A passive multi-monitoring station particle filter direct positioning and tracking method

Abstract

The present invention discloses a passive multi-monitoring station particle filter direct positioning and tracking method, which realizes the positioning and tracking of the information source by configuring the monitoring stations with uniform linear array and located in different positions. First, a state model of the information source is established, and according to the particle filter knowledge, a certain amount of initial particles are generated by the state model, and the particles are updated by the state transfer model; secondly, the PM algorithm spectrum function is constructed by the monitoring station measurement model; the angle information contained in the particle is obtained by the monitoring station position and the position information in the particle, and is brought into the spectrum function, and the spectrum function value is used as the particle weight and normalized. Subsequently, the particles are processed by resampling, and the particles are resampled multiple times using the positions and measurement information of multiple monitoring stations. After sampling, the particle weights are set to equal values, and the estimated value of the source position at that moment is obtained by the cumulative summation of the particles and their weights, and the particle state is continuously iterated and updated to realize positioning and tracking in the entire time period.

Description

Technical Field

[0001] The invention relates to a passive multi-monitoring station particle filtering direct positioning and tracking method, belonging to the field of direct positioning and tracking. Background Art

[0002] As a main method for studying the position, velocity, acceleration and characteristics of targets, target tracking has always attracted extensive attention from researchers. Its development can be traced back to before World War II. With the emergence of the first radar SCR-28, various sonar, radar and infrared light detection systems have been widely used in military and civilian fields. In recent years, with the continuous development of science and technology and the application of target tracking in engineering, target tracking technology has been studied more deeply. At the same time, target tracking also faces greater challenges. First, target tracking further highlights its multi-scale, random, nonlinear and non-Gaussian characteristics. Secondly, due to the continuous development of target stealth, camouflage and other technologies, target detection has encountered great obstacles, which makes the data obtained by the sensor have the characteristics of fuzziness and randomness. Finally, due to the randomness of the target motion process and the coordinate system conversion in the measurement modeling process and the error of the sensor itself, measurement errors are inevitable. Target tracking technology is still a very complex research topic. Its main purpose is to obtain target information through sensors, so as to realize the monitoring of targets. In the civilian field, target tracking is mainly used in video surveillance, positioning and navigation, etc. In the military field, target tracking is a necessary function for early warning and strike systems in air (radar) and ocean (sonar) environments. Since a single array element as a scalar sensor can obtain less source information and has a large error in the estimation of the direction of arrival, the positioning accuracy is low. Compared with this, multi-array element positioning can obtain richer source parameter information, which can greatly improve the positioning performance. In traditional positioning algorithms, such as subspace fusion algorithms and two-step positioning algorithms, there is a big problem in that the complexity is high. For moving targets, the signals received by the monitoring station are changing all the time. Algorithms with high algorithm complexity will result in too long calculation time. Therefore, for the above-mentioned direct positioning algorithm, it is impossible to achieve real-time tracking of a moving target. Summary of the invention

[0003] Based on the above problems, the present invention provides a passive multi-monitoring station particle filter direct positioning and tracking method, which realizes direct positioning and tracking of the source by combining multiple monitoring stations, particle filtering and spectrum peak search PM algorithm.

[0004] A passive multi-monitoring station particle filter direct positioning and tracking method realizes the positioning and tracking of a signal source by establishing monitoring stations including a uniform linear array and located at different positions. Specifically, the method comprises the following steps:

[0005] Step 1: Establish several monitoring stations containing uniform linear arrays and located at different positions to track the source and obtain the source signal. Step 2: Establish the state model of the source. According to the knowledge of particle filtering, the particle filtering algorithm includes a prediction stage and an update stage. A certain amount of initial particles are generated in the prediction stage. The source signal obtained in step 1 is input into the state model. In the update stage, the initial particles are updated based on the source state model. Step 3: Calculate the covariance matrix of the received signal based on the measurement model of the monitoring station, and construct the PM algorithm spectrum function according to the spectrum peak search PM algorithm. Step 4: Obtain the angle of the particle relative to the monitoring station from the position of the monitoring station and the position information contained in the particle, bring the angle value into the PM algorithm spectrum function, and obtain the spectrum function value corresponding to the particle. Using the characteristic that the spectrum function value is approximately equal to the particle weight, the spectrum function value is used as the particle weight and all particle weights are normalized. Step 5: In order to avoid particle degradation, the resampling method is used to repeatedly sample particles with larger weights. Step 6: Since the particle weight is not accurately judged by a single monitoring station, the particle is resampled twice at the monitoring station, and the particle is resampled multiple times by multiple monitoring stations to eliminate the particle weight error generated by a single monitoring station. After sampling, the particle weight is set to an equal value, and the estimated value of the source position at that moment is obtained by summing the particle and its weight. Then, the state of the resampled particle at this moment is updated at the next moment, and the previous steps are repeated to finally achieve positioning and tracking in the entire time period.

