A radar target tracking method based on transport mapping

Abstract

The present invention discloses a radar target tracking method based on transport mapping, which adopts a pulse-by-pulse estimation strategy, including initial construction, initial particle generation and pulse-by-pulse update steps. The target transfer equation and radar measurement equation are initially constructed to generate an equal-weighted particle set to represent the initial probability distribution. At each pulse moment, the transfer equation is first used to predict the target state and calculate the measurement prediction value, the high-dimensional measurement signal and the prediction value are downsampled, and then the transport mapping is constructed based on the state prediction and downsampled data, the particles are updated to obtain the posterior state set, and finally the average output target state estimate is calculated. This method solves the problems of particle degradation and computational complexity of high-dimensional measurement data processing in existing methods, improves the target state estimation accuracy and tracking efficiency, and is particularly suitable for complex nonlinear and high-noise environments. This method can process high-dimensional measurement data and multimodal distributions, avoid particle degradation, enhance real-time and robustness, and has important application value.

Description

Technical Field

[0001] The present invention relates to radar target tracking technology, and in particular to a radar target tracking method based on transport mapping. Background Art

[0002] Radar target tracking technology has widespread applications in a variety of fields, including low-altitude economics, communications and perception, and public safety. Its core task is to estimate the target's dynamic state, such as position, velocity, and acceleration, in real time using radar echo signals. Existing methods typically employ a frame-by-frame approach, accumulating multiple pulse echoes into coherent data frames and processing them in the range-Doppler domain to extract target motion information. Subsequently, Bayesian filtering methods are used to estimate the target state frame by frame. Common filtering methods can be categorized into Kalman filtering and particle filtering.

[0003] Kalman filter methods assume that the posterior distribution of the target state is Gaussian, and can effectively handle state estimation problems in linear and Gaussian scenarios. However, when the measurement model is highly nonlinear or the noise is non-Gaussian, the performance of Kalman filtering will degrade significantly. In contrast, particle filter methods use weighted particles to approximate the posterior distribution of the target state. They are suitable for nonlinear and non-Gaussian environments, but are prone to particle degradation during the filtering process, that is, the weights of some particles decay to zero, resulting in a decrease in the accuracy of the distribution approximation. Especially in radar target tracking scenarios, because the measurement model is usually high-dimensional and highly nonlinear, and non-Gaussian noise is present, these problems further exacerbate the performance degradation of commonly used filtering methods.

[0004] In recent years, Bayesian filtering methods based on transport mapping have gained increasing attention. This method represents the prior distribution of the target state via a set of equally weighted particles and updates the particle state values ​​at each time step using a transport mapping. This method effectively avoids particle degeneracy and supports the approximation of multimodal distributions.

[0005] While existing target tracking technologies have met the dynamic state estimation requirements of radar systems to a certain extent, they still have limitations when dealing with high-dimensional measurement data and complex environments. First, existing radar target tracking methods typically rely on frame-by-frame processing, which reduces the real-time performance and accuracy of target tracking in highly dynamic target scenarios. Second, Kalman filtering, a commonly used Bayesian filter, can lead to large estimation errors when dealing with highly nonlinear measurement models or non-Gaussian noise. While particle filtering methods are suitable for nonlinear and non-Gaussian environments, they are prone to particle degeneration during the filtering process in high-dimensional systems, resulting in a decrease in the accuracy of the posterior distribution approximation. For example, Snyder et al. constructed a high-dimensional simplified Gaussian model to analyze the performance of particle filtering. In this experiment, the system dimension was 100, the observation error and state error both followed an independent and identically distributed unit Gaussian distribution, and 1000 particles were used for estimation. Simulation results show that under this configuration, the average particle weight exceeds 0.8, indicating that the particle weights are severely concentrated, with only a very small number of particles making a substantial contribution to the posterior estimation result, resulting in significant particle degeneration. Further analysis shows that due to particle degradation, the mean squared error of the posterior mean estimation reaches 127, far exceeding the theoretical expected value of 50 for the posterior mean error, and thus failing to effectively characterize the true posterior distribution. It is important to note that in practical engineering applications such as radar target tracking, measurement models often have high dimensionality and highly nonlinear characteristics. This not only exacerbates the particle degradation problem but also significantly increases the difficulty of the filtering algorithm in balancing computational complexity and estimation accuracy, placing higher demands on the stability and robustness of the filter.

