Extended target joint tracking method and system, computer equipment and storage medium

Through the combination of Poisson space measurement model and multiplication error model, efficient and accurate tracking of the expansion target is achieved, solving the problems of complexity and inefficiency of the expansion target tracking in the existing technology, and improving tracking stability and accuracy sex.

CN118915047APending Publication Date: 2024-11-08STATE GRID ZHEJIANG ELECTRIC POWER CO LTD ZHOUSHAN POWER SUPPLY CO +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410836478.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-26
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In extended target tracking, prior art has difficulty effectively dealing with the association between multiple measurement values ​​and targets, especially in case of change in shape and dynamic changes in kinematics, resulting in increased tracking complexity and inefficiency.

Method used

Using the extended target joint tracking method based on the Poisson spatial measurement model, the target is measured and modeled through the Poisson point process, combined with the multiplication error measurement model and the joint probability data correlation, the kinematic state and shape parameters of the target are updated to achieve the goal. Efficient and accurate tracking of expansion goals.

Benefits of technology

Improves the accuracy and real-time processing capabilities of data associations, can handle uncertain target counts and changes in shape, reduce false positives and missed reports, and improves tracking stability and robustness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915047A_ABST
    Figure CN118915047A_ABST
Patent Text Reader

Abstract

The invention discloses an extended target joint tracking method and system, computer equipment and a storage medium, and relates to the technical field of extended target tracking. At present, a data association problem exists in extended target tracking. According to the method, the Poisson point process is utilized to carry out measurement modeling on an extended object, a multiplicative error measurement model is adopted for space distribution of the extended object, joint probability data association is carried out on a measurement value set, the state vector of each object is updated in a probability data association mode, and the updating process is based on a single extended object tracker. And the state vector of the target is independently updated according to the measured original point. According to the technical scheme, compared with a gamma-Gaussian inverse Wishart probability hypothesis density algorithm, good data association degree can be kept, especially when multiple extended targets are close, tracking can still be kept stable, and the data association problem in extended target tracking is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of extended target tracking, and in particular to an extended target joint tracking method, system, computer equipment and storage medium. Background Art

[0002] In modern radar and sensor systems, especially high-resolution LiDAR, due to the improvement of detection accuracy, a single target often occupies multiple sensor resolution units. This phenomenon is called extended target. Different from traditional point target tracking, extended target tracking not only focuses on the kinematic dynamics of the target, but also considers the spatial distribution information of the target, which greatly increases the complexity of tracking.

[0003] For extended target tracking, an effective method is to use the Poisson point process to model the target measurement. This method was first proposed by Gilholm et al. and subsequently adopted by many scholars. The Poisson point process can simulate the spatial range and number of measurement targets, providing a powerful tool for understanding the performance of targets in sensor space. However, relying solely on the Poisson point process for modeling is not enough to solve all problems in extended target tracking, especially when it involves the association between multiple measurements and targets.

[0004] Data association is a critical and complex step in extended target tracking. Since multiple measurements may originate from the same target, traditional data association methods such as probability hypothesis density filter, probabilistic multi-hypothesis tracking filter or joint probabilistic data association filter can handle this situation to a certain extent, but they often assume that the target is rigid and has a fixed shape, such as rectangle, ellipse, etc. This assumption is often not true in practical applications because the shape of the target may change with movement.

[0005] In addition, the clustering method is another intuitive way to deal with the association between multiple measurements and the target. However, this method requires the transformation of all possible combinations of measurements in the gated region, assuming that each combination is uniformly distributed and the benefit of each combination is known. Although the computational load can be improved through iteration, the complexity of calculating candidate measurement combinations is still very high, making real-time processing difficult.

[0006] In summary, how to effectively associate multiple measurement data is an urgent problem to be solved in the process of extended target tracking. Summary of the invention

[0007] The technical problem to be solved and the technical task proposed by the present invention are to improve and perfect the existing technical solutions and provide a joint tracking method of extended targets based on the Poisson space measurement model to achieve the purpose of efficient and accurate tracking of extended targets. To this end, the present invention adopts the following technical solutions.

