Robust distributed target tracking methods, storage media and devices for radar networking

Abstract

This invention relates to a robust distributed target tracking method, storage medium, and device for radar networks, belonging to the field of target tracking technology for multi-radar networks. To address the problem of poor tracking performance caused by non-Gaussian noise in multi-radar network target tracking scenarios, this invention first accurately models the target's state equation, radar measurement equation, and non-Gaussian scintillation noise; based on maximum correlation entropy and the CKF algorithm, it determines the maximum correlation entropy (CKF) method with variance compensation, and then estimates the target's state, obtaining the estimated state value and the corresponding estimation error covariance matrix. This estimation is based on all radar nodes and is further fused with neighboring nodes using a CI strategy.

Description

Technical Field

[0001] This invention belongs to the field of target tracking technology for multi-radar networking, and relates to a target tracking method, storage medium and device for radar networking. Background Technology

[0002] With the continuous advancement of science and manufacturing technologies, radar systems have become increasingly diversified, intelligent, and networked. Single radar systems can no longer meet the actual needs of target tracking, making multi-radar network target tracking a core technology. Multi-radar systems fully utilize the measurement information from each radar, creating an information resource advantage and significantly improving radar tracking capabilities. A radar network consists of multiple radars with different topologies. Compared to centralized processing methods, distributed structures offer advantages in fault tolerance, flexibility, and reliability, making them the preferred choice.

[0003] In real-world target tracking scenarios, both the state equation and measurement equation are nonlinear, making it essentially a nonlinear state estimation problem. Designing a suitable tracking method is crucial for ensuring tracking performance. Common nonlinear estimation methods include Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), and Cumulative Kalman Filter (CKF). Compared to other algorithms, CKF does not require calculating the Jacobian matrix, achieving higher accuracy with fewer sampling points, and offers advantages in stability and computational complexity. Furthermore, interference from target position reflections and environmental changes can cause significant deviations in radar target detection, leading to noise that no longer follows a Gaussian distribution and exhibits non-Gaussian flicker noise. The aforementioned nonlinear estimation methods are derived under the Gaussian assumption; direct application can result in a sharp decrease in accuracy or even divergence. Therefore, more reasonable nonlinear non-Gaussian estimation methods are needed for accurate target tracking. Summary of the Invention

[0004] This invention aims to address the problem of poor tracking performance caused by non-Gaussian noise in target tracking scenarios involving multi-radar networks.

[0005] A robust distributed target tracking method for radar networking includes the following steps:

[0006] Step 1: Construct the target's state equation, radar tracking measurement equation, and non-Gaussian scintillation noise model;

[0007] The state equation of the target:

[0008] x k =s(x k-1 )+ω k-1

[0009] Where, x k Let s(x) represent the state quantity at time k. k-1) represents the nonlinear transfer equation for the target's state vector as a function of time, ω k-1 Zero-mean Gaussian white noise, with noise variance matrix Q k-1 ;

[0010] The measurement equation for radar target tracking is as follows:

[0011]

[0012] Where N represents the number of radars. Indicates the measured value. It is non-Gaussian flicker noise; g(x) k ) is the measurement vector;

[0013] flicker noise model, i.e., non-Gaussian flicker noise The flicker noise model is as follows:

[0014]

[0015] Where, α (i) A random variable between 0 and 1, used to represent the probability of flashing; and They represent two zero-mean values ​​and noise variances, respectively. Gaussian noise distribution generally With ω k-1 Unrelated, take a nominal noise variance Approximate flicker noise Generally acceptable

[0016] Step 2: Based on the state equation, measurement equation, and non-Gaussian scintillation noise model, the radar estimates the target's state and obtains x. k The estimated value and the corresponding estimation error covariance matrix

[0017] Step 3: Estimation based on all radar nodes and The CI strategy is used to diffuse and merge with neighboring nodes.

