A method and device for double-station multi-target tracking based on WMFS and UKF

Abstract

The application provides a double-station multi-target tracking method and device based on WMFS and UKF, comprising: constructing a WMFS model, and reconstructing a state transition function by using a pre-trained WMFS model; combining the optimized WMFS model with a UKF model to obtain a double-station multi-target tracking model; and tracking a target by using the double-station multi-target tracking model. The method fully utilizes the advantages of WMFS in processing system uncertainty and complex modeling, adopts an unscented (UT) framework to solve the nonlinear measurement problem in the WMFS-UKF derivation process, and uses a pre-trained WMFS to reconstruct the state transition function. Through the historical state and expected output of the target, the WM fuzzy reasoning system realizes more accurate state prediction and accurate covariance estimation, improves the performance and stability of target tracking, and verifies the effectiveness and robustness of the algorithm in the target tracking scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of sensing, and particularly relates to a double-station multi-target tracking method and device based on WMFS and UKF. BACKGROUND

[0002] The passive positioning and tracking technology has the characteristics of strong concealment, strong anti-interference ability, wide detection range, etc., and can effectively improve the survivability, combat capability and detection capability of the system. Common passive positioning methods include direction of arrival (DOA), frequency difference of arrival (FDOA) and time difference of arrival (TDOA).

[0003] Among them, DOA is a relatively simple and reliable passive positioning and tracking method. The position of the target is determined by measuring the arrival time or arrival angle of the target signal at different receivers. The arrival time or arrival angle of the signal at different receivers determines the position of the target. DOA can be traced back to submarine sonar tracking, in which method, the position and velocity of the target are determined according to the bearing angle obtained by sonar measurement. The bearing angle obtained by sonar measurement is used to solve the position and velocity of the target. However, the traditional DOA method can only provide position information, and has limitations in tracking high-maneuverability targets. In order to solve this problem, the direction of arrival with Doppler (DOAD) method is proposed.

[0004] The Doppler (DOAD) method uses the Doppler shift of the target signal to obtain velocity information, thereby improving the tracking performance. DOAD is a relatively novel passive positioning and tracking method with high research significance and practical value. Passive positioning and tracking method, which has high research significance and practical value in this field. In the field of target tracking, processing nonlinear measurements is a big challenge for filtering problems, especially for complex measurements such as azimuth-Doppler. The classic Kalman filter (KF) performs well in simple linear systems with Gaussian noise, but performs poorly in the nonlinearity of measurement, which may reduce the estimation accuracy. In order to solve this problem, various filtering algorithms are introduced to better handle nonlinear measurements.

[0005] The existing technology such as the classic Kalman filter (KF) performs well in processing linear systems and Gaussian noise, but its estimation accuracy decreases significantly in the case of nonlinear measurement, especially when dealing with complex multi-target tracking, which becomes a significant bottleneck. In order to improve this point, methods such as extended Kalman filter (EKF) are proposed, however, these methods still have limitations in dealing with high-maneuverability targets, especially in the case of large measurement noise, the tracking stability and accuracy will decrease significantly. SUMMARY

[0006] The embodiment of the application provides a double-station multi-target tracking method based on WMFS and UKF, comprising:

[0007] A WMFS model is constructed, and a pre-trained WMFS model is used to reconstruct a state transition function;

[0008] The optimized WMFS model is combined with a UKF model to obtain a double-station multi-target tracking model;

[0009] The double-station multi-target tracking model is used to track a target.

[0010] Further, the WMFS model is constructed, comprising:

[0011] A WMFS model is trained by using historical data, and fuzzy rules and parameters of the trained WMFS model are determined, wherein the WMFS model is a multi-layer WMFS, each layer of the WMFS takes an output result of an upper layer as input data of the layer, and a last layer of the WMFS integrates a result.

[0012] Further, the pre-trained WMFS model is used to reconstruct a state transition function, comprising:

[0013] WMFS is used to generate state estimation and uncertainty information through internally defined fuzzy rules, and the state estimation and the uncertainty information are fused into a framework of UKF to perform target state prediction and updating;

[0014] Target state and covariance are estimated through propagation and conversion of Sigma points.

