A dual-entropy fusion multi-random matrix maneuvering extended target robust tracking method

Abstract

This invention discloses a robust target tracking method using a dual-entropy fusion multi-random matrix maneuvering extension. The method employs an interactive multi-model fusion architecture. In the fusion step, a cost function is established using the target motion state estimate of the previous time-instance sub-model as the independent variable, based on the correlation entropy criterion. Maximizing this cost function yields the fused target state estimate. Using relative entropy as the criterion, the sum of information gains of each sub-model regarding the target motion state covariance and target morphological distribution parameters is minimized. Then, a correlation entropy cost function is established using the filtered target motion state estimate of the current time-instance sub-model as the independent variable. Maximizing this cost function achieves target motion state estimation fusion. Simultaneously, the total information gain of each sub-model regarding the target motion state covariance and target morphological distribution parameters is minimized again to obtain a weighted estimate, ultimately yielding a robust joint estimate of the target motion state and morphology at the current time.

Description

Technical Field

[0001] This invention relates to the field of target tracking, and more particularly to a robust target tracking method using dual-entropy fusion multi-random matrix maneuvering. Background Technology

[0002] Traditional target tracking research only needs to focus on the basic dynamic information of a point mass. However, with the development of sensor technology, its resolution is constantly improving, breaking through the limitation of only receiving one measurement in each scanning cycle due to the precision constraints of point target tracking. Multiple measurements can now be obtained in each cycle, making the target's contour or shape features non-negligible. This type of tracking problem is called extended target tracking. On the one hand, the improved quality of target observation makes it possible to estimate extended information beyond target dynamics, such as size, shape, and orientation. On the other hand, the increased sensor resolution makes it exceptionally sensitive to disturbances outside the target in the monitoring environment. Furthermore, due to the unknown dynamic characteristics of the target, the target maneuvering problem existing in point target tracking still exists in extended target tracking applications. These highly maneuvering behaviors with randomness and unpredictability introduce more severe and complex non-Gaussian noise interference to the sensor, thus posing challenges to effective observation and accurate analysis.

[0003] Jumping Markov models are typical models for solving maneuvering target tracking. These models do not have an optimal general solution, but various suboptimal solution methods exist, such as pseudo-Bayesian, interactive multiple model (MMM), and particle filtering. Among these, the MMM is widely studied and applied due to its good performance in terms of estimation accuracy and computational cost. An inherent drawback of the MMM method is that it can only retain at most second-order relevant information, ignoring higher-order information. In the problem of tracking maneuvering extended targets, the existence of target maneuvering behavior inevitably introduces clutter into the target observation. This causes the MMM method to discard useful higher-order information during model fusion, resulting in a failure to obtain robust estimation results with high reliability. Furthermore, for extended target tracking, it is necessary to consider not only the target dynamic state variables but also the target extension. The target's maneuvering behavior affects not only the normal target state estimation but also the estimation of the target extension, i.e., the target morphology. This means that the MMM method must robustly fuse both the target state variables and the target morphology variables, which is something that the classical MMM method system cannot directly achieve.

[0004] The concepts of relevant entropy and relative entropy used in this invention both originate from information theory. The former has been successfully applied to robust machine learning based on information theory in kernel adaptive filtering problems, while the latter, also known as Kullback-Leibler (KL) divergence, is often used to compare the KL distance between two probability distributions. This invention focuses on solving the robust tracking problem of maneuvering extended targets based on stochastic matrix modeling using an interactive multi-model approach. It uses these two concepts from information theory as criteria for multi-model information fusion, enabling the efficient fusion of relevant model information on target state and morphology. This improves upon the robustness degradation problem caused by neglecting and discarding higher-order information in the model during model fusion when using traditional methods for maneuvering extended target tracking. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a robust target tracking method based on dual-entropy fusion and multi-random matrix maneuvering.