[0006] It should be noted that the information disclosed in the above background technology section is only used to understand the background of this application, and therefore may include information that does not constitute prior art known to ordinary technicians in this field. Summary of the Invention

[0007] The main purpose of the present invention is to overcome the defects existing in the above-mentioned background technology and provide a radar target tracking method based on transport mapping.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A radar target tracking method based on transport mapping adopts a pulse-by-pulse estimation strategy, including an initial construction step, an initial particle generation step, and a pulse-by-pulse update step. The real-time update of the target state is achieved through the following process:

[0010] Initial build steps:

[0011] S1. Construct target transfer equations and radar measurement equations to describe the dynamic changes in target state and the measurement characteristics of the radar system, respectively.

[0012] Initial particle generation steps:

[0013] S2. Generate a set of equally weighted particles based on the target initial state to represent the probability distribution of the target state;

[0014] The pulse-by-pulse update step is executed at each pulse moment:

[0015] S3. Use the target transfer equation to predict the state of the particle set, generate the target state prediction particle set at the next moment, and calculate the measurement prediction value of each particle based on the radar measurement equation;

[0016] S4. Downsampling the high-dimensional measurement signal received by the radar and the particle measurement prediction value, extracting the elements with the largest amplitude to form a downsampled measurement vector and a downsampled measurement prediction;

[0017] S5. Construct a transport map through joint distribution mapping based on the target state prediction particle set, downsampled measurement prediction, and downsampled measurement vector;

[0018] S6. Apply the transport mapping to all target state prediction particles to update the posterior state particle set with equal weights;

[0019] S7. Average the posterior state particle set and output the target state estimate at the current moment.

[0020] Furthermore, in step S3, the state prediction is achieved by superimposing process noise on the target transfer equation, and the measurement prediction is achieved by superimposing measurement noise on the radar measurement equation.

[0021] Furthermore, step S4 specifically includes:

[0022] S41. Arrange the measurement signal elements in descending order of amplitude to generate an index vector;

[0023] S42. Selecting the elements with the largest amplitude according to the index vector to form a downsampled measurement vector;

[0024] S43. Downsample the measurement prediction value of each particle according to the same index vector to generate a downsampled measurement prediction.

[0025] Furthermore, the downsampling dimension in step S4 is dynamically selected according to the nonlinearity degree and probability distribution characteristics of the measurement model.

[0026] Furthermore, step S5 specifically includes:

[0027] S51. Combining the target state prediction value of each particle with the corresponding downsampled measurement prediction into a joint sample;

[0028] S52. Constructing a mapping from a joint sample to a standard normal distribution via the Knothe-Rosenblatt rearrangement;

[0029] S53. Optimize the mapping function dimension by dimension, minimize the KL divergence between the joint sample distribution and the standard normal distribution, and extract the mapping component of the state dimension;

[0030] S54. Construct a transport map based on the downsampled measurement vector and the inverse mapping of the mapping component.

[0031] Furthermore, in step S53, the optimization mapping function adopts a linear combination form of nonlinear basis functions, and the basis function coefficients are solved by a numerical optimization method.

[0032] Furthermore, in step S6, the particle states are directly updated through transport mapping to keep the particle weights equal.

[0033] Furthermore, the pulse-by-pulse updating step directly utilizes the radar echo signal at each pulse moment to perform real-time status update.

[0034] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the radar target tracking method based on transport mapping.

