A Multi-Target Tracking Method, System, Electronic Device and Medium for Over-the-Horizon Radar

Abstract

The present invention provides a method, system, electronic device and medium for multi-target tracking of over-the-horizon radar, belonging to the field of multi-target tracking. The method includes: establishing an observation model based on the ionospheric heights of the transmitting side and the receiving side; mapping each measurement information received by the over-the-horizon radar at the current moment to the target state space, and associating each measurement information with the observation model to obtain a measurement-model associated random finite set; according to the measurement-model associated random finite set at the current moment and the state information of the existing targets at the previous moment, using the Gaussian mixture probability hypothesis density algorithm to determine the state information of the existing targets at the current moment so as to track the existing targets; according to the measurement-model associated random finite sets at the current moment and the previous moment, using the time-extended target detection algorithm to determine the random finite set of the state of the newly born targets so as to track the newly born targets. The present invention realizes the tracking of multiple targets with multiple observation paths under dense clutter and reduces the computational burden.

Description

Technical Field

[0001] The present invention relates to the field of multi-target tracking, and particularly to a multi-target tracking method, system, electronic device and medium for over-the-horizon radar. Background Art

[0002] Multi-target tracking technology is one of the key research directions in the field of target tracking and has been widely applied in civilian and military fields. In complex tracking scenarios with clutter, target disappearance and target emergence, multi-target tracking technology accurately estimates the states of multiple targets through sensor observations and prior information of multiple targets, thereby providing necessary information for subsequent decision-making control and other tasks. Therefore, it is necessary to conduct research on multi-target tracking methods.

[0003] When observing and tracking maritime targets, since maritime targets are usually far from the observation base station, over-the-horizon radar is generally used for observation and tracking. The over-the-horizon radar obtains target observations by reflecting off the ionosphere in the atmosphere. However, there are multiple ionospheres with different heights in the atmosphere, so the over-the-horizon radar has multiple observation paths. This will result in the situation where the same target corresponds to multiple measurements from different observation paths. At the same time, the receiver of the over-the-horizon radar will receive measurements from all observation paths, but the observation paths generated by the measurements cannot be directly obtained. Therefore, compared with single-observation-path multi-target tracking that only needs to solve the "measurement-target" data association problem, multi-observation-path multi-target tracking needs to solve the three-dimensional data association problem of "measurement-observation model-target". Achieving the optimal three-dimensional data association of "measurement-observation model-target" will cause a huge computational burden, and the computational burden increases in a quasi-exponential growth trend with the increase in the number of observation models, the number of targets, and the number of clutters, which is difficult to achieve in practical applications. Therefore, conducting research on multi-observation-path multi-target tracking methods with lower computational complexity has important theoretical value and practical engineering application value.

[0004] The commonly used multi-target tracking methods mainly include data association and random finite set. Among them, the method based on random finite set is a probability-based method. Compared with the data association method that requires complex data association calculations, it can effectively reduce the computational amount, so it has become a research hotspot in the field of multi-target tracking. However, the multi-target tracking method based on random finite set is proposed for single-observation-path scenarios and can only solve the "measurement-target" data association and is not applicable to multi-detection-path situations.

[0005] Based on the above problems, there is an urgent need for a multi-target tracking method that can be applied to multi-detection-path scenarios. Summary of the Invention

[0006] The object of the present invention is to provide a method, a system, an electronic device and a medium for multi-target tracking of an over-the-horizon radar, which can achieve multi-observation-path multi-target tracking under dense clutter and reduce the computational burden.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] An over-the-horizon radar multi-target tracking method, comprising:

[0009] Obtaining the ionospheric height on the transmitting side and the ionospheric height on the receiving side of each observation path of the over-the-horizon radar;

[0010] For any observation path, an observation model corresponding to the observation path is established according to the ionospheric height on the transmitting side and the ionospheric height on the receiving side of the observation path;

[0011] Obtaining in real time a set of measurement information received by the over-the-horizon radar; the set of measurement information includes a plurality of pieces of measurement information;

[0012] Mapping each piece of measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to the target state space, and associating each piece of measurement information with the observation model, to obtain a measurement-model associated random finite set at the current moment; the measurement-model associated random finite set includes each piece of measurement information and the observation model corresponding to each piece of measurement information;

[0013] According to the measurement-model associated random finite set at the current moment and the state information of the existing targets at the previous moment, filtering is performed by using the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment, to determine the state information of the existing targets at the current moment, so as to track the existing targets;

[0014] According to the measurement-model associated random finite set at the current moment and the measurement-model associated random finite set at the previous moment, a time-extended target detection algorithm is used to determine a random finite set of newborn target states, so as to track the newborn targets; the newborn targets become existing targets at the next moment.

