Space target attitude fuzzy classification method and device based on GBISAR imaging and medium

Abstract

The invention discloses a space target attitude fuzzy classification method and device based on GBISAR imaging and a medium. The method comprises the following steps: firstly, establishing an HCS, UNW-based coordinate system, an airframe coordinate system and a connection process between the coordinate systems, thereby establishing a new space observation model; secondly, equivalently decomposing the position and posture change of the space target into position and posture change and posture change, and providing a basis for subsequent analysis and establishment of a space target posture change mode and development of feature design; in the aspect of feature design, the attitude change is decomposed by introducing Fourier series, so that a polynomial decomposition result of the attitude change of the space target is given; and the attitude change is approximated by keeping a first-order item. And finally, establishing a criterion-based attitude classification method on a single judgment condition, and establishing a fuzzy set-based attitude classification method on multiple judgment conditions by introducing related contents of a fuzzy set theory, establishing a membership function based on an attitude change mode and giving a fuzzy membership degree of attitude classification.

Description

Technical Field

[0001] The present invention relates to the technical field of space target attitude classification, and in particular to a space target attitude fuzzy classification method, equipment and medium based on GBISAR imaging. Background Art

[0002] As human demand for space exploration and development continues to expand, the development of space is progressing rapidly. On the one hand, it makes an indispensable contribution to the development of human civilization. On the other hand, it also inevitably generates and brings about space security threats. The space situational security is becoming increasingly complex and severe and faces great challenges.

[0003] As a prerequisite for conducting space activities, responding to space security threats, and ensuring space situational safety, space situational awareness (SSA) and the construction of SSA systems have received more attention than ever before. SSA refers to the ability to detect, track, and predict the position of objects in Earth orbit, as well as the ability to monitor the broader space environment to identify potential risks or dangers. Within SSA, monitoring the motion state of space targets is a crucial research area. For example, monitoring the motion state of in-orbit satellites is a prerequisite for target attitude estimation, geometric inversion, and attitude classification. It also reflects the target's motion intent and level of danger, and is crucial for monitoring and understanding the operational status of space targets, especially non-cooperative ones, and thus for assessing space situational awareness.

[0004] Currently, there is a practical need to further refine the research on space target attitude classification. On the one hand, with the increasing complexity of space missions, especially the increasing demand for detailed monitoring of space targets in low-Earth orbit, the space target attitude classification task faces the challenge of the diversity and subtlety of target attitudes brought about by dynamic and complex environments. Traditional space target attitude classification only focuses on coarse attitude categories such as attitude stability and adjustment, which cannot fully meet the new needs. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method, device, and medium for fuzzy classification of space target attitudes based on GBISAR imaging.

[0006] A first aspect of the present invention provides a method for fuzzy classification of space target attitudes based on GBISAR imaging, comprising the following steps:

[0007] Build a space observation model based on the GBISAR observation station and the space targets to be observed;

[0008] Observing the on-orbit state of the space target using the GBISAR observation station, and inverting the observed on-orbit state of the space target into the space observation model;

[0009] Describing, in the space observation model, attitude changes of the space target during on-orbit operation;

[0010] The attitude changes of the space target during its on-orbit operation are projected onto a two-dimensional plane through GBISAR imaging to obtain an ISAR imaging result of the space target;

[0011] Performing polynomial decomposition on the ISAR imaging result of the space target using Fourier series to obtain a two-dimensional feature representation of the space target in the ISAR imaging result;

[0012] Fuzzy classification is performed based on the two-dimensional feature representation of the space target.

[0013] Furthermore, the construction of a space observation model based on the GBISAR observation station and the space target to be observed specifically includes the following steps:

[0014] With the GBISAR observation station as the origin, the HCS coordinate system is established Where, γ represents the radar line of sight LOS, represents the horizontal latitude, and θ represents the horizontal longitude;

[0015] With the space target to be observed as the origin, establish the UNW coordinate system (U, W, N) and the body coordinate system (X, Y, Z); where U represents the velocity vector direction of the orbit, N represents the normal direction in the orbital plane that is perpendicular to the U direction and points outside the orbit, and the W direction forms a right-handed system with the U and N directions; X represents the roll direction of the space target's rotation, Y represents the pitch direction of the space target's rotation, and the Z axis represents the yaw direction of the space target's rotation;

[0016] Establish the connection between the UNW coordinate system and the HCS coordinate system;

[0017] Establish the connection between the body coordinate system and the UNW coordinate system;

[0018] The connection relationship between the UNW coordinate system and the HCS coordinate system is expressed by the following coordinate matrix:

[0019]

[0020] The connection relationship between the body coordinate system and the UNW coordinate system is expressed by the following coordinate matrix:

[0021]

[0022] Furthermore, the space target is composed of a head, a solar panel portion, a pitch characteristic vector portion, a reference region portion, a body portion, and a foot;

[0023] The solar panel portion and the body portion are used to determine the current orientation of the space target;

[0024] The head and the foot are used to cooperate with the UNW coordinate system and the body coordinate system to calculate the current orientation of the space target;

[0025] The pitch eigenvector portion is used to represent the current orientation of the space target;

[0026] The reference area portion is used to provide an external reference benchmark for the current orientation of the space target.

[0027] Furthermore, inverting the observed on-orbit state of the space target into the space observation model specifically includes the following steps:

[0028] Decompose the on-orbit state of space targets observed by the GBISAR observation station into position change and attitude change;

[0029] The position change of the space target is expressed in the HCS coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the UNW coordinate system and the HCS coordinate system;

[0030] The attitude change of the space target is expressed in the body coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the body coordinate system and the UNW coordinate system;

[0031] The on-orbit status of space targets is expressed in the space observation model using the UNW coordinate system.

[0032] Furthermore, the description of the attitude change of the space target during on-orbit operation in the space observation model specifically includes:

[0033] A point per unit length of each coordinate axis in the UNW coordinate system is selected as a target point; a three-dimensional rotation matrix is ​​constructed through the three-dimensional motion of the target point to express the attitude change of the space target during on-orbit operation;

[0034] The attitude change of the space target during on-orbit operation is specifically expressed by the following coordinate matrix:

[0035]

[0036] in, represents the initial attitude of the space target when it is first observed, It represents the posture change of the space target at any time compared to the initial observation. Each element in the matrix represents the component generated by the target point on each coordinate axis after three-dimensional motion.

[0037] Furthermore, projecting the attitude changes of the space target during on-orbit operation onto a two-dimensional plane through GBISAR imaging specifically includes the following steps:

[0038] According to the radar line of sight direction during GBISAR imaging, the preserved dimension of the three-dimensional rotation matrix is ​​determined and the dimension-preserving matrix is ​​constructed.

[0039] The attitude change of the space target during on-orbit operation is multiplied by the dimension-preserving matrix to obtain the ISAR imaging result of the space target.

[0040] Furthermore, the polynomial decomposition of the ISAR imaging result of the space target using Fourier series specifically includes the following steps:

[0041] The ISAR imaging result of the space target is equivalently represented as the periodic circular motion generated by three target points in a plane, and is expressed using periodic functions x1(t), x2(t), and x3(t);

[0042] The motion trajectories of the three target points are decomposed into polynomial form using fast Fourier transform:

[0043]

[0044] Where Ω0 is the angular frequency of circular motion, T is the corresponding rotation period of the target point, j is the sign of the imaginary number, k is the sequence number of the Fourier series, and t is time;

[0045] Keeping the polynomial to order 1 and removing the order 0 result, we get the following two-dimensional attitude matrix:

[0046]

[0047] Rewrite the two-dimensional attitude matrix using the trigonometric cosine function to obtain the following two-dimensional cosine attitude matrix:

[0048]

[0049] The two-dimensional cosine attitude matrix is ​​used as a two-dimensional feature representation of the space target in the ISAR imaging result.

[0050] Furthermore, the fuzzy classification based on the two-dimensional feature representation of the space target specifically includes the following steps:

[0051] Designing special eigenvectors; the special eigenvectors include row eigenvectors, pitch eigenvectors, roll eigenvectors, yaw eigenvectors, and column eigenvectors; wherein the row eigenvectors and column eigenvectors are used to describe the reference benchmark of the range dimension in the ISAR imaging results; the pitch eigenvectors, roll eigenvectors, and yaw eigenvectors are used to describe the attitude change of the space target;

[0052] Designing a criterion feature vector; the criterion feature vector is used to match various posture change categories of the space target;

[0053] According to the variance of the space target on each criterion feature vector, the fuzzy membership of the space target in each posture category is calculated;

[0054] A fuzzy set is generated according to the fuzzy membership of the space target in each posture category, and the fuzzy set is used as a fuzzy classification result of the space target.