[0018] Furthermore, when the tracking system is two-dimensional, the measurement vector includes the slant range and azimuth angle between the radar station and the target, i.e. When the tracking system is three-dimensional, the measurement vector includes the radar station's slant range, elevation angle, and azimuth angle from the target. in These are the position coordinates in the x, y, and z directions at time k, respectively.

[0019] Furthermore, based on the state equation, measurement equation, and non-Gaussian scintillation noise model, the radar's process of estimating the target's state includes time updates and measurement updates.

[0020] For radar i, the time update process uses the CKF method to obtain the state prediction estimate at time k. The prediction estimation error covariance matrix at time k The measurement update process includes the following steps:

[0021]

[0022] in, Indicates by The matrix obtained by Cholesky decomposition;

[0023]

[0024]

[0025] in, This represents the volume point obtained from the prediction estimate and the prediction estimate error covariance matrix. This represents the new volume point obtained through the measurement equation.

[0026]

[0027] in, Indicates the predicted value of the measurement; This represents the covariance matrix of the measurement prediction estimation error. This represents the nominal noise variance of the measurement;

[0028] The state estimate at time k is obtained based on the cost function. And based on The corresponding estimation error covariance matrix is ​​obtained.

[0029] The cost function is:

[0030]

[0031] in, δ represents the state estimate at time k; σ Represents the Gaussian kernel function; x k This represents the actual value of the state variable at time k; Represents the actual value of the measurement at time k; operator A represents or

[0032] Furthermore, the state prediction estimate at time k is obtained using the CKF processing method. The prediction estimation error covariance matrix at time k The process includes the following steps:

[0033]

[0034] in, Let represent the estimation error covariance matrix at time k-1; It is by The matrix obtained by Cholesky decomposition;

[0035]

[0036] Where n is the dimension of the state, This represents the volume point obtained from the state estimate and the estimation error covariance matrix, where the subscript μ indicates the μ-th column element of the corresponding matrix. This represents the state estimate of the target at time k-1; [1] μ Represents the μ-th column of [1];

[0037]

[0038] in, s(·) represents the new volume point obtained through the nonlinear transfer function of the system state, and s(·) represents the nonlinear transfer function of the state vector.

[0039]

[0040] Where α = 1 / m; This represents the state prediction estimate at time k; Let Q represent the covariance matrix of the prediction estimation error at time k. k-1 This represents the noise variance of the system.

[0041] Furthermore, the state estimate at time k is obtained based on the cost function. During the process, first, based on the abnormal measurement information, After correction, (20) is calculated again to obtain

[0042] Furthermore, the state estimate at time k is obtained based on the cost function. During the process, first, based on the abnormal measurement information, After correction, (20) is calculated again to obtain The process includes the following steps:

[0043] The new information at time k is

[0044]

[0045] in, Indicates the predicted value of the measurement;

[0046] Under the Gaussian noise assumption, the theoretical value of the expectation of the new information covariance is expressed as:

[0047]

[0048] Based on the principle of orthogonality of new and new, (22) is generalized to

[0049]

[0050] in

[0051]

[0052] Introducing the attenuation matrix To compensate get

[0053]

[0054] If the elements on both diagonals are equal, we get

[0055]

[0056] Where n is the dimension of the state, d represents the variance of the actual information. (l) (·) indicates taking the element in the l-th row and l-th column of the matrix;

[0057] And thus obtain

[0058]

[0059] When k=1 Calculated as

[0060]

[0061] When k > 1 Calculated as

[0062]

[0063] Where λ is the forgetting factor;

[0064] The compensated variance is

[0065]

[0066] Therefore, after compensation in, This represents the covariance matrix of the prediction estimation error of the measurement after compensation.

