A multi-target tracking method, device, equipment and storage medium

Abstract

The application discloses a multi-target tracking method and device, equipment and a storage medium, and relates to the technical field of target tracking. The method comprises the following steps: constructing an original measurement model based on a Gaussian model and a Student's t model, predicting a predicted target probability density at a current moment based on a first target prediction result at a previous moment and a preset probability density prediction formula; obtaining current measurement information, updating the target probability density of each Gaussian element corresponding to each target by using the measurement information, and obtaining updated information; determining the final updated information of each target based on the calculated Gaussian element weight, the updated information and a preset target information determination method; and when the final updated information meets a preset condition, updating the original mixed parameter in the original measurement model by using the corresponding current mixed parameter, obtaining an updated measurement model and performing multi-target tracking. The application adjusts the model mixed parameter adaptively to cope with different measurement environments, improves the tracking precision and expands the applicable environment.

Description

Technical Field

[0001] This invention relates to the field of target tracking, and in particular to a multi-target tracking method, apparatus, device, and storage medium. Background Technology

[0002] Multi-target tracking has attracted much attention due to its civilian and military applications, requiring the simultaneous estimation of the state and number of multiple targets. Data association is the core idea of ​​traditional multi-target tracking methods. However, when there are a large number of targets and a lot of clutter and false alarms, association can lead to problems such as combinatorial explosion and a sharp increase in computational load.

[0003] Multi-target tracking methods based on random finite sets (RFS) are non-associative multi-target tracking methods. In the RFS framework, target states and measurements are represented as random sets, and through ensemble integration and differentiation, estimates of the number of targets and their states at each time step are obtained, enabling joint detection and tracking of targets. However, due to the complex high-dimensional integral operations involved, this framework is difficult to implement directly in practice. Therefore, various approximation algorithms have emerged, such as Probability Hypothesis Density (PHD) filters, Cardinality Probability Hypothesis Density (CPHD) filters, and Labeled Multi-Bernoulli (LMB) filters.

[0004] Furthermore, traditional target tracking relies on Gaussian noise distribution for measurement. While this yields optimal estimation results in normal measurement environments, practical applications often encounter measurement anomalies due to interference, making it difficult for algorithms based on Gaussian noise assumptions to produce stable tracking results. Additionally, some advanced algorithms utilize "thick-tailed" distributions to achieve tracking under abnormal measurement conditions, leading to a decrease in tracking accuracy within normal measurement ranges. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a multi-target tracking method, apparatus, device, and storage medium that can improve target tracking accuracy and expand the applicable environment. The specific solution is as follows:

[0006] Firstly, this application discloses a multi-target tracking method, including:

[0007] The original measurement model is constructed based on the Gaussian model and the Student's t model, and the prediction result of the first target at the previous time step is obtained based on the original measurement model.

[0008] Based on the first target prediction result and the preset probability density prediction formula, predict the target probability density at the current moment;

[0009] Obtain the current measurement information, and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element;

[0010] Gaussian weights are calculated using the prediction likelihood method, and the final updated information of each target is calculated based on the Gaussian weights, the updated information corresponding to each Gaussian element, and the preset target information determination method.

[0011] When the final updated information meets the preset conditions, the current mixing parameters corresponding to the final updated information are obtained, and the original mixing parameters in the original measurement model are updated using the current mixing parameters to obtain the updated measurement model. The updated measurement model is then used to perform multi-target tracking operations on the received tracking task.

[0012] Optionally, before obtaining the first target prediction result for the previous time step based on the original measurement model, the method further includes:

[0013] The posterior probability density of the previous time step is obtained based on the original measurement model.

[0014] The posterior probability density of the previous time step is predicted using a first prediction method to obtain the prediction result of the first target; the prediction result of the first target is the prediction result of the previous time step.

[0015] Optionally, the step of predicting the probability density of the target at the current moment based on the first target prediction result and a preset probability density prediction formula includes:

[0016] Obtain the LMB parameter set of the newborn target at the initial moment, and use the second prediction method to predict the LMB parameter set of the newborn target to obtain the birth target prediction result;

[0017] Based on the first target prediction result, the birth target prediction result, and the preset probability density prediction formula, the prediction target probability density at the current moment is predicted.