[0055] A second aspect of the present invention discloses an electronic device, comprising a processor and a memory;

[0056] The memory is used to store programs;

[0057] The processor executes the program to implement the space target posture fuzzy classification method based on GBISAR imaging described in the first aspect.

[0058] A third aspect of the present invention is a computer-readable storage medium storing a program, wherein the program is executed by a processor to implement the method for fuzzy classification of spatial target postures based on GBISAR imaging described in the first aspect.

[0059] The present invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the above method.

[0060] The embodiments of the present invention have the following beneficial effects: The present invention provides a method, device and medium for fuzzy classification of space target attitudes based on GBISAR imaging. Starting from establishing a GBISAR space observation model based on HCS, UNW coordinate system and body coordinate system, a new space observation model is established by constructing GBISAR, the coordinate system of the space target on-orbit and the space target body and the connection process between them, and some important conclusions are given, especially the equivalent decomposition of the space target posture change into posture change and attitude change, which provides a basis for subsequent analysis and establishment of a model of space target posture change and feature design. In the process of analyzing the posture change of space targets on-orbit, the factors of the space target posture change are first analyzed to refine the definition and interpretation of the posture category, and the posture change is decomposed by introducing the Fourier series to give the polynomial decomposition result of the space target posture change; and in further formal discussion, the posture change is approximated by retaining the first-order term. By proposing and discussing the concepts and definitions of special and characteristic vectors and criteria, a criterion-based posture classification method based on a single decision condition is established. By introducing relevant contents of fuzzy set theory, membership functions are established based on posture change patterns and the fuzzy membership degree of posture classification is given, and a fuzzy set-based posture classification method based on multiple decision conditions is established.

[0061] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0063] Figure 1 This is a basic implementation flow chart of a method for fuzzy classification of space target posture based on GBISAR imaging in the present invention;

[0064] Figure 2 It is a schematic diagram of the space observation model constructed by the present invention;

[0065] Figure 3 It is a schematic diagram of the HCS coordinate system in the space observation model of the present invention;

[0066] Figure 4 It is a schematic diagram of the UNW coordinate system in the space observation model of the present invention;

[0067] Figure 5 It is a schematic diagram of the basic model of the space target observed by the present invention;

[0068] Figure 6 It is a schematic diagram of the body coordinate system in the space observation model of the present invention;

[0069] Figure 7 This is a schematic diagram of the effect of projecting the three-dimensional posture change of a space target onto a two-dimensional plane;

[0070] Figure 8 This is a schematic diagram of the effect of the present invention on the change of the spatial target posture based on Fourier series decomposition;

[0071] Figure 9 It is a schematic diagram of designing a special feature vector based on a space target in the present invention;

[0072] Figure 10 3D point trace diagram of a space target in a specific embodiment;

[0073] Figure 11 is a schematic diagram of a two-dimensional point trace of a space target in a specific embodiment;

[0074] Figure 12 Schematic diagram of characteristic vectors of space targets in a specific embodiment;

[0075] Figure 13 Schematic diagram of the criterion feature vector used in the present invention;

[0076] Figure 14 This is a schematic structural diagram of an electronic device of the present invention;

[0077] Figure 15 It is a schematic diagram of the structure of a computer-readable storage medium of the present invention. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0079] Before describing the embodiments of the present invention, the Ground Based Inverse Synthetic Aperture Radar (GBISAR) imaging technology is first explained:

[0080] As a space observation radar system, GBISAR differs significantly from optical systems. In the optical imaging model, the imaging projection plane is perpendicular to the optical sensor's line of sight. A three-dimensional target is projected into a two-dimensional optical image, where the width and height dimensions are preserved, while the depth dimension (i.e., the radar's range dimension) is lost. In the radar imaging model, the imaging projection plane is parallel to the radar sensor's line of sight. A three-dimensional target is projected into a two-dimensional radar image, where the range and Doppler (azimuth) dimensions are preserved, while the altitude dimension (i.e., the radar's elevation dimension) is lost.

[0081] When discussing GBISAR, the measurable dimensions are generally considered to be the range and Doppler dimensions. Information about the elevation dimension is generally not discussed. Furthermore, GBISAR's range and azimuth measurements are relatively independent and can generally be discussed separately. Specifically, within an ISAR imaging system, the motion of an observed target can be decomposed into translation and rotation. Movement along an iso-Doppler line is the target's translation relative to the ISAR, while movement along an iso-distance line is the target's rotation relative to the ISAR. Based on the above definitions, target translation along an iso-Doppler line only affects the range dimension and has no effect on the azimuth dimension; conversely, target rotation only affects the azimuth dimension. The conventional process of ISAR imaging technology involves first using motion compensation to make the target's motion pattern relative to the ISAR radar equivalent to that of a turntable model. Then, through imaging processing, the compensated equivalent radar echo data is used to reconstruct an image to outline the spatial distribution of electromagnetic wave reflections from scattering points. ISAR imaging technology lays the foundation for the GBISAR spatial observation model, and the spatial observation model of the present invention is constructed based on GBISAR imaging.

[0082] Based on the above concepts, the first embodiment of the present invention provides a method for fuzzy classification of space target attitudes based on GBISAR imaging, comprising the following steps:

[0083] S1. Build a space observation model based on the GBISAR observation station and the space targets to be observed;

[0084] S2. Use the GBISAR observation station to observe the on-orbit state of space targets and invert the observed on-orbit state of space targets into the space observation model;

[0085] S3. Describe the attitude changes of space targets during on-orbit operation in the space observation model;

[0086] S4. Projecting the attitude changes of the space target during its on-orbit operation onto a two-dimensional plane through GBISAR imaging to obtain the ISAR imaging results of the space target;

[0087] S5. Use Fourier series to perform polynomial decomposition on the ISAR imaging results of the space target to obtain the two-dimensional feature representation of the space target in the ISAR imaging results;

[0088] S6. Perform fuzzy classification based on the two-dimensional feature representation of spatial targets.

[0089] The present invention is expounded from three perspectives: space observation model, space target attitude pattern and feature design. By establishing a new space observation model, the posture change of the space target is equivalently decomposed into position change and attitude change, and the polynomial decomposition of the attitude change is given based on the Fourier series. By retaining the order to establish an equivalence relationship, the attitude change is equivalent to three-dimensional rotational motion, and a complete space target attitude change pattern is established in combination with anomalies and processing solutions. Then, the feature design of space target attitude classification is carried out. By proposing and discussing the concepts of special feature vectors, criteria and fuzzy sets, the space target attitude classification method of the present invention is established.

[0090] The following describes in detail the implementation process of each step of the present invention:

[0091] S1. Build a space observation model based on the GBISAR observation station and the space targets to be observed.

[0092] The space observation model of the embodiment of the present invention is constructed based on the HCS coordinate system (Horizon Coordinate System, horizontal coordinate system), the UNW coordinate system and the body coordinate system (Body-Fixed Reference Frame, BFRF). The characteristic of this new space observation model is stronger connectivity, especially the connectivity between the various functional modules of the model itself and with the actual observation. By analyzing the properties of the various functional modules of the actual observation and the UNW coordinate system, a new space observation model composed of interconnected sub-processes is constructed, which provides a basis for the subsequent proposal of fuzzy classification technology for space target postures. The effect diagram is shown as follows: Figure 2 shown.

[0093] In an embodiment of the present invention, a space observation model is constructed based on a GBISAR observation station and a space target to be observed, specifically including the following steps:

[0094] S1-1. Establish the HCS coordinate system with the GBISAR observation station as the origin

[0095] The horizontal coordinate system is a type of celestial coordinate system. Its construction method is as follows: a celestial sphere is constructed with the observer at its center, and the horizontal plane is the horizontal plane. The horizontal plane and the celestial sphere intersect at the cardinal circle. A plumb line is drawn through the center, intersecting the celestial sphere at two points. The point above the cardinal circle is called the zenith, and the point below the cardinal circle is called the nadir. The Earth's axis intersects the celestial sphere at the north and south celestial poles, forming the meridian circle through the zenith. It intersects the cardinal circle at the north and south points. Starting from the north point, the dihedral angle coordinates (horizontal longitude, Az) are recorded by rotating clockwise in a left-handed coordinate system. Starting from the cardinal circle, the horizontal latitude (Ait) is recorded as the linear angle coordinate. The horizontal coordinate system uses two degrees of freedom (horizontal latitude and longitude) to describe any point on the celestial sphere, making it convenient for observers to quickly record astronomical observations.

