t2), y(t2), z(t2), dx(t2), dy(t2), dz(t2)}, …}, wherein x, y and z are three-axis space coordinates of the space target relative to the monitored space target body;

[0018] When the behavior characteristic type is rotation motion, the characteristic data information is a sequence of attitude change information: Raq={{q1(t1), q2(t1), q3(t1), q4(t4), dq1(t1), dq2(t1), dq3(t1), dq4(t1)}, {q1(t2), q2(t2), q3(t2), q4(t2), dq1(t2), dq2(t2), dq3(t2), dq4(t2)}, …}, wherein q1-q4 are attitude quaternions;

[0019] When the behavior characteristic type is load motion, the characteristic data information is a time sequence of load action information: Laq={{Loadt(t1), Loada(t2)}, {Loadt(t2), Loada(t2)}, …}, wherein Loadt is a current load type, and Loada is a current load action type.

[0020] In the space target behavior feature library construction method, the initial feature database is constructed according to the value of each type of space target and each type of feature data information, and the specific method is as follows:

[0021] The principle of equal variance test is adopted to detect the orbit change by using the posterior residual sequence, and the value of the feature data information Oaq of the behavior feature type of the orbit anomaly is obtained.

[0022] The value of the feature data information Maq of the behavior feature type of relative motion of each type of target is obtained by measuring the position vector of the target.

[0023] The value of the feature data information Raq of the behavior feature type of orbit rotation motion of each type of target is obtained by identifying the edge and relative pose information of the space target and using second-order linear differentiation.

[0024] According to the load type and the load state, the value of the feature data information Laq of the behavior feature type of target load action of each type of target is obtained.

[0025] According to the number of space target types, the data area is divided, in each data area, the target sample is divided into sub-data areas, and in each target sample sub-data area, the value of the feature data information Oaq of the orbit anomaly, the value of the feature data information Maq of the relative motion, the value of the feature data information Raq of the orbit rotation motion, and the value of the feature data information Laq of the target load action are sequentially placed to form the initial feature database.

[0026] In the space target behavior feature library construction method, the value of the feature data information Oaq of the behavior feature type of orbit anomaly of each type of target is obtained by the following method:

[0027] If the orbit root residual is close to 0 compared with the target orbit determination result at the previous sampling time, the orbit determination result obtained by the conventional least square estimation orbit determination method is not changed, the anomaly flag Flagyd is set to the no-change flag, otherwise the anomaly flag Flagyd is set to the change flag.

[0028] The three-axis anomaly velocity components Vex, Vey, and Vez are obtained by comparing the orbit roots before and after the orbit change.

[0029] In the space target behavior feature library construction method, the value of the feature data information Maq of the behavior feature type of relative motion of each type of target is obtained by the following method:

[0030] The target relative velocity information is obtained by using second-order linear differentiation, and the formula is as follows:

[0031]

[0032] In the formula, ω(t) is the position of the space target at t, Ts is the sampling time, and y(t) is the target relative velocity information; ω(t) is obtained by measuring the position vector of the space target.

[0033] In the space target behavior feature library construction method, the value of the characteristic data information Raq of each type of target behavior feature type is an orbit rotation motion, and the specific method is:

[0034] The time-stamped current attitude quaternions q1(t), q2(t), q3(t), and q4(t) of the space target are obtained by recognizing the edge of the space target.

[0035] The target attitude rotation speed information is obtained by using second-order linear differentiation, and the target rotation motion information dq1(t), dq2(t), dq3(t), and dq4(t) is obtained.

[0036] In the space target behavior feature library construction method, the behavior features of each type of space target in the initial feature database are modeled to construct a behavior feature model library, and the specific steps are:

[0037] (1) The behavior feature model of each type of space target is constructed using the hidden Markov algorithm, and the expression is:

[0038]

[0039] In the formula, λ i is the i-th type of space target, Z is the set of all possible hidden states in the entire hidden Markov model, N is the number of hidden states, M is the set of all observable system states in the entire hidden Markov space target classification model, m1={Oaq(1), Maq(1), Raq(1), Laq(1)}, m2={Oaq(2), Maq(2), Raq(2), Laq(2)}, and so on; K is the number of visible states; π is the state probability distribution of the first set of data {Oaq(1), Maq(1), Raq(1), Laq(1)} of the feature vector sequence, and the sum of the elements in π is 1; A is the hidden state probability transition matrix, and B is the emission probability matrix of the hidden Markov space target classification model.

