Target tracking method based on periodic pulse signal, recording medium and system

Abstract

The present application belongs to the technical field of target tracking, and particularly relates to a target tracking method based on periodic pulse signals, which calculates state data observation values of observed pulse signal emission points at each time by fusing a plurality of kinematic models of a target and an observation model based on periodic pulse signals of the target "DOA-TOA-TDOA", and obtains state data estimation values at a starting time by using a PLE algorithm after SG filtering and intercepting state data at previous time points; the state data estimation values at the next time are gradually derived from the starting time by using an unscented Kalman filtering algorithm, and the tracking of the target is completed. The method can significantly improve the positioning accuracy, the reliability of the estimation values and the stability of the tracking, and is particularly suitable for tracking a target emitting periodic pulse signals underwater. The present application also provides a non-transient readable recording medium storing a program of the method and a system containing the medium, and the program can be called by a processing circuit to execute the above method.

Description

Technical Field

[0001] The invention belongs to the technical field of target tracking and discloses a target tracking method, recording medium and system based on periodic pulse signals. Background Art

[0002] Underwater target positioning and tracking technology is an important research direction in the field of underwater acoustics. In the application of marine resource development and marine ecological protection, accurate positioning and stable tracking of targets are of great significance. Figure 1 This is a schematic diagram of the target passive positioning observation system. The target carries an acoustic beacon, which emits an acoustic signal at a fixed period. The emitted signal can be measured in advance, so the signal frequency band and other sound source parameters can be regarded as known information. In addition, it is equipped with a depth sensor, so the target working depth can be regarded as a known quantity. The schematic diagram of the target tracking model scenario is shown in Figure 2. Figure 2 As shown in the figure, the observation platform carries an acoustic positioning array consisting of a four-element short baseline array (SBL) as the receiving end. Each array element receives the periodic pulse signal transmitted by the target, extracts observation information from the received signal, and establishes a robust target positioning and tracking model to achieve precise positioning and tracking of the target. Therefore, the observation model and positioning and tracking model are the decisive factors affecting target positioning and tracking performance.

[0003] In recent years, scholars both domestically and internationally have conducted extensive research on observation estimation and tracking algorithms in the field of underwater target positioning and tracking. Common observation estimation methods include time delay estimation, frequency estimation, and bearing estimation. However, these methods often collect only one or two observations during positioning. The limited amount of information results in inaccurate positioning, while excessive information collection often leads to data redundancy, consumes computing resources, prolongs the time to obtain prediction results, and delays tracking. Due to interference and the influence of the propagation medium during the measurement process, observation noise is high, resulting in low data reliability and unstable tracking.

[0004] Moreover, the accuracy of tracking a moving object depends on the estimation of its motion state. The motion state of underwater targets is often changing. Due to the lack of prior information, a single motion model often deviates greatly from the actual state, which ultimately leads to unreliable tracking. Summary of the Invention

[0005] In view of the above problems, the present invention provides a target tracking method based on a periodic pulse signal, comprising the following steps:

[0006] S1. Construct multiple kinematic models of the target and an observation model for a periodic pulse signal based on the target's "direction of arrival (DOA)-pulse signal propagation time (TOA)-time difference of arrival (TDOA)" relationship. The observation model describes the correspondence between "direction of arrival (DOA)-pulse signal propagation time-time difference of arrival (TDOA)" and state data observations, including but not limited to the target's three-dimensional coordinates, motion state, and measurement noise.

[0007] S2. Searching for the peak of the multiple signal classification spectrum function to estimate the direction of arrival of the signal, and calculating the azimuth of the target based on the direction of arrival;

[0008] The pulse signal propagation time is determined by the spatial position relationship of different observation points and the time difference of receiving the same period pulse signal;

[0009] Perform cross-correlation calculation on adjacent periodic pulse signals at the same observation point to find the peak position of the correlation function, and use the delay corresponding to this position as the arrival time difference of the next periodic pulse;

[0010] S3. Substitute the pulse signal azimuth, pulse signal propagation time, and next cycle pulse arrival time difference data from the start to the previous moment, as well as the pulse signal azimuth and pulse signal propagation time observed at the last moment, into the observation model to calculate the state data observation value of the pulse signal emission point at each moment;

[0011] S4. Smoothe the observed data using the SG filtering algorithm, then extract the state data from the previous few moments and use the pseudo-linear initial value estimation algorithm to obtain the estimated state data value at the starting time;

[0012] S5. Starting from the starting moment, multiple kinematic models are integrated, the estimated state data value and the observed state data value at the current moment are substituted, and the estimated state data value at the next moment is gradually derived through the unscented Kalman filter algorithm to complete the tracking of the target.

