Target detection and tracking method and system based on fuzzy control optimized particle filter

Abstract

The application relates to a target detection pre-tracking method and system based on fuzzy control optimized particle filtering, and belongs to the field of target tracking. Particle states are predicted, and particle weights are calculated, the particle with the largest weight is regarded as an optimal particle, and the space near the optimal particle is regarded as an optimal region; the distance between the particle and the optimal particle and the particle ratio in the optimal region are calculated and taken as inputs of a fuzzy controller, a moving coefficient is output after a process of fuzzification, fuzzy reasoning and defuzzification; the particle state is updated according to the moving coefficient, and the particle with a lower weight is moved to a high-likelihood region; the iteration is performed for multiple times until the number of particles in the optimal region exceeds a threshold value; and the state of the target is estimated according to the particle state. The application can effectively alleviate the particle impoverishment problem, improve tracking accuracy, control particle movement through the fuzzy controller, be adaptively applied to different environments and scenes, and has good robustness and interpretability, and the rules can be expressed in natural language.

Description

Technical Field

[0001] This invention belongs to the field of target tracking, and in particular relates to a target detection pre-tracking method and system based on fuzzy control optimized particle filtering. Background Technology

[0002] In real-world environments, target information detected by radar often contains significant noise interference, and the target's motion is influenced by various factors. Therefore, radar target tracking is a typical nonlinear, non-Gaussian model. The particle filter algorithm, based on the Monte Carlo method, is a nonlinear filtering algorithm widely used in radar target tracking to handle nonlinear, non-Gaussian problems. Classical particle filters rely on resampling techniques, essentially replicating particles with larger weights and discarding those with smaller weights. When using classical particle filters for target tracking, over time, a large number of identical particles appear in the particle swarm, impairing particle diversity, reducing search capability, and making the target prone to local optima. This leads to decreased target tracking accuracy, deviations from the true trajectory, and even issues like missed tracking and false alarms.

[0003] To address the particle depletion problem inherent in classical particle filtering, particle filtering algorithms based on intelligent optimization are a major development direction. Genetic algorithms (GPAs) are stochastic search algorithms that simulate biological natural selection and genetic mechanisms, treating feasible solutions to a problem as chromosomes carrying multiple genes, producing a superior population through selection, crossover, and mutation. However, GPAs are highly random, potentially leading to uneven particle distribution, blank areas, and lost tracking targets. Furthermore, as an optimization algorithm, GPAs require significant time to search for the optimal solution, resulting in poor real-time tracking performance. In particle swarm optimization (PSO), each particle represents a possible solution to a problem. The position and velocity of particles are updated based on their individual optimal values ​​and the population optimal values ​​to achieve optimization. However, PSOs are sensitive to the initial state; an inappropriate initial state can cause the algorithm to get trapped in local optima, preventing the attainment of the global optimum. This is particularly problematic in radar target tracking scenarios, where prior information about the target is often lacking, making PSOs prone to local optima. Additionally, PSOs require substantial computation, resulting in poor real-time tracking performance. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a Particle Filter Based on Fuzzy Control Optimization For Track Before Detect (FCO-PF-TBD) method for target detection before tracking. In the resampling stage of the particle filter-based tracking algorithm, a dual-input single-output fuzzy controller is used to control the particles to move towards the high likelihood region. After multiple iterations, most particles are placed in the high likelihood region while maintaining particle diversity, thereby solving the problem of particle depletion and improving tracking accuracy.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a target detection pre-tracking method based on fuzzy control optimized particle filtering, comprising the following steps:

[0006] S1, determine the number of particles, initialize the particle set, the particles in the particle set represent the possible states of the target to be tracked;

[0007] S2, predict particle state, calculate the weight of each particle based on the actual measurement value of the target to be tracked at the current moment;

[0008] S3: Select the particle with the largest weight as the optimal particle, determine the optimal region, calculate the normalized distance between other particles and the optimal particle, and obtain the particle ratio within the optimal region.

[0009] S4 takes the result obtained from S3 as the input of the fuzzy controller. After fuzzification, fuzzy inference and defuzzification, the output is the movement coefficient. The fuzzy controller is a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle. Input 2 is the particle ratio in the optimal region.