[0040] The state probability distribution π, the hidden state probability transition matrix A, and the emission probability matrix B are iteratively calculated using the feature database and the Baum-Welch algorithm, wherein B is calculated using a Gaussian mixture model, π and A are calculated using a Lagrange method, and the training process of the hidden Markov model ends until the parameters π, A, and B converge, and the behavior feature model corresponding to each type of space target is obtained.

[0041] Put the behavior characteristic model corresponding to each space target type into the database to form a behavior characteristic model library.

[0042] In the above space target behavior characteristic library construction method, the value of the feature data information of the to-be-evaluated space target is acquired, each type of space target behavior model in the behavior characteristic model library is matched, the target type of the to-be-evaluated space target is determined according to the matching result, and the specific method is as follows:

[0043] The space target type of the to-be-evaluated space target is determined, the sequence X of the feature data information of the to-be-evaluated space target is acquired by using the behavior characteristic model, X={X(1), X(2), X(3), …}, wherein X(i)={Oaq(i), Maq(i), Raq(i), Laq(i)}, i=1, 2, 3, …;

[0044] The likelihood probability P(X|λi) of the sequence X under the behavior characteristic model is calculated by using a Markov model forward algorithm, the probability Pmax(X|λi) with the maximum likelihood probability value is acquired, and the corresponding λi is the space target type corresponding to the to-be-evaluated space target.

[0045] Compared with the prior art, the present application has the following beneficial effects:

[0046] (1) The present application realizes the quantification of the behavior characteristics of space targets by constructing a space target behavior characteristic library and setting the feature data information corresponding to the target behavior characteristics, and a new type of space target characteristic is added.

[0047] (2) The present application proposes a space target behavior characteristic library construction method, realizes the automatic and effective classification of the input behavior characteristics of targets by constructing a feature model library in the feature library, solves the problem that the traditional space target database does not have the function of automatically classifying and analyzing the behavior characteristics of space targets, and lays a foundation for realizing a space target behavior characteristic database with high accuracy, wide coverage and comprehensive characteristics. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a space target behavior characteristic library framework schematic diagram provided by an embodiment of the present application;

[0049] Figure 2 is a space target motion characteristic classification schematic diagram provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is to be understood that the present disclosure can be embodied in various forms without being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art. It should be noted that the embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0051] As shown in Figure 1 The present disclosure discloses a method for constructing a space target behavior feature library, the steps of which include:

[0052] Step (1), determining the type of space target; the type of space target includes remote sensing, communication, navigation, reconnaissance, and attack and defense.

[0053] Step (2), determining the type of behavior feature of space target; the type of behavior feature includes orbit anomaly, relative motion, rotational motion, and load action.

[0054] Step (3), determining the feature data information for each type of behavior feature; the specific method is:

[0055] When the type of behavior feature is orbit anomaly, the feature data information is the time sequence including anomaly flag and anomaly speed increment: Oaq={{Flagyd(t1), Vex(t1), Vey(t1), Vez(t1)}, {Flagyd(t2), Vex(t2), Vey(t2), Vez(t2)}, …}, wherein Flagyd is anomaly flag, and Vex, Vey, and Vez are three-axis anomaly speed components;

[0056] When the type of behavior feature is relative motion, the feature data information is the time sequence of relative motion information: Maq={{x(t1), y(t1), z(t1), dx(t1), dy(t1), dz(t1)}, {x(t2), y(t2), z(t2), dx(t2), dy(t2), dz(t2)}, …}, wherein x, y, and z are three-axis space coordinates of the space target relative to the monitored space target body.

[0057] When the behavior characteristic type is a rotational motion, the characteristic data information is a sequence of attitude change information: Raq = { {q1(t1), q2(t1), q3(t1), q4(t4), dq1(t1), dq2(t1), dq3(t1), dq4(t1)}, {q1(t2), q2(t2), q3(t2), q4(t2), dq1(t2), dq2(t2), dq3(t2), dq4(t2)},...}, wherein q1-q4 are attitude quaternions;

[0058] When the behavior characteristic type is a load motion, the characteristic data information is a time sequence of load action information: Laq = { {Loadt(t1), Loada(t2)}, {Loadt(t2), Loada(t2)},...}, wherein Loadt is a current load type, and Loada is a current load action type.