[0006] Preferably, by assuming the state equation of the information source to update the particle state, a certain amount of noise is added when modeling the information source, so that the resampled particles become diversified. It is not necessary to know the specific state equation of the information source, but only to assume its state equation. The assumption of the state equation is to add noise to the particles during particle update.

[0007] As a preference, the motion equation of the information source is expressed as: t =f t|t-1 (p t-1 )+ω t , where p t Represents the state vector of the target, including the position coordinates, velocity and acceleration information of the target at time t, f t|t-1 represents the state transfer function from time t to time t-1, ω t represents the state noise of the target at time t.

[0008] As a preferred method, since the signal source is a moving target, in order to reduce the complexity of the algorithm, a particle filter combined with a spectrum peak search PM algorithm is used to achieve positioning and tracking. The one-dimensional PM algorithm is used to achieve two-dimensional direct positioning without the need for spectrum peak search, which greatly reduces the complexity of the algorithm and improves the accuracy of the algorithm through resampling.

[0009] As a preferred embodiment, when the motion state of the information source is a coordinated turning motion, the motion equation of its discrete time state is expressed as: t =Fp t-1 +G CV ω t , where the state vector is p t =[x t ,v x,t ,y t ,v y,t ,w t ],(x t ,y t ) is the position coordinate of the source, (v x,t ,v y,t ) is the speed of the source in the x and y directions, w t Indicates the target constant speed at the tth moment.

[0010] As a preference, in step 2, N is generated in the prediction phase. s The initial particles are uniform distributions in a range that includes the true value, where is the coordinate value of the ith particle at time t, is the particle velocity component, is the angular velocity of the particle; the update of the initial particle is expressed as: Among them, F is the state transfer matrix, G CV is the input matrix, ω t is the noise matrix, the subscripts t and t-1 represent the time t and t-1 respectively, and the superscript i represents the i-th particle.

[0011] Preferably, step 3 is specifically as follows: step 3.1, calculating the covariance matrix of the signal X received by the monitoring station: And divide the matrix into blocks: Where L is the number of snapshots, X is the received signal, G and H are M×K and M×(MK) matrices respectively, M is the number of array elements, and K is the number of signal sources; Step 3.2, construct the PM algorithm spectrum function: F PM (θ) = a H (θ)QQ H a(θ), where θ is the angle, a is the direction vector, Q = [P H , -I M-K ] H , where P = (G H G) -1 G H H, I M-K is the MK-dimensional identity matrix.

[0012] As a preference, the angle θ of the source relative to the monitoring station tn It is expressed as: Among them, (x t ,y t ) is the coordinate position information of the source at time t, (X n ,Y n ) is the coordinate of the nth monitoring station; the updated angle of the particle relative to the monitoring station is is the coordinate value of the ith particle at time t.

[0013] Preferably, step 4 is specifically as follows: step 4.1, using the spectral function value corresponding to the particle as the weight of the particle: Step 4.2, normalize the particle weights:

[0014] As a preferred embodiment, step 6 is specifically as follows: use multiple monitoring stations to calculate the normalized weight of the particles and perform several resamplings, and after sampling, set the particle weight to be equal to 1 / N s , the estimated value of the source position at that moment is obtained by accumulating and summing the particles and their weights, and then the state of the resampled particles at this moment is updated at the next moment, and steps 3.1 to 5 are repeated to finally achieve positioning and tracking in the entire time period. Resampling is a way for a single monitoring station to process particles through the weight information obtained by the monitoring station during sampling. Because each monitoring station receives different signals, even for the same particle, the calculated weights are different. The reason is that the weight determines the angle, not the position, so each monitoring station needs to resample the particles once. In simple terms, the first monitoring station calculates the weight, and after resampling, the resampled particles use the second monitoring station to receive the signal to calculate the second weight and perform the second resampling, and the processing of all monitoring stations is completed in sequence.