[0010] S5 updates the particle state based on the movement coefficient, moving particles with lower weights towards the high likelihood region.

[0011] S6 executes multiple iterations from S3 to S5 until the number of particles in the optimal region exceeds the set threshold. Based on the state of the particle swarm, the state of the target to be tracked is estimated, and the pre-detection tracking of the radar target at the current moment is completed.

[0012] In S2, based on the system's state equation and the particle's state vector at time k-1, the particle's state vector at time k is predicted. ,

[0013]

[0014] In the formula, Here are the state equations of the system. The noise during the target's motion process.

[0015] System measurements at time k The weights of the particles at time k are calculated using a recursive estimation method. ,as follows:

[0016]

[0017] in, For importance density function, Let be the probability density function of the model. Let be the state transition probability density function.

[0018] In a radar scenario, the target's state vector at time k is:

[0019]

[0020] in , For the goal direction and The position of direction, , For the goal direction and velocity in the direction;

[0021] Radar in resolution unit The measured intensity at time k for:

[0022]

[0023] in, The measurement noise of the radar satisfies a mean of zero and a variance of... Gaussian white noise, independent of sampling time and each resolution unit; For the target in the resolution unit The echo intensity value at time k is measured by the radar as follows:

[0024]

[0025] in , , These are the radar's range, azimuth, and number of Doppler elements, respectively.

[0026] In S3, the particle with the largest weight is taken as the optimal particle. When determining the optimal region, the optimal particle is denoted as... ,by Centered on, with radius as The space is the optimal region. The number of particles within the optimal region is counted and denoted as . The specific expression is:

[0027] ;

[0028] Calculate the normalized distance between other particles and the optimal particle to obtain the particle ratio within the optimal region as follows:

[0029] The remaining particles in the particle set, excluding the optimal particle, are called other particles. The spatial normalized distance between the other particles and the optimal particle is calculated. Particle ratio within the optimal region ;

[0030] The spatially normalized distance between other particles in the particle swarm and the optimal particle The calculation is as follows:

[0031]

[0032] in, Indicates the maximum possible distance between particles;

[0033] Particle ratio within the optimal region for:

[0034]

[0035] in, It is the threshold for the number of particles in the optimal region.

[0036] The result obtained from S3 is used as the input to the fuzzy controller. After fuzzification, fuzzy inference, and defuzzification, the shift coefficient is output. The fuzzy controller has a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle; input 2 is the optimal region particle ratio, as detailed below:

[0037] Normalized distance Divided into three fuzzy subsets: far, moderate, and near, with corresponding fuzzy sets as follows: , , ; Fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a triangular function, and the fuzzy subset is... The membership function is a right trapezoidal function;

[0038] Normalized optimal region particle ratio The fuzzy subsets are divided into two categories: low clustering and moderate clustering. The fuzzy set corresponding to low clustering is... The fuzzy set corresponding to moderate clustering is fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a right trapezoidal function;

[0039] Moving coefficient Divide into three fuzzy subsets: large, medium, and small. The corresponding fuzzy sets are: , , fuzzy subset The membership function is a Z-type function, and the membership function is:

[0040]

[0041] in, and These represent the left and right tangent points of the Z-type membership function, respectively.

[0042] Fuzzy subsets The membership function is a Gaussian function, and the membership function is:

[0043]

[0044] in, c Let σ be the center value of the Gaussian membership function, and σ be the standard deviation of the Gaussian membership function.

[0045] Fuzzy subsets The membership function is an sigmoid function, and the membership function is:

[0046]

[0047] in, and These represent the S-type membership functions respectively. ;

[0048] Fuzzy reasoning follows six fuzzy rules:

[0049] (1) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0050] (2) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0051] (3) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0052] (4) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0053] (5) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0054] (6) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0055] The central mean method is used to defuzzify the data based on the center value of the membership function, thus obtaining the movement coefficient. .