[0096] The HCS coordinate system established by the above method is as follows Figure 3 As shown, it can be expressed in the following form:

[0097]

[0098] Where γ represents the radar line of sight (LOS), represents the horizontal latitude, and θ represents the horizontal longitude.

[0099] When GBISAR is used as an observer, the position of a space target in the sky can also be recorded using horizontal coordinates. When using the horizontal coordinate system, the space target is sufficiently small compared to the celestial sphere to be equivalent to a point mass. In this case, the position change of the space target is reflected only as a position change. By observing the position change of the space target using the horizontal coordinate system and recording it in real time, the observation of the space target's position change is divided into horizontal observations. Therefore, the present invention uses the HCS coordinate system to equate the space target to a point target on the celestial sphere, ignoring attitude changes and recording the position change of the space target using two coordinates: horizontal latitude and horizontal longitude. This effectively decomposes the position change of the space target into position change and attitude change. Its advantages are: first, as a method for determining the radar search area during actual observation, horizontal observation connects with the functional module that first determines the airspace in which the space target is located during space target observation and provides a model for this functional module. Second, when using a coordinate reduction algorithm with different station centers, the accuracy of space target position observation can be increased through multi-station observations, providing a new application for multi-station observations.

[0100] S1-2. With the space target to be observed as the origin, establish the UNW coordinate system (U, W, N) and the body coordinate system (X, Y, Z).

[0101] The UNW coordinate system is a coordinate system for the orbit of a space target. It was first proposed to address the problem of spacecraft collisions. The UNW coordinate system is defined by the on-orbit motion of the space target. The target's velocity is defined as the U direction, and the orbital normal vector W is obtained by cross-producting the observed line of sight. Finally, the N direction is determined according to the right-hand rule, and is generally considered to be the line of sight. It is worth noting that, while the origin of the UNW coordinate system is the center of gravity of the space target, the establishment of the UNW coordinate system is only related to the positional motion of the space target and the GBISAR line of sight, and not to changes in the target's attitude. Furthermore, the center of gravity of the space target is always stationary within the UNW coordinate system, and the target is sufficiently large relative to the UNW coordinate system that changes in position and posture manifest only as changes in attitude. Therefore, it is advisable to use the UNW coordinate system as a reference for attitude changes. The attitude of the space target is determined using the UNW coordinate system and the subsequently established target body coordinate system. The attitude and its changes are recorded in the UNW coordinate system, and observations of the target are referred to as UNW observations.

[0102] The UNW coordinate system established by the above method is as follows Figure 4 As shown, it can be expressed in the following form:

[0103]

[0104] Where v is the on-orbit velocity vector of the space target.

[0105] By observing and recording the attitude changes of space targets in real time using the UNW coordinate system, the observation of space target attitude changes is divided among UNW observations. On the one hand, as the reference coordinate system for determining the attitude changes of space targets during actual observations, horizon observations connect to and provide a model for the functional module that determines the reference baseline for determining the attitude changes of space targets during space target observations. On the other hand, when using attitude change equations based on rotation matrices, the attitude changes of space targets can be described and modeled through rotations, providing a new approach to describing attitude changes.

[0106] Before establishing the body coordinate system, the embodiment of the present invention first defines a basic model of the space target.

[0107] Space objects observed by GBISAR typically consist of two distinct components: the trunk and the solar array. Actual space objects vary in form, but the differences in these two components are relatively small. Therefore, a basic model can be extracted and constructed to represent the space object. Actual space objects are then refined based on this basic model, supplemented with additional details to construct specific models for each space object. The basic model is distinguished by using minimal elements to construct a model that characterizes important properties, including the distinction between the space object's three-view images, its attitude changes, and its reference and attitude change characteristics. This provides a foundation for subsequent feature design and classification methods.

[0108] The basic model of the space target extracted by the embodiment of the present invention is as follows Figure 5 As shown, the system consists of six components: the head, solar array, pitch eigenvector, reference region, body, and feet. The head's function is to calculate the current orientation of a space target. Combined with the body and UNW coordinate systems, it can be used to preliminarily determine the target's performance in ISAR imaging, particularly for predicting first-order attitude changes. However, the head alone is insufficient; the feet must be combined to calculate orientation. As the most observable component of a space target, the solar array's imaging performance significantly influences attitude classification, and this often depends on its orientation relative to the imaging projection plane during ISAR imaging. By combining the solar array's orientation with the head's orientation, the target's two-dimensional imaging result can be determined, and its three-dimensional pose can be restored to a certain extent. Directly determining the solar array's orientation is inconvenient in practice, so the more user-friendly pitch eigenvector is used to characterize the orientation. This involves calculating the solar array's normal vector using the pitch eigenvector and body components to determine the orientation. The body, another easily observable component besides the solar array, is responsible for determining the orientation of space targets. The head and feet, included in the body, are used for actual orientation calculations. Furthermore, the reference region, unrelated to space targets and relevant only to ISAR imaging, is an essential component of the model, providing a reference for calculating the orientation of space targets. Of the six components, the feet function identically to the head and are therefore not repeated here.

[0109] As a spacecraft, a space target is considered an aircraft. Combining the aforementioned basic space target model, a body coordinate system can be established for the space target. This establishment method is as follows: the head direction of the space target is the X-axis, the pitch eigenvector direction is the Y-axis, and the Z-axis direction is determined according to the right-hand rule. The origin of the coordinate system is fixed at the center of gravity of the space target. By definition, the body coordinate system is independent of the space target's motion state and shares the same origin as the UNW coordinate system.

[0110] The body coordinate system established by the above method is as follows Figure 6 As shown, it can be expressed in the following form:

[0111]

[0112] S1-3. Establish the connection between the UNW coordinate system and the HCS coordinate system.

[0113] The horizontal coordinate system and the UNW coordinate system, respectively, divide the functions of observing position and attitude changes of spatial objects. The resynthesis of these changes represents the connection between horizontal and UNW observations, that is, the connection between the two coordinate systems. This connection is a complex issue. This section provides a perspective on this connection by presenting a direct conversion relationship between the horizontal and UNW coordinate systems. This conversion relationship provides a foundation for aligning coordinate systems and correcting observation errors. It also provides a foundation for further derivative algorithms.

[0114] In this embodiment of the present invention, the radar line of sight is aligned with the N direction, and the horizontal latitude and longitude are used to represent the current directions U and W of the UNW coordinate system, respectively, to define the connection relationship between the observation station and the orbital level as follows:

[0115]

[0116] Through the above connections, the present invention completes the process from GBISAR space observation to space target on-orbit. By representing the space target differently in the horizontal coordinate system and the UNW coordinate system, the position change of the space target is equivalently decomposed into position change and attitude change, which are recorded separately by horizontal observation and UNW observation.

[0117] S1-4. Establish the connection between the body coordinate system and the UNW coordinate system;

[0118] By definition, the establishment of the body coordinate system is independent of the motion state of the space target and shares the same origin as the UNW coordinate system. Therefore, it is useful to combine the UNW coordinate system and the body coordinate system to define the attitude changes of the space target. Specifically, with the UNW coordinate system as the reference, the body coordinate system and the UNW coordinate system are aligned in the initial state. Thereafter, attitude changes at each moment are expressed as three-dimensional rotation changes of the body coordinate system relative to the UNW coordinate system. By establishing the UNW coordinate system and the body coordinate system, the attitude changes of the space target can be described by the relative three-dimensional rotation between the two coordinate systems. This allows the attitude changes to be modeled as the rotation equations of the coordinate systems and further established as a set of state equations that can be updated in real time. This provides a basis for describing the attitude adjustment of space targets, observation and imaging, and more.

[0119] The embodiment of the present invention uses the alignment between the UNW coordinate system and the BFRF to define the connection relationship between the orbit and the space target based on the initial attitude state, as shown below:

[0120]

[0121] It should be noted that, by definition, the UNW coordinate system and the body coordinate system have the same origin and belong to the same rectangular coordinate system, which allows the alignment of the two coordinate systems to be described by a one-to-one correspondence of the coordinate axes. According to the arrangement principle, there can be up to 6 alignment schemes. However, for the convenience of describing posture changes, the 6 schemes have priority. In the embodiment of the present invention, the optimal alignment scheme is used as the connection between the UNW coordinate system and the BFRF. In other embodiments, there may also be cases where other alignment schemes are used as the connection between the UNW coordinate system and the BFRF.