[0015] Optionally, the observation model is:

[0016]

[0017] Wherein, is the observation vector of the jth observation path, is the observation equation of the observation model corresponding to the jth observation path, x t is the target state vector at time t, is the observation noise of the observation model corresponding to the jth observation path, and Ω is the number of observation paths.

[0018] Optionally, the measurement information is mapped to the target state space using the following formula:

[0019]

[0020] where x is the target state vector after mapping the measurement information to the target state space, l1 is the first intermediate variable, l2 is the second intermediate variable, l2 = r - l1, r is half of the observed path distance, H t is the ionospheric height on the transmitting side, H r is the ionospheric height on the receiving side, D is the distance between the transmitter and the receiver of the over-the-horizon radar, is the deviation angle between the distance from the receiver to the ionosphere on the receiving side and the Y-axis, ρ is the distance of the target from the coordinate origin, is the change rate of the distance ρ of the target from the coordinate origin, θ is the angle by which the target deviates from the polar coordinate X-axis, is the measurement information, is the first derivative of r.

[0021] Optionally, the measurement-model associated random finite set at time t is:

[0022]

[0023] where Z L,t is the measurement-model associated random finite set at time t, is the measurement random finite set corresponding to the j-th observed path observation model at time t, z i is the i-th measurement information, Z j,t is the observation vector random finite set matching the j-th observed path observation model at time t, and Ω is the number of observed paths.

[0024] Optionally, based on the measurement-model associated random finite set at the current time and the state information of the existing targets at the previous time, the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current time are used for filtering to determine the state information of the existing targets at the current time, specifically including:

[0025] Based on the state information of the existing targets at the previous time, the prediction step of the Gaussian mixture probability hypothesis density algorithm is used to predict the state information of the existing targets and the newly born targets at the current time, obtaining the predicted state of the targets at the current time;

[0026] Based on the measurement-model associated random finite set at the current time, the update step of the Gaussian mixture probability hypothesis density algorithm is used to correct the predicted state of the targets at the current time, obtaining the state information of the existing targets at the current time.

[0027] Optionally, according to the measurement-model association random finite set at the current moment and the measurement-model association random finite set at the previous moment, the time-extended target detection algorithm is adopted to determine the random finite set of the states of newly born targets, specifically including:

[0028] Map the measurement information in the measurement-model association random finite set at the current moment to the target state space to obtain the intermediate variable at the current moment;

[0029] Determine the finite set of suspected newly born targets at the current moment according to the intermediate variable at the current moment;

[0030] Map the measurement information in the measurement-model association random finite set at the previous moment to the target state space to obtain the intermediate variable at the previous moment;

[0031] Determine the finite set of suspected newly born targets at the previous moment according to the intermediate variable at the previous moment;

[0032] Determine the random finite set of the states of newly born targets according to the finite set of suspected newly born targets at the current moment and the finite set of suspected newly born targets at the previous moment.

[0033] Optionally, the following formula is adopted to determine the random finite set of the states of newly born targets at time t:

[0034] X B,t ' = {x b,t | if || x b,t - f t (x b,t-1 ) || < δ, x b,t ∈ X B,t , x b,t-1 ∈ X B,t-1};

[0035] Wherein, X B,t ' is the random finite set of the states of newly born targets at time t, x b,t is the state vector of the newly born target at time t, x b,t-1 is the state vector of the newly born target at time t - 1, f t () is the target motion equation, δ is a preset parameter, X B,t is the finite set of suspected newly born targets at the current moment, and X B,t-1 is the finite set of suspected newly born targets at the previous moment.

[0036] To achieve the above object, the present invention also provides the following solution:

[0037] An over-the-horizon radar multi-target tracking system, including:

[0038] An altitude acquisition unit, configured to acquire the ionospheric altitude on the emission side and the ionospheric altitude on the reception side on each observation path of the over-the-horizon radar;

[0039] A model establishment unit, connected to the height acquisition unit, is configured to establish an observation model corresponding to any observation path according to the ionospheric height on the emission side and the ionospheric height on the reception side on the observation path.

[0040] A measurement acquisition unit is configured to acquire in real time a set of measurement information received by an over-the-horizon radar; the set of measurement information includes a plurality of pieces of measurement information.

[0041] An association unit, connected to the model establishment unit and the measurement acquisition unit respectively, is configured to map each piece of measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to a target state space, and associate each piece of measurement information with the observation model to obtain a measurement-model association random finite set at the current moment; the measurement-model association random finite set includes each piece of measurement information and the observation model corresponding to each piece of measurement information.

[0042] A first tracking unit, connected to the association unit, is configured to perform filtering according to the measurement-model association random finite set at the current moment and the state information of existing targets at the previous moment, and use the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment to determine the state information of existing targets at the current moment, so as to track existing targets.