[0035] A computer program product includes a computer program, wherein when the computer program is executed by a processor, the radar target tracking method based on transport mapping is implemented.

[0036] The present invention has the following beneficial effects:

[0037] The present invention proposes a radar target tracking method based on transport mapping. By downsampling and particle mapping under the transport mapping framework, it can solve the particle degradation problem existing in the prior art and the computational complexity problem of high-dimensional measurement data processing, thereby improving the accuracy and efficiency of radar target tracking, especially the radar target tracking performance in complex nonlinear and high-noise environments. As a Bayesian filtering method based on transport mapping, this method effectively addresses the above-mentioned problems existing in the prior art by adopting a pulse-by-pulse estimation strategy, introducing an equal-weighted particle update mechanism and downsampling processing. It can not only process high-dimensional measurement data and multimodal distributions, but also achieve high-precision target tracking in complex nonlinear and high-noise environments, providing a new and efficient technical solution for radar target tracking and has important application value. In practical applications, compared with traditional particle filtering methods, the method of the present invention has more advantages in estimation accuracy and tracking stability, especially in the case of non-Gaussian noise interference, showing stronger robustness and lower estimation error.

[0038] Compared with the prior art, the significant advantages of the present invention are embodied in the following aspects:

[0039] 1. Implement pulse-by-pulse estimation to enhance real-time target tracking. Unlike traditional frame-by-frame processing, this invention uses a pulse-by-pulse estimation strategy to directly use the radar echo signal at each pulse moment, improving real-time target tracking performance and being particularly suitable for highly dynamic target scenarios.

[0040] 2. Addressing particle degradation and improving the accuracy of target state estimation. This invention introduces transport mapping technology to update equally weighted particles at each time step, avoiding particle degradation and maintaining the diversity of the particle set. This improves the accuracy of the posterior distribution approximation, demonstrating greater robustness under multimodal distributions.

[0041] 3. Reduce the computational complexity of high-dimensional measurement data. This method processes high-dimensional measurement data through a downsampling strategy to extract key information, effectively mitigating the impact of noise and clutter on tracking accuracy. It also significantly reduces computational complexity compared to transport mapping-based methods before downsampling.

[0042] 4. Adaptability to complex nonlinear and non-Gaussian noise environments. This method can accurately characterize the posterior distribution of the target state in highly nonlinear and non-Gaussian noise environments, avoiding the performance degradation of existing filtering methods in complex scenarios and improving the robustness of the system.

[0043] In general, the present invention improves the real-time performance of the radar target tracking method, solves the particle degradation and high-dimensional measurement data processing problems in the existing filtering method, shows significant advantages in estimation accuracy and adaptability to complex environments, and has important technical value and application prospects.

[0044] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flowchart of a radar target tracking method based on transport mapping according to an embodiment of the present invention.

[0046] Figure 2 4 is a flow chart of a measurement downsampling module according to an embodiment of the present invention.

[0047] Figure 3 It is a flowchart of constructing a transport mapping module according to an embodiment of the present invention.

[0048] Figure 4 3 is a comparison diagram of the root mean square error between the embodiment of the present invention (method 1) and the particle filter algorithm (method 2). DETAILED DESCRIPTION

[0049] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope of the present invention and its application.

[0050] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0051] The present invention proposes a pulse-by-pulse radar target tracking method based on transport mapping to effectively solve the particle degradation problem and the computational complexity problem of high-dimensional measurement data processing in existing methods, improve the accuracy of target state estimation and the efficiency of target tracking, especially the radar target tracking performance in complex nonlinear and high-noise environments. The present invention is a Bayesian filtering method based on transport mapping, which can effectively solve the above-mentioned problems by adopting a pulse-by-pulse estimation strategy, introducing an equal-weighted particle update mechanism and downsampling processing. This method can not only process high-dimensional measurement data and multimodal distributions, but also achieve high-precision target tracking in complex nonlinear and high-noise environments, providing an efficient solution for radar target tracking and has important application value.

