Distributed multi-sensor multi-target tracking method based on event triggering mechanism

Abstract

The application discloses a kind of distributed multi-sensor multi-target tracking methods based on event triggering mechanism, it is related to target tracking field, the method is for the sensor in sensor network, based on GM-PHD filtering, the independent prediction of multi-target state is carried out, the Gaussian mixture form of the predicted target intensity function of k time is obtained;Determine whether the sensor will pass the measured value to the estimator based on event triggering mechanism;If delivery, the estimator updates the measured value of k time, and obtains the posterior target intensity function of k time;The estimator carries out pruning and merging to the Gaussian component in posterior target intensity function, obtains the weighted Gaussian component as initial value and carries out distributed multi-sensor information fusion, obtains the fused Gaussian component, according to which the multi-target number estimate value and multi-target state estimate value of k time are calculated to carry out multi-target tracking.The method of the application can reduce the energy consumption of sensor network while ensuring the accuracy of multi-target tracking, improve the concealment of sensor network.

Description

Technical Field

[0001] This invention relates to the field of target tracking technology, and in particular to a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism. Background Technology

[0002] With the rapid development of electronic information technology, target tracking technology has become a research hotspot in both civilian and military fields and has been widely applied. In the civilian field, target tracking has significant practical value in intelligent transportation and visual surveillance; in the military field, target tracking is beneficial for accurately tracking and striking enemy targets on the battlefield. In complex tracking environments, real-time accurate target tracking while shielding against external interference has always been a challenging and critical research issue. Target tracking technology was proposed in the 1930s, and later Wax defined the concept of multi-target tracking and initially applied it to air defense systems. Based on different mathematical theories, common multi-target tracking technologies are divided into two categories: tracking strategies based on data association and tracking strategies based on random finite sets.

[0003] Multi-target tracking technology often requires hardware-based wireless sensor networks. Since these networks are battery-powered, their continuous operating time is limited. Furthermore, in electronic warfare environments, large-scale data transmission between sensors increases the likelihood of detection by the enemy, reducing the network's survivability. The uncertainty of target maneuverability leads to problems such as packet loss, intermittent observation, noise interference, and communication network congestion during target tracking. Most existing multi-target tracking algorithms involve periodic data transmission, which consumes a lot of energy and is easily detected by the enemy. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism, which can reduce the energy consumption of the sensor network and improve the concealment of the sensor network while ensuring the tracking accuracy of multiple targets.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] On one hand, the present invention provides a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism, comprising:

[0007] For each sensor in the wireless sensor network, the state of multiple targets is independently predicted based on GM-PHD filtering to obtain a Gaussian mixture form of the target intensity function predicted at time k; the sensor and the estimator are separate and correspond one-to-one.

[0008] The event-triggered mechanism determines whether the sensor transmits the measurement value at time k to the estimator.

[0009] If the sensor transmits the measurement value at time k to the estimator, the estimator updates the measurement value at time k according to the Gaussian mixture form of the target intensity function predicted at time k, and obtains the posterior target intensity function at time k.

[0010] The estimator prunes and merges the Gaussian components in the posterior target intensity function at time k to obtain the weighted Gaussian components;

[0011] The weighted Gaussian components are used as initial values ​​for distributed multi-sensor information fusion to obtain fused Gaussian components.

[0012] The estimated number of targets and the estimated state of targets at time k are calculated based on the fused Gaussian components.

[0013] Multi-target tracking is performed based on the estimated number of targets and the estimated state of the targets.

[0014] Optionally, for each sensor in the wireless sensor network, the multi-target state is independently predicted based on GM-PHD filtering to obtain a Gaussian mixture form of the predicted target intensity function at time k, specifically including:

[0015] For each sensor in the wireless sensor network, the state of multiple targets is independently predicted based on GM-PHD filtering, resulting in a Gaussian mixture form of the predicted target intensity function at time k.