[0067] Mutual covariance for

[0068]

[0069] By statistical linearization, the observation equation can be rewritten as follows:

[0070]

[0071] in

[0072]

[0073] in, This represents the pseudo-measurement matrix obtained by statistical linearization. This represents the measurement noise obtained from statistical linearization. This represents the measurement noise variance obtained from statistical linearization;

[0074] Substituting (34) into (20) yields

[0075]

[0076] Differentiating (38) and setting it to 0, we get

[0077]

[0078] in

[0079]

[0080] Then get

[0081]

[0082] Iterate using prediction information get Formula (42) can be rewritten as

[0083]

[0084] in

[0085]

[0086] Furthermore, based on The corresponding estimation error covariance matrix is ​​obtained. The process includes the following steps:

[0087] From (46), we can obtain

[0088]

[0089] in, It is the estimation error, I n This represents the identity matrix with 1s on the diagonal, where n is the dimension of the system state;

[0090] The state estimation error covariance matrix is ​​obtained.

[0091]

[0092] Furthermore, estimates obtained based on all radar nodes and The process of diffusion and fusion with neighboring nodes using the CI strategy includes:

[0093]

[0094]

[0095] Here, trace(·) represents the matrix trace operation.

[0096] A computer storage medium storing at least one instruction, which is loaded and executed by a processor to implement the robust distributed target tracking method for radar networking.

[0097] A robust distributed target tracking device for radar networking is disclosed. The device includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the robust distributed target tracking method for radar networking.

[0098] The beneficial effects of this invention are:

[0099] This invention first accurately models the target's state equation, radar measurement equation, and non-Gaussian scintillation noise. Second, a variance compensation method is designed, and a maximum correlation entropy (CKF) method with variance compensation is proposed based on maximum correlation entropy and the CKF algorithm. This method effectively addresses the problem of non-Gaussian noise in radar and achieves higher accuracy than other existing algorithms. Finally, the method is extended to multi-radar network systems, employing a fast CI fusion algorithm for diffusion fusion, avoiding the calculation of cross-covariance and significantly improving tracking accuracy and detection range. Attached Figure Description

[0100] Figure 1 A flowchart illustrating robust distributed target tracking for radar networking;

[0101] Figure 2 This is a network structure diagram of four radars;

[0102] Figure 3 Here are the RMSE curves for the three methods;

[0103] Figure 4 The average RMSE curves for the four radars are shown. Detailed Implementation

[0104] Specific implementation method one: Combining Figure 1 This implementation method is described below.

[0105] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0106] This implementation method is a robust distributed target tracking method for radar networking, including the following steps:

[0107] Step 1: Construct the target's state equation, radar tracking measurement equation, and non-Gaussian scintillation noise model;

[0108] The state equation of the target:

[0109] x k =s(x k-1 )+ω k-1 (3)

[0110] Where, x k Let s(x) represent the state quantity at time k. k-1 ) represents the nonlinear transfer equation for the target's state vector as a function of time, ω k-1 Zero-mean Gaussian white noise, with noise variance matrix Q k-1 , used to represent the deviation caused by modeling;

[0111] Measurement modeling for radar tracking:

[0112] The communication topology between radar networks is represented by an undirected connected graph G = (V, E), where V = {1, 2, ..., N} represents the nodes corresponding to the radars. This represents the set of all links. If radar i and radar j can communicate with each other, then radar i and radar j are neighboring nodes. The neighboring nodes of radar i can be represented as N. i ={j∈V:(j,i)∈E}∪{i}.

[0113] When radar is used to track a target, the measurement equation is:

[0114]

[0115] Where N represents the number of radars. Indicates the measured value. It is non-Gaussian flicker noise;

[0116] When the tracking system is two-dimensional, the measurements include the slant range and azimuth angle between the radar station and the target, i.e. When the tracking system is three-dimensional, the measurements include the radar station's slant range, elevation angle, and azimuth angle from the target. in These are the position coordinates in the x, y, and z directions at time k, respectively.