[0018] Optionally, after predicting the target probability density at the current moment based on the first target prediction result and the preset probability density prediction formula, the method further includes:

[0019] The prediction mixture parameters corresponding to the original measurement model are obtained by using the Markov transition matrix; the prediction mixture parameters include a first prediction mixture parameter corresponding to the Gaussian model and a second prediction mixture parameter corresponding to the Student's t model.

[0020] Optionally, the calculation of Gaussian weights using the prediction likelihood method includes:

[0021] The target mixing parameters are calculated based on the predicted mixing parameters and the preset mixing parameter calculation formula;

[0022] The Gaussian weights are calculated based on the target mixing parameters and the preset weight calculation formula.

[0023] Optionally, the step of calculating the final updated information of each target based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and the preset target information determination method includes:

[0024] Obtain all Gaussian elements corresponding to each target, and determine the weight information corresponding to each Gaussian element in sequence based on the Gaussian element weights and a preset weight determination method;

[0025] The final updated information of each target is calculated based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, and the preset target information determination method; the final updated information includes the updated target state and the updated error covariance.

[0026] Optionally, the step of calculating the final updated information of each target based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, and the preset target information determination method includes:

[0027] The final updated information of each target is calculated based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, the feature function, the affine function, and the preset target information determination method.

[0028] Secondly, this application discloses a multi-target tracking device, comprising:

[0029] The model building module is used to build the original measurement model based on the Gaussian model and Student's t model;

[0030] The prediction result acquisition module is used to acquire the prediction result of the first target at the previous moment based on the prediction of the original measurement model;

[0031] The prediction probability density acquisition module is used to predict the prediction target probability density at the current moment based on the first target prediction result and the preset probability density prediction formula.

[0032] The updated information acquisition module is used to acquire the current measurement information and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element.

[0033] The Gaussian weight calculation module is used to calculate Gaussian weights using the prediction likelihood method.

[0034] The final updated information acquisition module is used to calculate the final updated information of each target based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and a preset target information determination method.

[0035] The current mixing parameter acquisition module is used to acquire the current mixing parameters corresponding to the final updated information when the final updated information meets the preset conditions.

[0036] The parameter update module is used to update the original mixed parameters in the original measurement model using the current mixed parameters to obtain the updated measurement model, and to use the updated measurement model to perform multi-target tracking operation on the received tracking task.

[0037] Thirdly, this application discloses an electronic device, including:

[0038] Memory, used to store computer programs;

[0039] A processor is configured to execute the computer program to implement the steps of the multi-target tracking method disclosed above.

[0040] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the multi-target tracking method disclosed above.

[0041] As can be seen, this application provides a multi-target tracking method, including: constructing an original measurement model based on a Gaussian model and a Student's t model, and obtaining a first target prediction result for the previous time step based on the original measurement model; predicting the target probability density for the current time step based on the first target prediction result and a preset probability density prediction formula; obtaining current measurement information, and updating the predicted target probability density for each Gaussian element corresponding to each target using the measurement information to obtain updated information for each Gaussian element; calculating Gaussian element weights using a prediction likelihood method, and calculating the final updated information for each target based on the Gaussian element weights, the updated information for each Gaussian element, and a preset target information determination method; when the final updated information meets a preset condition, obtaining the current mixing parameters corresponding to the final updated information, updating the original mixing parameters in the original measurement model using the current mixing parameters to obtain an updated measurement model, and performing multi-target tracking operations on the received tracking task using the updated measurement model. As can be seen, this application makes predictions by constructing the original measurement model, updates the probability density of the predicted target using the current measurement information, obtains the final updated information, and then uses the final updated information to adaptively adjust the model's hybrid parameters to cope with different measurement environments, thereby improving tracking accuracy and expanding the applicable environment. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 This is a flowchart of a multi-target tracking method disclosed in this application;

[0044] Figure 2 This is a schematic diagram of a specific multi-target tracking method disclosed in this application;

[0045] Figure 3 This is a flowchart of a specific multi-target tracking method disclosed in this application;

[0046] Figure 4 A schematic diagram of the multi-target tracking device provided in this application;

[0047] Figure 5 This application provides a structural diagram of an electronic device. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Currently, when there are many targets and a large amount of clutter and false alarms, correlation can lead to problems such as combinatorial explosion and a sharp increase in computational load. Traditional target tracking is based on Gaussian distribution measurement noise, which yields optimal estimation results in normal measurement environments. However, in practical applications, due to interference and measurement anomalies, algorithms based on Gaussian noise assumptions struggle to produce stable tracking results. Furthermore, some advanced algorithms utilize "thick-tailed" distributions to achieve tracking under abnormal measurement conditions, resulting in decreased tracking accuracy in normal measurement intervals. Therefore, this application provides a multi-target tracking method that adaptively adjusts model mixing parameters to cope with different measurement environments, improving tracking accuracy and expanding its applicability.