[0056] In S5, based on the movement coefficient output by the fuzzy controller... To control the particle to move towards the high likelihood region, the specific expression for the updated particle state is:

[0057]

[0058] in, Represents particles exist Time passes The state after the next iteration. Represents particles exist Time passes The state after the next iteration. It is the optimal particle at time k.

[0059] In S3, the threshold for the number of particles in the optimal region is set as follows: When the number of particles in the optimal region exceeds the threshold Then, based on the state of the particle swarm, the state of the system at time k is estimated. ;

[0060] ;

[0061] in, For particles i exist k Weight of time; when the number of particles in the optimal region is less than a threshold At that time, execute S2.

[0062] Based on the concept of the method, a target detection tracking system based on fuzzy control optimized particle filtering is also provided, including an initialization module, a particle state prediction module, a particle parameter acquisition module, a movement coefficient acquisition module, a particle movement module, and a loop iteration module.

[0063] The initialization module is used to determine the number of particles and initialize the particle set. The particles in the particle set represent the possible states of the target to be tracked.

[0064] The particle state prediction module is used to predict the particle state and calculate the weight of each particle based on the actual measurement value of the target to be tracked at the current moment.

[0065] The particle parameter acquisition module is used to select the particle with the largest weight as the optimal particle, determine the optimal region, calculate the normalized distance between other particles and the optimal particle, and obtain the particle ratio within the optimal region.

[0066] The movement coefficient acquisition module is used to take the normalized distance between a particle and the optimal particle and the particle ratio in the optimal region as inputs to the fuzzy controller. After fuzzification, fuzzy inference and defuzzification processes, the movement coefficient is output. The fuzzy controller has a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle; input 2 is the particle ratio in the optimal region.

[0067] The particle movement module updates the particle state based on the movement coefficient, moving particles with lower weights towards the high likelihood region;

[0068] Based on the iterative operation of the particle parameter acquisition module, the movement coefficient acquisition module, and the particle movement module, the number of particles in the optimal area exceeds the set threshold. The state of the target to be tracked is estimated based on the state of the particle swarm, and the pre-detection tracking of the radar target at the current moment is completed.

[0069] Another computer device is provided, including a processor and a memory. The memory is used to store a computer-executable program. The processor reads the computer-executable program from the memory and executes it. When the processor executes the program, it can implement the target detection pre-tracking method based on fuzzy control optimized particle filtering described in this invention.

[0070] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the target detection pre-tracking method based on fuzzy control optimized particle filtering described in the present invention.

[0071] Compared with the prior art, the present invention has at least the following beneficial effects:

[0072] This invention proposes a target detection tracking method based on fuzzy control-optimized particle filtering. First, the normalized distance between other particles and the optimal particle is calculated. An optimal region radius is set, and the proportion of particles within that region is calculated. Particle movement is controlled by a fuzzy control system, continuously optimizing the spatial distribution of the particle swarm during iteration, ensuring most particles cluster near the high-likelihood region. Simultaneously, a particle count threshold is set within the optimal region. When the number of particles within the optimal region reaches the threshold, particle movement stops to prevent excessive particle aggregation and maintain particle swarm diversity. This method avoids simple particle duplication and discarding, controls the proportion of particles within the optimal region, maintains particle swarm diversity, and achieves higher tracking accuracy. Compared to current mainstream particle filtering improvement algorithms based on intelligent optimization, this method uses a fuzzy controller to control particle movement, allowing for adaptive application to different environments and scenarios. Furthermore, its rules can be expressed in natural language, conforming to human thinking, resulting in better robustness and interpretability, while requiring less computation. Attached Figure Description

[0073] Figure 1 Flowchart of a target detection pre-tracking method based on fuzzy control and optimized particle filtering;

[0074] Figure 2 A fuzzy rule relationship diagram for the normalized distance between particles and the optimal particle;

[0075] Figure 3 This is a fuzzy rule relationship diagram of the particle ratio within the optimal region;

[0076] Figure 4 A fuzzy rule relationship diagram for the moving coefficients;

[0077] Figure 5 The system function relationship diagram for input 1 of the fuzzy controller (input 1 refers to the normalized distance between the particle and the optimal particle).