[0117] flicker noise model, i.e., non-Gaussian flicker noise The probability density function is expressed as

[0118]

[0119] Where, α (i) A random variable between 0 and 1, used to represent the probability of flashing; and They represent two zero-mean values ​​and noise variances, respectively. Gaussian noise distribution generally With ω k-1 Since the parameters in expression (3) are not relevant and cannot be accurately known in practice, a nominal noise variance is taken. To approximate flicker noise Generally acceptable

[0120] Step 2: Based on the state equation, measurement equation and non-Gaussian scintillation noise model (3), design a robust state estimation method to obtain the state estimates of all radars for the target.

[0121] The process of radar state estimation of a target includes time update and measurement update. Time update can generally be achieved using the CKF processing method. In measurement update, the measurement noise variance is compensated using the attenuation matrix. Based on the compensated measurement noise variance and the maximum correlation entropy criterion, each radar obtains the target state estimate and its error covariance matrix.

[0122] Maximum correlation entropy criterion:

[0123] For two random variables X and Y, their joint probability density function is F. X,Y For (x, y), its correlation entropy is defined as

[0124] V(X,Y)=∫∫κ(x,y)dF X,Y (x,y) (6)

[0125] Where x and y are variables used to represent probability densities; κ(x,y) represents a positive definite kernel function.

[0126] Commonly used Gaussian kernel functions have the following forms:

[0127]

[0128] Where e = xy is the difference between two variables x and y, and σ is the kernel width.

[0129] F X,Y (x,y) is usually not directly available; the relevant entropy can be estimated using a finite number of sampling points.

[0130]

[0131] Where e(i) = x(i) - y(i); N represents the number of sampling points, δ σ Represents the Gaussian kernel function; This represents an estimate of the correlation entropy, which reaches its maximum value if and only if X = Y.

[0132] Therefore, the estimate based on the maximum correlation entropy can be expressed as:

[0133]

[0134] in, For x k The estimated value, J(x) k ) for x k The cost function.

[0135] When radar is affected by the environment, the noise no longer follows a Gaussian distribution, resulting in flicker noise. Traditional nonlinear estimation methods such as CKF cannot directly handle non-Gaussian problems. Therefore, this paper will address non-Gaussian problems based on the maximum correlation entropy criterion and variance compensation strategy. Thus, for radar i,

[0136] Updated in time:

[0137]

[0138] in, Let represent the estimation error covariance matrix at time k-1; It is by The matrix obtained by Cholesky decomposition;

[0139]

[0140] Where n is the dimension of the state, This represents the volume point obtained from the state estimate and the estimation error covariance matrix, where the subscript μ indicates the μ-th column element of the corresponding matrix. This represents the state estimate of the target at time k-1;

[0141]

[0142]

[0143] Among them, [1] μ Represents the μ-th column of [1]; s(·) represents the new volume point obtained through the nonlinear transfer function of the system state, and s(·) represents the nonlinear transfer function of the state vector.

[0144]

[0145] Where α = 1 / m; This represents the state prediction estimate at time k;

[0146]

[0147] Let Q represent the covariance matrix of the prediction estimation error at time k. k-1 This represents the noise variance of the system;

[0148] Measurement Update:

[0149]

[0150] in, Indicates by The matrix obtained by Cholesky decomposition;

[0151]

[0152] in, This represents the volume point obtained from the prediction estimate and the prediction estimate error covariance matrix. This represents the new volume point obtained through the measurement equation, and g(·) is the same as the expression of the measurement equation and formula (2);

[0153]

[0154] in, This indicates the predicted value of the measurement.

[0155]

[0156] in, This represents the covariance matrix of the measurement prediction estimation error. This represents the new volume point obtained from the radar measurement equation. This represents the nominal noise variance of the measurement;

[0157] Define the cost function

[0158]

[0159] in, δ represents the state estimate at time k; σ Represents the Gaussian kernel function; x k This represents the actual value of the state variable at time k; Represents the actual value of the measurement at time k; operator A represents or

[0160] When the noise is non-Gaussian Mainly affected by noise variance The impact, and It cannot directly reflect the characteristics of abnormal measurements. If we first base our analysis on the abnormal measurement information... Correcting the algorithm and then recalculating (22) will inevitably further improve the accuracy and robustness of the algorithm and eliminate the impact of non-Gaussian noise on the estimation performance.