[0013] Preferably, the multiple kinematic models include a uniform linear motion model, a uniform acceleration motion model, a uniform angular velocity motion model, and segmented combinations of the above motion models.

[0014] Preferably, the method of determining the pulse signal propagation time through the spatial position relationship between different observation points and the time difference of receiving the same periodic pulse signal is a triangulation positioning algorithm.

[0015] Preferably, when the target is equipped with a depth sensor, the observation model of the target is constructed by the following steps:

[0016]

[0017] Where, θ k is the azimuth of the signal at the kth moment, (x k ,y k ,z k ) is the three-dimensional coordinate of the target at the kth moment; t k,i is the propagation time of the pulse signal when the signal reaches the observation point i at the kth moment, where the coordinates of the observation point i are expressed as (x i ,y i ,z i ), i=1,2…; △T k is the arrival time difference of the next cycle pulse at the kth moment, and the observation model at the last observation moment only includes θ k and t k,i ; T r is the period of the target transmitting pulse signal, d x 、d y is the displacement of the target along the x and y axes within the Tr time; n k is the observation noise at time k.

[0018] Preferably, the step of fusing multiple kinematic models in S5 includes: setting an initial state and initial probability for each kinematic model based on prior knowledge or historical data; predicting the mixed state and covariance at the next moment based on the probability and state of each model at the previous moment; updating the predicted state of the model using the state data observation value, calculating the updated mixed state and covariance and the likelihood function value of each model and updating the model probability, and performing weighted aggregation based on the updated model probability to obtain the motion state model of the target after fusion at each moment.

[0019] Another embodiment of the present invention provides a non-transitory readable recording medium for storing one or more programs including a plurality of instructions. When the instructions are executed, the processing circuit will execute the target tracking method based on the periodic pulse signal.

[0020] Another embodiment of the present invention is to provide a target tracking system based on a periodic pulse signal, comprising a processing circuit and a memory electrically coupled thereto, wherein the memory is configured to store at least one program, wherein the program includes a plurality of instructions, and the processing circuit runs the program to execute the above-mentioned target tracking method based on a periodic pulse signal.

[0021] Compared with the prior art, the target tracking method, recording medium, and system based on periodic pulse signals provided by the present invention have the following beneficial effects: the method of the present invention selects multiple underwater periodic pulse signal parameter types for target positioning and tracking, which can improve positioning accuracy compared to a single type;

[0022] The target positioning and tracking model in the method described in the present invention first processes the observation data through SG (Savitzky-Golay) smoothing filtering and then performs pseudo-linear initial value estimation (PLE / Pseudolinear Estimate). Combined with the initial estimate, the unscented Kalman filter (UKF / Unscented Kalman Filter) is used to achieve state estimation. Compared with the traditional Kalman filter algorithm, it can significantly improve tracking stability.

[0023] The method described in the present invention designs an Interacting Multiple Model (IMM) algorithm to study the target motion model and designs a multi-motion state fusion algorithm based on the original positioning and tracking algorithm to achieve accurate estimation of underwater targets.

[0024] The present invention studies a special positioning and tracking algorithm for the characteristics of underwater periodic pulse signals. TDOA is different from previous definitions. In the present invention, it innovatively represents the time difference between adjacent periodic pulse signals arriving at a reference array element (observation point), breaking through the limitation of traditional TDOA that only represents the time difference between signals arriving at different array elements. Without increasing the signal source, it provides more comprehensive information for the observation point, which is beneficial to the accuracy and timeliness of tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Schematic diagram of the acoustic beacon passive positioning observation system in the background technology;

[0026] Figure 2 Schematic diagram of the spatial position relationship between the four-element short baseline array observation platform and the target in the background technology;

[0027] Figure 3 This is a schematic diagram of the multi-model collaborative interaction process in an embodiment of the present invention;

[0028] Figure 4 : This is a target trajectory tracking effect diagram under the CV-CA model in an embodiment of the present invention;

[0029] Figure 5 is a curve showing the change of the normalized root mean square error (RMSE) of target tracking over time under the CV-CA model in an embodiment of the present invention;

[0030] Figure 6is a target trajectory tracking effect diagram under the CV-CT model in the embodiment of the application;

[0031] Figure 7 is a target tracking normalized root mean square error (RMSE) curve with time under the CV-CT model in the embodiment of the application;

[0032] Figure 8 is a schematic diagram of the arrangement structure of each observation point in the embodiment of the application. DETAILED DESCRIPTION

[0033] In order to make the objects, technical solutions and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described below with reference to the drawings in the embodiments of the application. The described embodiments are part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without making innovative labor fall within the protection scope of the application.