[0050] This invention discloses a multi-target tracking method, see [link to relevant documentation]. Figure 1 As shown, the method includes:

[0051] Step S11: Construct the original measurement model based on the Gaussian model and Student's t model, and obtain the first target prediction result of the previous time step based on the original measurement model.

[0052] In this embodiment, an original measurement model is constructed based on a Gaussian model and a Student's t model, and the prediction result of the first target at the previous time step is obtained based on the original measurement model. It is understood that, as... Figure 2 As shown, a multi-measurement model approach is proposed. A measurement model set is constructed using Gaussian and Student's t distributions, and a Gaussian-Student's t mixture multi-objective likelihood model based on LMB is built. The updated target state is calculated by minimizing the relative entropy method, and the model mixing coefficients are adaptively adjusted to cope with different measurement environments. LMB prediction is performed using the state at time K (obtaining the first target prediction result). Measurement is performed at time K+1, and an LMB update operation is executed. Specifically, a Gaussian-Student's t mixture multi-objective likelihood model is established (i.e., the original measurement model is constructed), Gaussian meta-parameters are calculated, and the likelihood is predicted through the multi-measurement model. Specifically, Gaussian-Student's t mixture coefficients and Gaussian meta-weights are calculated, and the state at time K+1 is obtained based on the calculation results.

[0053] For example, suppose the measurement model at time k+1 is:

[0054]

[0055] Where p g (z k+1 |x k+1 )=N(z k+1 H k+1 x k+1 ,R k+1 St(z) is a Gaussian measurement model. k+1 |x k+1 () represents the Student's t model. and Let represent the mixing coefficients (i.e., mixing parameters) of the Gaussian model and the Student's t model, respectively, and let represent the mixing measurement coefficients (i.e., ). and The transfer process follows a Markov transfer chain Π.

[0056] Step S12: Based on the first target prediction result and the preset probability density prediction formula, predict the target probability density at the current moment.

[0057] In this embodiment, an original measurement model is constructed based on a Gaussian model and a Student's t model. After obtaining the first target prediction result for the previous time step based on the original measurement model, the predicted target probability density for the current time step is predicted based on the first target prediction result and a preset probability density prediction formula. Specifically, the LMB parameter set of the newborn target at the initial time step is obtained, and the second prediction method is used to predict the LMB parameter set of the newborn target to obtain the birth target prediction result. The predicted target probability density for the current time step is predicted based on the first target prediction result, the birth target prediction result, and the preset probability density prediction formula.

[0058] For example, the initial LMB parameter set of the newly generated target can be represented as:

[0059]

[0060] in, Let this be the initial target label space. This represents the probability of the existence of a target with an initial label of l. This represents the probability density distribution of the target (i.e., the prior probability density at time k). The LMB parameter set of the newborn target at the initial time is predicted to obtain the birth target prediction result.

[0061] Predict the target at the current time based on the target probability density function of the previous time step. The predicted target probability density can be expressed as:

[0062] (i.e., the preset probability density prediction formula)

[0063] The first item is the prediction result of the target at the previous moment (the prediction result of the first target), and the second item is the prediction result of the birth target.

[0064] Furthermore, assume that the objective transfer function can be expressed as a Gaussian function:

[0065] f k+1|k (x k+1 |x k )=N(x k+1 ;F k x k Q k );

[0066] Where F k Let Q represent the transition matrix. k The process noise covariance (a preset fixed value). The target posterior density at the previous time step. The probability density of newly generated targets can be represented by a Gaussian mixture, as follows:

[0067]

[0068]

[0069] In the formula, ω, m(·), and P(·) represent the weights, mean, and variance of the Gaussian distribution, respectively (which can be obtained from the posterior probability density of the previous time step). Therefore, the predicted target probability density... It can be expressed in Gaussian mixture form. The Gaussian mixture form of each part in the predicted target probability density formula is as follows:

[0070]

[0071]

[0072]

[0073]

[0074] Step S13: Obtain the current measurement information, and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element.