[0052] See Figure 1 The embodiment of the present invention provides a radar target tracking method based on transport mapping, which adopts a pulse-by-pulse estimation strategy, including an initial construction step, an initial particle generation step, and a pulse-by-pulse update step. The real-time update of the target state is achieved through the following process:

[0053] Step S1. Initial construction step: construct the target transfer equation and the radar measurement equation to describe the dynamic change of the target state and the measurement characteristics of the radar system respectively.

[0054] Step S2. Initial particle generation step: Generate an equal-weighted particle set according to the target initial state to represent the probability distribution of the target state.

[0055] Update the steps pulse by pulse, and perform the following steps at each pulse moment:

[0056] Step S3: Use the target transfer equation to predict the state of the particle set, generate the target state prediction particle set at the next moment, and calculate the measurement prediction value of each particle based on the radar measurement equation.

[0057] In some embodiments, the state prediction in step S3 is achieved by superimposing process noise on the target transfer equation, and the measurement prediction is achieved by superimposing measurement noise on the radar measurement equation.

[0058] Step S4. Downsampling the high-dimensional measurement signal received by the radar and the particle measurement prediction value, extracting the elements with the largest amplitude to form a downsampled measurement vector and a downsampled measurement prediction;

[0059] In some embodiments, step S4 specifically includes:

[0060] S41. Arrange the measurement signal elements in descending order of amplitude to generate an index vector.

[0061] S42. Select several elements with the largest amplitude according to the index vector to form a downsampled measurement vector.

[0062] S43. Downsample the measurement prediction value of each particle according to the same index vector to generate a downsampled measurement prediction.

[0063] In some embodiments, the downsampling dimension in step S4 is dynamically selected according to the nonlinearity and probability distribution characteristics of the measurement model.

[0064] Step S5. Constructing a transport map through joint distribution mapping based on the target state prediction particle set, the downsampled measurement prediction, and the downsampled measurement vector;

[0065] In some embodiments, step S5 specifically includes:

[0066] S51. Combine the target state prediction value of each particle with the corresponding downsampled measurement prediction to form a joint sample. Specifically, the dimension of the particle state prediction is , the dimension of downsampling measurement prediction is .

[0067] S52. Construct a mapping of the joint sample to the standard normal distribution via the Knothe-Rosenblatt rearrangement. Specifically, the mapping is Mapping of dimensions .

[0068] S53. Optimize the mapping function dimension by dimension, minimize the KL divergence between the joint sample distribution and the standard normal distribution, and extract the mapping component of the state dimension. The optimized mapping function can be in the form of a linear combination of nonlinear basis functions, and the basis function coefficients are solved by a numerical optimization method. Specifically, an optimization method such as Newton's method is used to minimize the Kullback-Leibler divergence between the joint sample and the standard normal distribution, solve the linear combination coefficients of the nonlinear basis functions of the mapping dimension by dimension, and calculate the coefficients according to the numerical optimization method. No. Dimensional building blocks .

[0069] S54. Construct a transport map based on the downsampled measurement vector and the inverse mapping of the mapping component. Specifically, according to the downsampled measurement vector in step S42 and the mapping component in step S53, The inverse mapping of .

[0070] Step S6. Apply the transport mapping to all target state prediction particles to update the posterior state particle set with equal weights;

[0071] In some embodiments, in step S6, the particle states are directly updated through the transport map to keep the particle weights equal.

[0072] Step S7: average the posterior state particle set and output the target state estimate at the current moment.

[0073] In some embodiments, the pulse-by-pulse updating step directly utilizes the radar echo signal at each pulse moment to perform real-time status update.