[0043] A second tracking unit, connected to the association unit, is configured to determine a random finite set of newborn target states according to the measurement-model association random finite set at the current moment and the measurement-model association random finite set at the previous moment, and use the time-extended target detection algorithm to track newborn targets; the newborn targets become existing targets at the next moment.

[0044] To achieve the above object, the present invention also provides the following solution:

[0045] An electronic device includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program so that the electronic device executes the above-mentioned over-the-horizon radar multi-target tracking method.

[0046] To achieve the above object, the present invention also provides the following solution:

[0047] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned over-the-horizon radar multi-target tracking method is implemented.

[0048] According to the specific embodiments provided by the present invention, the following technical effects of the present invention are disclosed:

[0049] The present invention establishes a corresponding number of observation models according to different ionospheric heights on the transmitting side and ionospheric heights on the receiving side. Then, the measurement information received by the over-the-horizon radar is mapped to the target state space, and the measurement information is associated with the observation model to obtain the measurement-model association information at the current moment. The three-dimensional data association of "measurement-observation model-target" is decomposed into two low-dimensional data association problems of "measurement-observation model" data association and "measurement-target" data association, reducing the computational complexity of the association process. Then, according to the measurement-model association information at the current moment and the state vector of the existing targets at the previous moment, the Gaussian mixture probability hypothesis density algorithm is used to filter the measurement information at the current moment to determine the state vector of the existing targets at the current moment for tracking the existing targets. Finally, according to the measurement-model association information at the current moment and the measurement-model association information at the previous moment, the time-extended target detection algorithm is used to determine the random finite set of the states of the newborn targets for tracking the newborn targets. Finally, the tracking of multiple observation paths and multiple targets under dense clutter is realized, and the computational burden is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0051] Figure 1 is a flowchart of the multi-target tracking method for the over-the-horizon radar of the present invention;

[0052] Figure 2 is a schematic diagram of the multi-target tracking process of the over-the-horizon radar of the present invention;

[0053] Figure 3 is a schematic diagram of the observation principle of the over-the-horizon radar;

[0054] Figure 4 is a flowchart of the "measurement-observation model" association;

[0055] Figure 5 is a flowchart of the time-extended target detection algorithm based on measurement;

[0056] Figure 6 is a schematic diagram of the simulation result of the multi-target number prediction;

[0057] Figure 7 is a schematic diagram of the simulation result of the optimal sub-pattern assignment distance;

[0058] Figure 8 is a schematic diagram of the modules of the multi-target tracking system for the over-the-horizon radar of the present invention.

[0059] Symbol Explanation:

[0060] Emitter - 11, Receiver - 12, Target - 13, Altitude Acquisition Unit - 21, Model Establishment Unit - 22, Measurement Acquisition Unit - 23, Association Unit - 24, First Tracking Unit - 25, Second Tracking Unit - 26. Detailed Implementation Manner

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0062] The object of the present invention is to provide a multi - target tracking method, system, electronic device and medium for over - the - horizon radar in a multi - detection - path and multi - target tracking scenario in a dense clutter environment. On the one hand, the three - dimensional data association of "measurement - observation model - target" is decomposed into two low - dimensional data association problems of "measurement - observation model" data association and "measurement - target" data association to reduce the computational complexity of the association process. On the other hand, while performing the "measurement - observation model" data association, the measurements from clutter are excluded to further reduce the computational burden of the "measurement - target" association based on the random finite set. In addition, considering that the initial states of targets and newly - born targets cannot be directly obtained in many application scenarios, a measurement - based extended - time target detection algorithm is designed to obtain the initial states of targets to improve the applicability of the algorithm in scenarios where the initial states of targets are unknown.

[0063] To make the above - mentioned objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0064] Embodiment 1

[0065] As Figure 1 and Figure 2 shown, this embodiment provides a multi - target tracking method for over - the - horizon radar, including:

[0066] S1: Obtain the ionospheric height on the transmitting side and the ionospheric height on the receiving side of each observation path of the over - the - horizon radar.

[0067] S2: For any observation path, establish an observation model corresponding to the observation path according to the ionospheric height on the transmitting side and the ionospheric height on the receiving side of the observation path.

[0068] S3: Obtain the set of measurement information received by the over-the-horizon radar in real time. The set of measurement information includes multiple pieces of measurement information.

[0069] S4: Map each piece of measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to the target state space, and associate each piece of measurement information with the observation model to obtain the measurement-model associated random finite set at the current moment. The measurement-model associated random finite set includes each piece of measurement information and the corresponding observation model of each piece of measurement information.

[0070] S5: According to the measurement-model associated random finite set at the current moment and the state information of the existing targets at the previous moment, use the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment for filtering to determine the state information of the existing targets at the current moment, so as to track the existing targets.