[0161] The new information at time k is defined as

[0162]

[0163] in, Indicates the predicted value of the measurement;

[0164] Under the Gaussian noise assumption, the theoretical value of the expectation of the new information covariance is expressed as:

[0165]

[0166] Based on the principle of orthogonality of new and new information, (24) can be generalized to

[0167]

[0168] in

[0169]

[0170] Formula (25) is obtained under the Gaussian assumption. When the noise variance is non-Gaussian, the resulting innovation variance is: Compared with theoretical value There is a significant deviation; therefore, an attenuation matrix is ​​introduced. To compensate This reduces the interference of outlier observations on the system, thereby improving the robustness of the estimation. Therefore, there are...

[0171]

[0172] By taking the elements on both diagonals as equal, we can obtain...

[0173]

[0174] Where n is the dimension of the state, d represents the variance of the actual information. (l) (·) indicates taking the element in the l-th row and l-th column of the matrix.

[0175] so

[0176]

[0177] definition Therefore there is

[0178]

[0179] When k=1 Calculated as

[0180]

[0181] When k > 1 Calculated as

[0182]

[0183] Where λ is the forgetting factor, which is usually taken as 0.95.

[0184] The compensated variance is

[0185]

[0186] Therefore, after compensation

[0187]

[0188] in, This represents the covariance matrix of the prediction estimation error of the measurement after compensation.

[0189] Cross-covariance of x and z for

[0190]

[0191] By statistical linearization, the observation equation can be rewritten as follows:

[0192]

[0193] in

[0194]

[0195] in, This represents the pseudo-measurement matrix obtained by statistical linearization. This represents the measurement noise obtained from statistical linearization. This represents the measurement noise variance obtained from statistical linearization;

[0196] Substituting (34) into (22) yields

[0197]

[0198] Differentiating (40) and setting it to 0, we get

[0199]

[0200] in

[0201]

[0202] Then we can obtain

[0203]

[0204] Since (44) is a condition where both sides are about x k The function is a fixed-point equation, which can be solved using methods such as Newton's iteration, or as follows:

[0205] Iterate using prediction information Satisfactory estimation accuracy can be obtained, therefore, Formula (44) can be rewritten as

[0206]

[0207] in

[0208]

[0209] From (46), we can obtain

[0210]

[0211] in, It is the estimation error, I n Let n be the identity matrix with 1s on the diagonal, and n be the dimension of the system state; therefore, the state estimation error covariance matrix is...

[0212]

[0213] Step 3: Obtain local estimates based on all radar nodes from Step 2. and Diffusion and fusion with neighboring nodes using CI strategy

[0214]

[0215]

[0216] Here, trace(·) represents the matrix trace operation.

[0217] Therefore, the robust distributed target tracking method for radar networking proposed in this invention can be summarized as Method 1.

[0218]

[0219] Example:

[0220] To verify the effectiveness of this invention, a scenario of two-dimensional radar tracking a turning target was selected, with the state equation and measurement equation being as follows:

[0221]

[0222] in x k , y k , Ω k These represent the position and velocity in the x-direction, and the position, velocity, and turning rate in the y-direction at time k, respectively. A distributed network consisting of four sensors is used for tracking. The sensor locations are (1000m, 300m), (1500m, 300m), (1000m, -300m), and (1500m, -300m), respectively. Their topology is as follows: Figure 2 As shown.

[0223] Simulation time T = 0.1s, initial state x0 = [1000, 20, 1000, 0, -π / 60] T , P0=diag([1000,100,1000,100,2 / 100]), q1=0.2, q2=1.75×10- 5 , Process noise ω k variance Q k for

[0224] Q k =diag([q1M; q1M; q2T]) (54)

[0225] Measurement noise The nominal noise is Measurement noise It is non-Gaussian flicker noise, i.e.