[0075] In this embodiment, after predicting the predicted target probability density at the current moment based on the first target prediction result and a preset probability density prediction formula, current measurement information is obtained. The predicted target probability density of each Gaussian element corresponding to each target is updated using the measurement information to obtain the updated information corresponding to each Gaussian element. It is understood that measurement information is continuously acquired in real time, and the prediction result is updated based on the measurement information to obtain the updated target state and error covariance; the real-time updated target state and error covariance are collectively referred to as the updated information. Since this application involves multi-target tracking, and each target has multiple Gaussian elements, the updated information corresponding to each Gaussian element is obtained at this time. It is understood that the corresponding updated multi-target probability density is obtained with each update. The updated information is obtained from the updated multi-target probability density and includes the updated target state (i.e., mean) and the updated error covariance.

[0076] Step S14: Calculate the Gaussian weights using the prediction likelihood method, and calculate the final updated information of each target based on the Gaussian weights, the updated information corresponding to each Gaussian element, and the preset target information determination method.

[0077] In this embodiment, current measurement information is acquired, and the predicted target probability density of each Gaussian element corresponding to each target is updated using the measurement information to obtain updated information for each Gaussian element. Then, the Gaussian element weights are calculated using a prediction likelihood method, and the final updated information for each target is calculated based on the Gaussian element weights, the updated information for each Gaussian element, and a preset target information determination method. Specifically, target mixing parameters are calculated based on the predicted mixing parameters and a preset mixing parameter calculation formula; Gaussian element weights are calculated based on the target mixing parameters and a preset weight calculation formula. All Gaussian elements corresponding to each target are acquired, and the weight information for each Gaussian element is determined sequentially based on the Gaussian element weights and the preset weight determination method; the final updated information for each target is calculated based on the weight information for each Gaussian element, the updated information for each Gaussian element, and the preset target information determination method; the final updated information includes the updated target state and the updated error covariance.

[0078] Step S15: When the final updated information meets the preset conditions, obtain the current mixing parameters corresponding to the final updated information, update the original mixing parameters in the original measurement model using the current mixing parameters to obtain the updated measurement model, and use the updated measurement model to perform multi-target tracking operation on the received tracking task.

[0079] In this embodiment, after calculating the final updated information of each target based on the Gaussian weights, the updated information corresponding to each Gaussian element, and the preset target information determination method, when the final updated information meets the preset conditions, the current mixing parameters (i.e., ...) corresponding to the final updated information are obtained. as well as The original mixture parameters (i.e., predicted mixture parameters) in the original measurement model are updated using the current mixture parameters. and The updated measurement model is obtained by using the updated measurement model to perform multi-target tracking on the received tracking task. It is understood that the updated information is determined to meet the preset conditions corresponding to the current measurement environment. If so, the current mixed parameters are obtained to replace the original mixed parameters in the original measurement model, resulting in the updated measurement model. Therefore, under the same environmental conditions, using the updated measurement model for multi-target tracking improves the accuracy of tracking.

[0080] This invention improves upon the traditional single Gaussian measurement model of the LMB filter by establishing a multi-measurement model. By utilizing the Gaussian distribution and Student's t to construct the LMB measurement model set, the algorithm's adaptability to outlier measurement environments is enhanced. During target state estimation, the mean and covariance of each Gaussian element are calculated by minimizing the relative entropy, while the weights of the Gaussian elements are calculated using the predicted likelihood. Furthermore, the mixing coefficients of the Gaussian-Student's t model are adaptively adjusted to address different measurement environments, ultimately yielding more reasonable target state information.

[0081] As can be seen, this application provides a multi-target tracking method, including: constructing an original measurement model based on a Gaussian model and a Student's t model, and obtaining a first target prediction result for the previous time step based on the original measurement model; predicting the target probability density for the current time step based on the first target prediction result and a preset probability density prediction formula; obtaining current measurement information, and updating the predicted target probability density for each Gaussian element corresponding to each target using the measurement information to obtain updated information for each Gaussian element; calculating Gaussian element weights using a prediction likelihood method, and calculating the final updated information for each target based on the Gaussian element weights, the updated information for each Gaussian element, and a preset target information determination method; when the final updated information meets a preset condition, obtaining the current mixing parameters corresponding to the final updated information, updating the original mixing parameters in the original measurement model using the current mixing parameters to obtain an updated measurement model, and performing multi-target tracking operations on the received tracking task using the updated measurement model. As can be seen, this application makes predictions by constructing the original measurement model, updates the probability density of the predicted target using the current measurement information, obtains the final updated information, and then uses the final updated information to adaptively adjust the model's hybrid parameters to cope with different measurement environments, thereby improving tracking accuracy and expanding the applicable environment.