[0006] The objective of this invention is achieved through the following technical solution: a robust tracking method for extended targets using a dual-entropy fusion multi-random matrix maneuver, comprising the following steps: S1. Initialize the parameters of the linear jumping Markov model for tracking maneuvering targets. The model contains several sub-models. S2. Based on the prior probabilities of the sub-models in the previous time step, calculate the mixture probability among the sub-models; S3. Based on the maximum correlation entropy criterion, combined with the target motion state covariance and mixture probability of the previous time step, the cost function of the target motion state estimation in the sub-model of the previous time step is solved by iterative method to obtain the mixture estimate of the target motion state of the previous time step. Based on the minimum weighted relative entropy criterion and mixture probability, the fusion value of the target motion state covariance of each sub-model and the fusion value of the target morphological parameters are obtained. S4. Predict the current time value of the fused values ​​of the mixed target motion state parameters, the target motion state covariance, and the target morphology parameters. S5. Update the predicted values ​​using the latest observations as filtered estimates; S6. Update the posterior probabilities of the sub-model; S7. Based on the maximum correlation entropy criterion and posterior probability, the cost function of the target motion state filtering estimate in the sub-model at the current moment is solved by iterative method to obtain the final estimate of the target state at the current moment; based on the minimum weighted relative entropy criterion and posterior probability, the filtered estimate of the target motion state covariance and the filtered estimate of the target morphological distribution parameters are respectively fused. S8. Use the filtered estimates of each sub-model at the current time as the initial values ​​of the filtered estimates and variance matrix of the target motion state and target shape of each sub-model at the next time. Return to S2 and repeat the process sequentially until the extended target tracking process ends.

[0007] Furthermore, the linear jumping Markov model for tracking maneuvering targets includes: ; Among them, subscript Representing the Discrete time moments Represents the target state variable. Represents the sensor's observations. and These represent the state transition matrix and the observation matrix, respectively. and These represent the Gaussian process noise during system state transition and the Gaussian observation noise during sensor observation, respectively. and Each has a known process noise variance and observation noise variance ; superscript This means that the data belongs to the first The total number of sub-models, the jumping Markov model has There are several sub-models, and the switching between each sub-model is determined by the sub-model state transition matrix.

[0008] Furthermore, the calculation of the mixture probability between the various sub-models specifically includes: Using the prior probabilities of the sub-model from the previous time step Sub-model state transition matrix Using Bayes' theorem, the mixture probability of the sub-model is obtained. as follows: ; Where the superscript 'l' represents the 'l'-th sub-model, and there are a total of Sub-model.

[0009] Furthermore, S3 specifically includes: The iterative solution for the maximum correlation entropy is: ; superscript This means the data comes from the first... iteration loop Represents the Gaussian kernel function in the correlation entropy, the letter... and Number the model. Representative sub-model and The mixed probability between and Each represents a sub-model exist The state estimate and covariance matrix at time t are both known quantities; Let the unknown be the solution, and let its solution be denoted as... ; If the last iteration is denoted as Then the final state estimation expression should be: ; And matrix The definition is as follows: ; The covariance matrices of each sub-model are mixed using an information fusion strategy based on minimum weighted relative entropy. ; The resulting covariance matrix; Given The two morphological parameters of the target at any given time are the degrees of freedom. and scale matrix The two target morphological parameters are then mixed using a weighted relative entropy strategy, resulting in the following weighted relative entropy morphological parameters: ; .

[0010] Furthermore, the current-time prediction of the fused values ​​of the mixed target motion state parameters, the target motion state covariance, and the target morphology parameters specifically includes: For the target motion state respectively and its covariance matrix Make a prediction: ; ; in, Representing Kronecker, The dimension extended for the target, The dimension is The identity matrix; Let be the process noise variance matrix for the target state transition. The target state transition matrix; Degrees of freedom of the morphological parameters of the target and scale matrix Make a prediction: ; ; in, The time decay constant, The sampling time is used for observation.

[0011] Furthermore, updating the predicted values ​​using the latest observations includes: Obtain the latest extended target observation dataset Then, update the target motion state respectively. and its covariance matrix : ; ; in, The dimension extended for the target, The dimension is The identity matrix; intermediate function , and for: ; ; ; ; in, for The base number represents the quantity measured; Then update the target shape parameters, i.e., degrees of freedom. and scale matrix : ; .