[0078] Figure 6 The system function relationship diagram for input 2 of the fuzzy controller (input 2 refers to the particle ratio in the optimal region);

[0079] Figure 7 The overall system function relationship diagram of the fuzzy controller;

[0080] Figure 8 Comparison chart of target tracking trajectory effects;

[0081] Figure 9 This is a comparison chart of target tracking errors. Detailed Implementation

[0082] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings.

[0083] Figure 1 This is a flowchart of a target detection pre-tracking method based on fuzzy control optimized particle filtering according to an embodiment of the present invention.

[0084] like Figure 1 As shown, the target detection pre-tracking method based on fuzzy control optimized particle filtering includes:

[0085] Step S1: Initialize the particle set, where each particle is represented by a state vector. and the corresponding weights Composition. Assuming the number of particles is N, sampled from the initial probability distribution. denoted as weighted particle set In the initial state, each particle has the same weight, which is... ;

[0086] Step S2: Based on the system's state equation and the particle's state vector at time k-1, predict the particle's state vector at time k. As shown in equation (1):

[0087]

[0088] In the formula, Here are the state equations of the system. The noise during the motion of the target.

[0089] System measurements at time k The weights of the particles at time k are calculated using a recursive estimation method. As shown in equation (2):

[0090]

[0091] In equation (2), For importance density function, Let be the probability density function of the model. Let be the state transition probability density function;

[0092] In the radar scenario, the target's state vector at time k is shown in equation (3):

[0093]

[0094] in , For the goal direction and The position of direction, , For the goal direction and Velocity in a certain direction.

[0095] Based on the target's position and velocity information, the distance to the target at time k can be calculated. Doppler ,position :

[0096]

[0097]

[0098]

[0099] Radar measurements consist of a series of three-dimensional images, each frame of which includes... 1 resolving unit, of which , , These are the radar's range, azimuth, and number of Doppler elements, respectively. Each resolution element includes one... A rectangular region, in which , , This refers to the radar's range, azimuth, and unit range of the Doppler cells. The radar's resolution in the cells... The measured intensity at time k for:

[0100]

[0101] in, The measurement noise of the radar satisfies a mean of zero and a variance of... Gaussian white noise, independent of sampling time and each resolution unit; For the target in the resolution unit The echo intensity value on the radar signal can be represented by the point spread function, as shown in equation (8):

[0102]

[0103] in, , , The attenuation coefficient; , , For radar in resolution unit Distance, Doppler, and orientation information; , , For the target's distance, Doppler, and orientation information.

[0104] The radar measurement at time k is shown in equation (9):

[0105]

[0106] Step S3: Select the particle with the highest weight as the reference for the movement of other particles, and denote it as the optimal particle. .

[0107] Step S4: Calculate the number of particles within the optimal region. , determined to Centered on, with radius as The space is the optimal region, and the number of particles in the optimal region is... The calculation is shown in equation (10):

[0108]

[0109] When the number of particles in the optimal region is too large, the particles become overly clustered, severely impacting the search range for the next time step. The threshold for the number of particles in the optimal region is set as follows: When the number of particles within the optimal region exceeds this threshold, particle movement stops to ensure particle swarm diversity. That is: if... If the condition is met, proceed to step S5; otherwise, proceed to step S8.

[0110] Step S5: The remaining particles in the particle set, excluding the optimal particle, are called other particles. Calculate the spatial normalized distance between the other particles and the optimal particle. Particle ratio within the optimal region .

[0111] The spatially normalized distance between other particles in the particle swarm and the optimal particle The calculation is shown in equation (11):

[0112]

[0113] in, This represents the maximum possible distance between particles.

[0114] The particle ratio within the optimal region is calculated as shown in equation (12):

[0115]

[0116] in, It is the threshold for the number of particles in the optimal region.

[0117] Step S6, calculate the normalized distance in step S5. Particle ratio in the optimal region As input to the fuzzy controller, after fuzzification, fuzzy inference, and defuzzification operations, the output is the movement coefficient. Moving average It directly affects the distance a particle moves towards the optimal particle. This is achieved through the movement coefficient. This method moves particles with smaller weights to the high likelihood region, avoiding the simple duplication and discarding of particles by the classic resampling algorithm. Therefore, this method does not lose particle diversity and can avoid the problem of particle impoverishment.