[0016] Where V k|k-1 (x k V is the target intensity function predicted at time k; S,k|k-1 (x k ) is the survival target intensity function at time k; γ k (x k V is a function of the newly generated target RFS at time k; β,k|k-1 (x k J is the derived target intensity function at time k; k|k-1 V represents k|k-1 (x k The number of Gaussian components in the equation; V represents k|k-1 (x k The i-th Gaussian component in ), where x k Let k be the target state. The mean, For variance; For the i-th Gaussian component The corresponding weights.

[0017] Optionally, determining whether the sensor transmits the measurement value at time k to the estimator based on the event-triggered mechanism specifically includes:

[0018] Based on event triggering mechanism Determine whether sensor l transmits the measurement value at time k to the estimator; where τ k-1 The time is the last time sensor l was triggered; Is sensor l at τ k-1 Measurement of time A matrix composed of vectors; It is the measurement value of sensor l at the current time k. The matrix composed of vectors in δ; 2×2 It is the trigger threshold matrix.

[0019] Optionally, the estimator updates the measurement at time k based on the Gaussian mixture form of the target intensity function predicted at time k, to obtain the posterior target intensity function at time k, specifically including:

[0020] The estimator updates the measurement at time k based on the Gaussian mixture form of the target intensity function predicted at time k, thus obtaining the posterior target intensity function at time k. Where (1-p) D,k V k|k-1 (x k ) represents the intensity function of the missed detection target; p D,k For detection probability; The intensity function of the detected target; Let be the i-th Gaussian component of the target intensity function detected at time k, where The mean, For variance; for The corresponding weights.

[0021] Optionally, the estimator prunes and merges the Gaussian components in the posterior target intensity function at time k to obtain a weighted Gaussian component, specifically including:

[0022] The estimator is based on the formula For the posterior target intensity function V at time k k (x k The Gaussian components in z are pruned to obtain the pruned weights. Where T th This is the pruning threshold;

[0023] Based on the weight after pruning For V k (x k The Gaussian components in z are merged to obtain the weighted Gaussian components.

[0024] A distributed multi-sensor multi-target tracking system based on an event-triggered mechanism includes:

[0025] A multi-target state prediction module is used to independently predict the state of multiple targets for each sensor in the wireless sensor network based on GM-PHD filtering, and obtain the Gaussian mixture form of the target intensity function predicted at time k; the sensor and the estimator are separate and correspond one-to-one.

[0026] The event triggering mechanism module is used to determine whether the sensor should transmit the measurement value at time k to the estimator based on the event triggering mechanism;

[0027] The measurement update module is used to update the measurement value at time k according to the Gaussian mixture form of the target intensity function predicted at time k if the sensor transmits the measurement value at time k to the estimator, so as to obtain the posterior target intensity function at time k.

[0028] The Gaussian component pruning and merging module is used to prune and merge the Gaussian components in the posterior target intensity function at time k using an estimator to obtain the weighted Gaussian components.

[0029] A distributed multi-sensor information fusion module is used to perform distributed multi-sensor information fusion with the weighted Gaussian component as the initial value to obtain the fused Gaussian component.

[0030] The multi-target state estimation module is used to calculate the estimated number of multiple targets and the estimated state of multiple targets at time k based on the fused Gaussian components.

[0031] A multi-target tracking module is used to perform multi-target tracking based on the estimated number of multi-targets and the estimated state of multi-targets.

[0032] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned distributed multi-sensor multi-target tracking method based on an event-triggered mechanism.

[0033] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the aforementioned distributed multi-sensor multi-target tracking method based on an event-triggered mechanism.