[0015] Compared with the prior art, the present invention adopts the above technical solution and has the following effects:

[0016] (1) Achieve higher performance direct positioning tracking with fewer snapshots by combining the particle filter algorithm with the peak search PM algorithm;

[0017] (2) It has low algorithm complexity and good tracking accuracy, especially in environments with poor signal-to-noise ratio, it can still obtain good tracking and positioning results;

[0018] (3) Resampling particles multiple times through multiple monitoring stations avoids the error in weight calculation at a single monitoring station and improves the accuracy of the algorithm;

[0019] (4) When the number of monitoring stations increases, the complexity will not increase significantly when the positioning accuracy increases;

[0020] (5) Compared with the subspace fusion algorithm, this algorithm only needs to perform uniform sampling in the interval containing the true initial position when setting the initial particles, while the spectrum peak search range of the subspace fusion algorithm must include the range of all positions, which greatly reduces the complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a two-dimensional positioning scene model of the present invention;

[0022] Figure 2 It is a schematic diagram of the array structure configured in the monitoring station of the present invention;

[0023] Figure 3 This is a diagram of the information source tracking result of the present invention;

[0024] Figure 4 It is a comparison diagram of the root mean square error of the tracking under different particle numbers of the present invention;

[0025] Figure 5 is a comparison diagram of the root mean square error of the tracking of the present invention under different snapshots;

[0026] Figure 6 : is a comparison diagram of the root mean square error of the positioning results under different signal-to-noise ratios when the number of monitoring stations is different in the present invention;

[0027] Figure 7 is a flow chart of the present invention; DETAILED DESCRIPTION

[0028] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0029] Symbols: In the present invention, (·) T ,(·) H , (·) -1 They are represented as transpose, conjugate transpose, and inverse respectively.

[0030] like Figure 7As shown, the present invention designs a passive multi-monitoring station particle filter direct positioning and tracking method, establishes a monitoring station with a known position, uses a uniform linear array as a signal receiving array in each monitoring station, establishes a state model of the information source, and generates a certain amount of initial particles, and the initial particles are updated by the state model of the information source; the covariance matrix is ​​calculated by the monitoring station measurement model, and the PM algorithm spectrum function is constructed according to the spectrum peak search PM algorithm; the angle of the particle relative to the monitoring station is obtained by the position information contained in the monitoring station and the particle, and the angle value is brought into the spectrum function of the spectrum peak search PM algorithm. Each particle corresponds to a spectrum function value, and the spectrum function value is used as the particle weight and all particle weights are normalized. The particle degradation problem of the particle filter algorithm is avoided by using the resampling method. Since the judgment of the particle weight by a single monitoring station is inaccurate, multiple monitoring stations are used to resample the particles multiple times, and the particle weight is set to be equal after sampling. The estimated value of the source position at this moment is obtained by the cumulative summation of the particle and its weight. Then, the state of the resampled particle at this moment is updated at the next moment, and the previous steps are repeated to finally achieve positioning and tracking in the entire time period.

[0031] On the one hand, the present invention provides a passive multi-monitoring station direct positioning and tracking method based on a particle filter algorithm. In the positioning scenario, the method assumes that there are two known monitoring stations to track the position of the signal source, and each monitoring station uses a uniform linear array as a signal receiving array element. As a further optimization scheme of the present invention, since the signal source is a moving target, in order to reduce the complexity of the algorithm, a particle filter is combined with a spectrum peak search PM algorithm to achieve positioning and tracking. On the other hand, the present invention uses the PM algorithm spectrum function as a particle weight for particle resampling and particle update, thereby obtaining a higher positioning accuracy.