[0122] S2. Use the GBISAR observation station to observe the on-orbit status of space targets and invert the observed on-orbit status of space targets into the space observation model.

[0123] GBISAR has all-day, all-weather, active space detection capabilities, and provides continuous tracking and observation support for space targets through narrowband precision tracking and broadband high-resolution imaging working modes. More importantly, unlike optical images, the ISAR imaging process is mostly based on the Doppler analysis of the target echo signal. Therefore, the imaging image itself also contains information about the relative motion between the space target and the ground-based radar, which helps the observer to achieve parametric estimation of the target's on-orbit state. The embodiment of the present invention uses the GBISAR observation station to observe the on-orbit state of space targets, which can effectively track, identify and monitor space targets.

[0124] After observing the on-orbit state of the space target, the observed on-orbit state of the space target is inverted into the space observation model, which specifically includes the following steps:

[0125] S2-1. Decompose the on-orbit state of the space target observed by the GBISAR observation station into position change and attitude change.

[0126] S2-2. The position change of the space target is expressed in the HCS coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the UNW coordinate system and the HCS coordinate system.

[0127] The positional elements of a space object's posture describe its relative position to Earth, observation stations, and nearby targets while in orbit. From a ground-based perspective, these include latitude, longitude, and distance from the observation station. From an in-orbit perspective, these include altitude, velocity, and distance to nearby targets.

[0128] The attitude element of a space target's position describes the imaging angle presented by the space target when observed by an observation station while the space target is in orbit. From the perspective of ground observation, the space target's attitude can be ignored. However, from the perspective of the space target in orbit, starting from the UNW observation, the attitude element includes the content of the U-axis, N-axis, and W-axis.

[0129] S2-3. The posture change of the space target is expressed through the body coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the body coordinate system and the UNW coordinate system;

[0130] S2-4. Express the on-orbit status of space targets in the space observation model using the UNW coordinate system.

[0131] Based on the space observation model of the present invention, the use of HCS more directly connects to actual observations, dividing the work of position observations and conveniently recording the celestial position of space targets. The use of UNW and BFRF also more directly connects to actual observations, dividing the work of attitude observations and conveniently recording the attitude changes of space targets. The attitude change state equations are used to predict attitude changes, facilitating actual observations. Decomposing position and attitude changes allows for relatively convenient recording of space target observations within their respective coordinate systems. Furthermore, the equivalent decomposition of position and attitude changes allows for error correction using higher-precision algorithms within their respective coordinate systems, providing a way to increase overall observation accuracy and facilitating the next stage of three-dimensional motion description.

[0132] S3. Describe the attitude changes of space targets during on-orbit operation in the space observation model;

[0133] The decomposed attitude change can be defined by the coordinate system transformation relationship between the static UNW coordinate system and the dynamic BFRF. In view of the fact that the common attitude change of space targets in engineering practice is periodic spin adjustment, the embodiment of the present invention models the attitude change of space targets as a three-dimensional rotational motion.

[0134] Specifically, the space observation model describes the attitude changes of space targets during on-orbit operation, including:

[0135] The points of unit length of each coordinate axis in the UNW coordinate system are selected as target points; the three-dimensional rotation matrix is ​​constructed through the three-dimensional motion of the target point to express the attitude changes of the space target during on-orbit operation.

[0136] To describe the three-dimensional rotation of a space object, embodiments of the present invention define the following directions of three-dimensional motion: roll (when the X-axis is used as the rotation vector), pitch (when the Y-axis is used as the rotation vector), and yaw (when the Z-axis is used as the rotation vector). Rotational motion is measured using rotation angles, which are based on the UNW coordinate system and are the roll, pitch, and yaw angles.

[0137] like Figure 7 As shown, after establishing the connection between the body coordinate system and the UNW coordinate system, any point of a given space target, after any attitude change, can be given a unique corresponding point and its coordinates in the UNW coordinate system. However, for convenience, it is advisable to select unit vectors on the U, N, and W coordinate axes to represent the attitude change of the space target. Specifically, in the initial state, the points at the unit 1 position on the coordinate system are selected in the UNW coordinate system to form the initial coordinate matrix. After three-dimensional rotation, the coordinate matrix of the space target at any time can be obtained. The attitude of the space target can be uniquely determined through the coordinate matrix of the space target in the UNW coordinate system.

[0138] Based on the above definition, the attitude change of a space target during on-orbit operation is specifically expressed by the following coordinate matrix:

[0139]

[0140] in, represents the initial attitude of the space target when it is first observed, It represents the posture change of the space target at any time compared to the initial observation. Each element in the matrix represents the component generated by the target point on each coordinate axis after three-dimensional motion.

[0141] S4. Projecting the attitude changes of the space target during its on-orbit operation onto a two-dimensional plane through GBISAR imaging to obtain the ISAR imaging results of the space target;

[0142] Due to the different imaging projection planes in optical imaging and radar imaging, the dimensions retained when projecting into a two-dimensional image are different. According to the UNW coordinate system and coordinate matrix, this projection process is equivalent to retaining a coordinate matrix with a rank of 3 into a new coordinate matrix with a rank of 2. On the one hand, during radar imaging, the line of sight is parallel to the imaging projection plane, and the retained dimensions are U and N. On the other hand, during optical imaging, the line of sight is perpendicular to the imaging projection plane, and the retained dimensions are U and W. According to the definition of the coordinate matrix, the mathematical descriptions of the imaging projection matrix and the projected coordinate matrix can be given respectively:

[0143]

[0144] Through the imaging process of the space target's attitude changes, these two matrices construct the imaging projection equations for the space target's attitude changes. It is useful to introduce an intermediate recording result to describe the space target's attitude stored during the observation station's imaging of the space target. From the perspective of the radar observation system, this embodiment of the present invention selects GBISAR as the observation station for space observation. The imaging projection matrix and the projected coordinate matrix mentioned hereafter refer to the radar imaging results.

[0145] Finally, the two-dimensional range-Doppler (RD) imaging plane equations for the projection of the three-dimensional posture change of the space target are defined as follows:

[0146]

[0147] At this point, the space observation model of the embodiment of the present invention provides a complete model of the process from the attitude change of the space target to the ISAR image. Starting from the GBISAR observation station, the position change of the space target is captured and observed through the horizontal coordinate system, and the aligned UNW coordinate system and body coordinate system are established. The attitude change of the space target is observed through the UNW coordinate system, and it is projected to the RD plane according to the GBISAR imaging characteristics to obtain the ISAR image of the space target. This new space observation model has a clear division of labor and close connection. The connection with actual observations is strengthened, which provides a basis for the subsequent analysis of the pattern of space target attitude change and the development of feature design.

[0148] S5. Use Fourier series to perform polynomial decomposition on the ISAR imaging results of the space target to obtain the two-dimensional feature representation of the space target in the ISAR imaging results.

[0149] The attitude adjustment of a space target refers to adjusting the attitude, including maintaining a stable state, roll, pitch and yaw. It is worth pointing out that maintaining a stable state is also a kind of attitude adjustment, and its function is consistent with that of position adjustment and is no longer repeated. Roll, pitch and yaw constitute all the degrees of freedom of three-dimensional rotation, providing a basis for establishing a set of state equations for the attitude change of a space target. It is worth mentioning that the order of these three attitude adjustment methods does not affect the final attitude adjustment result, but it will change the mathematical form of the set of state equations. Starting from the space observation model of the embodiment of the present invention, the classification and interpretation of attitude and attitude adjustment are further discussed, and the three attitude adjustment categories of roll, pitch and yaw are refined, which makes innovative content for the classification of space target attitude. The discussion and understanding of attitude adjustment provide a basis for the subsequent feature design of space target attitude classification and the establishment of space target attitude fuzzy classification technology.

[0150] like Figure 7As shown, the embodiment of the present invention introduces Fourier series to perform polynomial decomposition of the attitude changes of the space target when it is in orbit, and retains the most significant first-order terms to approximately describe the attitude changes, thereby constructing a Fourier series-based space target attitude model based on a new space observation model.