[0015] Further, the target state and the covariance are estimated through propagation and conversion of Sigma points, comprising:

[0016] A signal state is determined according to a state equation, a measurement equation, a double-station measurement equation and an observation equation of an unscented Kalman filter.

[0017] Sigma points are generated according to the signal state, and the sigma point set is transformed by using a state function;

[0018] A priori estimation of states and covariance of the transformed sigma points is calculated, and the sigma points are estimated from the priori estimation of states and covariance;

[0019] The sigma points are converted by using a measurement function, and a priori estimation and covariance are calculated.

[0020] Further, the pre-trained WMFS model is used to reconstruct a state transition function, comprising:

[0021] Determine the input-output data pair of the Wang-Mendel fuzzy model;

[0022] Determine the weight and weight output parameter of the fuzzy set according to the input-output data pair;

[0023] Calculate the nonlinear state function of the UKF model according to the weight and weight output parameter of the fuzzy set.

[0024] The application provides a dual-station multi-target tracking device based on WMFS and UKF, comprising:

[0025] An acquisition module is configured to construct a WMFS model and reconstruct a state transition function by using a pre-trained WMFS model;

[0026] A processing module is configured to combine the optimized WMFS model with a UKF model to obtain a dual-station multi-target tracking model;

[0027] An execution module is configured to track a target by using the dual-station multi-target tracking model.

[0028] Further, the WMFS model is constructed by:

[0029] A first acquisition module is configured to train a WMFS model by using historical data and determine the fuzzy rules and parameters of the trained WMFS model, wherein the WMFS model is a multi-layer WMFS, each layer of the WMFS takes the output result of the upper layer as the input data of the layer, and the last layer of the WMFS integrates the results.

[0030] Further, the acquisition module comprises:

[0031] A second acquisition module is configured to generate state estimation and uncertainty information by using the WMFS through internally defined fuzzy rules, and fuse the state estimation and the uncertainty information into the framework of the UKF to predict and update the target state.

[0032] A first processing module is configured to estimate the target state and covariance by propagation and conversion of sigma points.

[0033] Further, the first processing submodule comprises:

[0034] A second processing submodule is configured to determine the signal state according to the state equation, the measurement equation, the dual-station measurement equation and the observation equation of the unscented Kalman filter.

[0035] A third processing submodule is configured to generate sigma points according to the signal state and transform the sigma point set by using the state function.

[0036] A fourth processing submodule is configured to calculate a priori estimation of state and covariance of the transformed sigma points, and reconstruct the sigma points from the previous state and covariance estimation;

[0037] A first processing submodule is configured to transform the sigma points by a measurement function, and calculate the priori estimation and covariance.

[0038] Further, the processing module comprises:

[0039] A second processing submodule is configured to determine input-output data pairs of the Wang-Mendel fuzzy model;

[0040] A third processing submodule is configured to determine weight and weight output parameters of the fuzzy set according to the input-output data pairs;

[0041] A fourth processing submodule is configured to calculate a nonlinear state function of the UKF model according to the weight and weight output parameters of the fuzzy set.

[0042] The present application provides a computer device comprising a memory and a processor, the memory storing computer readable instructions, the computer readable instructions being executed by the processor to make the processor execute the steps of the above-mentioned dual-station multi-target tracking method based on WMFS and UKF.

[0043] The present application provides a storage medium storing computer readable instructions, the computer readable instructions being executed by one or more processors to make the one or more processors execute the steps of the above-mentioned dual-station multi-target tracking method based on WMFS and UKF.