[0059] In step (4), an initial characteristic database is constructed according to the value of each type of characteristic data information of each type of space object, and the specific method is as follows:

[0060] The principle of equal variance test using the posterior residual sequence is used to detect the orbit change, and the value of the characteristic data information Oaq of the behavior characteristic type of the orbit change is obtained. The target orbit determination result obtained by using the conventional least square estimation orbit determination method is compared with the target orbit determination result at the previous sampling time. If the residual of the orbit elements is close to 0, there is no orbit change, and the change flag Flagyd is set to the no-change flag. Otherwise, the change flag Flagyd is set to the change flag. The three-axis change velocity components Vex, Vey, and Vez are obtained according to the comparison of the orbit elements before and after the orbit change.

[0061] The value of the characteristic data information Maq of the behavior characteristic type of relative motion of each type of target is obtained through the position vector measurement result of the target, and the specific method is as follows:

[0062] The second-order linear differential is used to obtain the target relative velocity information, and the formula is as follows:

[0063]

[0064] In the formula, ω(t) is the position of the space object at time t, Ts is the sampling time, and y(t) is the target relative velocity information. ω(t) is obtained by measuring the position vector of the space object.

[0065] The value of the characteristic data information Raq of the behavior characteristic type of orbit rotational motion of each type of target is obtained by recognizing the edge and relative pose information of the space object and using the second-order linear differential, and the specific method is as follows:

[0066] By identifying the spatial target edge, the time-stamped current attitude quaternion q1(t), q2(t), q3(t), q4(t) of the spatial target is obtained;

[0067] The second-order linear differential is used to obtain the target attitude rotation speed information, and the target rotation motion information dq1(t), dq2(t), dq3(t), dq4(t) is obtained.

[0068] According to the load type and the load state, the characteristic data information Laq of the target load action is obtained.

[0069] According to the number of spatial target types, the data area is divided, in each data area, the target sample is divided into a sub-data area, and in each target sample sub-data area, the values of the characteristic data information Oaq of the orbit anomaly, the values of the characteristic data information Maq of the relative motion, the values of the characteristic data information Raq of the orbit rotation motion, and the values of the characteristic data information Laq of the target load action are sequentially placed to form an initial feature database.

[0070] Step (5), modeling the behavior characteristics of each type of spatial target in the initial feature database, and constructing a behavior characteristic model library

[0071] (1) The behavior characteristic model of each type of spatial target is constructed by using the hidden Markov algorithm, and the expression is:

[0072]

[0073] In the formula, λ i is the i-th type of spatial target, Z is the set of all possible hidden states in the entire hidden Markov model, N is the number of hidden states, M is the set of all observable system states in the entire hidden Markov spatial target classification model,'m1={Oaq(1), Maq(1), Raq(1), Laq(1)}, m2={Oaq(2), Maq(2), Raq(2), Laq(2)}, and so on; K is the number of explicit states; π is the state probability distribution of the first group of data {Oaq(1), Maq(1), Raq(1), Laq(1)} of the feature vector sequence, and the sum of each element in π is 1; A is the hidden state probability transition matrix, and B is the emission probability matrix of the hidden Markov spatial target classification model.

[0074] The state probability distribution pi, the hidden state probability transition matrix A and the emission probability matrix B are iteratively calculated by using the feature database and the Baum-Welch algorithm, wherein the Gaussian mixture model is used to calculate B, the Lagrange method is used to calculate pi and A, until the parameters pi, A and B are converged, the training process of the hidden Markov model is ended, and the behavior feature model corresponding to each space target type is obtained.

[0075] The behavior feature model corresponding to each space target type is put into the database to form a behavior feature model library.