[0151] Fourier series can decompose any periodic function, so we first need to discuss the periodicity and functional form of the attitude changes of space targets. On the one hand, any non-periodic function can be approximately considered to have an infinite period, and the attitude changes of space targets can be considered to be motions with such an infinite period, with a certain functional form. According to the new space observation model, this motion is the three-dimensional rotation of the space target and the circular motion after two-dimensional projection. Under the condition of a short observation sequence, the rotational and circular motions can be made fully periodic by extending the period. On the other hand, the attitude changes of space targets can be equivalently represented as the motion of specific points on the space target, and the trajectories of these points can be described as functional forms, including direct three-dimensional motion trajectories and projected two-dimensional motion trajectories. According to the new space observation model, the points can be selected as points with a unit length of 1 on each axis of the UNW coordinate system. The point motion is a three-dimensional rotation, and when projected onto the RD plane, it is a circular motion. Therefore, the attitude change of the space target can be equivalently expressed as three periodic circular motions on the imaging projection plane. By considering the observation sequence of a given shorter length as a periodic extension, Fourier series decomposition can be performed, and the attitude change of the space target can be approximately described by retaining a limited number of terms.

[0152] Specifically, the ISAR imaging results of the space target are decomposed into polynomials using Fourier series, which includes the following steps:

[0153] The ISAR imaging results of space targets are equivalently represented as the periodic circular motion of three target points in a plane, which is expressed using periodic functions x1(t), x2(t), and x3(t).

[0154] By superimposing high frequencies, two-dimensional circular motion can be decomposed into multiple harmonic circular motions. Based on the comprehensive formula, the description method of circular motion using complex exponentials is introduced to establish a polynomial decomposition of two-dimensional circular motion. Specifically, the embodiment of the present invention uses fast Fourier transform to decompose the motion trajectories of the three target points into a polynomial form:

[0155]

[0156] Where Ω0 is the angular frequency of circular motion, T is the corresponding rotation period of the target point, j is the imaginary number sign, k is the sequence number of the Fourier series, and t is time;

[0157] Keeping the polynomial to order 1 and removing the order 0 result, we get the following two-dimensional pose matrix:

[0158]

[0159] The polynomial decomposition of space target attitude changes based on Fourier series can be divided into three categories: 0th-order results, 1st-order results, and higher-order results, each with a clear physical meaning. The 0th-order term is commonly known as the DC term and can be considered to describe the position change of the space target. According to the new space observation model, horizontal observations account for the observation of the space target's position change. Therefore, under conditions where position change is not involved in space target attitude classification, the 0th-order term should be discarded. The 1st-order term is the fundamental frequency result and can be considered to describe the most significant component of the space target's attitude change. According to the new space observation model, this significant component is the space target's 3D rotation, which is imaged as 2D circular motion. As the basis for space target attitude classification, the 1st-order term should be retained. Higher-order terms, including 2nd-order and higher-order terms, can be considered to describe detailed components of the high-harmonic frequencies in the attitude change, including the space target's on-orbit jitter and imaging errors in the observation image. They do not contribute to classification and should therefore be discarded.

[0160] According to the properties of real even functions, the attitude changes of space targets displayed by the observed imaging in actual observations are considered to be real even and rewritten as trigonometric cosine functions. Using trigonometric cosine functions to rewrite the two-dimensional attitude matrix, the following two-dimensional cosine attitude matrix is ​​obtained:

[0161]

[0162] The amplitude term represents the radius of the circular motion trajectory in the RD plane. On the one hand, different attitude changes with different centers will produce two-dimensional circles with different radii and centers, so the pattern formed by all the circles can serve as a discriminative feature. On the other hand, different attitude changes of the target with the same center can produce different amplitude terms, so the amplitude term itself can also serve as a discriminative feature. In summary, the attitude change characteristics of space targets can be characterized based on the characteristic patterns of the two-dimensional circular motion trajectory generated by the target's key points in ISAR images.

[0163] S6. Perform fuzzy classification based on the two-dimensional feature representation of spatial targets.

[0164] In an embodiment of the present invention, fuzzy classification is performed based on the two-dimensional feature representation of the space target, specifically including the following steps:

[0165] S6-1. Design special feature vectors. Feature vectors are a type of multidimensional feature and are often used to construct feature spaces. For two-dimensional images, feature vectors can be geometric vectors. Feature vectors suitable for posture classification may be referred to as special feature vectors. The special feature vectors designed in this embodiment of the present invention are shown in Table 1 below:

[0166] Table 1

[0167]

[0168]

[0169] The following describes the special feature vectors involved in the embodiments of the present invention:

[0170] 1) Row eigenvector

[0171] The row eigenvector describes the reference benchmark of the azimuth dimension in the ISAR image. From the perspective of understanding the change in the attitude of a space target, the change in the angle between each component of the space target and a certain reference benchmark is a common feature. Therefore, an embodiment of the present invention proposes to construct a unit vector in the directional dimension as a special eigenvector as this reference benchmark to calculate the relevant angles. For example, the unit vector pointing from the upper left corner origin to the right in the ISAR image can be taken as the row eigenvector. As the first of the two reference benchmarks, the row eigenvector provides a way to calculate the angle between each eigenvector and the equidistant line by being parallel to the equidistant line in the ISAR imaging, and provides a reference for judging the change in the attitude of the space target. It is worth noting that the establishment of the row eigenvector as a reference benchmark is independent of the space target, but directly comes from the image imaging result, which provides a reference benchmark that is independent of observation errors and annotation errors.

[0172] 2) Roll eigenvector

[0173] The roll eigenvector describes the X-axis and its variations in the space target's body coordinate system. Construction Method: With the exception of Aura, which uses eigenvectors for key points 2 to 1, all other methods use eigenvectors for key points 3 to 6. As discussed in Section 3.2.3, the attitude change of a space target is equivalent to a three-dimensional rotational motion, and the three coordinate axes of the body coordinate system each describe a rotational motion in one dimension. The Y-axis is perpendicular to the orientation of the space target's body and perpendicular to the X-axis, thus orthogonally representing X-axis variations. Considering the geometry of the space target, the layout of the solar panels shares similar characteristics with the Y-axis. Therefore, constructing a special eigenvector describing the X-axis and its variations based on the key points of the solar panels is relatively reasonable. When selecting key points, it is desirable to construct eigenvectors with relatively large moduli and easy identification. Therefore, the roll eigenvector is constructed from the corner points on both sides of the solar panels, located near the head. It is worth noting that, due to the cross product orientation, the roll eigenvector is oriented from the right shoulder to the left shoulder. The roll eigenvector provides characteristic quantities such as modulus and angle, providing a means for determining the attitude of the space target.

[0174] 3) Pitch eigenvector

[0175] The pitch eigenvector describes the Y-axis and its variations in the space target's body coordinate system. Construction Method: With the exception of Aura, which uses eigenvectors for key points 6 to 5, all other methods use eigenvectors for key points 10 to 9. Similarly, the X-axis is parallel to the space target's body orientation and perpendicular to the Y-axis, thus orthogonally representing Y-axis variations. Considering the geometry of the space target, the body layout characteristics are similar to those of the X-axis, making it relatively reasonable to construct a specific eigenvector describing the Y-axis and its variations based on the body's key points. It's worth noting that, unlike solar panels, the body of a space target is generally a non-negligible three-dimensional object with a wide variety of shapes. Due to annotation difficulty and error considerations, only the foot-to-head orientation is selected to construct the pitch eigenvector. The head and foot annotations are based on a specific side of the space target. In this section, the right front side of the space target is chosen. The pitch eigenvector provides characteristic quantities such as modulus and angle, providing a means for determining the pose of the space target.

[0176] 4) Yaw eigenvector

[0177] The yaw eigenvector describes the Z axis and its changes in the coordinate system of the space target body. The characteristic of the Z axis is that the direction is determined by the right-hand rule and the cross product. Considering the motion characteristics of the space target, when the Z axis is selected as the rotation vector, the eigenvector with obvious changes in modulus and direction in the eigenvector composed of the key points of the space target is used as a special eigenvector. Therefore, the embodiment of the present invention selects the foot pointing to the shoulder to construct a special eigenvector. It is worth mentioning that, due to the difficulty and error of annotation, the right foot of the space target is selected to point to the right shoulder to establish a special eigenvector. The yaw eigenvector can provide characteristic quantities such as modulus and angle, providing a way to judge the posture of the space target.