[0118] This method employs a fuzzy controller to control particle movement, which can be adaptively applied to different environments and scenarios. Furthermore, its rules can be expressed in natural language, conforming to human thinking and exhibiting better robustness and interpretability. As a feasible example, the fuzzy controller of this invention is designed with a dual-input, single-output structure. One input is the normalized distance between other particles and the optimal particle. Input 2 is the optimal region particle ratio. The output is the moving average coefficient. Based on the application scenario and data conditions of this embodiment, the basic parameters of the fuzzy controller were determined through model verification and optimization adjustments as follows:

[0119] In this embodiment, the normalized distance Divided into three fuzzy subsets: far, moderate, and near, with corresponding fuzzy sets as follows: , , ; Fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a triangular function, and the fuzzy subset is... The membership function is a right trapezoidal function; in this embodiment, the normalized distance... The membership function graph is as follows Figure 2 As shown.

[0120] In this embodiment, the normalized optimal region particle ratio The fuzzy subsets are divided into two categories: low clustering and moderate clustering. The fuzzy set corresponding to low clustering is... The fuzzy set corresponding to moderate clustering is fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a right trapezoidal function; in this embodiment, the optimal region particle ratio The membership function graph is as follows Figure 3 As shown.

[0121] In this embodiment, the movement coefficient Divide into three fuzzy subsets: large, medium, and small. The corresponding fuzzy sets are: , , fuzzy subset The membership function is a Z-type function, and the membership function is:

[0122]

[0123] in, and These represent the left and right tangent points of the Z-type membership function, respectively.

[0124] Fuzzy subsets The membership function is a Gaussian function, and the membership function is:

[0125]

[0126] in, c Let σ be the center value of the Gaussian membership function, and σ be the standard deviation of the Gaussian membership function.

[0127] Fuzzy subsets The membership function is an sigmoid function, and the membership function is:

[0128]

[0129] in, and These represent the S-type membership functions respectively. ;

[0130] Fuzzy reasoning follows six fuzzy rules:

[0131] (1) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0132] (2) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0133] (3) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0134] (4) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0135] (5) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0136] (6) If the normalized distance is The optimal region particle ratio is The moving average is then... ;

[0137] The central average method is used to defuzzify the data based on the center value of the membership function.

[0138] In this embodiment, the system function of input one is as follows: Figure 5 As shown in the system function graph, the farther a particle is from the optimal particle, the greater its output movement coefficient.

[0139] In this embodiment, the system function of input two is as follows: Figure 6 As shown in the system function graph, the lower the particle ratio in the optimal region, the larger the output movement coefficient.

[0140] In this embodiment, the overall coefficient function of the fuzzy controller is as follows: Figure 7 As shown in the system function graph, for particles with a larger normalized distance, the system outputs a larger movement coefficient, while for particles with a smaller normalized distance, the system outputs a smaller movement coefficient. As the number of particles in the optimal region gradually increases to the threshold, the system outputs a smaller movement coefficient. When the number of particles in the optimal region approaches the threshold, the system outputs a moderate movement coefficient for particles with a larger normalized distance, and a smaller movement coefficient for other particles. In summary, in this embodiment, the fuzzy controller basically satisfies the particle movement rules.

[0141] Step S7, based on the movement coefficient output by the fuzzy controller The particle state is controlled to move towards the high likelihood region. The particle state update is shown in equation (16):

[0142]

[0143] in, Represents particles exist Time passes The state after the next iteration. Represents particles exist Time passes The state after the next iteration. for k Find the optimal particle at any given time. Let the maximum number of iterations be Q. If the current number of iterations is less than Q, proceed to step S4; otherwise, proceed to step S8.

[0144] Step S8: Estimate based on the state of the particle swarm k The state of the time system As shown in equation (17):

[0145]

[0146] in, For particles i exist k The weight of time. Let Execute step S2 to estimate the next time step.