[0071] S6: According to the measurement-model associated random finite set at the current moment and the measurement-model associated random finite set at the previous moment, use the time-extended target detection algorithm to determine the random finite set of newborn target states, so as to track the newborn targets. The newborn targets will become existing targets at the next moment.

[0072] Further, between S2 and S3, the over-the-horizon radar multi-target tracking method further includes: establishing a single-target motion model: x t = f t-1 (x t-1 ) + w t-1 ; where x t is the target state vector at time t, f t-1 (·) represents the target motion equation from time t - 1 to time t, and w t-1 represents the system noise, which follows a Gaussian distribution with a mean of 0 and a variance of Q t-1 .

[0073] For the convenience of calculation, the target state vector is represented in polar coordinate form: where ρ is the distance of the target from the origin of the coordinate, θ is the angle by which the target deviates from the polar coordinate X-axis, and represent the change rates of ρ and θ.

[0074] The observation principle of the over-the-horizon radar is as Figure 3 shown. In the figure, H A , H B , H C are the heights of the ionospheres A, B, and C respectively, r 11 , r 12 , r 13 , r 14 , r 21 , r22 , r 23 , r 24 is the reflection distance. The over-the-horizon radar uses the reflection of the ionosphere in the atmosphere to observe distant targets. The signal sent by the over-the-horizon radar transmitter is reflected in sequence by the ionosphere on the transmitting side, the target, and the ionosphere on the receiving side, and finally reaches the receiver. However, there are multiple ionospheres with different heights in the atmosphere at the same time and place, so there will be multiple observation paths, that is, there are multiple observation models. Assuming that the over-the-horizon radar has Ω observation models, then the observation model can be expressed as:

[0075]

[0076] where, is the observation vector of the j-th observation path, is the observation equation of the observation model corresponding to the height of the j-th observation path, x t is the target state vector at time t, is the observation noise of the observation model corresponding to the j-th observation path, which follows a Gaussian distribution with a mean of 0 and a variance of Ω is the number of observation paths.

[0077] In this embodiment, the measurement vector is also represented in the polar coordinate system as where r = r1 + r2 represents half of the distance of the observation path, r1 is the distance from the transmitter to the ionosphere on the transmitting side, and r2 is the distance from the receiver to the ionosphere on the receiving side, is the deviation angle between r2 and the Y-axis.

[0078] Specifically, the observation equation is:

[0079]

[0080] where z is the measurement vector, x is the target state vector,

[0081] If there are three ionospheres (A, B, and C) with different heights in the scene, then there are a total of 9 observation models (A-A, A-B, A-C, B-A, B-B, B-C, C-A, C-B, C-C):

[0082]

[0083] ​Furthermore, both the observation model corresponding to the measurement information received by the over-the-horizon radar and the target are uncertain. The present invention uses a dimensionality reduction method to first determine the "measurement-observation model" correlation relationship. The "measurement-observation model" correlation is carried out in the observation space and the target state space. All the measurement information received by the receiver is mapped from the same observation space to the target state space. In S4, the following formula is used to map the measurement information to the target state space:

[0084]

[0085] where x is the target state vector after mapping the measurement information to the target state space, l1 is the first intermediate variable, and l2 is the second intermediate variable. l2 = r - l1, r is half of the observation path distance, H t is the ionospheric height on the transmitting side, H r is the ionospheric height on the receiving side, D is the distance between the transmitter and the receiver of the over-the-horizon radar. is the deviation angle between the distance from the receiver to the ionosphere on the receiving side and the Y-axis, ρ is the distance of the target from the coordinate origin. is the change rate of the distance ρ of the target from the coordinate origin, θ is the angle by which the target deviates from the polar coordinate X-axis. is the measurement information. is the first derivative of r.

[0086] After that, in the target state space, the "measurement-target" states mapped from different observation model spaces can be obtained, that is, the measurement-model correlation random finite set.

[0087] After mapping the measurement information to the target state space, the present invention also includes filtering out clutter. Multiple observation models of the over-the-horizon radar will generate multiple observations of the real target. Therefore, in the target state space, there will be multiple "measurement-target" states from different observation model spaces around the real target state. And the clutter observed by each observation model is different, so there are no multiple "measurement-target" states from different observation models around the clutter state. Reflected in the mathematical form, it is that after mapping the measurement information from different observation spaces to the target state space, some mapping states will gather together, and these gathered mapping states are the correct target measurement mappings, and the remaining mapping states are clutter. Therefore, as long as these measurement information and the corresponding observation models are recorded, the measurement-observation model matching can be realized.