[0012] Furthermore, updating the posterior probability of the sub-model specifically includes: Based on likelihood probability function Update sub-model posterior probability : ; in and These are the weight coefficients of the sub-model's state transition matrix. Let M be the posterior probability at time k-1, and M be the total number of sub-models. Where the likelihood probability function for: ;in, for The base number represents the quantity measured. The dimension is The identity matrix.

[0013] Further, S7 includes: The iterative estimation solution for the maximum correlation entropy is as follows: ; in, Let S6 be the posterior probability of the sub-model obtained. For the variable to be determined, For the final solution; superscript This means the data comes from the first... iteration loop Define matrix : ; The last iteration is denoted as The final solution expression is: ; A weighted relative entropy-based information fusion strategy is used to seek the fusion covariance matrix. : ; The two morphological parameters of the target are fused using a weighted relative entropy strategy, and the final fused relative entropy weighted morphological parameters are as follows: ; .

[0014] On the other hand, the specification also provides a robust tracking device for extended targets using a dual-entropy fusion multi-random matrix maneuver, including a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements the aforementioned robust tracking method for extended targets using a dual-entropy fusion multi-random matrix maneuver.

[0015] On the other hand, the specification also provides a computer-readable storage medium on which a program is stored, which, when executed by a processor, implements the aforementioned dual-entropy fusion multi-random matrix maneuvering extended target robust tracking method.

[0016] The beneficial effects of this invention are: in the interaction process of multiple sub-models of the extended target, more high-order information is preserved and transmitted, rather than being limited to the transmission of at most second-order information in traditional methods. Therefore, when the maneuvering extended target moves under sensor observation conditions with outlier interference, the tracking method of this invention can reduce the sensitivity to outliers without additional robust filtering estimation, thereby improving the robustness of extended target tracking under these conditions using the interactive multi-model framework. Since this invention still relies on the interactive multi-model framework and does not introduce other robust data processing techniques, and can integrate target morphology information represented by ellipses, it has good scalability. It can be further combined with robust filtering techniques to further improve the robustness of target tracking in strongly non-Gaussian scenarios, or the extended target morphology can be modeled using rectangular, star-convex, or other models and jointly estimated with the motion state. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method execution flow provided in the embodiments of the present invention; Figure 2 This is a schematic diagram of the device provided in an embodiment of the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0019] like Figure 1 As shown, the proposed dual-entropy fusion multi-random matrix maneuvering extended target robust tracking method includes the following steps: Step 1: Initialize the parameters of each target motion model; First, we present a linear, skipping Markov model for tracking maneuvering targets: (1) Among them, subscript Representing the Discrete time moments This represents the target's state variables, including physical quantities such as the target's position and velocity. Represents the sensor's observations. and These represent the state transition matrix and the observation matrix, respectively. and These represent the Gaussian process noise during system state transition and the Gaussian observation noise during sensor observation, respectively. and Each has a known process noise variance and observation noise variance ,use Representing variables Follow the mean variance is If it is a Gaussian distribution, then , Superscript This means that the data belongs to the first There are a total of sub-models. Let the jumping Markov model have a total of... Each sub-model has a state transition matrix, and the transitions between sub-models are determined by the sub-model state transition matrix. Decision, among which and Representing the first and the Sub-models, satisfying Weight coefficients satisfy It represents the previous moment. From the model Transfer to model The probability of.

[0020] If we consider a maneuvering extended target, at the initial time zero, in addition to the variables mentioned above... , , , , and Assigning values ​​also requires initialization to conform to the inverse Wittsaud distribution. Target expansion ,in and These are the degrees of freedom parameters and the scale matrix, respectively, used to initialize the prior of any sub-model (relative to...). (Probability in terms of time) .