[0147] Reference Figure 8 The same target was tracked using three different methods: the classical particle filter-based target detection tracking method (PF-TBD), the particle swarm optimization-based particle filter-based target detection tracking method (PSO-PF-TBD), and the fuzzy control-optimized particle filter-based target detection tracking method proposed in this invention. Particle swarm optimization is a swarm intelligence algorithm where particles search for the optimal solution in a multi-dimensional space based on both their historical best position and their neighborhood best position. Applying particle swarm optimization to the particle filter-based target detection tracking algorithm can improve detection accuracy and has strong robustness, making it one of the main improvements to particle filter-based target detection tracking. However, its drawbacks include the need for extensive computation, poor real-time performance, and a tendency to get trapped in local optima.

[0148] like Figure 8 As shown, the estimated trajectories labeled PF-TBD represent the tracking performance of the classical particle filter algorithm, the estimated trajectories labeled PSO-PF-TBD represent the classical particle filter algorithm, and the estimated trajectories labeled FCO-PF-TBD represent the method proposed in this invention. The PSO-PF-TBD algorithm sets a learning factor. The inertia weight has a maximum value of 0.9 and a minimum value of 0.3, with the weight decreasing linearly. The maximum particle velocity is 1, the minimum velocity is -1, and the maximum number of iterations is 10. The FCO-PF-TBD algorithm is set to a maximum of 40 iterations, an optimal region radius of 3, and a threshold of 0.8 for the particle ratio within the optimal region.

[0149] like Figure 9 As shown, compared with the classic PF-TBD method and the current mainstream PSO-PF-TBD method, the target detection tracking algorithm based on fuzzy control optimized particle filter proposed in this invention has higher tracking accuracy. Taking the root mean square error as the standard for tracking accuracy, the root mean square error of the classic particle filter is 2.1846, the root mean square error of the particle filter detection tracking algorithm based on particle swarm optimization is 1.2154, and the root mean square error of the method proposed in this invention is 0.9751.

[0150] The particle filter detection-before-tracking algorithm based on particle swarm optimization needs to recalculate the fitness of each particle at the beginning of each iteration to update the individual's historical best position and the population's best position. This process takes more time, resulting in a slower response speed for the entire tracking system.

[0151] Table 1

[0152]

[0153] Referring to Table 1, the average computation time per frame for the three methods is compared under different particle numbers. As shown in Table 1, the method proposed in this invention outperforms the particle swarm optimization algorithm in terms of running speed, and is comparable to the traditional particle filtering algorithm in terms of running speed, exhibiting good real-time performance.

[0154] This embodiment introduces the application of the optimization method proposed in this invention in radar target tracking scenarios. The method proposed in this invention is applicable to various target tracking scenarios, including but not limited to radar target tracking, infrared target tracking, and visible light target tracking. When using different tracking techniques, the method optimizes the tracking based on the target state parameters or monitoring information corresponding to the particle state vector in the proposed method.

[0155] Based on the concept of the method, a target detection tracking system based on fuzzy control optimized particle filtering is also provided, including an initialization module, a particle state prediction module, a particle parameter acquisition module, a movement coefficient acquisition module, a particle movement module, and a loop iteration module.

[0156] The initialization module is used to determine the number of particles and initialize the particle set. The particles in the particle set represent the possible states of the target to be tracked.

[0157] The particle state prediction module is used to predict the particle state and calculate the weight of each particle based on the actual measurement value of the target to be tracked at the current moment.

[0158] The particle parameter acquisition module is used to select the particle with the largest weight as the optimal particle, determine the optimal region, calculate the normalized distance between other particles and the optimal particle, and obtain the particle ratio within the optimal region.

[0159] The movement coefficient acquisition module is used to take the normalized distance between a particle and the optimal particle and the particle ratio in the optimal region as inputs to the fuzzy controller. After fuzzification, fuzzy inference and defuzzification processes, the movement coefficient is output. The fuzzy controller has a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle; input 2 is the particle ratio in the optimal region.