[0178] 5) Column feature vector

[0179] The column eigenvector describes the reference for the range dimension in ISAR images. As the other of the two references, the column eigenvector, by being parallel to the iso-Doppler lines in ISAR imaging, provides a method for calculating the angle between each eigenvector and the iso-Doppler line, and thus provides a reference for determining the attitude changes of spatial targets. For example, a unit vector pointing downward from the origin in the upper left corner of the ISAR image can be used. It is worth noting that the column eigenvector, as a reference, is established independently of the spatial target and is derived directly from the image results. This provides a reference that is independent of observation errors and annotation errors.

[0180] In an embodiment of the present invention, a special feature vector refers to a feature vector that provides a characteristic quantity for classifying the attitude of a space target. A special feature vector is constructed based on the key point standard according to the pattern of the space target's attitude change. Two of the feature vectors used as reference bases are unrelated to the change in the space target's attitude, while the remaining three feature vectors describe each coordinate axis in the space target's body coordinate system and the resulting rotational motion, respectively, and represent feature changes based on ISAR imaging. The features that the special feature vector can provide provide a basis for subsequently establishing criteria for determining changes in the space target's attitude.

[0181] S6-2. Design a criterion feature vector; the criterion feature vector is used to match various posture change categories of space targets.

[0182] In the embodiments of the present invention, a criterion refers to a basis for classifying the posture of a space target, obtained through calculations and transformations based on a special feature vector. By way of example, the embodiments of the present invention provide two criterion feature vectors: a first criterion and a second criterion. These criterion vectors are used from different perspectives to generate the required feature quantities and establish the criteria for the fuzzy classification technology for the posture of a space target.

[0183] The first criterion is shown in Table 2:

[0184] Table 2

[0185]

[0186]

[0187] The first criterion involved in the embodiment of the present invention is described below:

[0188] 1) Differentiate between roll, pitch, and yaw judgments

[0189] Regarding the attitude changes of space targets, according to the attitude model in 3.3.2, they are classified into three types: roll, pitch, and yaw. The type of attitude change determines the type of judgment, which is based on the criteria. Therefore, the feature design of the criteria is based on the feature quantities provided by the special feature vectors and is designed for each of the three types of attitude changes. In particular, calculations such as dot products, cross products, and modulo are used to extract feature quantities such as length, angle, and direction, and then establish criteria based on these feature quantities. From a practical design perspective, compared to establishing criteria directly derived from a single operation, some criteria are more suitable for constructing using a system of equations. This ensures non-overlapping partitioning and enables the reuse of some operations. It is worth noting that conventional attitude classification divides attitude into a set of stable and unstable states. Refining the judgments and criteria can not only enhance the interpretation of the classification semantics of attitude classification, increasing its value in practical applications, but also provide estimates of some attitude parameters during classification, bridging attitude classification and attitude estimation. From the perspective of practical use, the steps of posture classification should be: first determine the subdivided posture subcategories, and then summarize and classify them into the original larger posture categories.

[0190] Based on the above discussion, the definition of the criterion can be given. It is worth noting that the definition of the criterion is divided into Definition 1 and Definition 2. On the one hand, this distinction can reduce the difficulty of criterion design while ensuring the orthogonality of the criterion. On the other hand, by designing reusable calculation content in Definition 2, the overall calculation scale of the criterion is reduced. From the perspective of the composition of the criterion, the criterion is a one-dimensional sequence of the same length as the image sequence, and the value of each component comes from the calculation of the coordinates obtained by the key point annotation of the current single image. Therefore, it can be considered that the first criterion is a one-dimensional time domain feature sequence, which provides a basis for the subsequent establishment of the second criterion.

[0191] Having completed the feature design for the first criterion, we now discuss the second criterion. On the one hand, in actual space observations, space target posture classification faces the complexities of posture variations and the randomness of observation anomalies. Using only the first criterion to determine and classify space target posture variations is far from sufficient. On the other hand, the first criterion still contains untapped information, and further extracting this information is valuable in increasing classification accuracy and reducing classification costs. Therefore, it is necessary to use transformation methods such as FFT to process the first criterion to generate the second criterion.

[0192] The second criterion is shown in Table 3:

[0193] Table 3

[0194]

[0195]

[0196] The second criterion involved in the embodiment of the present invention is described below:

[0197] 1) Reprocessing based on the first criterion

[0198] The second criterion processes the discrete sequence formed by the first criterion. The first criterion is further processed using various methods to extract further information, which is then used to determine the change in the spatial target's posture. It is worth noting that the second criterion can be processed using more than the four methods mentioned above. Methods suitable for processing discrete sequences, such as spline interpolation, wavelet transform, and FIR, can theoretically also serve as second criterion processing methods.

[0199] 2) Simple processing and transformation solutions

[0200] Simple processing, as opposed to transformation schemes, refers to processing the first criterion in the time domain, including operations such as interpolation, sampling, fitting, and time-domain filtering. The simple processing scheme employed in the embodiments of the present invention is a polynomial fitting scheme. On the one hand, simple processing can enhance the prominent properties required by the first criterion and reduce unnecessary properties. For example, when the first criterion only requires monotonicity and periodicity for determination, polynomial fitting can be used to remove burrs from the sequence, facilitating subsequent extraction of periodicity and observation of monotonicity. On the other hand, time-domain processing is affected by the time-domain performance of the first criterion. In some cases, the time-domain performance of the first criterion is not suitable for determining the attitude of a space target. In these cases, simple processing is not suitable as a processing scheme for the second criterion. Therefore, the transformation scheme provides a basis for understanding and utilizing the first criterion in the transform domain. Commonly used transforms include Fourier transform, Laden transform, and wavelet transform.

[0201] 3) Machine Learning Solutions

[0202] Solutions based on surface learning generally only involve the above two processing solutions, while further utilization methods involve machine learning. The embodiment of the present invention selects clustering technology as the processing solution for the second criterion. The idea of ​​clustering is to establish a feature space through the first criterion, and then perform clustering in the feature space to complete the posture transformation judgment and classification tasks of the space target. On the one hand, clustering does not require obtaining the classification of the space target posture in advance, but self-learns the posture category during the clustering process, which has advantages in segmentation and understanding complex situations. On the other hand, clustering depends on the establishment of the feature space, and the selection of feature quantities determines the final clustering effect. In addition, the clustering solution can also adopt the method of multiple clustering, that is, the idea of ​​clustering-division-reclustering, and through detailed classification, a posture classification method for space targets under the conditions of processing massive data is established.

[0203] This completes the construction of a single-decision posture classification method. By designing a special feature vector and constructing the first and second criteria, a scheme for determining and classifying the posture of space targets in a single-decision situation has been established. This provides a foundation for proposing a fuzzy classification technique for space target postures and establishing a surface learning-based scheme.

[0204] S6-3. Calculate the fuzzy membership of the space target in each posture category based on the variance of the space target in each criterion feature vector;

[0205] Before establishing the fuzzy classification method, the present embodiment first explains the relevant content of fuzzy set theory. Define the membership function as any mapping from the domain X to the number domain [0,1]. Then, the membership function gives the fuzzy membership degree of any element x of X to the fuzzy set A:

[0206] μ A :X→[0,1];

[0207] x→μ A (x);

[0208] According to the definition of membership function and fuzzy membership, a fuzzy set A on X can be determined:

[0209] A={(x,μ A (x))|x∈X};

[0210] Based on fuzzy set theory, the embodiment of the invention provides posture classification results in the form of fuzzy sets, thereby establishing fuzzy classification.

[0211] This embodiment of the present invention defines fuzzy membership for target pose fuzzy classification based on saliency. Saliency is defined as the severity of target pose changes in ISAR images, and severity can be expressed as the value of the criterion variance. This embodiment of the present invention defines the saliency of a pose category as the sum of the variances of the corresponding criteria, and designs a membership function based on the variance:

[0212]

[0213] where σ i is the variance of the i-th criterion, σ j (x) is the variance of the j-th criterion of posture x, w A is the final weight.

[0214] S6-4. Generate a fuzzy set according to the fuzzy membership of the space target in each posture category, and use the fuzzy set as the fuzzy classification result of the space target.

[0215] In practice, GBISAR will establish a continuous tracking observation database in the form of ISAR image sequences, and combine it with inversion and restoration methods to generate a three-dimensional coordinate sequence of space targets. Based on this, the embodiment of the present invention further processes the coordinate sequence to achieve fuzzy classification of refined attitude categories.