[0226]

[0227] To verify the effectiveness of the invention, the root mean square error (RMSE) of the location was selected as a performance indicator, defined as follows:

[0228]

[0229] Among them, (x k ,y k )and These represent the actual and estimated positions of the i-th node and the j-th experiment, respectively, where M is the number of Monte Carlo experiments. This is the simulated step size.

[0230] Taking radar 1 as the comparison object, the CKF and MCKF methods are compared with the present invention. The RMSE curves of the three methods are as follows: Figure 3 As shown. From Figure 3 As can be seen from the results, the RMSE of this invention is lower and the tracking accuracy is higher. This is because the method introduces a variance compensation strategy and a distributed fusion strategy, thereby improving the tracking accuracy. Figure 4 The average RMSE value of the four radars shows that all radars achieved consistent tracking accuracy. Therefore, this invention is more suitable for scenarios involving networked multi-radar target tracking and has greater application value. Specific Implementation Method Two:

[0232] This embodiment is a computer storage medium that stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the robust distributed target tracking method for radar networking.

[0233] It should be understood that the instructions include computer program products, software, or computerized methods corresponding to any method described in this invention; the instructions can be used to program computer systems or other electronic devices. Computer storage media may include readable media on which instructions are stored, and may include, but are not limited to, magnetic storage media, optical storage media; magneto-optical storage media include read-only memory (ROM), random access memory (RAM), erasable programmable memory (e.g., EPROM and EEPROM), and flash memory layers, or other types of media suitable for storing electronic instructions. Specific implementation method three:

[0235] This embodiment is a robust distributed target tracking device for radar networking. The device includes a processor and a memory. It should be understood that it includes any device including a processor and a memory described in this invention. The device may also include other units and modules that perform display, interaction, processing, control and other functions through signals or instructions.

[0236] The memory stores at least one instruction, which is loaded and executed by the processor to implement the robust distributed target tracking method for radar networking.

[0237] Those skilled in the art will understand that at least one stored instruction constitutes a computer program product corresponding to a method or system. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0238] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application, and can also be used with corresponding devices. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0239] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0240] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0241] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0242] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

[0243] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A robust distributed target tracking method for radar networking, characterized in that, Includes the following steps: Step 1: Construct the target's state equation, radar tracking measurement equation, and non-Gaussian scintillation noise model; The state equation of the target: in, This represents the state quantity at time k. The nonlinear transfer equation representing the state vector of the target as a function of time. Zero-mean Gaussian white noise, with noise variance matrix as follows: ; The measurement equation for radar target tracking is as follows: in, Indicates the number of radars, Indicates the measured value. It is non-Gaussian flicker noise; For measurement vectors; flicker noise model, i.e., non-Gaussian flicker noise The flicker noise model is as follows: in, A random variable between 0 and 1, used to represent the probability of flashing; and They represent two zero-mean values ​​and noise variances, respectively. Gaussian noise distribution ;generally and Unrelated, take a nominal noise variance Approximate flicker noise It is usually advisable ; Step 2: Based on the state equation, measurement equation, and non-Gaussian scintillation noise model, the radar estimates the target's state, obtaining... The estimated value and the corresponding estimation error covariance matrix ; The process of radar estimating the state of a target includes both time updates and measurement updates. For radar The time update process uses the CKF processing method to obtain the state prediction estimate at time k. The prediction estimation error covariance matrix at time k The measurement update process includes the following steps: in, Indicates by The matrix obtained by Cholesky decomposition; in, This represents the volume point obtained from the prediction estimate and the prediction estimate error covariance matrix. This represents the new volume point obtained through the measurement equation. in, Indicates the predicted value of the measurement; This represents the covariance matrix of the measurement prediction estimation error. This represents the nominal noise variance of the measurement; Based on the cost function, we obtain State estimation at time 1 and based on The corresponding estimation error covariance matrix is ​​obtained. ; The cost function is: （20） in, express State estimation at time; Represents the Gaussian kernel function; This represents the actual value of the state variable at time k; Represents the actual value of the measurement at time k; operator , represent or ; The state prediction estimate at time k is obtained using the CKF processing method. The prediction estimation error covariance matrix at time k The process includes the following steps: in, express The covariance matrix of the estimation error at time; It is by The matrix obtained by Cholesky decomposition; in, Let be the dimension of the state. The subscript represents the volume point obtained from the state estimate and the estimation error covariance matrix. Represents the first element of the corresponding matrix Column elements, express State estimation of the target at any given time; express The μth column; in, This represents the new volume point obtained through the transfer of the system state nonlinear function. A nonlinear transfer function representing a state vector; in, ; This represents the state prediction estimate at time k; This represents the prediction estimation error covariance matrix at time k. This represents the noise variance of the system; Step 3: Estimation based on all radar nodes and It utilizes the CI strategy to diffuse and merge with neighboring nodes.