[0008] A method for joint tracking of an extended target based on a Poisson space measurement model, characterized by comprising the following steps:

[0009] 1) Using the Poisson point process to model the measurement of the extended target, where each extended target generates multiple measurement values, and the number of measurement values ​​follows the Poisson distribution;

[0010] 2) A multiplicative error measurement model is used for the spatial distribution of the extended target, where the state vector of the extended target includes the kinematic state and shape parameters, both of which obey the normal distribution;

[0011] 3) Perform joint probability data association of the measurements with the extended target to obtain the marginal association probability of the target-associated measurements, and distinguish the measurement set assigned to the target from the clutter measurement set by defining association events;

[0012] 4) After determining the edge association probability, the state vector of each target is updated in a probabilistic data association manner. The updating process is based on a single extended target tracker and independently updates the target's state vector, including kinematic state and shape parameters, according to the measured origin.

[0013] This technical solution not only uses the Poisson point process to measure and model the extended target, but also adopts a multiplicative error measurement model to describe the spatial distribution of the extended target. At the same time, by combining joint probabilistic data association and probabilistic data association, effective tracking of the extended target is achieved, including the update of its kinematic state and shape parameters. This technical solution not only improves the accuracy of data association, but also can process a large amount of measurement data in real time, effectively solving the problem of low efficiency of traditional data association methods and improving tracking stability.

[0014] Specifically, this technical solution uses the Poisson point process to measure and model the extended targets, which can handle the situation where the number of targets changes over time. It does not need to know the exact number of targets in advance. Especially in complex scenarios, the number of targets is often uncertain. This method can be better applied to complex scenarios with uncertain target numbers and large measurement uncertainties, effectively improving the effectiveness of target tracking.

[0015] The use of a multiplicative error measurement model makes the method robust to shape changes and kinematic dynamics of the target, and can maintain high tracking accuracy even when the target shape changes significantly or the motion state is unstable.

[0016] By combining joint probabilistic data association and probabilistic data association, high frame rate video data can be processed in a real-time environment. After determining the edge association probability, the state vector update process based on a single extended target tracker can quickly and accurately update the target's kinematic and shape parameters.

[0017] By defining association events to distinguish the measurement set assigned to the target from the clutter measurement set, the accuracy of data association is effectively improved, which helps to reduce false positives and false negatives and improve the overall performance of the tracking system.

[0018] This technical solution is not only applicable to the tracking of a single extended target, but also to the joint tracking of multiple extended targets. When multiple targets exist at the same time, this method can accurately identify and track each target and maintain stable tracking performance.

[0019] As a preferred technical means: in step 1), each expansion target produce measurements, where is a function with parameter λ i of the Poisson distribution, that is,

[0020] in, Represents the state of a given extension target In the case of The probability of a measurement value.

[0021] The Poisson distribution naturally handles the uncertainty in the number of measurements. In practical applications, it is difficult to accurately predict how many measurements a stretched target will produce due to factors such as sensor noise, target occlusion, or signal attenuation. The Poisson distribution provides a flexible framework to estimate the distribution of the number of measurements based on the characteristics of the target and the environmental conditions. The Poisson distribution requires only one parameter, λ i to describe the quantity distribution of the measured values, which makes the model simple and easy to implement. In addition, the parameter λ iIt can be adjusted according to the characteristics of the extended target (such as size, shape, reflectivity, etc.) and sensor performance to adapt to different application scenarios. By modeling the number of measurements as a Poisson distribution, the association probability between each measurement and the target can be estimated more accurately, which helps to distinguish between true measurements and clutter (i.e., measurements that do not belong to any target) during the data association process, thereby improving the accuracy of target tracking. In multi-extended target tracking scenarios, overlapping or intersecting measurements may be generated between different targets. By independently modeling the number distribution of measurements for each extended target, this situation can be better handled and confusion between measurements between different targets can be reduced. The measurement modeling method based on Poisson distribution is scalable and flexible. With the development of sensor technology and changes in application scenarios, the model parameters can be adjusted according to actual needs or other distributions can be introduced to describe the number distribution of measurements, thereby maintaining the adaptability and accuracy of the model.

[0022] As a preferred technical means: in step 2), the state vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: Where: y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij .

[0023] State Vector By kinematic state and shape parameters This allows a more comprehensive description of the extended target. By simultaneously considering the target's motion state and shape, this method can more accurately track the dynamic changes of the target, which is particularly important when dealing with target tracking problems in complex scenes. and All are modeled as normal distributions, and statistical knowledge and methods can be used to estimate and predict the state of the target. The characteristics of normal distribution allow us to quantify the uncertainty of the target state and accurately describe this uncertainty through mean and covariance. The data association expression G of the joint probability data association filter ijConsidering the measured value y j The association between the predicted measurement value of target i at time k can accurately associate the measurement value with a specific target, thereby improving the tracking accuracy. In addition, by using the spatial correlation rate function λ of the non-homogeneous Poisson process i G ij , which can further optimize the data association process to adapt to the dynamically changing number of targets and environmental conditions. Due to the use of the non-homogeneous Poisson process, it can adapt to the target distribution in different time and space. The characteristics of the non-homogeneous Poisson process allow the model to adjust the target occurrence rate according to the actual situation, thereby showing stronger robustness when dealing with complex situations such as occlusion, clutter and changes in target appearance. Although this technical solution takes into account multiple complex factors, it can still maintain efficient computing performance through reasonable mathematical models and algorithm design, which enables the solution to run in a real-time environment, process high frame rate video data, and meet the real-time requirements in practical applications.

[0024] As a preferred technical means: in step 3), the associated events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying

[0025] By explicitly assigning measurements to targets or clutter sets, it helps reduce false positives and negatives and improves the accuracy of target tracking. Since the attribution of each measurement value (to a certain target or clutter) is clear, the subsequent data processing and state update process can be more efficient. For example, when updating the target state, only the measurement value assigned to the target needs to be considered without traversing all the measurement values. This technical solution can adapt to different numbers of targets and measurement values. No matter how the number of targets in the scene changes, it can adapt by adjusting the number of targets and the total number of measurements; in addition, since the Poisson distribution is used to describe the number of measurement values, it is also highly adaptable to the uncertainty of the number of measurement values.

[0026] As a preferred technical means: in step 4), when updating the target state vector, if the origin of the measurement is known, the tracker updates the kinematic state and shape parameters of the target sequentially and independently.

[0027] By decoupling the update process of kinematic and shape parameters, the tracker can focus on optimizing the estimates of these two aspects separately, which helps to improve the accuracy and efficiency of parameter updates, because the update of each parameter can be optimized according to its specific statistical characteristics and models. Since the updates of kinematic and shape parameters are independent, the tracker can process these two processes in parallel, or at least continuously in a pipelined manner, thereby reducing the overall computation time, which is critical in real-time target tracking applications because it helps ensure that the tracker can respond to new measurements and update the target state within a limited time. Since the update process is separate, the tracker can flexibly choose different update strategies or algorithms for different parameters. For example, kinematic parameters may be more suitable for updating using filtering or smoothing-based methods, while shape parameters may be more suitable for updating using optimization-based methods. By updating kinematic and shape parameters independently, the tracker is better able to cope with the uncertainty of the target state; even if there is an error or noise interference in the estimate of one of the parameters, the update of the other parameter will not be significantly affected. This helps to improve the robustness and stability of the tracker in complex environments. Since the updating process is based on independent modules, the tracker can easily extend its updating strategy when new parameters or features need to be introduced, which helps to keep the tracker flexible and adaptable to meet the needs of different application scenarios.

[0028] A second object of the present invention is to provide an extended target joint tracking system based on a Poisson space measurement model, and the extended target joint tracking system based on a Poisson space measurement model comprises:

[0029] A measurement modeling module, used for performing measurement modeling on the extended target by using a Poisson point process, wherein each extended target generates a plurality of measurement values, and the number of the measurement values ​​follows a Poisson distribution;

[0030] A multiplicative error measurement model module is used to adopt a multiplicative error measurement model for the spatial distribution of the extended target, wherein the state vector of the extended target includes a kinematic state and a shape parameter, both of which obey a normal distribution;

[0031] A probabilistic data association module is used to perform joint probabilistic data association of measurements with extended targets, obtain marginal association probabilities of measurements associated with targets, and distinguish between measurement sets assigned to targets and clutter measurement sets by defining association events;

[0032] The state update module is used to update the state vector of each target in a probabilistic data association manner after determining the edge association probability. The update process is based on a single extended target tracker and independently updates the state vector of the target according to the measured origin, including the kinematic state and shape parameters.