[0216] In some embodiments, various anomalies may occur during GBISAR space observations and imaging. Because anomaly detection is a deep and extensive topic, and the situations encountered during GBISAR observations and imaging of space targets are even more complex and diverse, the present invention only selects two anomalies for illustration: anomalies where the space target appears out of focus during GBISAR imaging and anomalies caused by the space target itself being occluded during GBISAR imaging. Solutions for these anomalies are discussed.

[0217] Defocus anomalies: Defocus anomalies are unique to radar observation systems and are common during radar imaging. From a GBISAR perspective, ISAR imaging results intermittently appear as blurred point clouds during continuous observations, with the degree of blur varying with the target's motion during imaging. When the blur is mild and the target's shape is still fully visible, the impact of the mass phenomenon on pose estimation and classification performance is manageable. However, when the blur is severe and the target image is distorted, performance deteriorates rapidly.

[0218] From the perspective of ISAR imaging, achieving high-precision imaging requires focusing the target echo. Motion compensation and other related compensation methods in ISAR imaging aim to address this issue. For example, when a target moves relative to the ISAR, it also undergoes translational motion, resulting in echo misalignment and additional Doppler phase in the range dimension. Imaging without translational compensation will result in low-precision ISAR images due to defocus. Regarding GBISAR space observations, on the one hand, the motion of space targets changes during their return, from range migration to their own jitter, all of which affect GBISAR imaging. Furthermore, the complex environmental conditions encountered by space targets while in orbit also produce complex and diverse motion changes. On the other hand, compensation algorithms cannot fully compensate for these complex variations and often perform poorly in some situations, making image defocus unavoidable. On the one hand, defocus occurs intermittently and continuously, and there will always be more or less out-of-focus images in a set of image sequences. On the other hand, the degree of defocus is distinctive, that is, among the out-of-focus images, completely blurred images are only a minority, and the vast majority of blurred images can still be reduced in blur through image processing and other technologies.

[0219] Occlusion anomalies: Compared to defocus anomalies, occlusion anomalies are encountered in both radar and optical imaging processes. Their characteristic is that they are unrelated to the imaging mechanism and are solely related to the imaging perspective of the imager relative to the target during the imaging process. Occlusion anomalies can be divided into two types depending on the number of targets in the image: occlusion between components of a single target, and occlusion between multiple targets. This section only covers occlusion between components of a single target and provides a preliminary discussion. The unavoidable conditions for occlusion anomalies in imaging are that the target is non-transparent in optical imaging and that the target is non-electromagnetic reflective in radar imaging.

[0220] From the perspective of ISAR imaging, target occlusion occurs in the range and azimuth dimensions, unlike optical imaging. Therefore, when imaging the same target at the same moment, optical and radar imaging utilize different imaging dimensions, reducing the probability of target occlusion anomalies in both imaging modes. Based on this approach, joint radar-optical observations offer advantages in reducing occlusion anomalies, and related research is in full swing. From an application perspective, various estimation and classification algorithms exhibit varying robustness to occlusion anomalies, but are generally adaptable. On the one hand, most processing algorithms incorporate design considerations for occlusion anomalies, such as redundant features. On the other hand, given the constantly changing motion of a target, the frequency, degree, range, and duration of occlusion anomalies will constantly change, and the occlusions that actually cause impact only account for a minority of all occlusion anomalies.

[0221] In the embodiment of the present invention, the abnormal situation is effectively handled by the first criterion and the second criterion. On the one hand, the normalization of the first criterion can reduce the influence of noise factors in judgment and classification, thereby providing an exception handling capability for noise anomalies represented by defocus anomalies. On the other hand, the second criterion has time domain processing and frequency domain processing schemes, and by providing the functions of time domain filtering and frequency domain filtering, it provides an exception handling capability for noise addition, phase distortion, scaling distortion and jitter generated by defocus anomalies and occlusion anomalies. Therefore, the first criterion and the second criterion functional modules give the space target posture fuzzy classification scheme the main exception handling capability. In some other schemes, the exception handling capability of the scheme can also be improved from the perspectives of key point labeling and feature vector design of space targets, such as removing the key point coordinates of other irrelevant factors through limited key point labeling, reducing the occurrence of anomalies through isolation, etc.

[0222] The following is a specific example of fuzzy classification of space targets. The three-dimensional point trace, two-dimensional point trace, special feature vector and criterion feature vector of the space target are respectively as follows: Figure 10 、 11 , 12, and 13.

[0223] In this embodiment, the imaging angle includes an azimuth angle and a polar angle based on spherical coordinates, and a value range is specified.

[0224]

[0225] On the one hand, when the control azimuth angle remains unchanged, the same rotation axis will produce different attitude category components as the polar angle changes, which should cause the classification results to change significantly. It can be seen that as the polar angle increases, the yaw component of the same attitude adjustment gradually shifts to the roll component, while the pitch component remains basically unchanged. The fuzzy classification results when the control azimuth angle remains unchanged are shown in Table 4:

[0226] Table 4

[0227]

[0228] On the other hand, when the control polar angle remains unchanged, the two-dimensional projection of the target pose remains essentially unchanged, and thus the classification result should remain essentially unchanged. It can be seen that as the polar angle increases, the various posture components of the same pose remain essentially unchanged. The fuzzy classification results when the control polar angle remains unchanged are shown in Table 5:

[0229] Table 5

[0230]

[0231] The image sequence length is equal to the length of the input coordinate sequence. On the one hand, a shorter length results in insufficient information, affecting classification accuracy. On the other hand, a longer length may lead to information redundancy and affect real-time classification. As can be seen, the minimum required input sequence length for the algorithm tested is 6 frames. Above this lower bound, the algorithm performs well and maintains stability. The fuzzy classification results for controlled sequence lengths are shown in Table 6:

[0232]

[0233] The embodiment of the present invention defines anomalies as defocus anomalies and occlusion anomalies. Defocus anomaly refers to the anomaly that the space target appears to be out of focus during GBISAR imaging, which is related to the range migration, effective echo and relative motion speed of the space target. The embodiment of the present invention uses an equivalent experiment when the signal-to-noise ratio accounts for 10% (10dB) and 90% (1dB) of the noise level. It can be seen that the algorithm performs well at the conventional noise level of 10dB in engineering practice, and the algorithm shows effective anomaly handling capabilities at the common sudden defocus level of 1dB. The fuzzy classification results when controlling the defocus anomaly level are shown in Table 7:

[0234] Table 7

[0235]

[0236] Occlusion anomalies refer to anomalies caused by the occlusion of a space target during GBISAR imaging. These anomalies are related to imaging angle, back-face effects, and mutual occlusion between components. This embodiment of the present invention employs random loss of image sequences for equivalent experiments. It can be seen that the impact of occlusion anomalies is similar to that of image sequence length. When the minimum input sequence length is guaranteed, the algorithm is capable of handling occlusion anomalies. The fuzzy classification results when controlling the occlusion anomaly level are shown in Table 8:

[0237] Table 8

[0238]

[0239] The embodiment of the present invention combines the space observation model, the space target attitude change pattern, feature design and fuzzy set theory to propose a fuzzy classification method for space target attitude. Compared with the traditional inversion and restoration method, the fuzzy classification method replaces the parameter estimation capability with the target data further processing and interpretation capability, and achieves high interpretability of the target attitude adjustment intention interpretation based on the fuzzy set. Compared with the traditional feature matching method, the fuzzy classification method does not rely on the continuous tracking observation database on the one hand, and achieves high real-time performance through extremely low computing scale. On the other hand, it enhances the adaptability to the diversity and meticulousness of space target attitude classification tasks based on refined categories and fuzzy classification. A series of tests based on real space target data verified the effectiveness of the fuzzy classification method, thereby adding a new classification method to the space target attitude classification technology based on the GBISAR platform.

[0240] Figure 14 : is a structural diagram of the electronic device proposed in the second embodiment of the present invention. The memory of this embodiment stores program instructions for implementing the space target attitude fuzzy classification method based on GBISAR imaging of any of the above embodiments. The processor is used to execute the program instructions stored in the memory to perform space target attitude fuzzy classification based on GBISAR imaging. The processor can also be called a CPU (Central Processing Unit). The processor may be an integrated circuit chip with signal processing capabilities. The processor can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0241] The contents of the method in the first embodiment of the present invention are all applicable to this electronic device embodiment. The functions specifically implemented by this electronic device embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method.