[0032] Specifically, a passive multi-monitoring station particle filter direct positioning and tracking method includes the following steps:

[0033] Step 1, establish several monitoring stations including uniform linear arrays and located at different positions to track the information source and obtain the information source signal; Step 2, establish the state model of the information source, generate a certain amount of initial particles, input the information source signal obtained in step 1 into the state model, and update the initial particles based on the state model of the information source; Step 3, calculate the covariance matrix of the received signal based on the measurement model of the monitoring station, and construct the PM algorithm spectrum function according to the spectrum peak search PM algorithm; Step 4, obtain the angle information contained in the particle from the position of the monitoring station and the position information contained in the particle, bring the angle information into the PM algorithm spectrum function, obtain the spectrum function value corresponding to the particle, use the spectrum function value as the particle weight and normalize all particle weights; Step 5, resample the particles at the monitoring station to avoid particle degradation of the particle filter algorithm; Step 6, perform normalized weight calculation and resampling in different monitoring stations in turn, and finally obtain the estimated value of the information source position at that moment by weighted summation of the particles; Continuously iterate and update the particles to finally achieve positioning and tracking within the entire time period.

[0034] Specifically, the method of the present invention is theoretically explained based on four modules:

[0035] 1. Data Model

[0036] like Figure 1 The example shown is a two-dimensional positioning scenario. A plane rectangular coordinate system XOY is established in the two-dimensional space. Assume that there are N fixed monitoring stations with precisely known positions in the space. Ignore the position error of the monitoring station itself. The position of the monitoring station is U n =[X n ,Y n ] T , n=1,2,…,N, each monitoring station is equipped with a uniform linear array with M array elements. All array elements are indifferent and the mutual coupling problem between array elements is ignored. The structure of the uniform linear array is as follows: Figure 2 In two-dimensional space, there is a moving source. Assuming that it satisfies a certain motion model, its motion equation can be expressed by p t =f t|t-1 (p t-1 )+ω t , where p t represents the state vector of the target, which includes the position coordinates, velocity, acceleration and other information of the target at time t. t|t-1 is the state transfer function representing the movement from time t to time t-1, ω t represents the state noise of the target at time t, and its coordinate position information is (x t ,y t ).

[0037] Figure 1The schematic diagram of the algorithm model shows that the true azimuth angle θ of the source relative to the observation station is tn It can be expressed as

[0038]

[0039] Assuming that the signal source generates a far-field narrowband uncorrelated non-Gaussian signal, which is incident on the array of each monitoring station, the received signal can be expressed as

[0040] X(t)=As(t)+n(t) (2)

[0041] in, is the direction matrix,

[0042] represents the direction vector, θ k is the incident angle of the kth source. is a non-Gaussian source vector, J represents the number of snapshots, n(t) has a mean of 0 and a variance of Gaussian noise, where k represents the number of information sources.

[0043] 2. Particle Filter Algorithm

[0044] The particle filter algorithm is a widely used algorithm based on Bayesian theory and importance sampling. The algorithm is widely used in linear and nonlinear system filtering methods. Its basic principle is to fit the prior and posterior probability density distribution of the target through a series of random samples with weights, also known as random particles.

[0045] The goal of the particle filter algorithm is to recursively estimate the motion process of the target based on the measurement model and state model. It mainly includes prediction and update steps. In the prediction stage, a set of particles with target state information is generated. N s Represents the number of particles, assuming that the initial weight of each particle is 1 / N s , let the tracking time be T, and the output of the array at each moment be Y 1:t ={Y1,…,Y t} as observation information to update the predicted particles. At time t, a series of predicted particles are generated through the prior distribution.

[0046]

[0047] In the update phase, the particle weights are updated. Since the importance probability density is approximately the state transition probability density in sequential importance sampling, the particle weight is

[0048]

[0049] The particle weights are normalized, that is

[0050]

[0051] However, a series of iterations will lead to particle degradation, that is, only some particles have large weights, and the weights of other particles become very small, resulting in increased estimation errors. To solve this problem, a resampling algorithm is used, that is, particles with large weights are replicated multiple times to solve the above problem. After resampling, the weights of each particle are set to 1 / N again. s Finally, the particles are summed up according to their weights to get the tracking result at that moment.