[0044] In order to solve the problems of excessive estimation error and filtering instability caused by process noise uncertainty and the problems of excessive estimation error and filtering instability caused by system model uncertainty, a non-gain Kalman filter (UKF) algorithm based on Wang-Mendel fuzzy system (WMFS) is proposed for Doppler dual-station multi-target tracking algorithm. The algorithm fully utilizes the advantages of WMFS in dealing with system uncertainty and complex modeling. In the derivation process of WMFS-UKF, a non-gain transformation (UT) framework is used to solve the nonlinear measurement problem, and a pre-trained WMFS is used to reconstruct the state transition function. By utilizing the historical state and expected output of the target, the WM fuzzy reasoning system realizes more accurate state prediction and accurate covariance estimation, which greatly improves the performance and stability of target tracking. The simulation experiment verifies the effectiveness and robustness of the algorithm in the target tracking scene. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0046] Figure 1 A flowchart of a double-station multi-target tracking method based on WMFS and UKF provided by the embodiments of the present application is shown in the figure.

[0047] Figure 2 A Wang-Mendel fuzzy system diagram provided by the embodiments of the present application is shown in the figure.

[0048] Figure 3 A flowchart of a deep fuzzy inference system based on WMFS is shown in the figure. DETAILED DESCRIPTION

[0049] In order to make the technical problems, technical solutions and beneficial effects of the present application more clearly understood, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0050] As shown in the figure is a double-station multi-target tracking method based on WMFS and UKF provided by the embodiments of the present application, comprising: Figure 1

[0051] S1, constructing a WMFS model, and reconstructing a state transition function using a pre-trained WMFS model;

[0052] The WMFS model is trained using historical data, and the fuzzy rules and parameters of the trained WMFS model are determined. The WMFS generates state estimates and uncertainty information through internally defined fuzzy rules, and fuses the state estimates and uncertainty information into the framework of UKF to predict and update the target state. The WMFS model is a multi-layer WMFS, each layer of the WMFS takes the output of the upper layer as the input data of the layer, and the last layer of the WMFS integrates the results to predict the target state.

[0053] The historical state data and Sigma points used to represent state estimates and uncertainties.

[0054] As shown in the figure is a double-station multi-target tracking method based on WMFS and UKF provided by the embodiments of the present application, comprising: Figure 2 ​As shown, the Wang-Mendel Fuzzy System (WMFS) is constructed using the Wang-Mendel method, which designs low-dimensional fuzzy systems in a bottom-up, layer-by-layer manner, ultimately forming the entire WMFS model. The size *m* of the moving window controls the number of inputs to the low-dimensional fuzzy system; *m* can be 3, 4, or 5. Figure 2 As shown, the input vector for WMFS is This vector has high dimension, where, The output is Y L In the first layer (l = 1, 2, ..., L-1), there are several fuzzy systems. Its output is denoted as And it serves as the input to the (l+1)th layer. In the Lth layer, there is only one fuzzy system, WMFS. L It combines the outputs of the (L-1)th layer into the final output x. L .

[0055] Define p fuzzy sets A 1 A 2 ,...,A p The rules of the fuzzy system WMFS are as follows:

[0056]

[0057] x and y represent the input and output, respectively. Let be a fuzzy set defined on the output variable. This set represents the output of the fuzzy set based on the input variable.

[0058] Fuzzy set A j This can be represented using fuzzy membership functions; in this embodiment, triangular membership functions are used. The output is:

[0059]

[0060] parameter For fuzzy sets At its core, the key to the WMFS fuzzy system is determining the parameters. These are membership functions of a fuzzy set, which define the degree to which an input variable corresponds to the fuzzy set.

[0061] S2. Combine the reconstructed WMFS model with the UKF model to obtain a bi-station multi-target tracking model;

[0062] In this embodiment, the reconstructed WMFS model is combined with an unscented Kalman filter for bistatic azimuth and Doppler target tracking. Specifically:

[0063] Step one, determine the signal state according to the state equation, measurement equation, double station measurement equation and observation equation of the unscented Kalman filter.

[0064] Define a nonlinear system, whose state equation and measurement equation are respectively:

[0065] x k+1 =f(x k ,w k )

[0066] Z k+1 =h(x k+1 )+v k+1

[0067] wherein w k represents the process noise affecting the state of the system, v k+1 represents the measurement noise affecting the observation.