[0074] The method of the present invention is particularly suitable for high-dynamic target scenarios and non-Gaussian noise environments, and reduces computational complexity and suppresses noise interference through downsampling. The main technical advantages are: directly processing the radar echo signal at each pulse moment through a pulse-by-pulse estimation strategy, significantly improving the real-time tracking performance in high-dynamic target scenarios; innovatively combining transport mapping technology with an equal-weighted particle update mechanism, effectively overcoming the particle degradation problem of traditional particle filtering in high-dimensional systems, maintaining the diversity of particle sets, and greatly improving the approximate accuracy of the posterior distribution and robustness under multimodal distribution; adopting an amplitude downsampling strategy for high-dimensional measurement data, while retaining key state information, significantly reducing computational complexity and suppressing noise and clutter interference; in highly nonlinear measurement models and non-Gaussian noise environments, accurately characterizing the posterior distribution of the target state through joint distribution mapping, breaking through the performance bottleneck of existing filtering methods in complex scenarios, and comprehensively improving the system's anti-interference ability and tracking stability.

[0075] The following further describes specific embodiments of the present invention, its algorithm examples and experimental verification.

[0076] A radar target tracking method based on transport mapping solves the problems of particle degradation and high-dimensional measurement data processing in existing technologies by downsampling and particle mapping under the transport mapping framework, thereby improving the accuracy and efficiency of radar target tracking. The method includes the following specific steps:

[0077] Step S1: Construct target transfer equation and radar measurement equation

[0078] According to the target's motion prior information and radar signal model, the target transfer equation and radar measurement equation are constructed to describe the dynamic change of the target state and the measurement characteristics of the radar system respectively. The dimension of the target state is recorded as d , the dimension of radar measurement is recorded as N .

[0079] Step S11: Construct a target transfer equation to describe the state change of the target in continuous time or discrete time, in the form of:

[0080]

[0081] in, Indicates the k The target state at the pulse moment, is the state transition function, is the process noise that obeys a known Gaussian distribution, .

[0082] Step S12: Construct a radar measurement equation to describe the relationship between the target state and the radar measurement, in the form of:

[0083]

[0084] in, Indicates the k The measurement vector of the pulse time, is the measurement function, is the measurement noise, and the measurement noise corresponding to each element in the measurement vector is independent and identically distributed.

[0085] Step S2: Particle initialization

[0086] According to the initial state information of the target, a group of particles with equal weights are randomly generated , used to represent the probability distribution of the target state at the initial moment. Among them, J is the number of particles, which should be selected to ensure that the particle set can fully approximate the probability distribution of the target state, thereby ensuring the approximate accuracy of the estimate.

[0087] Step S3: State prediction and measurement prediction

[0088] Step S31: According to k The particle collection at the pulse time , using the target transfer equation, generate the k+ Particle set for target state prediction at moment 1 , the expression corresponding to each particle is:

[0089] .

[0090] Step S32: According to k+ Particle set for target state prediction at moment 1 , using the radar measurement equation, calculate the measurement prediction value corresponding to each particle , whose expression is:

[0091] .

[0092] Step S4: Measurement downsampling

[0093] Step S41: k+ The measurement signal received by the radar at 1 pulse time Process and calculate the index vector in descending order based on the magnitude , whose expression is:

[0094]

[0095] in, Express The index vector returned after sorting the elements of in descending order, Indicates taking the front indexes, The dimension after downsampling should be selected by comprehensively considering the nonlinearity and probability distribution characteristics of the measurement model to extract the most critical information for state estimation.

[0096] Step S42: Using the downsampling index vector , from the measured signal Extract the value with the largest amplitude Elements, get the downsampled measurement vector , the expression of each element is:

[0097] .

[0098] Step S43: Using the downsampling index vector , according to k+ The set of measured predicted values ​​at time 1 , calculate the downsampled measurement prediction corresponding to each particle , the expression of each element is:

[0099] .

[0100] Step S5: Constructing transport map

[0101] Step S51: Predict the target state of each particle point and its corresponding downsampled measurement prediction vector , forming a joint sample ,in Represents the transpose of a vector.