[0021] Step 2: Calculate the mixture probability between each sub-model; Known sub-model prior probabilities (i.e., the posterior probability of the sub-model at the previous time step), using the sub-model state transition matrix Using Bayes' theorem, the mixture probability of the sub-model can be obtained. as follows (2) Step 3: Based on the maximum correlation entropy criterion, use the iterative method to solve the cost function for the target motion state estimation in the sub-model of the previous time step, and obtain the mixed estimate of the target motion state of the previous time step; based on the minimum weighted relative entropy criterion, obtain the fusion value of the target motion state covariance and the fusion value of the target morphological distribution parameters respectively. The fusion step involves multi-model interaction and requires mixing the estimation results from the various sub-models at the previous time step. A relevant entropy cost function is established for the target state estimates from the sub-models at the previous time step. (3) in, Representative matrix Weighted vector The 2-norm, This represents the Gaussian kernel function, and its expression is: , Represents the error variable. This represents the Gaussian kernel bandwidth. and Each represents a sub-model in The state estimate and covariance matrix at time t are both known quantities. Let the unknown be the solution, and let its solution be denoted as... Maximizing the relevant entropy is equivalent to maximizing the cost function. Taking the derivative of this cost function using a gradient algorithm yields an iterative solution for the state estimate: (4) Wherein, the matrix is ​​defined. as follows: (5) superscript This means the data comes from the first... This is the 10th iteration cycle. If the last iteration is denoted as... Then the final state estimation expression should be: .

[0022] It's important to note that the solution obtained based on the maximum correlation entropy criterion is about the state variables and doesn't explicitly relate to the covariance matrix. Therefore, we consider using an information fusion strategy based on minimum weighted relative entropy to mix the covariance matrices of each sub-model. The minimum weighted relative entropy theorem is known as follows: (6) Where is a random variable , The probability density function after mixing. For any number A probability density function that needs to be mixed. Here, represents the weighting factor. In this invention, applying the above theorem, the resulting covariance matrix after mixing is: (7) Given The two morphological parameters of the target at any given time are the degrees of freedom. and scale matrix We continue to use the weighted relative entropy strategy to mix the two target morphological parameters, as shown in the following expressions: (8) (9) Step 4: Predict the mixed target motion state and target shape obtained in Step 3 respectively; Since this invention targets linear, jumping Markov models, it can obtain the target state and covariance matrix in... Predicted value at time , They are represented as follows: (10) (11) in, Representing Kronecker, The dimension extended for the target, The dimension is The identity matrix; Let be the process noise variance matrix for the target state transition. Let be the target state transition matrix.

[0023] Then, consider the morphological parameters and degrees of freedom of the target. and scale matrix Make a prediction: (12) (13) in, The time decay constant, The sampling time is used for observation.

[0024] Step 5: Update the target motion state and target shape from Step 4 using the latest observations to obtain the filtered estimate; Obtain the latest extended target measurement dataset Then, update the target motion state and its covariance matrix respectively: (14) (15) in (16) (17) (18) (19) in, for The base number represents the quantity of measurements.

[0025] Then update the target morphological parameters, namely the degrees of freedom and scale matrix: (20) (twenty one) Step 6: Update the model probabilities of the sub-models; Sub-model The posterior probability is: (twenty two) Where the likelihood probability function for: (twenty three) Step 7: Based on the maximum correlation entropy criterion, use an iterative method to solve the cost function for the filtered estimation of the target motion state in the sub-model at the current moment, and obtain the final estimate of the target state at the current moment; based on the minimum weighted relative entropy criterion, fuse the filtered estimate of the target motion state covariance and the filtered estimate of the target morphological distribution parameters respectively. This section again involves multi-model interaction, but unlike the interaction in step three, this involves the estimation fusion between sub-models. A relevant entropy cost function is established for the sub-model state estimates at the current time. as follows: (twenty four) in and The target state and covariance matrix obtained in step five are respectively... The update value at any given time. This represents the posterior probability of the sub-model obtained in step six. Let be the variable to be solved, and let its final solution be denoted as... The iterative state estimation solution obtained by differentiation is as follows: (25) Where a matrix is ​​defined : (26) The last iteration is denoted as The final solution expression is: .