[0082] See Figure 3 As shown, this embodiment of the invention discloses a multi-target tracking method. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution.

[0083] Step S21: Construct the original measurement model based on the Gaussian model and Student's t model.

[0084] Step S22: Obtain the posterior probability density of the previous time step based on the original measurement model, and use the first prediction method to predict the posterior probability density of the previous time step to obtain the first target prediction result, and obtain the first target prediction result of the previous time step based on the original measurement model.

[0085] In this embodiment, after constructing the original measurement model based on the Gaussian model and Student's t model, the posterior probability density of the previous time step is obtained based on the original measurement model. The first prediction method is then used to predict the posterior probability density of the previous time step to obtain the first target prediction result. It is understood that the posterior probability density at time k can be expressed in LMB form. At this point, the posterior probability density at time k is predicted using the first prediction method to obtain the first target prediction result, which is the prediction result of the previous time.

[0086] Step S23: Based on the first target prediction result and the preset probability density prediction formula, predict the target probability density at the current moment.

[0087] Step S24: Use the Markov transition matrix to make predictions to obtain the predicted mixture parameters corresponding to the original measurement model.

[0088] In this embodiment, while predicting the probability density of the target at the current moment based on the first target prediction result and the preset probability density prediction formula, a Markov transition matrix is ​​used for prediction to obtain the prediction mixing parameters corresponding to the original measurement model. The prediction mixing parameters include a first prediction mixing parameter corresponding to the Gaussian model and a second prediction mixing parameter corresponding to the Student's t model, that is, the predicted Gaussian-Student's t mixture model coefficients are calculated using the Markov transition matrix Π. and It should be noted that the process of predicting the mixture parameters and the process of predicting the target probability density are carried out simultaneously and do not overlap.

[0089] Step S25: Obtain the current measurement information, and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element.

[0090] Step S26: Calculate the Gaussian weights using the prediction likelihood method, and calculate the final updated information of each target based on the Gaussian weights, the updated information corresponding to each Gaussian element, and the preset target information determination method.

[0091] In this embodiment, the predicted target probability density of each Gaussian element corresponding to each target is updated using the measurement information to obtain the updated information corresponding to each Gaussian element. Then, the Gaussian element weights are calculated using the prediction likelihood method. Finally, the updated information of each target is calculated based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and the preset target information determination method. It can be understood that the final updated information of each target is calculated based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, the feature function, the affine function, and the preset target information determination method.

[0092] Based on the Gaussian-Student's t-measure model, the updated posterior probability density is... The parameters in can be represented as:

[0093]

[0094]

[0095]

[0096] in Represents the set of trajectory labels; ξ = (θ1,…,θ) t-1 )∈Ξ represents the previous association mapping (the trajectory label set and association mapping are constantly updated and can be obtained from the previous iteration); p D (·) represents the target detection probability (which is a preset fixed value), and q D (·)=1-p D (·); This is the clutter intensity function.

[0097] Then, by minimizing the relative entropy, the feature function can be obtained. Affine function Simultaneously, the Gaussian weights are updated using the mixed predictive likelihood. And the model mixing coefficient. The calculation process is as follows:

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104] (Used to calculate Gaussian weights)

[0105] Where λ represents the degrees of freedom of Student's t-distribution. The updated error covariance matrix... It can be calculated as follows:

[0106]

[0107] This application establishes a multi-measurement model, utilizing a measurement model set composed of Gaussian distribution and Student's t distribution, and establishes a corresponding likelihood model; it uses the minimization of relative entropy to calculate the mean and covariance of each Gaussian element, and uses the predicted likelihood to calculate the weight of the Gaussian element; it adapts the mixing coefficients of the Gaussian-Student's model to cope with different measurement environments by adaptively adjusting the mixing coefficients.

[0108] Step S27: When the final updated information meets the preset conditions, obtain the current mixing parameters corresponding to the final updated information, update the original mixing parameters in the original measurement model using the current mixing parameters to obtain the updated measurement model, and use the updated measurement model to perform multi-target tracking operation on the received tracking task.