[0160] The particle movement module updates the particle state based on the movement coefficient, moving particles with lower weights towards the high likelihood region;

[0161] Based on the iterative operation of the particle parameter acquisition module, the movement coefficient acquisition module, and the particle movement module, the number of particles in the optimal area exceeds the set threshold. The state of the target to be tracked is estimated based on the state of the particle swarm, and the pre-detection tracking of the radar target at the current moment is completed.

[0162] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the target detection pre-tracking method based on fuzzy control optimized particle filtering described in the present invention when executing the computer executable program.

[0163] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the target detection pre-tracking method based on fuzzy control optimized particle filtering described in the present invention.

[0164] The computer device may be a laptop, a desktop computer, or a workstation.

[0165] The processor described in this invention may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).

[0166] The memory described in this invention can be an internal storage unit of a laptop, desktop computer, or workstation, such as memory or hard disk; or it can be an external storage unit, such as a portable hard disk or flash memory card.

[0167] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0168] Finally, it should be noted that the above description is only for illustrating specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that any modifications or variations made based on the technical solutions and inventive concepts of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A target detection pre-tracking method based on fuzzy control optimized particle filtering, characterized in that, Includes the following steps: S1, determine the number of particles, initialize the particle set, the particles in the particle set represent the possible states of the target to be tracked; S2, predict particle state, calculate the weight of each particle based on the actual measurement value of the target to be tracked at the current moment; S3: Select the particle with the largest weight as the optimal particle, determine the optimal region, calculate the normalized distance between other particles and the optimal particle, and obtain the particle ratio within the optimal region. S4 takes the result obtained from S3 as the input of the fuzzy controller. After fuzzification, fuzzy inference and defuzzification, the output is the movement coefficient. The fuzzy controller is a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle. Input 2 is the particle ratio in the optimal region. S5 updates the particle state based on the movement coefficient, moving particles with lower weights towards the high likelihood region. S6 executes multiple iterations from S3 to S5 until the number of particles in the optimal region exceeds the set threshold. Based on the state of the particle swarm, the state of the target to be tracked is estimated, and the pre-detection tracking of the radar target at the current moment is completed.

2. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, In S2, based on the system's state equation and the particle's state vector at time k-1, the particle's state vector at time k is predicted. , ， In the formula, Here are the state equations of the system. The noise during the target's motion process. System measurements at time k The weights of the particles at time k are calculated using a recursive estimation method. ,as follows: ， in, For importance density function, Let be the probability density function of the model. Let be the state transition probability density function.

3. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, In a radar scenario, the target's state vector at time k is: ， in , For the goal direction and The position of direction, , For the goal direction and velocity in the direction; Radar in resolution unit The measured intensity at time k for: ， in, The measurement noise of the radar satisfies a mean of zero and a variance of... Gaussian white noise, independent of sampling time and each resolution unit; For the target in the resolution unit The echo intensity value at time k is measured by the radar as follows: ， in , , These are the radar's range, azimuth, and number of Doppler elements, respectively.

4. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, In S3, the particle with the largest weight is taken as the optimal particle. When determining the optimal region, the optimal particle is denoted as... ,by Centered on, with radius as The space is the optimal region. The number of particles within the optimal region is counted and denoted as . The specific expression is: ； Calculate the normalized distance between other particles and the optimal particle to obtain the particle ratio within the optimal region as follows: The remaining particles in the particle set, excluding the optimal particle, are called other particles. The spatial normalized distance between the other particles and the optimal particle is calculated. Particle ratio within the optimal region ; The spatially normalized distance between other particles in the particle swarm and the optimal particle The calculation is as follows: in, Indicates the maximum possible distance between particles; Particle ratio within the optimal region for: in, It is the threshold for the number of particles in the optimal region.

5. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, The result obtained from S3 is used as the input to the fuzzy controller. After fuzzification, fuzzy inference, and defuzzification, the shift coefficient is output. The fuzzy controller has a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle; input 2 is the optimal region particle ratio, as detailed below: Normalized distance Divided into three fuzzy subsets: far, moderate, and near, with corresponding fuzzy sets as follows: , , ; Fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a triangular function, and the fuzzy subset is... The membership function is a right trapezoidal function; Optimal region particle ratio The fuzzy subsets are divided into two categories: low clustering and moderate clustering. The fuzzy set corresponding to low clustering is... The fuzzy set corresponding to moderate clustering is fuzzy subset The membership function is a left trapezoidal function, and the fuzzy subset is... The membership function is a right trapezoidal function; Moving coefficient Divide into three fuzzy subsets: large, medium, and small. The corresponding fuzzy sets are: , , fuzzy subset The membership function is a Z-type function, and the membership function is: in, and These represent the left and right tangent points of the Z-type membership function, respectively. Fuzzy subsets The membership function is a Gaussian function, and the membership function is: in, c Let σ be the center value of the Gaussian membership function, and σ be the standard deviation of the Gaussian membership function. Fuzzy subsets The membership function is an sigmoid function, and the membership function is: in, and These represent the S-type membership functions respectively. ; Fuzzy reasoning follows six fuzzy rules: (1) If the normalized distance is The optimal region particle ratio is The moving average is then... ; (2) If the normalized distance is The optimal region particle ratio is The moving average is then... ; (3) If the normalized distance is The optimal region particle ratio is The moving average is then... ; (4) If the normalized distance is The optimal region particle ratio is The moving average is then... ; (5) If the normalized distance is The optimal region particle ratio is The moving average is then... ; (6) If the normalized distance is The optimal region particle ratio is The moving average is then... ; The central mean method is used to defuzzify the data based on the center value of the membership function, thus obtaining the movement coefficient. .

6. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, In S5, based on the movement coefficient output by the fuzzy controller... To control the particle to move towards the high likelihood region, the specific expression for the updated particle state is: in, Represents particles exist Time passes The state after the next iteration. Represents particles exist Time passes The state after the next iteration. Let be the optimal particle at time k.

7. The target detection pre-tracking method based on fuzzy control optimized particle filtering according to claim 1, characterized in that, In S3, the threshold for the number of particles in the optimal region is set as follows: When the number of particles in the optimal region exceeds the threshold Then, based on the state of the particle swarm, the state of the system at time k is estimated. ; ； in, For particles i exist k The weight of each moment; When the number of particles in the optimal region is less than the threshold At that time, execute S2.

8. A target detection pre-tracking system based on fuzzy control optimized particle filtering, characterized in that, It includes an initialization module, a particle state prediction module, a particle parameter acquisition module, a movement coefficient acquisition module, a particle movement module, and a loop iteration module; The initialization module is used to determine the number of particles and initialize the particle set. The particles in the particle set represent the possible states of the target to be tracked. The particle state prediction module is used to predict the particle state and calculate the weight of each particle based on the actual measurement value of the target to be tracked at the current moment. The particle parameter acquisition module is used to select the particle with the largest weight as the optimal particle, determine the optimal region, calculate the normalized distance between other particles and the optimal particle, and obtain the particle ratio within the optimal region. The movement coefficient acquisition module is used to take the normalized distance between a particle and the optimal particle and the particle ratio in the optimal region as inputs to the fuzzy controller. After fuzzification, fuzzy inference and defuzzification processes, the movement coefficient is output. The fuzzy controller has a dual-input single-output structure. Input 1 is the normalized distance between other particles and the optimal particle; input 2 is the particle ratio in the optimal region. The particle movement module updates the particle state based on the movement coefficient, moving particles with lower weights towards the high likelihood region; Based on the iterative operation of the particle parameter acquisition module, the movement coefficient acquisition module, and the particle movement module, the number of particles in the optimal area exceeds the set threshold. The state of the target to be tracked is estimated based on the state of the particle swarm, and the pre-detection tracking of the radar target at the current moment is completed.

9. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer-executable program, the processor reading the computer-executable program from the memory and executing it, and the processor executing the program being able to implement the target detection pre-tracking method based on fuzzy control optimized particle filtering as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer-readable storage medium stores a computer program that, when executed by a processor, enables the target detection pre-tracking method based on fuzzy control optimized particle filtering as described in any one of claims 1-7.

Images