[0242] Figure 15 This is a structural diagram of the computer-readable storage medium of the third embodiment of the present invention. The computer-readable storage medium of the fourth embodiment of the present invention stores program instructions that can implement the above-mentioned space target attitude fuzzy classification method based on GBISAR imaging, wherein the program instructions can be stored in the above-mentioned storage medium in the form of a software product, including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to execute all or part of the steps of the methods of each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0243] The contents of the method in the first embodiment of the present invention are all applicable to this computer-readable storage medium embodiment. The functions specifically implemented by this computer-readable storage medium embodiment are the same as those of the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method.

[0244] This embodiment further provides a computer program product. When the computer program product is run on a computer, the computer is caused to execute the above-mentioned related steps to implement the space target attitude fuzzy classification method based on GBISAR imaging provided by the above embodiment.

[0245] Those skilled in the art will appreciate that the modules in the devices in the embodiments of the present invention can be adaptively changed and set in one or more devices different from the embodiments. The modules or units or components in the embodiments of the present invention can be combined into one module or unit or component, and in addition they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including corresponding claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including corresponding claims, abstracts and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0246] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0247] In addition, each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. In particular, for embodiments such as devices and equipment, since they are basically similar to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The embodiments of the devices and equipment described above are merely schematic, wherein the modules, units, etc. described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed to multiple places, such as nodes in a system network. Specifically, some or all of the modules and units may be selected according to actual needs to achieve the purpose of the above-mentioned embodiment scheme. Those skilled in the art can understand and implement it without paying any creative work.

[0248] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0249] In addition, the terms "first" and "second" used in the embodiments of the present invention are only used for descriptive purposes and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in this embodiment. Therefore, the features defined by the terms "first" and "second" in the embodiments of the present invention can explicitly or implicitly indicate that the embodiment includes at least one of such features. In the description of the present invention, the word "plurality" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0250] In the embodiments of the present invention, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or apparatus comprising the element. In addition, components, features, and elements with the same name in different embodiments of the present invention may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.

[0251] Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention, and those of ordinary skill in the art may change, modify, replace, and modify the above embodiments within the scope of the present invention. Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the present invention. This application is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the present invention. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present invention are indicated by the claims below.

Claims

1. A fuzzy classification method for space target attitude based on GBISAR imaging, characterized in that: The following steps are involved: Build a space observation model based on the GBISAR observation station and the space targets to be observed; Observing the on-orbit state of the space target using the GBISAR observation station, and inverting the observed on-orbit state of the space target into the space observation model; Describing, in the space observation model, attitude changes of the space target during on-orbit operation; The attitude changes of the space target during its on-orbit operation are projected onto a two-dimensional plane through GBISAR imaging to obtain an ISAR imaging result of the space target; Performing polynomial decomposition on the ISAR imaging result of the space target using Fourier series to obtain a two-dimensional feature representation of the space target in the ISAR imaging result; Fuzzy classification is performed based on the two-dimensional feature representation of the space target.

2. A method for fuzzy classification of space target posture based on GBISAR imaging according to claim 1, characterized in that: The space observation model is constructed based on the GBISAR observation station and the space target to be observed, and specifically includes the following steps: With the GBISAR observation station as the origin, the HCS coordinate system is established ; where γ represents the radar line of sight LOS, represents the horizontal latitude, and θ represents the horizontal longitude; With the space target to be observed as the origin, establish the UNW coordinate system (U, W, N) and the body coordinate system (X, Y, Z); where U represents the velocity vector direction of the orbit, N represents the normal direction in the orbital plane that is perpendicular to the U direction and points outside the orbit, and the W direction forms a right-handed system with the U and N directions; X represents the roll direction of the space target's rotation, Y represents the pitch direction of the space target's rotation, and the Z axis represents the yaw direction of the space target's rotation; Establish the connection between the UNW coordinate system and the HCS coordinate system; Establish the connection between the body coordinate system and the UNW coordinate system; The connection relationship between the UNW coordinate system and the HCS coordinate system is expressed by the following coordinate matrix: The connection relationship between the body coordinate system and the UNW coordinate system is expressed by the following coordinate matrix:

3. A space target attitude fuzzy classification method based on GBISAR imaging according to claim 2, characterized in that: The space target consists of a head, a solar panel part, a pitch feature vector part, a reference area part, a body part and a foot; The solar panel portion and the body portion are used to determine the current orientation of the space target; The head and the foot are used to cooperate with the UNW coordinate system and the body coordinate system to calculate the current orientation of the space target; The pitch eigenvector portion is used to represent the current orientation of the space target; The reference area portion is used to provide an external reference benchmark for the current orientation of the space target.

4. A method for fuzzy classification of space target posture based on GBISAR imaging according to claim 2, characterized in that: The inversion of the observed on-orbit state of the space target into the space observation model specifically includes the following steps: Decompose the on-orbit state of space targets observed by the GBISAR observation station into position change and attitude change; The position change of the space target is expressed in the HCS coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the UNW coordinate system and the HCS coordinate system; The attitude change of the space target is expressed in the body coordinate system, and is inverted into the UNW coordinate system through the connection relationship between the body coordinate system and the UNW coordinate system; The on-orbit status of space targets is expressed in the space observation model using the UNW coordinate system.

5. A method for fuzzy classification of space target posture based on GBISAR imaging according to claim 2, characterized in that: The description of the attitude change of the space target during on-orbit operation in the space observation model specifically includes: A point per unit length of each coordinate axis in the UNW coordinate system is selected as a target point; a three-dimensional rotation matrix is ​​constructed through the three-dimensional motion of the target point to express the attitude change of the space target during on-orbit operation; The attitude change of the space target during on-orbit operation is specifically expressed by the following coordinate matrix: in, represents the initial attitude of the space target when it is first observed, It represents the posture change of the space target at any time compared to the initial observation. Each element in the matrix represents the component generated by the target point on each coordinate axis after three-dimensional motion.

6. A method for fuzzy classification of space target posture based on GBISAR imaging according to claim 5, characterized in that: Projecting the attitude changes of the space target during on-orbit operation onto a two-dimensional plane through GBISAR imaging specifically includes the following steps: According to the radar line of sight direction during GBISAR imaging, the preserved dimension of the three-dimensional rotation matrix is ​​determined and the dimension-preserving matrix is ​​constructed. The attitude change of the space target during on-orbit operation is multiplied by the dimension-preserving matrix to obtain the ISAR imaging result of the space target.

7. The method for fuzzy classification of space target posture based on GBISAR imaging according to claim 1, characterized in that: The polynomial decomposition of the ISAR imaging result of the space target using the Fourier series specifically includes the following steps: The ISAR imaging result of the space target is equivalently represented as the periodic circular motion generated by three target points in a plane, and is expressed using periodic functions x1(t), x2(t), and x3(t); The motion trajectories of the three target points are decomposed into polynomial form using fast Fourier transform: Where Ω0 is the angular frequency of circular motion, T is the corresponding rotation period of the target point, j is the sign of the imaginary number, k is the sequence number of the Fourier series, and t is time; Keeping the polynomial to order 1 and removing the order 0 result, we get the following two-dimensional attitude matrix: Rewrite the two-dimensional attitude matrix using the trigonometric cosine function to obtain the following two-dimensional cosine attitude matrix: The two-dimensional cosine attitude matrix is ​​used as a two-dimensional feature representation of the space target in the ISAR imaging result.

8. A method for fuzzy classification of space target posture based on GBISAR imaging according to claim 7, characterized in that: The fuzzy classification based on the two-dimensional feature representation of the space target specifically includes the following steps: Designing special eigenvectors; the special eigenvectors include row eigenvectors, pitch eigenvectors, roll eigenvectors, yaw eigenvectors, and column eigenvectors; wherein the row eigenvectors and column eigenvectors are used to describe the reference benchmark of the range dimension in the ISAR imaging results; the pitch eigenvectors, roll eigenvectors, and yaw eigenvectors are used to describe the attitude change of the space target; Designing a criterion feature vector; the criterion feature vector is used to match various posture change categories of the space target; According to the variance of the space target on each criterion feature vector, the fuzzy membership of the space target in each posture category is calculated; A fuzzy set is generated according to the fuzzy membership of the space target in each posture category, and the fuzzy set is used as a fuzzy classification result of the space target.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement a fuzzy classification method for space target posture based on GBISAR imaging according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The storage medium stores a program, and the program is executed by a processor to implement a fuzzy classification method for space target posture based on GBISAR imaging according to any one of claims 1 to 8.

Images