[0052] 3. Peak Search PM Algorithm

[0053] Calculate the covariance matrix of the signal received by the monitoring station

[0054]

[0055] Block the covariance matrix

[0056]

[0057] Where G and H are M×K and M×(MK) matrices respectively, M is the number of array elements, and K is the number of information sources.

[0058] Estimate the propagation operator matrix P

[0059] P=(G H G) -1 G H H (8)

[0060] Construct Q matrix

[0061] Q=[P H , -I M-K ] H (9)

[0062] Construct PM algorithm spectrum function

[0063] F PM (θ) = a H (θ)QQ H a(θ) (10)

[0064] 4. Direct positioning and tracking algorithm combining particle filtering with one-dimensional spectrum peak search PM

[0065] The algorithm can track and locate the signal source that meets any motion form. We assume that the motion state of the signal source is a coordinated turning motion, and its discrete time state model can be expressed as

[0066] p t =Fp t-1+G CV ω t (11)

[0067] The state vector p t =[x t ,v x,t ,y t ,v y,t ,w t ],(x t ,y t ) is the position coordinate of the source, (v x,t ,v y,t ) is the speed of the source in the x and y directions, w t represents the target constant speed at the tth moment. The state transfer matrix Input Matrix

[0068] First, N is generated in the prediction stage s Initial Particles in are uniform distributions in a range that includes the true value.

[0069] In the monitoring phase, the particles are updated through the state model of the information source.

[0070]

[0071] The updated particle angle relative to the monitoring station is

[0072]

[0073] Then at each monitoring station, each particle has a corresponding angle value Since the observation information is contained in the spectrum function of the PM algorithm, that is, the true angle value is contained in the spectrum function, the particle corresponding angle value The closer it is to the true value, the larger its spectral function is, and the larger the weight of the particle is. Then the spectral function value of the PM algorithm can be used instead of the likelihood function to update the weight of the particle, that is,

[0074]

[0075] Then the particle weights are normalized, that is,

[0076]

[0077] In order to avoid the problem of increased error caused by particle degradation, resampling is used to replicate particles with larger weights multiple times. After resampling, the weights of all particles are set to 1 / N. s .

[0078] Since equation (1) contains both x and y, its true weight cannot be directly determined by a single monitoring station. To avoid the problem of false weights caused by a single monitoring station, the resampled particles are subjected to secondary angle calculation, secondary particle weight calculation, and secondary resampling in the form of multiple stations to solve the problem that a single monitoring station cannot fully and accurately estimate.

[0079] Finally, the estimated value of the source position at this moment is

[0080] Most of the resampled particles are concentrated near the real position of the source, so the particle at this moment is used as the initial particle at the next moment to update the particle at the next moment, thereby realizing the positioning and tracking of a certain target within a period of time.

[0081] The algorithm flow is as follows:

[0082]

[0083] Figure 3 This is a result diagram of the information source tracking of the present invention when the signal-to-noise ratio is 5 dB, the number of snapshots is 200, the number of particles is 1000, and the tracking time is 60 seconds.

[0084] Figure 4 This is a comparison diagram of the root mean square error of the present invention during the entire tracking time when the number of snapshots is 200 and different numbers of particles are under different signal-to-noise ratios.

[0085] Figure 5 This is a comparison diagram of the root mean square error of the present invention during the entire tracking time when the number of particles is 1000 and the number of snapshots is different under different signal-to-noise ratios.

[0086] Figure 6 This is a comparison chart of the tracking performance of the present invention when the number of snapshots is 200, the number of particles is 1000, the number of monitoring stations is different, and the signal-to-noise ratio is different.

[0087] Depend on Figures 3 to 6 As shown, the particle filtering direct positioning and tracking method based on multiple monitoring stations of the present invention achieves higher performance direct positioning and tracking with fewer snapshots, and has better tracking accuracy, and can still obtain better tracking and positioning results in an environment with poor signal-to-noise ratio.