[0102] Step S52: Use Knothe-Rosenblatt (KR) rearrangement to construct a mapping from the sample joint distribution to the standard normal distribution .

[0103] Step S53: Using an optimization method, such as Newton's method, to minimize the Kullback-Leibler (KL) divergence between the sample joint distribution and the standard normal distribution, solve the mapping dimension by dimension The mapping The optimization problem expression of the dimension is:

[0104]

[0105] in, represents a linear combination of nonlinear basis functions, Indicates the The partial derivatives in dimensions. Specifically, The expression is:

[0106]

[0107] in, Represents a joint sample No. elements, represents the nonlinear basis function, Represents the coefficients of the basis function to be solved. The nonlinear basis function can be selected according to the characteristics of the joint distribution, and the radial basis function is one of the common implementation forms. Mapping No. Dimensions Components can be built together , where the superscript “~” indicates downsampling, and its expression is:

[0108] .

[0109] Step S54: Downsample the measurement vector according to step S42 and the inverse mapping of the mapping component in step S53 , constructing transport maps , whose expression is:

[0110]

[0111] The inverse mapping of a mapping component is essentially the inverse function of a linear combination of multiple nonlinear functions, which can be solved numerically. ":=" indicates "assignment to", "·" indicates that the position will be filled with a parameter in subsequent operations, and "○" indicates a composite operation of functions.

[0112] Step S6: Posterior state estimation

[0113] The target state prediction value of each particle in step S31 Applying transport mapping , we get the posterior estimate of the target state of each particle, which is expressed as follows:

[0114]

[0115] These posterior estimated particles are still equal-weighted particles.

[0116] Step S7: Target state estimation

[0117] The average of the posterior estimates of the target state of all particles is obtained. k Target state estimation at +1 pulse time:

[0118] .

[0119] Test example:

[0120] Consider a target moving at high speed in the radial direction of the radar. Initially, the target's radial distance from the radar is 1000 meters, its initial velocity is Mach 1, and it is accelerating at 5 times the acceleration of gravity. The radar system transmits a linear frequency modulated (LFM) signal with a pulse repetition interval of 0.005 seconds. In the simulation environment, the signal-to-noise ratio (SNR) is set to -10 dB, and the noise model is Rayleigh noise.

[0121] In this experiment, the proposed transport mapping Bayesian filtering method (Method 1) and the traditional particle filtering method (Method 2) were used to estimate and track the target state in real time. To ensure a fair comparison, both methods used a set number of 1000 particles in the simulation and employed a dimensionality downsampling strategy during processing, downsampling the observed data to three dimensions.

[0122] The simulation results are as follows Figure 4 As shown in the figure, the target position root mean square error (Position RMSE) of the two methods changes with time steps in the entire estimation time range. Figure 4 It can be seen that the method of the present invention is superior to the particle filter method in terms of estimation accuracy and tracking stability. In particular, in the case of non-Gaussian noise interference, the method of the present invention shows stronger robustness and lower estimation error.

[0123] In summary, the present invention effectively improves the real-time performance of the radar target tracking method, solves the particle degradation and high-dimensional measurement data processing problems in the existing filtering method, shows significant advantages in estimation accuracy and adaptability to complex environments, and has important technical value and application prospects.

[0124] An embodiment of the present invention further provides a storage medium for storing a computer program, which at least performs the above method when executed.

[0125] An embodiment of the present invention further provides a control device, comprising a processor and a storage medium for storing a computer program; wherein the processor is configured to execute at least the method described above when executing the computer program.

[0126] An embodiment of the present invention further provides a processor, which executes a computer program and at least performs the method described above.

[0127] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc or a read-only optical disc (CD-ROM); the magnetic surface memory can be a magnetic disk memory or a magnetic tape memory. The storage medium described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0128] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0129] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0131] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0132] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0133] The methods disclosed in the several method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0134] The features disclosed in several product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0135] The features disclosed in several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0136] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. Those skilled in the art will recognize that, without departing from the scope of the present invention, several equivalent substitutions or obvious variations can be made, and the performance or use of the same should be considered to fall within the scope of protection of the present invention.