[0088] After completing the "measurement-observation model" correlation, the measurement is labeled with its corresponding observation model. The measurement-model correlation random finite set at time t is:

[0089]

[0090] Among them, Z L,t is the measurement-model associated random finite set at time t, is the measurement random finite set corresponding to the observation model of the j-th observation path at time t, z i is the i-th measurement information, Z j,t is the observation vector random finite set matched with the observation model of the j-th observation path at time t, and Ω is the number of observation paths.

[0091] The process of "measurement-observation model" association is as Figure 4 shown. In the figure, and are both elements in X t and comes from the observation model a q , comes from the observation model a p .

[0092] Furthermore, S5 specifically includes:

[0093] (51) According to the state information of the existing targets and the state information of the newly born targets at the previous moment, use the prediction step of the Gaussian mixture probability hypothesis density algorithm to predict the state information of the existing targets and the newly born targets at the current moment, and obtain the predicted states of the existing targets and the newly born targets at the current moment.

[0094] In this embodiment, the prediction step of the Gaussian mixture probability hypothesis density algorithm used is:

[0095]

[0096] Among them, b t-1 is the intensity of the newly born target, X B,t-1 ' is the state random finite set of the newly born target obtained by the time extension method at the previous moment, is the state vector of the i-th newly born target in X B,t-1 ', D t|t-1 is the multi-target prior intensity, w0 is a preset weight constant, P0 is a preset covariance matrix, P S,t is the target survival probability at time t, J t-1 is the number of surviving targets, w i,t-1 is the weight corresponding to the Gaussian component, x i,t|t-1 = F t-1 x i,t-1 , P i,t|t-1 = F t-1 P i,t-1 (F t-1 ) T + Qt-1 , x i,t-1 is the state vector of the i-th Gaussian component at the previous moment, and P i,t-1 is the state covariance matrix of the i-th Gaussian component at the previous moment.

[0097] (52) According to the measurement-model association random finite set at the current moment, the update step of the Gaussian mixture probability hypothesis density algorithm is used to correct the predicted state of the target at the current moment, and the state information of the target at the current moment is obtained.

[0098] In this embodiment, the update step of the Gaussian mixture probability hypothesis density algorithm used is:

[0099]

[0100] where D t is the multi-target posterior intensity, P D is the observation probability of the observation model, J t|t-1 is the number of Gaussian components in the multi-target prior intensity, w i,t|t-1 is the weight of the i-th Gaussian component in the multi-target prior intensity, κ(z) is the distribution of clutter, x i,t|t-1 is the state vector of the i-th Gaussian component in the multi-target prior intensity, P i,t|t-1 is the state covariance matrix of the i-th Gaussian component in the multi-target prior intensity, R j,t is the observation noise covariance matrix of the j-th observation model, I is the identity matrix, is the observation function of the j-th observation model.

[0101] The state fusion and state extraction steps are the same as those of the traditional Gaussian mixture probability hypothesis density algorithm, and will not be elaborated here.

[0102] Furthermore, in the multi-target tracking scenario, target disappearance and emergence will occur, making multi-target tracking have stronger uncertainty. In the existing research on multi-target tracking algorithms, whether it is the method based on data association or the method based on random finite sets, prior information of the multi-target state is required. However, in the actual application process, it may be difficult to directly obtain the prior information of the multi-target state being tracked, and corresponding target detection algorithms need to be designed to obtain it. When a new target appears, it will exist in the tracking scenario for a period of time. Therefore, this new target will be continuously observed by the over-the-horizon radar, and the measurements of this new target will also gather around the state of the new target after the "measurement-observation model" association. Moreover, the Gaussian weights obtained when these measurements update the previous multi-target state will be very small (these measurements do not belong to the previously known multi-targets). Therefore, two consecutive measurements from the new target satisfy the following after being mapped to the target state space:

[0103]

[0104] Among them, X B is the newborn target random finite set, and X t is the multi-target state random finite set at time t.

[0105] S6 specifically includes:

[0106] (61) Initialization:

[0107] (62) Map the measurement information in the measurement-model association random finite set at the current moment to the target state space to obtain the intermediate variable X at the current moment pb,t : X pb,t = h -1 (Z L,t , H t , H r , D), where Z L,t is the measurement-model association random finite set at time t.

[0108] (63) Determine the finite set of suspected newborn targets at the current moment according to the intermediate variable at the current moment.

[0109] If then use the following formula to determine the finite set of suspected newborn targets X at the current moment B,t :

[0110]

[0111] (64) Map the measurement information in the measurement-model association random finite set at the previous moment to the target state space to obtain the intermediate variable X at the previous moment pb,t-1 : X pb,t-1 = h -1 (Z L,t-1 , H t , H r , D), where Z L,t-1 is the measurement-model association random finite set at time t-1.

[0112] (65) Determine the finite set of suspected newborn targets at the previous moment according to the intermediate variable at the previous moment.