[0068] The double station measurement equation is:

[0069]

[0070] wherein the observation equation is as follows:

[0071]

[0072] wherein the state of the observation station i = 1, 2 at K time is

[0073] Step two, generate sigma points according to the signal state, and transform the sigma point set by the state function;

[0074] Generate 2n+1 sigma points from the state and covariance at k time:

[0075]

[0076] wherein λ represents a scaling parameter, used to control the diffusion degree of Sigma point, and P(k|k) represents the covariance matrix.

[0077] Calculate the weight of each 2n+1 sigma point:

[0078]

[0079] Transform the sigma point set by the state function:

[0080] x (i) (k+1|k)=f(x (i) (k|k)),i=0,…,2n

[0081] Step three, calculate the priori estimation of state and covariance of transformed sigma points, and generate sigma points from the priori estimation of state and covariance;

[0082] The priori estimation of state and covariance is:

[0083]

[0084] Generate 2n+1 sigma points from the priori estimation of state and covariance:

[0085]

[0086] Step four, transform sigma points by measurement function, and calculate the priori estimation and covariance; transform the sigma point set by measurement function:

[0087] z (i) (k+1|k)=h(x (i) (k+1|k)),i=0,…2n

[0088] The priori estimation and covariance of measurement are:

[0089]

[0090] In the embodiment of the application, the reconstructed WMFS model is combined with the UKF model to obtain a double-station multi-target tracking model, and a flowchart of the double-station multi-target tracking model is as shown in Figure 3 , and specifically includes:

[0091] Step one, determine the input / output data pairs of the Wang-Mendel fuzzy model;

[0092] Given N input / output data pairs:

[0093]

[0094] Input recombination: the input vector is divided into T-2 segments by using a sliding window method. The window size m is set to 3, and the moving step is set to 1.

[0095] Input:

[0096]

[0097] The input / output data pairs for designing are:

[0098]

[0099] For each input of , consider p fuzzy sets A 1 ,A 2 ,...,Ap Select the endpoints as:

[0100] x min =min{x 0 (kt|kt)|t=1,...,T,k=T+1}

[0101] x max =max{x 0 (kt|kt)|t=1,...,T,k=T+1}

[0102] Step 2: Determine the weights and weight output parameters of the fuzzy set based on the input and output data;

[0103] For each input / output data pair starting from k, determine the fuzzy set. Obtain p fuzzy sets A at each location. 1 A 2 ,…,A p The largest membership value in the set, i.e., determining

[0104]

[0105] Update weights and weight output parameters

[0106]

[0107] Step 3: Calculate the nonlinear state function of the UKF model based on the weights and weight output parameters of the fuzzy set.

[0108] Deterministic fuzzy system Parameters in

[0109]

[0110] say Cells (j1,…,j m ), overwritten by data, and defined as:

[0111]

[0112] Define the neighborhood N((j1,…,j) m )) represents cells with a distance of no more than 1.

[0113] N((j1,…,j m ))={(j′1,…,j′ m |for all o=1,…,m|j i -j′ i |≤1,then(j′1,…,j′m )∈C(0)}

[0114] Calculate the average C(k)

[0115]

[0116] Calculate the new support set C(k+1), and calculate the To C(k+1)

[0117]

[0118] Where As a new input / output data pair, a level 1 fuzzy system is designed

[0119] Define the nonlinear state function as:

[0120]

[0121] The technical scheme of the application improves the performance of the dual-station multi-target tracking system in dealing with nonlinear and uncertain problems by combining WMFS and UKF.

[0122] S3, track the target by using the dual-station multi-target tracking model.

[0123] The non-gain Kalman filter (UKF) algorithm based on the Wang-Mendel fuzzy system (WMFS) provided by the application is used for a Doppler dual-station multi-target tracking algorithm. The algorithm fully utilizes the advantages of WMFS in dealing with system uncertainty and complex modeling. In the derivation process of WMFS-UKF, the unscented transformation (UT) framework is used to solve the nonlinear measurement problem, and the pre-trained WMFS is used to reconstruct the state transition function. By utilizing the historical state of the target and the expected output, the WM fuzzy reasoning system realizes more accurate state prediction and accurate covariance estimation, which greatly improves the performance and stability of target tracking. The simulation experiment verifies the effectiveness and robustness of the algorithm in the target tracking scene.