[0033] By measuring and modeling the extended target through the Poisson point process, the multiple measurement values ​​generated by each extended target can be accurately described, and the number of measurement values ​​follows the Poisson distribution; it can more accurately reflect the actual state of the extended target and provide a more reliable data basis for subsequent tracking processing. The multiplicative error measurement model module can fully consider the actual distribution of the extended target in space, as well as the uncertainty of kinematic and shape parameters, thereby improving the accuracy and robustness of tracking. The probabilistic data association module can make full use of the information in the measurement data, effectively distinguish between targets and clutter, and provide more accurate association information for subsequent state updates. The state update module can fully consider the impact of measurement data on the target state, achieve efficient state updates, and ensure the real-time and accuracy of tracking.

[0034] As a preferred technical means: measurement modeling module, a function that calculates the number of measurement values ​​generated by each extended target, where each extended target produce measurements, the number of measurements produced The parameter is λ i Poisson distribution, that is, where λ i is with the stretch goal The related parameters represent the average number of measurements generated by the extended target in a given state. Represents the state of a given extension target In the case of The probability of a measurement value. The probability density function of the classic Poisson distribution is:

[0035]

[0036] The parameter λ is the average number of random events per unit time (or unit area). Poisson distribution is suitable for describing the number of random events per unit time. and equal.

[0037] The multiplicative error measurement model module sets the state vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: where y jis the measured value, y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij ;

[0038] The probabilistic data association module sets association events, where the association events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying

[0039] The state update module independently updates the target state vector, wherein when updating the target state vector, the tracker sequentially and independently updates the kinematic state and shape parameters of the target if the origin of the measurement is known.

[0040] The third object of the present invention is to provide a computer device; the computer device includes one or more processors and one or more memories, and at least one program code is stored in the one or more memories. When the program code is executed by the one or more processors, the extended target joint tracking system based on the Poisson space measurement model is implemented.

[0041] The fourth object of the present invention is to provide a storage medium, in which at least one program code is stored. When the program code is executed by a processor, the steps of the extended target joint tracking method based on the Poisson space measurement model are implemented.

[0042] Beneficial effects: This technical solution not only uses the Poisson point process to measure and model the extended target, but also adopts a multiplicative error measurement model to describe the spatial distribution of the extended target. At the same time, by combining joint probabilistic data association and probabilistic data association, effective tracking of the extended target is achieved, including the update of its kinematic state and shape parameters. This technical solution not only improves the accuracy of data association, but also can process a large amount of measurement data in real time, effectively solves the problem of low efficiency of traditional data association methods, and improves tracking stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the present invention;

[0044] Figure 2 Provide examples of target and measurement associations;

[0045] Figure 3It is the simulation result of Gamma-Gaussian inverse Wishart probability hypothesis density method;

[0046] Figure 4 Simulation results of the joint probability data association method for the Poisson space measurement model. DETAILED DESCRIPTION

[0047] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0048] Embodiment 1: Figure 1 As shown, the present invention comprises the following steps:

[0049] S1: Use Poisson point process to measure and model the extended target, where each extended target generates multiple measurements and the number of measurements follows Poisson distribution;

[0050] Each stretch target produce measurements, where is a function with parameter λ i of the Poisson distribution, that is,

[0051] in, Represents the state of a given extension target In the case of The probability of a measurement value.

[0052] The specific steps include:

[0053] S1.1: is the set of available measurements at time k, is a set of unknown extended targets. Each measurement comes from an extended target or clutter, where the number of clutter is Poisson distributed with parameter λ0. It is assumed that the clutter is uniformly distributed in the radar surveillance area with a density of ρ. The difference between extended target tracking and traditional point target tracking is mainly reflected in the following two aspects: (1) The state vector It contains not only motion parameters but also shape (size and direction) information; (2) Extended targets produce fluctuating measurements. The goal of multi-extended target tracking is to solve the distribution of measurements to targets and estimate X k .

[0054] S1.2: Use Poisson point process to model the extended object measurement, each extended target produce measurements, where is a function with parameter λ i of the Poisson distribution, that is, Parameter λ iIt plays a role similar to detection probability in traditional tracking. When the target is not detected From Target The measured value y is the spatial distribution p(y|x i ). Assume that the associated event θ is a random sample of the measurement set Y i Assign to target x i ,have, in is given The measured value y j For multiple targets X k ,have,