[0026] A weighted relative entropy-based information fusion strategy is used to seek the fusion covariance matrix. : (27) The two morphological parameters of the target are fused using a weighted relative entropy strategy, and the final fused relative entropy weighted morphological parameters are as follows: (28) (29) The random matrix is ​​calculated using the following formula. (30) The matrix The eigenvalues ​​and eigenvectors are the squares of the semi-axis lengths and directions of the ellipse representing the target, respectively, thus determining the estimate of the extended target shape at the current moment.

[0027] Step 8: Use the filtered estimates of each sub-model at the current moment as the initial values ​​for the filtered estimates of the target motion state and target shape of each sub-model in Step 3 at the next moment, and then return to Step 2 to execute them sequentially until the extended target tracking process ends.

[0028] Retain the estimates of each sub-model at the current time step. and the corresponding variance matrix This allows their assignments to be passed to the next time step, and then... Return to step two and execute the remaining steps in sequence, repeating the above loop until target tracking stops.

[0029] Corresponding to the aforementioned embodiment of a robust tracking method for extended targets using a multi-random matrix maneuvering system based on dual-entropy fusion, the present invention also provides an embodiment of a robust tracking device for extended targets using a multi-random matrix maneuvering system based on dual-entropy fusion.

[0030] See Figure 2 The present invention provides a robust tracking device for extended targets using a dual-entropy fusion multi-random matrix maneuver, comprising a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it implements a robust tracking method for extended targets using a dual-entropy fusion multi-random matrix maneuver as described in the above embodiments.

[0031] The embodiment of the dual-entropy fusion multi-random matrix maneuvering extended target robust tracking device provided by this invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 2The diagram shown is a hardware structure diagram of any device with data processing capabilities, where the dual-entropy fusion multi-random matrix maneuvering extended target robust tracking device provided by the present invention is located. (Except for...) Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0032] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0033] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0034] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a robust target tracking method for dual-entropy fusion multi-random matrix maneuvering described in the above embodiments.

[0035] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0036] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned dual-entropy fusion multi-random matrix maneuvering extended target robust tracking method.

[0037] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0038] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. This application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for robust tracking of a maneuvering extended target with dual-entropy fusion and multiple random matrices, characterized in that, Includes the following steps: S1. Initialize the parameters of the linear jumping Markov model for tracking maneuvering targets. The model contains several sub-models. S2. Based on the prior probabilities of the sub-models in the previous time step, calculate the mixture probability among the sub-models; S3. Based on the maximum correlation entropy criterion, combined with the target motion state covariance and mixture probability of the previous time step, the cost function of the target motion state estimation in the sub-model of the previous time step is solved by iterative method to obtain the mixture estimate of the target motion state of the previous time step. Based on the minimum weighted relative entropy criterion and mixture probability, the fusion value of the target motion state covariance of each sub-model and the fusion value of the target morphological parameters are obtained. S4. Predict the current time value of the fused values ​​of the mixed target motion state parameters, the target motion state covariance, and the target morphology parameters. S5. Update the predicted values ​​using the latest observations as filtered estimates; S6. Update the posterior probabilities of the sub-model; S7. Based on the maximum correlation entropy criterion and posterior probability, the cost function of the target motion state filtering estimate in the sub-model at the current moment is solved by iterative method to obtain the final estimate of the target state at the current moment; based on the minimum weighted relative entropy criterion and posterior probability, the filtered estimate of the target motion state covariance and the filtered estimate of the target morphological distribution parameters are respectively fused. S8. Use the filtered estimates of each sub-model at the current time as the initial values ​​of the filtered estimates and variance matrix of the target motion state and target shape of each sub-model at the next time. Return to S2 and repeat the process sequentially until the extended target tracking process ends.

2. The method of claim 1, wherein The linear jumping Markov model for tracking maneuvering targets includes: ； Among them, subscript Representing the Discrete time moments Represents the target state variable. Represents the sensor's observations. and These represent the state transition matrix and the observation matrix, respectively. and These represent the Gaussian process noise during system state transition and the Gaussian observation noise during sensor observation, respectively. and Each has a known process noise variance and observation noise variance ; superscript This means that the data belongs to the first The total number of sub-models, the jumping Markov model has There are several sub-models, and the switching between each sub-model is determined by the sub-model state transition matrix.

3. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, The calculation of the mixture probability between the various sub-models specifically includes: Using the prior probabilities of the sub-model from the previous time step Sub-model state transition matrix Using Bayes' theorem, the mixture probability of the sub-model is obtained. as follows: ； Among them, subscript l Representing the l Sub-model, This represents the sub-model state transition matrix between the i-th and j-th sub-models, and has a total of Sub-model.

4. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, S3 specifically includes: The iterative solution for the maximum correlation entropy is: ； superscript This means the data comes from the first... iteration loop The subscript represents the Gaussian kernel function in the correlation entropy. and Number the model. Representative sub-model and The mixed probability between and Each represents a sub-model exist The state estimate and covariance matrix at time t are both known quantities; Let the unknown be the solution, and let its solution be denoted as... ; The last iteration is denoted as The final state estimation expression is: ; And matrix The definition is as follows: ； The covariance matrices of each sub-model are mixed using an information fusion strategy based on minimum weighted relative entropy. ； The resulting covariance matrix; Given The two morphological parameters of the target at any given time are the degrees of freedom. and scale matrix The two target morphological parameters are then mixed using a weighted relative entropy strategy to obtain the mixed relative entropy weighted morphological parameters. and They are as follows, in order: ； 。 5. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, The current-time prediction of the fused values ​​of the mixed target motion state parameters, the target motion state covariance, and the target morphology parameters specifically includes: For the target motion state respectively and its covariance matrix Make a prediction: ； ； in, Representing Kronecker, The dimension extended for the target, The dimension is The identity matrix; Let be the process noise variance matrix for the target state transition. The target state transition matrix; Degrees of freedom of the morphological parameters of the target and scale matrix Make a prediction: ； ； in, The time decay constant, The sampling time is used for observation.

6. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, The step of updating the predicted values ​​using the latest observations includes: Obtain the latest extended target observation dataset Then, update the target motion state respectively. and its covariance matrix : ； ； in, The dimension extended for the target, The dimension is The identity matrix; , They are in The predicted values ​​of the target state and covariance matrix at each time step; intermediate function , and for: ； ； ； ； in, for The base number represents the quantity measured; The observation matrix of the j-th sub-model Then update the target shape parameters, i.e., degrees of freedom. and scale matrix : ； 。 7. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, The update of the posterior probability of the sub-model specifically includes: Based on likelihood probability function Update sub-model posterior probability : ； in and These are the weight coefficients of the sub-model's state transition matrix. Let M be the posterior probability at time k-1, and M be the total number of sub-models. Represents the observation matrix. Where the likelihood probability function for: ;in, for The cardinality represents the number of measurements, and N is the probability density function obtained from a Gaussian distribution. The dimension is The identity matrix, Representing the target state and covariance matrix in Predicted value at time, This is the observation matrix.

8. The robust target tracking method for dual-entropy fusion multi-random matrix maneuvering according to claim 1, characterized in that, S7 includes: The iterative estimation solution for the maximum correlation entropy is as follows: ； in, Let S6 be the posterior probability of the sub-model obtained. The Gaussian kernel function represents the correlation entropy. This is the covariance matrix updated by S5. For the variable to be determined, For the final solution; superscript This means the data comes from the first... iteration loop Define matrix : ； The last iteration is denoted as The final solution expression is: ; A weighted relative entropy-based information fusion strategy is used to seek the fusion covariance matrix. : ； The two morphological parameters of the target are fused using a weighted relative entropy strategy, and the final fused relative entropy weighted morphological parameters are as follows: ； 。 9. A robust target tracking device for dual-entropy fusion multi-random matrix maneuvering, comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that... When the processor executes the executable code, it implements a dual-entropy fusion multi-random matrix maneuver extended target robust tracking method as described in any one of claims 1-8.

10. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a robust target tracking method for dual-entropy fusion multi-random matrix maneuvering as described in any one of claims 1-8.

Images