[0124] The application provides a dual-station multi-target tracking device based on WMFS and UKF, which comprises: an acquisition module, which is used for constructing a WMFS model and reconstructing a state transition function by using a pre-trained WMFS model; a processing module, which is used for combining the optimized WMFS model with a UKF model to obtain a dual-station multi-target tracking model; and an execution module, which is used for tracking a target by using the dual-station multi-target tracking model.

[0125] In some embodiments, the constructing the WMFS model comprises: a first obtaining module configured to train a WMFS model using historical data and determine fuzzy rules and parameters of the trained WMFS model, wherein the WMFS model is a multi-layer WMFS, each layer of the WMFS takes output results of an upper layer as input data of the layer, and a last layer of the WMFS integrates results.

[0126] In some embodiments, the obtaining module comprises: a second obtaining module configured to use the WMFS to generate state estimation and uncertainty information by using fuzzy rules defined internally, and fuse the state estimation and the uncertainty information into a framework of the UKF to perform prediction and update of a target state.

[0127] A first processing module is configured to estimate the target state and covariance by propagation and conversion of sigma points.

[0128] In some embodiments, the first processing submodule comprises: a second processing submodule configured to determine signal states according to a state equation, a measurement equation, a double station measurement equation and an observation equation of the unscented Kalman filter; a third processing submodule configured to generate sigma points according to the signal states and transform the set of sigma points by a state function; a fourth processing submodule configured to calculate a priori estimation of states and covariance of the transformed sigma points and estimate sigma points from the previous state and covariance; and a first processing submodule configured to transform the sigma points by a measurement function and calculate a priori estimation and covariance.

[0129] In some embodiments, the processing module comprises: a second processing submodule configured to determine input-output data pairs of the Wang-Mendel fuzzy model; a third processing submodule configured to determine weights and weight output parameters of fuzzy sets according to the input-output data pairs; and a fourth processing submodule configured to calculate a nonlinear state function of the UKF model according to the weights and weight output parameters of the fuzzy sets.

[0130] The application provides a non-gain Kalman filter (UKF) algorithm based on a Wang-Mendel fuzzy system (WMFS) and used for a Doppler double-station multi-target tracking algorithm.

[0131] To solve the above technical problems, the embodiment of the application further provides a computer device. The computer device comprises a processor, a non-volatile storage medium, a memory and a network interface connected through a system bus. The non-volatile storage medium of the computer device stores an operating system, a database and computer readable instructions, the database can store a control information sequence, and the computer readable instructions can enable the processor to implement a double-station multi-target tracking method based on WMFS and UKF when executed by the processor. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device can store computer readable instructions, which can enable the processor to execute a double-station multi-target tracking method based on WMFS and UKF when executed by the processor. The network interface of the computer device is used to connect and communicate with a terminal. In the embodiment, the processor is used to execute the specific contents of the acquisition module and the processing module, and the memory stores program codes and various data required for executing the above modules. The network interface is used for data transmission between the user terminal or the server. The memory in the embodiment stores program codes and data required for executing all sub-modules in the image processing method, and the server can call the program codes and data of the server to execute the functions of all sub-modules.

[0132] The application further provides a storage medium storing computer readable instructions, which enable one or more processors to execute the steps of the double-station multi-target tracking method based on WMFS and UKF according to any one of the above embodiments when executed by the one or more processors.

[0133] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a non-volatile storage medium such as a magnetic disc, an optical disc, a read-only memory (ROM), or a random access memory (RAM).

[0134] It should be understood that, although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0135] The above only describes some embodiments of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be considered as falling within the scope of protection of the present application.

[0136] The above only describes some embodiments of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be considered as falling within the scope of protection of the present application. The above only describes some embodiments of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be considered as falling within the scope of protection of the present application.