[0055] The Poisson distribution naturally handles the uncertainty in the number of measurements. In practical applications, it is difficult to accurately predict how many measurements a stretched target will produce due to factors such as sensor noise, target occlusion, or signal attenuation. The Poisson distribution provides a flexible framework to estimate the distribution of the number of measurements based on the characteristics of the target and the environmental conditions. The Poisson distribution requires only one parameter, λ i to describe the quantity distribution of the measured values, which makes the model simple and easy to implement. In addition, the parameter λ i It can be adjusted according to the characteristics of the extended target (such as size, shape, reflectivity, etc.) and sensor performance to adapt to different application scenarios. By modeling the number of measurements as a Poisson distribution, the association probability between each measurement and the target can be estimated more accurately, which helps to distinguish between true measurements and clutter (i.e., measurements that do not belong to any target) during the data association process, thereby improving the accuracy of target tracking. In multi-extended target tracking scenarios, overlapping or intersecting measurements may be generated between different targets. By independently modeling the number distribution of measurements for each extended target, this situation can be better handled and confusion between measurements between different targets can be reduced. The measurement modeling method based on Poisson distribution is scalable and flexible. With the development of sensor technology and changes in application scenarios, the model parameters can be adjusted according to actual needs or other distributions can be introduced to describe the number distribution of measurements, thereby maintaining the adaptability and accuracy of the model.

[0056] S2: A multiplicative error measurement model is used for the spatial distribution of the extended target, where the state vector of the extended target includes the kinematic state and shape parameters, both of which obey the normal distribution;

[0057] State Vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: Where: y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij .

[0058] State Vector By kinematic state and shape parameters This allows a more comprehensive description of the extended target. By simultaneously considering the target's motion state and shape, this method can more accurately track the dynamic changes of the target, which is particularly important when dealing with target tracking problems in complex scenes. and All are modeled as normal distributions, and statistical knowledge and methods can be used to estimate and predict the state of the target. The characteristics of normal distribution allow us to quantify the uncertainty of the target state and accurately describe this uncertainty through mean and covariance. The data association expression G of the joint probability data association filter ij Considering the measured value y j The association between the predicted measurement value of target i at time k can accurately associate the measurement value with a specific target, thereby improving the tracking accuracy. In addition, by using the spatial correlation rate function λ of the non-homogeneous Poisson process i G ij , which can further optimize the data association process to adapt to the dynamically changing number of targets and environmental conditions. Due to the use of the non-homogeneous Poisson process, it can adapt to the target distribution in different time and space. The characteristics of the non-homogeneous Poisson process allow the model to adjust the target occurrence rate according to the actual situation, thereby showing stronger robustness when dealing with complex situations such as occlusion, clutter and changes in target appearance. Although this technical solution takes into account multiple complex factors, it can still maintain efficient computing performance through reasonable mathematical models and algorithm design, which enables the solution to run in a real-time environment, process high frame rate video data, and meet the real-time requirements in practical applications.

[0059] S3: Perform joint probability data association between the measurement and the extended target to obtain the marginal association probability β of measurement j associated with target i ij;Distinguish the measurement set assigned to the target from the clutter measurement set by defining association events;

[0060] Related events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying

[0061] The specific steps include:

[0062] S3.1: At time step k, M k The measured values ​​are assigned to N k between measurements and extended targets, where the measurements are unknown and fluctuate over time, i.e., multiple measurements can be assigned to one target. The association hypothesis is expressed as 1×M k Random Vector in

[0063] S3.2: An example of θ is Figure 2 As shown. There are two elliptical objects and three measurements. If all measurements are from clutter, the corresponding association event is represented as [0,0,0]. If the association hypothesis θ = [2,1,2], it means that the first measurement and the third measurement are assigned to target 2, and the first measurement is assigned to target 1. Note that the joint likelihood can be simplified to, for example, the likelihood of θ = [2,1,2] is λ1G 12 λ2G 21 λ2G 23 Table 1 gives the likelihood of the associated event θ.

[0064] Table 1 Likelihood of associated event θ

[0065]

[0066] By explicitly assigning measurements to targets or clutter sets, it helps reduce false positives and negatives and improves the accuracy of target tracking. Since the attribution of each measurement value (to a certain target or clutter) is clear, the subsequent data processing and state update process can be more efficient. For example, when updating the target state, only the measurement value assigned to the target needs to be considered without traversing all the measurement values. This technical solution can adapt to different numbers of targets and measurement values. No matter how the number of targets in the scene changes, it can adapt by adjusting the number of targets and the total number of measurements; in addition, since the Poisson distribution is used to describe the number of measurement values, it is also highly adaptable to the uncertainty of the number of measurement values.