[0113] If then use the following formula to determine the finite set of suspected newborn targets at the previous moment:

[0114]

[0115] (66) Determine the newborn target state random finite set according to the finite set of suspected newborn targets at the current moment and the finite set of suspected newborn targets at the previous moment:

[0116] X B,t ' = {x b,t | if || x b,t - f t (x b,t-1 ) || < δ, x b,t ∈ X B,t , x b,t-1 ∈ X B,t-1}};

[0117] Wherein, X B,t ' is the random finite set of newborn target states at time t, x b,t is the vector of newborn target states at time t, x b,t-1 is the vector of newborn target states at time t - 1, f t () is the target motion equation, δ is a preset parameter, X B,t is the finite set of suspected newborn targets at the current time, X B,t-1 is the finite set of suspected newborn targets at the previous time.

[0118] As Figure 5 shown is the flowchart of the time-extended target detection algorithm based on measurements. In the figure, N1 is the number of measurements collected by the over-the-horizon radar receiver at time t, N2 is the number of measurements collected by the over-the-horizon radar receiver at time t + 1, X pre,t is the random finite set of target states obtained by mapping the measurement information in the measurement-model association random finite set at time t to the target state space, is X pre,t the M1-th target state vector in, is X pre,t+1 the M2-th target state vector in, and X' is the random finite set of newborn target states at time t + 1.

[0119] The present invention is directed to a dense clutter environment. For the application scenario of multi-target tracking with multiple observation paths using an over-the-horizon radar, a reduced-dimensional association multi-detection multi-target tracking method based on random finite sets is proposed. First, the initial multi-target states are obtained by using the time-extended target detection algorithm based on measurements, getting rid of the dependence on the prior knowledge of the target states. Then, the "measurement-observation model" association algorithm is used to match the measurements with the observation model and delete the clutter measurements, which can realize the association of the measurements with their corresponding observation models and the identification of the clutter measurements at the same time, further reducing the computational complexity. Then, using the measurement information associated with the observation model, the improved Gaussian mixture probability hypothesis density algorithm is used to realize the "measurement-target" association, and the three-dimensional association of "measurement-observation model-target" is completed by the reduced-dimensional association method. Finally, the tracking of multiple targets with multiple observation paths under dense clutter is realized, and the computational burden is reduced.

[0120] The effectiveness of the multi-target tracking method for over-the-horizon radar of the present invention is verified through simulation below. Consider a multi-target tracking scenario with 6 moving targets. The initial states, motion models, appearance and disappearance times of the 6 targets are shown in Table 1 and Table 2.

[0121] Table 1 Target Initial States

[0122]

[0123] Table 2 Target Appearance and Disappearance Times

[0124]

[0125]

[0126] where, x t = F t-1 x t-1 + w t , w t obeys Q = diag([0.1km 2 0.01km 2 / s 2 0.005rad 2 0.0005rad 2 / s 2 ).

[0127] There are three ionospheres with different heights in the atmosphere: H ion = [100km 800km 1500km],

[0128] Then there are 9 observation models, as shown in Table 3.

[0129] Table 3 Heights of the Transmitting and Receiving Sides of the Observation Model

[0130] Model 1 2 3 4 5 6 7 8 9 <![CDATA[H t \km]]> 100 100 100 800 800 800 1500 1500 1500 <![CDATA[H r \km]]> 100 800 1500 100 800 1500 100 800 1500

[0131] The observation noise obeys R = diag([0.001km 2 0.0001km 2 / s 2 0.0001rad 2 ), the observation probability of each observation model is set to P D = 0.6, the distance between the receiver and the transmitter of the over-the-horizon radar is 50km, the target survival probability is P S = 0.99, the simulation step size is 1s, the total number of simulation steps is 100, and the number of Monte Carlo simulations used is 100. The clutter distribution follows a Poisson distribution with a mean of 50\km2 The tracking effect is measured by the optimal secondary pattern assignment distance, the number of multi-target estimations, and the running time. The simulation results are as Figure 6 and Figure 7 multi-mode.

[0132] The simulation running time of the present invention is 0.225 s, and the simulation running time of the traditional SD-FUSION algorithm is 3.704 s. It can be determined from the simulation results that the present invention is superior to the SD-FUSION algorithm in terms of the optimal secondary pattern assignment distance, the number of multi-target estimations, and the algorithm simulation running time. From the simulation results of the optimal secondary pattern assignment distance, it can be seen that when a new target appears, both the present invention and the SD-FUSION algorithm will show a sudden peak. This is because at the first sampling moment when the new target appears, the time-extended target detection method cannot detect the target, and the target detection occurs at the second sampling moment. Therefore, a sudden peak will appear. The comparison between the two algorithms can highlight the effectiveness of the present invention in reducing the computational complexity in a dense clutter environment, and at the same time can improve the accuracy of multi-target tracking.