2. The robust distributed target tracking method for radar networking according to claim 1, characterized in that, When the tracking system is two-dimensional, the measurement vector includes the slant range and azimuth angle between the radar station and the target, i.e. When the tracking system is three-dimensional, the measurement vector includes the radar station's slant range, elevation angle, and azimuth angle from the target, i.e. ,in They are respectively The position coordinates in the x, y and z directions at time 1.

3. A robust distributed target tracking method for radar networking according to claim 2, characterized in that, Based on the cost function, we obtain State estimation at time 1 During the process, first, based on the abnormal measurement information, Perform corrections, and then calculate formula (20) to obtain... .

4. A robust distributed target tracking method for radar networking according to claim 3, characterized in that, Based on the cost function, we obtain State estimation at time 1 During the process, first, based on the abnormal measurement information, Perform corrections, and then calculate formula (20) to obtain... The process includes the following steps: The new information at time k is in, Indicates the predicted value of the measurement; Under the Gaussian noise assumption, the theoretical value of the expectation of the new information covariance is expressed as: （22） Based on the principle of orthogonality of new and new, (22) is generalized to in Introducing the attenuation matrix To compensate ,get If the elements on both diagonals are equal, we get in, Let be the dimension of the state. The variance of the actual information. Indicates taking the first matrix. Line number Column elements; And thus obtain when hour, Calculated as when hour, Calculated as in, Forgetting factor; The compensated variance is Therefore, after compensation ,in, This represents the covariance matrix of the prediction estimation error of the measurement after compensation. Mutual covariance for By statistical linearization, the observation equation can be rewritten as follows: （34） in in, This represents the pseudo-measurement matrix obtained by statistical linearization. This represents the measurement noise obtained from statistical linearization. This represents the measurement noise variance obtained from statistical linearization; Substituting (34) into (20) yields （38） Differentiating (38) and setting it to 0, we get in Then get （42） Iterate using prediction information get , Formula (42) can be rewritten as in （44）。 5. A robust distributed target tracking method for radar networking according to claim 4, characterized in that, based on The corresponding estimation error covariance matrix is ​​obtained. The process includes the following steps: From (44), we can obtain （45） in, It is the estimation error. This represents the identity matrix with 1s on the diagonal. Let be the dimension of the system state; The state estimation error covariance matrix is ​​obtained. 。 6. A robust distributed target tracking method for radar networking according to claim 5, characterized in that, Estimates based on all radar nodes and The process of diffusion and fusion with neighboring nodes using the CI strategy includes: in, This indicates the matrix trace operation.

7. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement a robust distributed target tracking method for radar networking as described in any one of claims 1 to 6.

8. A robust distributed target tracking device for radar networking, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement a robust distributed target tracking method for radar networking as described in any one of claims 1 to 6.

Images