[0067] S4: After determining the edge association probability, the state vector of each target is updated in a probabilistic data association manner. The updating process is based on a single extended target tracker and the state vector of the target is updated independently according to the measured origin, covariance.

[0068] When updating the target state vector, the tracker updates the target’s kinematic and shape parameters sequentially and independently if the origin of the measurements is known.

[0069] By decoupling the update process of kinematic and shape parameters, the tracker can focus on optimizing the estimates of these two aspects separately, which helps to improve the accuracy and efficiency of parameter updates, because the update of each parameter can be optimized according to its specific statistical characteristics and models. Since the updates of kinematic and shape parameters are independent, the tracker can process these two processes in parallel, or at least continuously in a pipelined manner, thereby reducing the overall computation time, which is critical in real-time target tracking applications because it helps ensure that the tracker can respond to new measurements and update the target state within a limited time. Since the update process is separate, the tracker can flexibly choose different update strategies or algorithms for different parameters. For example, kinematic parameters may be more suitable for updating using filtering or smoothing-based methods, while shape parameters may be more suitable for updating using optimization-based methods. By updating kinematic and shape parameters independently, the tracker is better able to cope with the uncertainty of the target state; even if there is an error or noise interference in the estimate of one of the parameters, the update of the other parameter will not be significantly affected. This helps to improve the robustness and stability of the tracker in complex environments. Since the updating process is based on independent modules, the tracker can easily extend its updating strategy when new parameters or features need to be introduced, which helps to keep the tracker flexible and adaptable to meet the needs of different application scenarios.

[0070] This technical solution not only uses the Poisson point process to measure and model the extended target, but also adopts a multiplicative error measurement model to describe the spatial distribution of the extended target. At the same time, by combining joint probabilistic data association and probabilistic data association, effective tracking of the extended target is achieved, including the update of its kinematic state and shape parameters. This technical solution not only improves the accuracy of data association, but also can process a large amount of measurement data in real time, effectively solving the problem of low efficiency of traditional data association methods and improving tracking stability.

[0071] Specifically, this technical solution uses the Poisson point process to measure and model the extended targets, which can handle the situation where the number of targets changes over time. It does not need to know the exact number of targets in advance. Especially in complex scenarios, the number of targets is often uncertain. This method can be better applied to complex scenarios with uncertain target numbers and large measurement uncertainties, effectively improving the effectiveness of target tracking.

[0072] The use of a multiplicative error measurement model makes the method robust to shape changes and kinematic dynamics of the target, and can maintain high tracking accuracy even when the target shape changes significantly or the motion state is unstable.

[0073] By combining joint probabilistic data association and probabilistic data association, high frame rate video data can be processed in a real-time environment. After determining the edge association probability, the state vector update process based on a single extended target tracker can quickly and accurately update the target's kinematic and shape parameters.

[0074] By defining association events to distinguish the measurement set assigned to the target from the clutter measurement set, the accuracy of data association is effectively improved, which helps to reduce false positives and false negatives and improve the overall performance of the tracking system.

[0075] This technical solution is not only applicable to the tracking of a single extended target, but also to the joint tracking of multiple extended targets. When multiple targets exist at the same time, this method can accurately identify and track each target and maintain stable tracking performance.

[0076] like Figure 3 , Figure 4 In the specific embodiment result diagram shown, Figure 3 is the simulation result of the Gamma-Gaussian inverse Wishart probability hypothesis density method, Figure 4 The simulation results of the joint probability data association method for the Poisson space measurement model. In this case, the clutter is all Poisson distributed with an average of 40. The actual semi-axis lengths of the ellipse are [40, 15], [20, 10], and [30, 12], respectively. The number of measurements of the target follows a Poisson distribution with averages of 9, 7, and 8, respectively. All three move according to the constant velocity model. It can be observed that the Gamma-Gaussian inverse Wishart probability hypothesis density method often loses track when the targets cross or approach due to obvious clustering. The joint probability data association method for the Poisson space measurement model shows good tracking.