[0034] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0035] This invention provides a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism. The method includes: for each sensor in a wireless sensor network, independently predicting the state of multiple targets based on GM-PHD filtering to obtain a Gaussian mixture form of the predicted target intensity function at time k; the sensors and estimators are separate and correspond one-to-one; determining whether the sensor transmits the measurement value at time k to the estimator based on the event-triggered mechanism; if the sensor transmits the measurement value at time k to the estimator, the estimator updates the measurement value at time k according to the Gaussian mixture form of the predicted target intensity function at time k to obtain the posterior target intensity function at time k; the estimator prunes and merges the Gaussian components in the posterior target intensity function at time k to obtain a weighted Gaussian component; using the weighted Gaussian component as an initial value for distributed multi-sensor information fusion to obtain a fused Gaussian component; calculating the estimated number of multiple targets and the estimated state of multiple targets at time k based on the fused Gaussian component; and performing multi-target tracking based on the estimated number of multiple targets and the estimated state of multiple targets. The method of this invention can reduce the energy consumption of sensor networks and improve the concealment of sensor networks while ensuring the accuracy of multi-target tracking. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism according to the present invention;

[0038] Figure 2 Two-dimensional trajectory diagram of a multi-target system provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the sensor communication topology provided in an embodiment of the present invention;

[0040] Figure 4 The diagram illustrates the OSPA distance at average communication savings of 10%, 20%, 30%, and 40% as provided in embodiments of the present invention.

[0041] Figure 5 This is a schematic diagram illustrating the target number estimation when the average communication saving rate is 10%, 20%, 30%, and 40% as provided in the embodiments of the present invention. Detailed Implementation

[0042] 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.

[0043] The purpose of this invention is to provide a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism, which can reduce the energy consumption of the sensor network and improve the concealment of the sensor network while ensuring the tracking accuracy of multiple targets.

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Figure 1 This is a flowchart illustrating a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism according to the present invention. See also... Figure 1 The present invention discloses a distributed multi-sensor multi-target tracking method based on an event-triggered mechanism, comprising:

[0046] Step 1: For each sensor in the wireless sensor network, independently predict the state of multiple targets based on GM-PHD filtering to obtain the Gaussian mixture form of the target intensity function predicted at time k.

[0047] The method of this invention targets a wireless sensor network comprising multiple sensor nodes, assuming that the sensors located in the air and the estimators located on the ground are separate and correspond one-to-one. An undirected graph G = (v, ε) is used to represent the communication topology of the estimator network, where v = {1, 2, ..., L} represents the set of estimators, and ε represents the communication channels between estimator nodes. In the undirected graph, if there exists (i, j) ∈ ε, then there exists (j, i) ∈ ε, meaning that nodes i and j are neighboring nodes and can exchange information. The neighboring nodes of node l are represented as a set {l' ∈ v | (l', l) ∈ ε}, with a total of L nodes in the set. l The number of nodes. L is the total number of nodes in the estimator. l Let l be the number of neighboring nodes of estimator l.

[0048] In this invention, the Gaussian distribution of the random variable x is denoted by N(x; m, P), where m represents the mean and P represents the covariance. Consider the linear Gaussian form f of the target state equation and the sensor observation equation. k|k-1 (x k |x k-1 )and They are respectively:

[0049] fk|k-1 (x k |x k-1 )=N(x k ;F k-1 x k-1 Q k-1 (1)

[0050]

[0051] Where zk is the sensor measurement at time k, x k Let x be the target state at time k. k-1 Let R be the target state at time k-1. k and Q k-1 These are the covariance matrices of measurement noise and process noise, respectively. k and F k-1 These are the measurement matrix and the state transition matrix, respectively.

[0052] This invention assumes that the sensor located in the air and the estimator located on the ground are separate and correspond one-to-one. The sensor is responsible for measuring the state (1) of multiple targets, obtaining the measured value zk, and passing the measured value to the corresponding estimator for GM-PHD filtering to obtain the state estimate of multiple targets for multi-target tracking.

[0053] This invention performs multi-target tracking using GM-PHD filtering. First, the GM-PHD filter is initialized at time k=0, and the observations at time k=0 are used to perform the initialization. The Gaussian mixture form V0(x0) of the target posterior intensity at time k=0 is initialized as follows:

[0054]

[0055] Where J0 is the number of Gaussian components in V0(x0), and These are the i-th Gaussian components in V0(x0). The weights, mean, and covariance.