Claims

1. A method for bi-static multi-target tracking based on WMFS and UKF, characterized in that, The method comprises the following steps: A WMFS model for determining a signal state is constructed, and a state transition function is reconstructed using a pre-trained WMFS model; An optimized WMFS model is combined with a UKF model to obtain a bistatic multi-target tracking model; The bistatic multi-target tracking model is used to track a target; The state transition function is reconstructed using the pre-trained WMFS model, which comprises the following steps: A WMFS is used to generate state estimation and uncertainty information through internally defined fuzzy rules, and the state estimation and the uncertainty information are fused into a UKF framework to predict and update a target state; Target state and covariance are estimated through propagation and conversion of Sigma points; The target state and covariance are estimated through propagation and conversion of Sigma points, which comprises the following steps: A signal state is determined according to a state equation, a measurement equation, a bistatic measurement equation and an observation equation of an unscented Kalman filter; Sigma points are generated according to the signal state, and the sigma point set is transformed by a state function; A priori estimation of state and covariance of the transformed sigma points is calculated, and sigma points are estimated from the previous state and covariance; The sigma points are converted by a measurement function, and a priori estimation and covariance are calculated; The state equation and the measurement equation are as follows: wherein, represents a process noise affecting the system state, represents a measurement noise affecting the observation; The bistatic measurement equation is as follows: The observation equation is as follows: where the state of the observation stations i = 1, 2 at time K is .

2. The method of claim 1, wherein, The WMFS model is constructed, which comprises the following steps: A WMFS model is trained using historical data, and fuzzy rules and parameters of the trained WMFS model are determined, wherein the WMFS model is a multi-layer WMFS, each layer of the WMFS takes the output result of the upper layer as input data of the layer, and the last layer of the WMFS integrates the results.

3. The method of claim 1, wherein, The reconstructed WMFS model is combined with the UKF model to obtain the bistatic multi-target tracking model, which comprises the following steps: Input and output data pairs of a Wang-Mendel fuzzy model are determined; Weight and weight output parameters of a fuzzy set are determined according to the input and output data pairs; A nonlinear state function of the UKF model is calculated according to the weight and weight output parameters of the fuzzy set.

4. A two-station multi-target tracking apparatus based on WMFS and UKF, characterized by, The method comprises the following steps: An acquisition module is configured to construct a WMFS model for determining a signal state, and to reconstruct a state transition function using a pre-trained WMFS model; A processing module is configured to combine an optimized WMFS model with a UKF model to obtain a bistatic multi-target tracking model; An execution module is configured to track a target using the bistatic multi-target tracking model; The acquisition module comprises the following steps: A second acquisition module is configured to use a WMFS to generate state estimation and uncertainty information through internally defined fuzzy rules, and to fuse the state estimation and the uncertainty information into a UKF framework to predict and update a target state; A first processing module is configured to estimate target state and covariance through propagation and conversion of Sigma points; The first processing module comprises the following steps: A second processing sub-module is configured to determine a signal state according to a state equation, a measurement equation, a bistatic measurement equation and an observation equation of an unscented Kalman filter; The third processing submodule is configured to generate sigma points according to the signal state, and transform the sigma point set by using a state function; The fourth processing submodule is configured to calculate a priori estimation of the state and covariance of the transformed sigma points, and estimate the sigma points from the previous state and covariance; The first processing submodule is configured to convert the sigma points by using a measurement function, and calculate the priori estimation and covariance; The state equation and the measurement equation are respectively as follows: wherein, represents a process noise affecting the system state, represents a measurement noise affecting the observation; The double-station measurement equation is as follows: The observation equation is as follows: where the state of the observation stations i = 1, 2 at time K is .

5. The apparatus of claim 4, wherein, The WMFS model is constructed, including: The first acquisition module is configured to train the WMFS model by using historical data, and determine fuzzy rules and parameters of the trained WMFS model, wherein the WMFS model is a multi-layer WMFS, each layer of the WMFS takes an output result of an upper layer as input data of the layer, and a last layer of the WMFS integrates a result. 6.A computer device, comprising a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute steps of the double-station multi-target tracking method based on the WMFS and the UKF according to any one of claims 1 to 3. 7.A storage medium storing computer readable instructions, wherein the computer readable instructions are executed by one or more processors to make the one or more processors execute steps of the double-station multi-target tracking method based on the WMFS and the UKF according to any one of claims 1 to 3.

Images