[0077] The joint probabilistic data association method for the Poisson space measurement model uses the predicted measurement likelihood for association and can distinguish targets in this challenging scenario. Note that the Gamma-Gaussian Inverse Wishart Probability Hypothesis Density method estimates the number of targets, while the joint probabilistic data association method for the Poisson space measurement model assumes that the cardinality is known. Therefore, the Gamma-Gaussian Inverse Wishart Probability Hypothesis Density method solves a much more complex problem and runs less efficiently. Both methods were implemented in Matlab 2022b and executed on a Windows 11 PC equipped with an Intel(R) i7 (4.8GHz) processor.

[0078] Embodiment 2:

[0079] This embodiment provides an extended target joint tracking system based on a Poisson space measurement model, including: a measurement modeling module, used to perform measurement modeling on extended targets using a Poisson point process, wherein each extended target generates multiple measurement values, and the number of measurement values ​​follows a Poisson distribution;

[0080] A multiplicative error measurement model module is used to adopt a multiplicative error measurement model for the spatial distribution of the extended target, wherein the state vector of the extended target includes a kinematic state and a shape parameter, both of which obey a normal distribution;

[0081] A probabilistic data association module is used to perform joint probabilistic data association of measurements with extended targets, obtain marginal association probabilities of measurements associated with targets, and distinguish between measurement sets assigned to targets and clutter measurement sets by defining association events;

[0082] The state update module is used to update the state vector of each target in a probabilistic data association manner after determining the edge association probability. The update process is based on a single extended target tracker and independently updates the state vector of the target according to the measured origin, including the kinematic state and shape parameters.

[0083] By measuring and modeling the extended target through the Poisson point process, the multiple measurement values ​​generated by each extended target can be accurately described, and the number of measurement values ​​follows the Poisson distribution; it can more accurately reflect the actual state of the extended target and provide a more reliable data basis for subsequent tracking processing. The multiplicative error measurement model module can fully consider the actual distribution of the extended target in space, as well as the uncertainty of kinematic and shape parameters, thereby improving the accuracy and robustness of tracking. The probabilistic data association module can make full use of the information in the measurement data, effectively distinguish between targets and clutter, and provide more accurate association information for subsequent state updates. The state update module can fully consider the impact of measurement data on the target state, achieve efficient state updates, and ensure the real-time and accuracy of tracking.

[0084] The measurement modeling module calculates the number of measurement values ​​generated by each extended target, where each extended target produce measurements, the number of measurements produced The parameter is λ i Poisson distribution, that is, where λ i is with the stretch goal The related parameters represent the average number of measurements generated by the extended target in a given state. Represents the state of a given extension target In the case of The probability of a measurement value;

[0085] The multiplicative error measurement model module sets the state vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: where y j is the measured value, y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij ;

[0086] The probabilistic data association module sets association events, where the association events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying

[0087] The state update module independently updates the target state vector, wherein when updating the target state vector, the tracker sequentially and independently updates the kinematic state and shape parameters of the target if the origin of the measurement is known.

[0088] It is understandable that the detailed functional implementation of each of the above modules can be found in the introduction of the aforementioned method embodiment, and no further elaboration is made here.

[0089] Embodiment three:

[0090] The functions that can be achieved by the extended target joint tracking system based on the Poisson space measurement model can all be completed by a computer device. The computer device includes one or more processors and one or more memories. At least one program code is stored in the one or more memories. The program code is loaded and executed by the one or more processors to implement the wireless signal processing method.

[0091] The processor takes out instructions from the memory one by one, analyzes the instructions, and then completes the corresponding operations according to the instruction requirements, generating a series of control commands, so that the various parts of the computer can automatically, continuously and coordinately move to become an organic whole, realize the input of programs, the input of data, and the calculation and output of results. The arithmetic operations or logical operations generated in this process are all completed by the operator; the memory includes a read-only memory (ROM), which is used to store computer programs, and a protection device is provided outside the memory.

[0092] Exemplarily, the computer program may be divided into one or more modules, one or more modules are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0093] Those skilled in the art will understand that the description of the above service equipment is merely an example and does not constitute a limitation on the terminal equipment. It may include more or fewer components than described above, or a combination of certain components, or different components, for example, it may include input and output devices, network access devices, buses, etc.

[0094] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire user terminal.