[0056] The target posterior strength function (3) is obtained after initialization using GM-PHD filtering. Each sensor in the wireless sensor network independently predicts the state of multiple targets based on GM-PHD filtering.

[0057] Assume that after k-1 update steps, the Gaussian mixture form of the target intensity function is V. k-1 (x k-1 )for

[0058]

[0059] J k-1 V representsk-1 (x k-1 The number of Gaussian components in the equation is expressed as follows: To represent V k-1 (x k-1 The i-th Gaussian component in ) where This represents its mean. For variance, For weights.

[0060] Based on equation (4), the target state at time k is predicted, and the Gaussian mixture form V of the predicted target intensity function at time k is obtained. k|k-1 (x k )for:

[0061]

[0062] Where V k|k-1 (x k V is the target intensity function predicted at time k; S,k|k-1 (x k ) is the survival target intensity function at time k; γ k (x k V is a function of the newly generated target RFS at time k; β,k|k-1 (x k J is the derived target intensity function at time k; k|k-1 V represents k|k-1 (x k The number of Gaussian components in the equation; V represents k|k-1 (x k The i-th Gaussian component in ), where x k Let k be the target state. The mean, For variance; For the i-th Gaussian component The corresponding weights.

[0063] After completing the prediction step of GM-PHD using formula (5), the obtained V k|k-1 (x k By substituting the measured values ​​selectively transferred according to formula (6) in step 2 into formula (7), the update step of GM-PHD can be realized.

[0064] Step 2: Determine whether the sensor transmits the measurement value at time k to the estimator based on the event triggering mechanism.

[0065] To reduce data transmission and energy consumption caused by the sensor transmitting measurement values ​​to the estimator, step 2 uses an event-triggered mechanism to determine whether the sensor should transmit the measurement value at time k to the estimator. The measurement value (2) obtained by the sensor is judged by the event-triggered method (6) proposed in step 2 and selectively output to the ground estimator corresponding to the sensor.

[0066] In order to τ k-1 By comparing the measurements at time k and time k, this invention proposes a method to make the dimensions of the two sets equal, thereby measuring the difference between the measurements at the two times and determining whether to transfer the measurements. Let F = {f1,...,f...} a Let} be a set of a vectors, and S = {s1,...,s2} b Let} be a set of b vectors, defined as... and These are matrices composed of vectors from sets F and S, respectively. When a ≠ b, a |ab|-dimensional zero matrix is ​​added to the smaller matrix to make the dimensions of the two matrices equal, and then execution can proceed. operate.

[0067] At each time k, whether sensor l transmits the measurement value of the current time to the estimator node depends on the following event triggering mechanism:

[0068]

[0069] Where τ k-1 The time is the last time sensor l was triggered; Is sensor l at τ k-1 Measurement of time A matrix composed of vectors in the middle, for The number of vectors contained in it. It is the measurement value of sensor l at the current time k. The matrix M composed of vectors in the matrix k for The number of vectors contained in it. This is a trigger threshold matrix, where the trigger thresholds a1≥0, a3≥0, and a2 is a sufficiently small scalar that has no effect on triggering. According to... The value of determines whether sensor l transmits the measurement value at time k to the estimator, i.e. The value of determines whether data transmission occurs. Specifically, when When the value is 1, sensor l transmits the observation value at time k to the corresponding estimator node. The estimator uses this measurement value to perform GM-PHD filtering in formula (7) of step 3. When the difference between the measurements at two times is small, If the value is 0, no data transfer occurs, and the estimator uses τ. k-1 The measured value at time is updated using the GM-PHD filter in formula (7) in step 3.