Claims

1. A radar target tracking method based on transport mapping, characterized in that: A pulse-by-pulse estimation strategy is adopted, including an initial construction step, an initial particle generation step, and a pulse-by-pulse update step, to achieve real-time update of the target state through the following process: Initial build steps: S1. Construct a target transfer equation to describe the target state changes in continuous time or discrete time, and construct a radar measurement equation to describe the relationship between the target state and the radar measurement; Initial particle generation steps: S2. Generate a set of equally weighted particles based on the target's initial state to represent the probability distribution of the target's state at the initial moment; The pulse-by-pulse update step is executed at each pulse moment: S3. Use the target transfer equation to predict the state of the particle set, generate the target state prediction particle set at the next moment, and calculate the measurement prediction value of each particle based on the radar measurement equation; S4. Downsampling the high-dimensional measurement signal received by the radar and the particle measurement prediction value, extracting the elements with the largest amplitude to form a downsampled measurement vector and a downsampled measurement prediction; S5. Construct a transport map through joint distribution mapping based on the target state prediction particle set, downsampled measurement prediction, and downsampled measurement vector; S6. Apply the transport mapping to all target state prediction particles to update the posterior state particle set with equal weights; S7. Average the posterior state particle set and output the target state estimate at the current moment; Step S5 specifically includes: S51. Combine the target state prediction value of each particle with the corresponding downsampled measurement prediction into a joint sample; the dimension of the particle state prediction is , the dimension of downsampling measurement prediction is ; S52. Construct a mapping of the joint sample to the standard normal distribution via the Knothe-Rosenblatt rearrangement, which is a mapping with Mapping function of dimensions ; S53. Optimizing the Mapping Function Dimensionally , minimize the Kullback-Leibler divergence of the joint sample distribution and the standard normal distribution, and extract the mapping component of the state dimension; among them, solve the mapping function dimension by dimension The linear combination coefficients of the nonlinear basis functions, and according to the mapping function No. Dimensions to build mapping components ; S54. Construct a transport map based on the downsampled measurement vector and the inverse mapping of the mapping component.

2. The radar target tracking method based on transport mapping according to claim 1, wherein: In step S3, the state prediction is achieved by superimposing the process noise on the target transfer equation, and the measurement prediction is achieved by superimposing the measurement noise on the radar measurement equation.

3. The radar target tracking method based on transport mapping according to claim 1 or 2, characterized in that: Step S4 specifically includes: S41. Arrange the measurement signal elements in descending order of amplitude to generate an index vector; S42. Selecting the elements with the largest amplitude according to the index vector to form a downsampled measurement vector; S43. Downsample the measurement prediction value of each particle according to the same index vector to generate a downsampled measurement prediction.

4. The radar target tracking method based on transport mapping according to claim 1 or 2, characterized in that: The downsampling dimension in step S4 is dynamically selected according to the nonlinearity and probability distribution characteristics of the measurement model.

5. The radar target tracking method based on transport mapping according to claim 1, wherein: In step S53: The optimization mapping function adopts the linear combination form of nonlinear basis functions, and the basis function coefficients are solved by a numerical optimization method.

6. The radar target tracking method based on transport mapping according to any one of claims 1 to 2, characterized in that: In step S6, the particle states are directly updated through the transport mapping to keep the particle weights equal.

7. The radar target tracking method based on transport mapping according to any one of claims 1 to 2, characterized in that: The pulse-by-pulse updating step directly utilizes the radar echo signal at each pulse moment to perform real-time status update.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the radar target tracking method based on transport mapping according to any one of claims 1 to 7 is implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the radar target tracking method based on transport mapping according to any one of claims 1 to 7 is implemented.

Images