[0076] In step (6), the value of the feature data information of the space target to be evaluated is obtained, and each type of space target behavior model in the behavior feature model library is matched, and the target type of the space target to be evaluated is determined according to the matching result, and the specific method is as follows:

[0077] In step (6), the value of the feature data information of the space target to be evaluated is obtained, and each type of space target behavior model in the behavior feature model library is matched, and the target type of the space target to be evaluated is determined according to the matching result, and the specific method is as follows:

[0078] The sequence X of the feature data information of the space target to be evaluated is obtained by using the behavior feature model, X={X(1), X(2), X(3), …}, wherein X(i)={Oaq(i), Maq(i), Raq(i), Laq(i)}, i=1, 2, 3, ….

[0079] In step (7), the target type of the space target to be evaluated and the feature data information thereof are added to the initial feature database, and the initial feature database is updated to form a space target behavior feature library. Thus, the construction of the space target behavior feature library can be realized, and the finally constructed database framework is as shown in Figure 2 .

[0080] Although the present application has been disclosed with the above preferred embodiments, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications to the technical solutions of the present application by using the disclosed methods and technical contents without departing from the spirit and scope of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, which does not depart from the content of the technical solutions of the present application, belongs to the protection scope of the technical solutions of the present application.

Claims

1. A method for constructing a behavior feature library of a space target, characterized in that, The method comprises the following steps: determining the type of a space target; determining the behavior characteristic type of the space target; determining the characteristic data information of each behavior characteristic type respectively; constructing an initial characteristic database according to the value of each characteristic data information of each type of space target; modeling the behavior characteristics of each type of space target in the initial characteristic database by using a Hidden Markov Model and a Gaussian Mixture Model to construct a behavior characteristic model library; obtaining the value of the characteristic data information of the space target to be evaluated, and matching the value with each type of space target behavior model in the behavior characteristic model library by using a forward algorithm, and determining the type of the space target to be evaluated according to the matching result; adding the type of the space target to be evaluated and its characteristic data information to the initial characteristic database to update the initial characteristic database to form a space target behavior characteristic library; constructing an initial characteristic database according to the value of each characteristic data information of each type of space target, and the specific method is as follows: detecting the variable orbit condition by using the principle of equal variance test of the posterior residual sequence to obtain the value of the characteristic data information Oaq of the behavior characteristic type of orbit anomaly; obtaining the value of the characteristic data information Maq of the behavior characteristic type of relative motion of each type of target by measuring the position vector of the target; obtaining the value of the characteristic data information Raq of the behavior characteristic type of orbit rotation motion of each type of target by recognizing the edge and relative pose information of the space target and using second-order linear differentiation; obtaining the value of the characteristic data information Laq of the behavior characteristic type of target load action of each type of target according to the load type and load state; dividing the data area according to the number of space target types, dividing a sub-data area according to the target sample in each data area, and sequentially placing the value of the characteristic data information Oaq of orbit anomaly, the value of the characteristic data information Maq of relative motion, the value of the characteristic data information Raq of orbit rotation motion, and the value of the characteristic data information Laq of target load action in each target sample sub-data area to form an initial characteristic database.

2. The method of claim 1, wherein: The space target type is a remote sensing type, a communication type, a navigation type, a reconnaissance type, or an attack and defense type.

3. The method of claim 1, wherein: The behavior characteristic type is orbit anomaly, relative motion, rotation motion, or load action.

4. The method of claim 1, wherein: The specific method for determining the characteristic data information of each behavior characteristic type is as follows: When the behavior feature type is track movement, the feature data information is a time sequence including movement flag and movement speed increment: Oaq = { {Flag yd (t1), Vex(t1), Vey(t1), Vez(t1)}, {Flag yd (t2), Vex(t2), Vey(t2), Vez(t2)}, …}, wherein Flag yd is the movement flag, and Vex, Vey, and Vez are three-axis movement speed components. when the behavior characteristic type is relative motion, the characteristic data information is a time sequence of relative motion information: Maq={{x(t1),y(t1),z(t1),dx(t1),dy(t1),dz(t1)}, {x(t2),y(t2),z(t2),dx(t2),dy(t2),dz(t2)},…}, wherein x, y, and z are three-axis space coordinates of the space target relative to the monitored space target body. When the behavior characteristic type is a rotating motion, the characteristic data information is a posture change information sequence: Raq={{q1(t1), q2(t1), q3(t1), q4(t1), dq1(t1), dq2(t1), dq3(t1), dq4(t1)}, {q1(t2), q2(t2), q3(t2), q4(t2), dq1(t2), dq2(t2), dq3(t2), dq4(t2)}, …}, wherein q1-q4 are posture quaternions; When the behavior feature type is load movement, the feature data information is a time sequence of load action information: Laq = { {Load t (t1), Load a (t2)}, {Load t (t2), Load a (t2)},...}, wherein Load t is a load type used at a current time, and Load a is an action type used by the load at the current time.