[0070] Most existing multi-target tracking algorithms perform periodic data transmission, which consumes a lot of energy and is easily detected by the enemy. However, this invention introduces an event triggering mechanism (6) to transmit data only when the target's motion state changes significantly. This reduces energy consumption and improves the concealment of the sensor network while ensuring the tracking effect.

[0071] Step 3: If the sensor transmits the measurement value at time k to the estimator, the estimator updates the measurement value at time k according to the Gaussian mixture form of the target intensity function predicted at time k, and obtains the posterior target intensity function at time k.

[0072] The posterior target intensity function V at time k is obtained by updating the Gaussian mixture form (5) obtained after step 1 at time k and the measurement value obtained through the event triggering mechanism (6) mentioned above. k (x k ;z) is

[0073]

[0074] Where (1-p) D,k V k|k-1 (x k ) represents the intensity function of the missed detection target; p D,k For detection probability; The intensity function of the detected target; Let be the i-th Gaussian component of the target intensity function detected at time k, where P is the mean. k i represents the variance; for The corresponding weights.

[0075] Step 4: The estimator prunes and merges the Gaussian components in the posterior target intensity function at time k to obtain the weighted Gaussian components.

[0076] The output V after performing update step 3 on a single estimator k (x k The process of pruning and merging the Gaussian components in z) consists of two steps.

[0077] First, the Gaussian components are pruned, and a pruning threshold T is set. th Make weight Satisfying (8), the pruned weights are obtained.

[0078]

[0079] Then, follow the steps in Algorithm 1 to process V. k (x k The Gaussian components in z are merged.

[0080]

[0081]

[0082] After performing step 4, the Gaussian component in the intensity function after pruning and merging operations is obtained. The corresponding Gaussian component weights are used to obtain the weighted Gaussian components.

[0083] Step 5: Use the weighted Gaussian components as initial values ​​to perform distributed multi-sensor information fusion to obtain the fused Gaussian components.

[0084] In step 4, the Gaussian components generated by a single sensor for multi-target tracking are pruned and merged. Step 5 then performs distributed multi-sensor information fusion. Specifically, the weighted Gaussian components obtained from step 4 of the l-th (l=1,...,L) estimator are... Represented as

[0085] First, initialize the distributed fusion algorithm, letting... in, and , respectively, are the weights, mean, and covariance of estimator l after the c-th Gaussian component iteration.

[0086] The estimator l communicates with its neighboring node l' based on the topology, receives the Gaussian component information from the neighboring nodes, and obtains L. l Gaussian components of +1 estimator (estimator l and its neighbor estimator l'), reset the threshold D for the number of Gaussian components. k And the merging threshold Y, for the obtained L l The Gaussian components of +1 estimator are merged using Algorithm 1.

[0087] The estimator communication process and algorithm 1 are repeated C times (C is a given number of iterations) to obtain the weights after distributed fusion C times. Gaussian components make The fused Gaussian components obtained after performing distributed multi-sensor fusion using each sensor as an estimator l The weights, mean, and covariance of the value.

[0088] Step 6: Calculate the estimated number of targets and the estimated state of targets at time k based on the fused Gaussian components;

[0089] The fused Gaussian component obtained from distributed multi-sensor information fusion in step 5 is used. The number estimate and state estimate of the estimator l for the multiple targets at time k are calculated using formulas (9) and (10).

[0090] The number of targets estimated by estimator l at time k The sum of the weights obtained from formula (9) is the sum of the weights.

[0091]

[0092] Gaussian components that satisfy condition (10) at time k The average value is taken, and this average value is the multi-objective state estimate X output by estimator l at time k. k .

[0093]

[0094] Step 7: Perform multi-target tracking based on the estimated number of targets and the estimated state of the targets.

[0095] Based on the multi-objective number estimate obtained in step 6 and multi-objective state estimate X k This allows for multi-target tracking.