[0133] Embodiment 2

[0134] In order to execute the method corresponding to the above Embodiment 1 to achieve the corresponding functions and technical effects, a multi-target tracking system for an over-the-horizon radar is provided below.

[0135] As Figure 8 shown, the multi-target tracking system for an over-the-horizon radar provided in this embodiment includes: an altitude acquisition unit 21, a model establishment unit 22, a measurement acquisition unit 23, an association unit 24, a first tracking unit 25, and a second tracking unit 26.

[0136] Among them, the altitude acquisition unit 21 is used to acquire the ionospheric altitude on the transmitting side and the ionospheric altitude on the receiving side of each observation path of the over-the-horizon radar.

[0137] The model establishment unit 22 is connected to the altitude acquisition unit 21. The model establishment unit 22 is used to establish an observation model corresponding to any observation path according to the ionospheric altitude on the transmitting side and the ionospheric altitude on the receiving side of the observation path.

[0138] The measurement acquisition unit 23 is used to acquire the set of measurement information received by the over-the-horizon radar in real time; the set of measurement information includes a plurality of measurement information.

[0139] The association unit 24 is respectively connected to the model establishment unit 22 and the measurement acquisition unit 23. The association unit 24 is configured to map each measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to the target state space, and associate each measurement information with the observation model, so as to obtain the measurement-model association random finite set at the current moment. The measurement-model association random finite set includes each measurement information and the corresponding observation model of each measurement information.

[0140] The first tracking unit 25 is connected to the association unit 24. The first tracking unit 25 is configured to perform filtering according to the measurement-model association random finite set at the current moment and the state information of the existing targets at the previous moment, and use the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment to determine the state information of the existing targets at the current moment, so as to track the existing targets.

[0141] The second tracking unit 26 is connected to the association unit 24. The second tracking unit 26 is configured to determine the random finite set of the states of the newly born targets according to the measurement-model association random finite set at the current moment and the measurement-model association random finite set at the previous moment, and use the time-extended target detection algorithm to track the newly born targets; the newly born targets will become existing targets at the next moment.

[0142] Compared with the prior art, the over-the-horizon radar multi-target tracking system provided in this embodiment has the same beneficial effects as the over-the-horizon radar multi-target tracking method provided in Embodiment 1, which will not be elaborated here.

[0143] Embodiment 3

[0144] This embodiment provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the over-the-horizon radar multi-target tracking method of Embodiment 1.

[0145] Optionally, the above electronic device may be a server.

[0146] In addition, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, it implements the over-the-horizon radar multi-target tracking method of Embodiment 1.

[0147] The various embodiments in this specification are described in a progressive manner. The key points of each embodiment are the differences from other embodiments. The same and similar parts among the various embodiments can be referred to each other.

[0148] In this text, specific examples are used to illustrate the principles and implementation modes of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation modes and application scopes. To sum up, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for multi-target tracking of over-the-horizon radar, characterized in that, The method for multi-target tracking of over-the-horizon radar includes: Obtaining the ionospheric height on the transmitting side and the ionospheric height on the receiving side of each observation path of the over-the-horizon radar; For any observation path, an observation model corresponding to the observation path is established according to the ionospheric height on the transmitting side and the ionospheric height on the receiving side of the observation path; Obtaining the set of measurement information received by the over-the-horizon radar in real time; the set of measurement information includes a plurality of measurement information; Mapping each measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to the target state space, and associating each measurement information with the observation model, to obtain the measurement-model associated random finite set at the current moment; the measurement-model associated random finite set includes each measurement information and the observation model corresponding to each measurement information; According to the measurement-model associated random finite set at the current moment and the state information of the existing targets at the previous moment, filtering is performed using the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment to determine the state information of the existing targets at the current moment, so as to track the existing targets; According to the measurement-model associated random finite set at the current moment and the measurement-model associated random finite set at the previous moment, the time-extended target detection algorithm is used to determine the random finite set of the states of newborn targets, so as to track the newborn targets; the newborn targets become existing targets at the next moment.

2. The over-the-horizon radar multi-target tracking method according to claim 1, characterized in that, The observation model is: Among them, is the observation vector of the j-th observation path, is the observation equation of the observation model corresponding to the j-th observation path, x t is the target state vector at time t, is the observation noise of the observation model corresponding to the j-th observation path, and Ω is the number of observation paths.