[0109] For details regarding steps S21, S23, S25, and S27, please refer to the corresponding content disclosed in the foregoing embodiments, which will not be repeated here.

[0110] As can be seen, this application embodiment constructs an original measurement model based on a Gaussian model and a Student's t model; obtains the posterior probability density of the previous time step based on the original measurement model, and uses a first prediction method to predict the posterior probability density of the previous time step to obtain the first target prediction result; obtains the first target prediction result of the previous time step based on the original measurement model; predicts the predicted target probability density of the current time step based on the first target prediction result and a preset probability density prediction formula; performs prediction using a Markov transition matrix to obtain the prediction mixing parameters corresponding to the original measurement model; obtains the current measurement information, and uses the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target to obtain the predicted target probability density of each Gaussian element corresponding to each target. The updated information corresponding to each Gaussian element is used; the weights of the Gaussian elements are calculated using the prediction likelihood method, and the final updated information of each target is calculated based on the weights of the Gaussian elements, the updated information corresponding to each Gaussian element, and the preset target information determination method; when the final updated information meets the preset conditions, the current mixing parameters corresponding to the final updated information are obtained, and the original mixing parameters in the original measurement model are updated using the current mixing parameters to obtain the updated measurement model. The updated measurement model is then used to perform multi-target tracking operations on the received tracking task, adaptively adjusting the model mixing parameters to cope with different measurement environments, improving tracking accuracy and expanding the applicable environment.

[0111] See Figure 4 As shown in the figure, this application also discloses a multi-target tracking device, including:

[0112] Model building module 11 is used to build the original measurement model based on the Gaussian model and Student's t model;

[0113] The prediction result acquisition module 12 is used to acquire the prediction result of the first target at the previous moment based on the prediction of the original measurement model;

[0114] The prediction probability density acquisition module 13 is used to predict the prediction target probability density at the current moment based on the first target prediction result and the preset probability density prediction formula.

[0115] The updated information acquisition module 14 is used to acquire the current measurement information and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element.

[0116] Gaussian weight calculation module 15 is used to calculate Gaussian weights using the prediction likelihood method;

[0117] The final updated information acquisition module 16 is used to calculate the final updated information of each target based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and the preset target information determination method.

[0118] The current mixing parameter acquisition module 17 is used to acquire the current mixing parameters corresponding to the final updated information when the final updated information meets the preset conditions.

[0119] The parameter update module 18 is used to update the original mixed parameters in the original measurement model using the current mixed parameters to obtain the updated measurement model, and to use the updated measurement model to perform multi-target tracking operation on the received tracking task.

[0120] As can be seen, this application includes: constructing an original measurement model based on a Gaussian model and a Student's t model, and obtaining a first target prediction result for the previous time step based on the original measurement model; predicting the target probability density for the current time step based on the first target prediction result and a preset probability density prediction formula; obtaining current measurement information, and updating the predicted target probability density for each Gaussian element corresponding to each target step using the measurement information to obtain updated information for each Gaussian element step; calculating Gaussian element weights using a prediction likelihood method, and calculating the final updated information for each target step based on the Gaussian element weights, the updated information for each Gaussian element step, and a preset target information determination method; when the final updated information meets a preset condition, obtaining the current mixing parameters corresponding to the final updated information, updating the original mixing parameters in the original measurement model using the current mixing parameters to obtain an updated measurement model, and performing multi-target tracking operations on the received tracking task using the updated measurement model. As can be seen, this application makes predictions by constructing the original measurement model, updates the probability density of the predicted target using the current measurement information, obtains the final updated information, and then uses the final updated information to adaptively adjust the model's hybrid parameters to cope with different measurement environments, thereby improving tracking accuracy and expanding the applicable environment.

[0121] In some specific embodiments, the model building module 11 specifically includes:

[0122] The model building unit is used to build the original measurement model based on the Gaussian model and Student's t model.

[0123] In some specific embodiments, the prediction result acquisition module 12 specifically includes:

[0124] The posterior probability density acquisition unit is used to acquire the posterior probability density of the previous time step based on the original measurement model.

[0125] The first prediction unit is used to predict the posterior probability density of the previous time step using a first prediction method to obtain the first target prediction result; the first target prediction result is the prediction result of the previous time step.

[0126] The first target prediction result acquisition unit is used to acquire the first target prediction result of the previous moment based on the original measurement model.