[0096] In the method of this invention, the sensor and its corresponding estimator perform GM-PHD filtering and data transmission based on an event-triggered mechanism. Then, the data is fused a certain number of times using a matrix reconstruction method. On the one hand, while ensuring the accuracy of multi-target tracking, the event-triggered mechanism reduces system costs and improves the concealment of the sensor network. On the other hand, the distributed algorithm only needs local information and does not require all estimators to obtain the global information of the target's motion trajectory, thereby further reducing data transmission and effectively improving computational efficiency.

[0097] This invention introduces an event-triggered mechanism, enabling sensors to transmit data only after triggering conditions are met. This reduces the power consumption of the sensor network and lowers the likelihood of detection by the enemy. Therefore, the distributed multi-sensor multi-target tracking based on the event-triggered mechanism of this invention has not only theoretical significance but also practical engineering implications.

[0098] The effectiveness of the method proposed in this invention is verified through a specific embodiment below. The specific implementation steps of this embodiment are as follows:

[0099] 1) Construction of simulation scenarios

[0100] The multi-target tracking environment is set as a two-dimensional space with a detection area of ​​[-1000, 1000] (m) × [-1000, 1000] (m). Let x... k ={x k,1 ,x k,2 ,x k,3 ,x k,4} T Let {x} represent the state of the target at time k, where {x} k,1 ,x k,3} T Indicates the target location, {x k,2 ,x k,4} T This represents the target velocity. The algorithm's simulation time is 100 seconds, the sampling interval T is 1 second, and the target state equation and observation equation are respectively... and

[0101] Measurement noise v k and process noise v′ k The noise is Gaussian white noise with a mean of 0, and their covariance matrices are Q. v = diag([0.5, 0.5]) and Q v′ =diag([0.5,0.5]). In the experiment, the detection probability was pd = 0.99, the survival probability was ps = 0.98, and the pruning threshold was T. th =10 -7 The merging threshold is Y=1.

[0102] like Figure 2 As shown, the two-dimensional space contains the trajectories of four targets throughout the detection process, where the amount of clutter during the sensor's measurement of the targets follows a Poisson distribution with a mean of λ = 3. The simulation environment includes two surviving targets, one derived target, and one newly formed target, with the covariance matrix of each target being P = diag(1,2,1,2).

[0103] Establish a wireless sensor network consisting of estimator nodes corresponding to 4 sensor nodes, with the following topology: Figure 3 As shown.

[0104] 2) Performance verification of the event triggering mechanism

[0105] All results presented in this embodiment are averages obtained from 30 independent Monte Carlo runs. The performance of the method of this invention was tested at different average trigger rates (90%, 80%, 70%, and 60%). Different trigger thresholds correspond to different average trigger rates, as shown in Table 1 below.

[0106] Table 1. Trigger thresholds and corresponding trigger rates

[0107]

[0108]

[0109] To conserve energy, the number of iterations for distributed information fusion is set to 1. The OSPA distance (m) and the estimated number of targets at average communication rate savings of 10%, 20%, 30%, and 40% are respectively as follows: Figure 4 and Figure 5 As shown. By Figure 4 and Figure 5 It can be seen that the method of the present invention can effectively estimate the target state and the number of targets while reducing the communication rate. This verifies that the method of the present invention can significantly reduce data transmission and energy consumption.

[0110] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: The method of the present invention is an event-triggered distributed multi-target tracking technology. While ensuring the tracking accuracy of multiple sensors for multiple targets, the transmission of measurement values ​​based on the event-triggered mechanism can significantly reduce data transmission and energy consumption, thereby reducing system costs and improving the concealment of sensor networks; The method of the present invention only uses the information of neighboring nodes for data fusion, has a simple structure, good scalability and self-organization, and can effectively improve computational efficiency.

[0111] Based on the method provided by this invention, this invention also provides a distributed multi-sensor multi-target tracking system based on an event-triggered mechanism, comprising:

[0112] A multi-target state prediction module is used to independently predict the state of multiple targets for each sensor in the wireless sensor network based on GM-PHD filtering, and obtain the Gaussian mixture form of the target intensity function predicted at time k; the sensor and the estimator are separate and correspond one-to-one.