[0095] The memory can be used to store computer programs and / or modules. The processor can realize various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as an information collection template display function, a product information release function, etc.); the data storage area can store data created according to the use of the berth status display system (such as product information collection templates corresponding to different product types, product information that different product providers need to release, etc.). In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0096] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable medium. Based on this understanding, the present invention implements all or part of the modules / units in the above-mentioned embodiment system, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable medium, and the computer program can realize the functions of the above-mentioned various system embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The storage medium can be a computer-readable medium, and the computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A joint tracking method for extended targets based on a Poisson space measurement model, characterized in that The following steps are involved: 1) Using the Poisson point process to model the measurement of the extended target, where each extended target generates multiple measurement values, and the number of measurement values ​​follows the Poisson distribution; 2) A multiplicative error measurement model is used for the spatial distribution of the extended target, where the state vector of the extended target includes the kinematic state and shape parameters, both of which obey the normal distribution; 3) Perform joint probability data association of the measurements with the extended target to obtain the marginal association probability of the target-associated measurements, and distinguish the measurement set assigned to the target from the clutter measurement set by defining association events; 4) After determining the edge association probability, the state vector of each target is updated in a probabilistic data association manner. The updating process is based on a single extended target tracker and independently updates the target's state vector, including kinematic state and shape parameters, according to the measured origin.

2. The extended target joint tracking method based on the Poisson space measurement model according to claim 1 is characterized in that: In step 1), each expansion target produce measurements, where is subject to the parameter λ i Poisson distribution, that is, in, Represents the state of a given extension target In the case of The probability of a measurement value.

3. The extended target joint tracking method based on the Poisson space measurement model according to claim 2 is characterized in that: In step 2), the state vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: Where: y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij .

4. The extended target joint tracking method based on the Poisson space measurement model according to claim 3 is characterized in that: In step 3), the associated events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying 5. The extended target joint tracking method based on the Poisson space measurement model according to claim 4 is characterized in that: In step 4), when updating the target state vector, if the origin of the measurement is known, the tracker updates the kinematic state and shape parameters of the target sequentially and independently.

6. An extended target joint tracking system based on Poisson space measurement model, characterized in that include: A measurement modeling module, used for performing measurement modeling on the extended target by using a Poisson point process, wherein each extended target generates a plurality of measurement values, and the number of the measurement values ​​follows a Poisson distribution; A multiplicative error measurement model module is used to adopt a multiplicative error measurement model for the spatial distribution of the extended target, wherein the state vector of the extended target includes a kinematic state and a shape parameter, both of which obey a normal distribution; A probabilistic data association module is used to perform joint probabilistic data association of measurements with extended targets, obtain marginal association probabilities of measurements associated with targets, and distinguish between measurement sets assigned to targets and clutter measurement sets by defining association events; The state update module is used to update the state vector of each target in a probabilistic data association manner after determining the edge association probability. The update process is based on a single extended target tracker and independently updates the state vector of the target according to the measured origin, including the kinematic state and shape parameters.

7. The extended target joint tracking system based on the Poisson space measurement model according to claim 6, characterized in that: The measurement modeling module calculates the function of the number of measurement values ​​generated by each extended target, where each extended target produce measurements, the number of measurements produced The parameter is λ i Poisson distribution, that is, where λ i is with the stretch goal The related parameters represent the average number of measurements generated by the extended target in a given state. Represents the state of a given extension target In the case of The probability of a measurement value; The multiplicative error measurement model module sets the state vector By kinematic state and shape parameters composition, and All are normally distributed, kinematic state and shape parameters The mean values ​​of and Their covariances are and In the joint probability data association filter, the data association expression is: where y j is the measured value, y j is the measured value, is the predicted measurement value of target i at time k, is its covariance; the spatial correlation function of the non-homogeneous Poisson process is λ i G ij ; The probabilistic data association module sets association events, where the association events include: that is, the set of measurements assigned to target i, and That is, the clutter measurement set; among them, yes The base number, is the number of clutter, satisfying The state update module independently updates the target state vector, wherein when updating the target state vector, the tracker sequentially and independently updates the kinematic state and shape parameters of the target if the origin of the measurement is known.

8. A computer device, characterized in that: The device includes one or more processors and one or more memories, wherein at least one program code is stored in the one or more memories. When the program code is executed by the one or more processors, an extended target joint tracking system based on a Poisson space measurement model as described in claim 6 or 7 is implemented.

9. A storage medium, wherein at least one program code is stored in the storage medium, characterized in that: When the program code is executed by a processor, the steps of the extended target joint tracking method based on the Poisson space measurement model as described in any one of claims 1 to 5 are implemented.