[0127] In some specific embodiments, the prediction probability density acquisition module 13 specifically includes:

[0128] The newborn target LMB parameter set acquisition unit is used to acquire the newborn target LMB parameter set at the initial moment;

[0129] The birth target prediction result acquisition unit is used to predict the newborn target LMB parameter set using the second prediction method to obtain the birth target prediction result;

[0130] The target probability density acquisition unit is used to predict the target probability density at the current moment based on the first target prediction result, the birth target prediction result, and the preset probability density prediction formula.

[0131] The predictive mixture parameter acquisition unit is used to perform prediction using the Markov transition matrix to obtain the predictive mixture parameters corresponding to the original measurement model; the predictive mixture parameters include a first predictive mixture parameter corresponding to the Gaussian model and a second predictive mixture parameter corresponding to the Student's t model.

[0132] In some specific embodiments, the updated information acquisition module 14 specifically includes:

[0133] The current measurement information acquisition unit is used to acquire the current measurement information;

[0134] The target probability density update unit is used to update the predicted target probability density of each Gaussian element corresponding to each target using the measurement information, so as to obtain the updated information corresponding to each Gaussian element.

[0135] In some specific embodiments, the Gaussian weight calculation module 15 specifically includes:

[0136] The target mixing parameter calculation unit is used to calculate the target mixing parameter based on the predicted mixing parameter and the preset mixing parameter calculation formula.

[0137] The Gaussian weight calculation unit is used to calculate the Gaussian weights based on the target mixing parameters and a preset weight calculation formula.

[0138] In some specific embodiments, the final updated information acquisition module 16 specifically includes:

[0139] A Gaussian element acquisition unit is used to acquire all the Gaussian elements corresponding to each of the targets;

[0140] The weight information determination unit is used to sequentially determine the weight information corresponding to each Gaussian element based on the Gaussian element weight and a preset weight determination method.

[0141] The final updated information calculation unit is used to calculate the final updated information of each target based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, the feature function, the affine function, and the preset target information determination method.

[0142] In some specific embodiments, the current mixing parameter acquisition module 17 specifically includes:

[0143] The current mixing parameter acquisition unit is used to acquire the current mixing parameters corresponding to the final updated information when the final updated information meets the preset conditions.

[0144] In some specific embodiments, the parameter update module 18 specifically includes:

[0145] The parameter update unit is used to update the original mixed parameters in the original measurement model using the current mixed parameters to obtain the updated measurement model, and to use the updated measurement model to perform multi-target tracking operation on the received tracking task.

[0146] Furthermore, embodiments of this application also provide an electronic device. Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0147] Figure 5This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the multi-target tracking method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0148] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0149] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0150] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the multi-target tracking method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program capable of performing other specific tasks.

[0151] Furthermore, this application also discloses a storage medium storing a computer program, which, when loaded and executed by a processor, implements the steps of the multi-target tracking method disclosed in any of the foregoing embodiments.

[0152] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0153] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0154] The foregoing has provided a detailed description of a multi-target tracking method, apparatus, device, and storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A multi-target tracking method, characterized in that, include: The original measurement model is constructed based on the Gaussian model and the Student's t model, and the prediction result of the first target at the previous time step is obtained based on the original measurement model. Based on the first target prediction result and the preset probability density prediction formula, predict the target probability density at the current moment; Obtain the current measurement information, and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element; Gaussian weights are calculated using the prediction likelihood method, and the final updated information of each target is calculated based on the Gaussian weights, the updated information corresponding to each Gaussian element, and the preset target information determination method. When the final updated information meets the preset conditions, the current mixing parameters corresponding to the final updated information are obtained, and the original mixing parameters in the original measurement model are updated using the current mixing parameters to obtain the updated measurement model. The updated measurement model is then used to perform multi-target tracking operations on the received tracking task. After predicting the target probability density at the current moment based on the first target prediction result and the preset probability density prediction formula, the method further includes: using a Markov transition matrix to perform prediction to obtain the prediction mixing parameters corresponding to the original measurement model; the prediction mixing parameters include the first prediction mixing parameters corresponding to the Gaussian model. and the second prediction mixture parameters corresponding to the Student's t model. ; The calculation of Gaussian weights using the predicted likelihood method includes: calculating a target mixture parameter based on the predicted mixture parameter and a preset mixture parameter calculation formula; and calculating the Gaussian weights based on the target mixture parameter and a preset weight calculation formula. The calculation formula for the target mixture parameter is as follows: ； ； in, This indicates the previous association mapping; This represents the target mixture parameters corresponding to the Gaussian model; This represents the objective mixture parameters corresponding to Student's t model; When the final updated information meets the preset conditions, the current mixing parameters corresponding to the final updated information are obtained, including: determining whether the final updated information meets the preset conditions corresponding to the current measurement environment; if so, the current mixing parameters corresponding to the final updated information are obtained.