[0088] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A passive multi-monitoring station particle filter direct positioning and tracking method, characterized in that: The following steps are involved: Step 1, establish a number of monitoring stations including uniform linear arrays and located at different positions to track the information source and obtain the information source signal; Step 2, establishing a state model of the information source, generating a certain amount of initial particles, inputting the information source signal obtained in step 1 into the state model, and updating the initial particles based on the state model of the information source; Step 3, based on the measurement model of the monitoring station, the covariance matrix of the received signal of the signal source incident on the monitoring station array is calculated, and the PM algorithm spectrum function is constructed according to the spectrum peak search PM algorithm; Step 4, obtaining the angle information contained in the particle from the position of the monitoring station and the position information contained in the particle, bringing the angle information into the PM algorithm spectrum function, obtaining the spectrum function value corresponding to the particle, using the spectrum function value as the particle weight and normalizing all particle weights; Step 5, resample the particles at the monitoring station to obtain resampled particles to avoid particle degradation of the particle filter algorithm; Step 6: Move to the next monitoring station to resample the particle, repeat steps 3 to 5, perform normalized weight calculation and resample at different monitoring stations, and continuously iterate and update the particles until all monitoring stations are traversed. Finally, perform weighted summation on the particles to obtain the estimated value of the source position at time t, thus achieving positioning and tracking in the entire time period.

2. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 1, characterized in that: A uniform linear array is used as the signal receiving array in each monitoring station, and the moving signal source satisfies a certain mathematical model in a two-dimensional plane monitoring scenario.

3. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 2, characterized in that: The motion equation of the source is expressed as: t =f t|t-1 (p t-1 )+ω t , where p t Represents the state vector of the target, including the position coordinates, velocity and acceleration information of the target at time t, f t|t-1 represents the state transfer function from time t to time t-1, ω t represents the state noise of the target at time t.

4. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 3, characterized in that: When the motion state of the information source is a coordinated turning motion, the motion equation of its discrete time state is expressed as: t =Fp t-1 +G CV ω t , where the state vector is p t =[x t ,v x,t ,y t ,v y,t ,w t ],(x t ,y t ) is the position coordinate of the source, (v x,t ,v y,t ) is the speed of the source in the x and y directions, w t represents the target constant speed at the tth moment, F is the state transfer matrix, G CV is the input matrix.

5. A passive multi-monitoring station particle filter direct positioning and tracking method according to any one of claims 1 to 4, characterized in that: In step 2, N is generated in the prediction phase s The initial particles are uniform distributions in a range that includes the true value, where is the coordinate value of the ith particle at time t, is the particle velocity component, is the angular velocity of the particle; The update of the initial particles is expressed as: Among them, F is the state transfer matrix, G CV is the input matrix, ω t is the noise matrix, the subscripts t and t-1 represent the time t and t-1 respectively, and the superscript i represents the i-th particle.

6. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 5, characterized in that: The step 3 is specifically as follows: Step 3.1, calculate the covariance matrix of the signal X received by the monitoring station: And divide the matrix into blocks: Where L is the number of snapshots, X is the received signal, G and H are M×K and M×(MK) matrices respectively, M is the number of array elements, and K is the number of signal sources; Step 3.2, construct the PM algorithm spectrum function: F PM (θ) = a H (θ)QQ H a(θ), where θ is the angle, a is the direction vector, Q = [P H , -I M-K ] H , where P = (G H G) -1 G H H, I M-K is the MK-dimensional identity matrix.

7. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 6, characterized in that: The angle θ of the source relative to the monitoring station tn It is expressed as: Among them, (x t ,y t ) is the coordinate position information of the source at time t, (X n ,Y n ) is the coordinate of the nth monitoring station; the updated angle of the particle relative to the monitoring station is is the coordinate value of the ith particle at time t.

8. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 7, characterized in that: The step 4 is specifically as follows: Step 4.1, use the spectral function value corresponding to the particle as the particle weight: Step 4.2, normalize the particle weights:

9. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 8, characterized in that: The step 6 is specifically as follows: using several monitoring stations to calculate the normalized weight of the particles and perform several resamplings, and after sampling, the particle weight is set to the same value 1 / N s , the estimated value of the source position at time t is obtained by accumulating and summing the particles and their weights, and then the state of the resampled particles at time t is updated at the next moment, and steps 3 to 5 are repeated to finally achieve positioning and tracking in the entire time period.

10. A passive multi-monitoring station particle filter direct positioning and tracking method according to claim 9, characterized in that: The estimated source position obtained in step 6 is

Images