[0113] The event triggering mechanism module is used to determine whether the sensor should transmit the measurement value at time k to the estimator based on the event triggering mechanism;

[0114] The measurement update module is used to update the measurement value at time k according to the Gaussian mixture form of the target intensity function predicted at time k if the sensor transmits the measurement value at time k to the estimator, so as to obtain the posterior target intensity function at time k.

[0115] The Gaussian component pruning and merging module is used to prune and merge the Gaussian components in the posterior target intensity function at time k using an estimator to obtain the weighted Gaussian components.

[0116] A distributed multi-sensor information fusion module is used to perform distributed multi-sensor information fusion with the weighted Gaussian component as the initial value to obtain the fused Gaussian component.

[0117] The multi-target state estimation module is used to calculate the estimated number of multiple targets and the estimated state of multiple targets at time k based on the fused Gaussian components.

[0118] A multi-target tracking module is used to perform multi-target tracking based on the estimated number of multi-targets and the estimated state of multi-targets.

[0119] Furthermore, the present invention also provides an electronic device, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The processor can call a computer program stored in the memory to execute the aforementioned event-triggered distributed multi-sensor multi-target tracking method.

[0120] Furthermore, when the computer program in the aforementioned memory is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0121] Furthermore, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed, can implement the aforementioned distributed multi-sensor multi-target tracking method based on an event-triggered mechanism.

[0122] In the distributed multi-sensor multi-target tracking method based on an event-triggered mechanism provided by this invention, the sensor and its corresponding estimator perform GM-PHD filtering and data transmission based on the event-triggered mechanism, and then use a matrix reconstruction method to fuse the data a certain number of times. The method of this invention has the following characteristics: 1. An event-triggered mechanism is designed, through which the sensor determines whether to transmit data; 2. The designed event-triggered mechanism is combined with the multi-sensor multi-target tracking method to reduce data transmission and energy consumption; 3. The designed multi-target tracking method is distributed, meaning that individual estimators only utilize information from their neighbors. Therefore, on the one hand, the method of this invention, by adopting an event-triggered mechanism, reduces system cost and improves the concealment of the sensor network while ensuring the accuracy of multi-target tracking; on the other hand, the distributed algorithm only requires local information and does not require all estimators to obtain the global information of the target's motion trajectory, thereby further reducing the energy consumption of data transmission and greatly improving computational efficiency.

[0123] 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 systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0124] This document uses specific examples 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. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A distributed multi-sensor multi-target tracking method based on an event-triggered mechanism, characterized in that, include: For each sensor in the wireless sensor network, the state of multiple targets is independently predicted based on GM-PHD filtering, resulting in... The Gaussian mixture form of the target intensity function for time-mapping prediction specifically includes: for each sensor in the wireless sensor network, independent prediction of the multi-target state is performed based on GM-PHD filtering, resulting in... Gaussian mixture form of the target intensity function predicted at each time step ;in for The target intensity function predicted at any given time; for The survival target strength function at any given time; for A function for the newly generated target RFS at time step; for The derived target intensity function at time step; express The number of Gaussian components; express The first in Gaussian components, of which for Constant target status, The mean, For variance; For the first Gaussian components The corresponding weights; the sensors and estimators are separate and correspond one-to-one; Determine whether the sensor will [activate / activate] based on an event-triggered mechanism. The measured value at time is passed to the estimator, specifically including: Based on event triggering mechanism Determine the sensor Whether to The measured value at time t is passed to the estimator; where Time is a sensor The last time it was triggered; It is a sensor exist Measurement of time A matrix composed of vectors; It is a sensor At the present moment Measured values A matrix composed of vectors in the matrix; It is the trigger threshold matrix; If the sensor will The measured value at time is passed to the estimator, which then calculates based on the... Gaussian mixture form of the target intensity function predicted at each time step The measured value at time is updated to obtain The posterior target intensity function at time step; The estimator for the The Gaussian components in the posterior target intensity function at time step are pruned and merged to obtain the weighted Gaussian components. The weighted Gaussian components are used as initial values ​​for distributed multi-sensor information fusion to obtain fused Gaussian components. Calculated based on the fused Gaussian components Estimates of the number of multiple targets and the state of multiple targets at any given time; Multi-target tracking is performed based on the estimated number of targets and the estimated state of the targets.