3. The over-the-horizon radar multi-target tracking method according to claim 1, characterized in that The following formula is used to map the measurement information to the target state space: Among them, x is the target state vector after mapping the measurement information to the target state space, l1 is the first intermediate variable, and l2 is the second intermediate variable. r is half of the observed path distance, and H t is the ionospheric height on the transmitting side, and H r is the ionospheric height on the receiving side, D is the distance between the transmitter and the receiver of the over-the-horizon radar, is the deviation angle between the distance from the receiver to the ionosphere on the receiving side and the Y-axis, ρ is the distance of the target from the coordinate origin, is the change rate of the distance ρ of the target from the coordinate origin, θ is the angle by which the target deviates from the polar coordinate X-axis, is the measurement information, is the first derivative of r.

4. The over-the-horizon radar multi-target tracking method according to claim 1, wherein The measurement-model associated random finite set at time t is: Among them, Z L,t is the measurement-model association random finite set at time t, is the measurement random finite set corresponding to the j-th observation path observation model at time t, z i is the i-th measurement information, Z j,t is the observation vector random finite set matching the j-th observation path observation model at time t, and Ω is the number of observation paths.

5. The over-the-horizon radar multi-target tracking method according to claim 1, characterized in that According to the measurement-model associated random finite set at the current moment and the state information of the existing targets at the previous moment, filtering is performed using the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment to determine the state information of the existing targets at the current moment, specifically including: According to the state information of the existing targets at the previous moment and the state information of the newborn targets, the prediction step of the Gaussian mixture probability hypothesis density algorithm is used to predict the state information of the existing targets and the newborn targets at the current moment, to obtain the predicted state of the targets at the current moment; According to the measurement-model associated random finite set at the current moment, the update step of the Gaussian mixture probability hypothesis density algorithm is used to correct the predicted state of the targets at the current moment, to obtain the state information of the targets at the current moment.

6. The over-the-horizon radar multi-target tracking method according to claim 1, characterized in that, According to the measurement-model associated random finite set at the current moment and the measurement-model associated random finite set at the previous moment, the time-extended target detection algorithm is used to determine the random finite set of the states of newborn targets, specifically including: Mapping the measurement information in the measurement-model associated random finite set at the current moment to the target state space, to obtain the intermediate variable at the current moment; Determining the finite set of suspected newborn targets at the current moment according to the intermediate variable at the current moment; Mapping the measurement information in the measurement-model associated random finite set at the previous moment to the target state space, to obtain the intermediate variable at the previous moment; Determining the finite set of suspected newborn targets at the previous moment according to the intermediate variable at the previous moment; Determining the random finite set of the states of newborn targets according to the finite set of suspected newborn targets at the current moment and the finite set of suspected newborn targets at the previous moment.

7. The over-the-horizon radar multi-target tracking method according to claim 6, wherein The following formula is used to determine the random finite set of the states of newborn targets at time t: X B,t ' = {x b,t | if || x b,t - f t (x b,t-1 ) || < δ, x b,t ∈ X B,t , x b,t-1 ∈ X B,t-1}; Among them, X B,t ' is the random finite set of newborn target states at time t, and x b,t is the newborn target state vector at time t, and x b,t-1 is the newborn target state vector at time t - 1, and f t () is the target motion equation, δ is a preset parameter, and X B,t is the finite set of suspected newborn targets at the current moment, and X B,t-1 is the finite set of suspected newborn targets at the previous moment.

8. An over-the-horizon radar multi-target tracking system, characterized in that, The over-the-horizon radar multi-target tracking system includes: An altitude acquisition unit for acquiring the ionosphere altitude on the transmitting side and the ionosphere altitude on the receiving side of each observation path of the over-the-horizon radar; A model establishment unit connected to the altitude acquisition unit for establishing an observation model corresponding to any observation path according to the ionosphere altitude on the transmitting side and the ionosphere altitude on the receiving side of the observation path; A measurement acquisition unit for acquiring in real time a set of measurement information received by the over-the-horizon radar; the set of measurement information includes a plurality of measurement information; An association unit connected to the model establishment unit and the measurement acquisition unit respectively for mapping each measurement information in the set of measurement information received by the over-the-horizon radar at the current moment to the target state space and associating each measurement information with the observation model to obtain a measurement-model association random finite set at the current moment; the measurement-model association random finite set includes each measurement information and the observation model corresponding to each measurement information; A first tracking unit connected to the association unit for filtering according to the measurement-model association random finite set at the current moment and the state information of the existing targets at the previous moment by using the Gaussian mixture probability hypothesis density algorithm and the measurement information at the current moment to determine the state information of the existing targets at the current moment so as to track the existing targets; A second tracking unit connected to the association unit for determining a random finite set of newborn target states by using the time-extended target detection algorithm according to the measurement-model association random finite set at the current moment and the measurement-model association random finite set at the previous moment so as to track the newborn targets; the newborn targets become existing targets at the next moment.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic device to execute the over-the-horizon radar multi-target tracking method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by the processor, it implements the over-the-horizon radar multi-target tracking method according to any one of claims 1 to 7.

Images