5. The method of claim 1, wherein: The value of the characteristic data information Oaq of each type of target behavior characteristic type being an orbit anomaly is obtained in the following manner: If the orbit root residual is close to 0 when compared with the target orbiting result at the previous sampling time, the target orbiting result obtained by using the conventional least square estimation orbiting method is not changed, the anomaly flag Flagyd is set as a no-change flag, otherwise, the anomaly flag Flagyd is set as a change flag; The three-axis anomaly velocity components Vex, Vey, and Vez are obtained according to the comparison of the orbit roots before and after the change.

6. The method of claim 1, wherein: The value of the characteristic data information Maq of each type of target behavior characteristic type being a relative motion is obtained in the following manner: The target relative velocity information is obtained by using a second-order linear differential, and the formula is: where ω(t) is the position of the space object at time t, T s is the sampling time, y(t) is the target relative velocity information; ω(t) is obtained by measuring the position vector of the space object.

7. The method of claim 1, wherein: The value of the characteristic data information Raq of each type of target behavior characteristic type being an orbit rotating motion is obtained in the following manner: The time-stamped current posture quaternions q1(t), q2(t), q3(t), and q4(t) of the space target are obtained by recognizing the edge of the space target. The target posture rotating velocity information is obtained by using a second-order linear differential, and the target rotating motion information dq1(t), dq2(t), dq3(t), and dq4(t) is obtained.

8. The method of claim 1, wherein: The behavior characteristics of each type of space target in the initial characteristic database are modeled, and a behavior characteristic model library is constructed in the following manner: (1) The behavior characteristic model of each type of space target is constructed by using an implicit Markov algorithm, and the expression is: where λ i is the i-th spatial target type, Z is the set of all possible hidden states in the whole hidden Markov model, N is the number of hidden states, M is the set of all possible observed system states in the whole hidden Markov spatial target classification model,'m1 = {Oaq(1), Maq(1), Raq(1), Laq(1)}, m2 = {Oaq(2), Maq(2), Raq(2), Laq(2)}, and so on; K is the number of visible states; π is the state probability distribution of the first set of data {Oaq(1), Maq(1), Raq(1), Laq(1)} of the feature vector sequence, the sum of each element in π is 1; A is the hidden state probability transition matrix, and B is the emission probability matrix of the hidden Markov spatial target classification model; (2) The state probability distribution π, the hidden state probability transition matrix A, and the emission probability matrix B are iteratively calculated by using the characteristic database and the Baum-Welch algorithm, wherein B is calculated by using a Gaussian mixture model, π and A are calculated by using a Lagrange method, until the parameters π, A, and B are all converged, the training process of the implicit Markov model is ended, and the behavior characteristic model corresponding to each type of space target is obtained; (3) The behavior characteristic model corresponding to each type of space target is placed in the database, and a behavior characteristic model library is formed.

9. The method of claim 1, wherein: The value of the characteristic data information of the space target to be evaluated is obtained, and each type of space target behavior model in the behavior characteristic model library is matched, and the target type of the space target to be evaluated is determined according to the matching result in the following manner: Determine the space target type of the space target to be evaluated, obtain the sequence X of the feature data information of the space target to be evaluated by using the behavior feature model, X={X(1), X(2), X(3), …}, wherein X(i)={Oaq(i), Maq(i), Raq(i), Laq(i)}, i=1, 2, 3, …; And use the Markov model forward algorithm to calculate the likelihood probability P(X|λi) of the sequence X under the behavior feature model, obtain the maximum probability Pmax(X|λi) of the likelihood probability value, and the corresponding λi is the space target type corresponding to the space target to be evaluated.

Images