2. The distributed multi-sensor multi-target tracking method based on an event-triggered mechanism according to claim 1, characterized in that, The estimator is based on the Gaussian mixture form of the target intensity function predicted at each time step The measured value at time is updated to obtain The posterior target intensity function at time step 1 includes: The estimator is based on the Gaussian mixture form of the target intensity function predicted at each time step The measured value at time is updated to obtain Post-hoc target intensity function ;in The target intensity function for missed detection; For detection probability; The detected target intensity function; for The first time of the target intensity function detected at time t Gaussian components, of which The mean, For variance; for The corresponding weights.

3. The distributed multi-sensor multi-target tracking method based on an event-triggered mechanism according to claim 2, characterized in that, The estimator is for the The Gaussian components in the posterior objective intensity function at time step 1 are pruned and merged to obtain a weighted Gaussian component, specifically including: The estimator is based on the formula Regarding the Post-hoc target intensity function Pruning is performed on the Gaussian components to obtain the pruned weights. ;in This is the pruning threshold; Based on the weight after pruning right The Gaussian components are merged to obtain the weighted Gaussian components.

4. A distributed multi-sensor multi-target tracking system based on an event-triggered mechanism, characterized in that, include: The multi-target state prediction module is used to independently predict the state of multiple targets based on GM-PHD filtering for each sensor in a wireless sensor network, thus obtaining... The Gaussian mixture form of the target intensity function for time-mapping prediction specifically includes: for each sensor in the wireless sensor network, independent prediction of the multi-target state is performed based on GM-PHD filtering, resulting in... Gaussian mixture form of the target intensity function predicted at each time step ;in for The target intensity function predicted at any given time; for The survival target strength function at any given time; for A function for the newly generated target RFS at time step; for The derived target intensity function at time step; express The number of Gaussian components; express The first in Gaussian components, of which for Constant target status, The mean, For variance; For the first Gaussian components The corresponding weights; the sensors and estimators are separate and correspond one-to-one; The event triggering mechanism module is used to determine whether the sensor should trigger based on the event triggering mechanism. The measured value at time is passed to the estimator, specifically including: Based on event triggering mechanism Determine the sensor Whether to The measured value at time t is passed to the estimator; where Time is a sensor The last time it was triggered; It is a sensor exist Measurement of time A matrix composed of vectors; It is a sensor At the present moment Measured values A matrix composed of vectors in the matrix; It is the trigger threshold matrix; The measurement value update module is used to update the measurement value if the sensor will... The measured value at time is passed to the estimator, which then calculates based on the... Gaussian mixture form of the target intensity function predicted at each time step The measured value at time is updated to obtain The posterior target intensity function at time step; The Gaussian component pruning and merging module is used to apply the estimator to the... The Gaussian components in the posterior target intensity function at time step are pruned and merged to obtain the weighted Gaussian components. A distributed multi-sensor information fusion module is used to perform distributed multi-sensor information fusion with the weighted Gaussian component as the initial value to obtain the fused Gaussian component. The multi-objective state estimation module is used to calculate based on the fused Gaussian components. Estimates of the number of multiple targets and the state of multiple targets at any given time; A multi-target tracking module is used to perform multi-target tracking based on the estimated number of multi-targets and the estimated state of multi-targets.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the distributed multi-sensor multi-target tracking method based on the event-triggered mechanism as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the distributed multi-sensor multi-target tracking method based on the event-triggered mechanism as described in any one of claims 1 to 3.

Images