2. The multi-target tracking method according to claim 1, characterized in that, Before obtaining the first target prediction result for the previous time step based on the original measurement model, the method further includes: The posterior probability density of the previous time step is obtained based on the original measurement model. The posterior probability density of the previous time step is predicted using a first prediction method to obtain the prediction result of the first target; the prediction result of the first target is the prediction result of the previous time step.

3. The multi-target tracking method according to claim 1, characterized in that, The step of predicting the probability density of the target at the current moment based on the first target prediction result and the preset probability density prediction formula includes: Obtain the LMB parameter set of the newborn target at the initial moment, and use the second prediction method to predict the LMB parameter set of the newborn target to obtain the birth target prediction result; Based on the first target prediction result, the birth target prediction result, and the preset probability density prediction formula, the prediction target probability density at the current moment is predicted.

4. The multi-target tracking method according to any one of claims 1 to 3, characterized in that, The calculation of the final updated information of each target based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and the preset target information determination method includes: Obtain all the Gaussian elements corresponding to each target, and determine the weight information corresponding to each Gaussian element in sequence based on the Gaussian element weights and a preset weight determination method; The final updated information of each target is calculated based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, and the preset target information determination method; the final updated information includes the updated target state and the updated error covariance.

5. The multi-target tracking method according to claim 4, characterized in that, The step of calculating the final updated information of each target based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, and the preset target information determination method includes: The final updated information of each target is calculated based on the weight information corresponding to each Gaussian element, the updated information corresponding to each Gaussian element, the feature function, the affine function, and the preset target information determination method.

6. A multi-target tracking device, characterized in that, include: The model building module is used to construct the original measurement model based on the Gaussian model and Student's t model. The prediction result acquisition module is used to acquire the prediction result of the first target at the previous moment based on the prediction of the original measurement model; The prediction probability density acquisition module is used to predict the prediction target probability density at the current moment based on the first target prediction result and the preset probability density prediction formula. The updated information acquisition module is used to acquire the current measurement information and use the measurement information to update the predicted target probability density of each Gaussian element corresponding to each target, so as to obtain the updated information corresponding to each Gaussian element. The Gaussian weight calculation module is used to calculate Gaussian weights using the prediction likelihood method. The final updated information acquisition module is used to calculate the final updated information of each target based on the Gaussian element weights, the updated information corresponding to each Gaussian element, and a preset target information determination method. The current mixing parameter acquisition module is used to acquire the current mixing parameters corresponding to the final updated information when the final updated information meets the preset conditions. The parameter update module is used to update the original mixed parameters in the original measurement model using the current mixed parameters to obtain the updated measurement model, and to use the updated measurement model to perform multi-target tracking operation on the received tracking task. The apparatus is further configured to perform prediction using a Markov transition matrix to obtain prediction mixing parameters corresponding to the original measurement model; the prediction mixing parameters include a first prediction mixing parameter corresponding to the Gaussian model. and the second prediction mixture parameters corresponding to the Student's t model. ; The Gaussian weight calculation module is specifically used to calculate the target mixing parameter based on the predicted mixing parameter and the preset mixing parameter calculation formula; and to calculate the Gaussian weight based on the target mixing parameter and the preset weight calculation formula. The formula for calculating the target mixing parameter is as follows: ； ； in, This indicates the previous association mapping; This represents the target mixture parameters corresponding to the Gaussian model; This represents the objective mixture parameters corresponding to Student's t model; The current hybrid parameter acquisition module is specifically used to determine whether the final updated information meets the preset conditions corresponding to the current measurement environment. If so, it acquires the current hybrid parameters corresponding to the final updated information.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the multi-target tracking method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, Used to store a computer program; wherein, when the computer program is executed by a processor, it implements the multi-target tracking method as described in any one of claims 1 to 5.

Images