[0034] The embodiment of the application provides a target tracking method based on a periodic pulse signal, including the following steps: the specific implementation process is as follows:

[0035] Step 1: constructing a target kinematic model and an observation model

[0036] The target kinematic model includes a constant velocity (CV) model, a constant acceleration (CA) model and a constant angular velocity (CT) model.

[0037] (1) CV model

[0038] The target state vector is: ;

[0039] The state transition matrix is: ;

[0040] (2) CA model

[0041] The target state vector is:

[0042] ;

[0043] The state transition matrix is:

[0044]

[0045] (3) CT model

[0046] The target state vector is:

[0047] ;

[0048] The state transition matrix is:

[0049]

[0050] The target observation model is determined by the following steps:

[0051]

[0052] wherein, θ k is the azimuth angle of the signal at the kth moment, (x k ,y k ,z k ) is the three-dimensional coordinate of the target at the kth moment; t k,i is the pulse signal propagation time of the signal arriving at the observation point i at the kth moment, wherein the coordinates of the observation point i are represented as (x i ,y i ,z i ), i = 1, 2, …; △T k is the time difference of the next period pulse arriving at the kth moment, and the observation model at the last observation moment only contains θ k and t k,i ; T r is the period of the target transmitting the pulse signal, d x 、d y is the displacement of the target along the x and y axes within the Tr time; n k is the observation noise at the kth moment.

[0053] Step 2: Design a periodic pulse signal parameter estimation method based on DOA-TOA-TDOA

[0054] According to the waveform form of the transmitted sound source signal, a corresponding DOA estimation algorithm is designed, and the multiple signal classification (MUSIC) spectrum function is represented as:

[0055] ;

[0056] wherein, θ represents the direction of arrival, Φ represents the pitch angle, a( θ,Φ ) is the array flow vector, E n is the noise subspace matrix, and H represents the conjugate transpose operation; the direction of arrival of the signal is estimated by searching the peak value of the spectrum function.

[0057] The time of arrival (TOA) of the pulse signal is determined by the spatial position relationship of different observation points and the time difference of receiving the same periodic pulse signal.

[0058] Perform cross-correlation calculations on adjacent periodic pulse signals at the same observation point to find the peak position of the correlation function, and use the delay corresponding to this position as the time difference of arrival (TDOA) of the next periodic pulse.

[0059] Step 3: Analyze the observable conditions of the target positioning and tracking system based on periodic pulse signal parameters

[0060] The comprehensive observation at the kth observation moment is defined as:

[0061] ;

[0062] Calculate the partial derivatives of the DOA and TOA parameters respectively:

[0063] ;

[0064] ;

[0065] The Jacobian matrix using only a single DOA-TOA observation is:

[0066] ;

[0067] Find the partial derivative of TDOA:

[0068] ;

[0069] where it is defined , the partial derivative of TDOA with respect to ( d x ,d y ) is not zero.

[0070] Therefore, the observation equation pair composed of the DOA-TOA-TDOA system ( d x ,d y )、(x k ,y k, ,d x ,d y ) can achieve full column rank (i.e., rank 4) under general positions and velocities. Therefore, only a single observation (measurement at the same time) is required to achieve complete observability of the target position and velocity state.

[0071] Step 4: Build a target positioning and tracking model

[0072] The construction of the target positioning and tracking model is determined by the following steps:

[0073] First, the initial value is obtained by pseudo linear initial value estimation (PLE). The true value calculation formula under noise-free conditions is as follows:

[0074] ;

[0075] Introducing pitch angle Convenient calculation:

[0076]

[0077] The measurement equation of the pseudo-linearization of the system can be obtained as follows:

[0078]

[0079] in, and are the measurement noise of DOA and TDOA respectively. It should be noted that the pitch angle It is not used as a landscape measurement for this field, but only used when the system is pseudo-linearized. The value is obtained by converting the time delay information contained in TOA into distance information and then combining it with the known depth information z. Therefore, the PLE model of DOA-TDOA actually requires the TOA observation, and from a broad perspective, it is also a PLE model of DOA-TOA-TDOA.

[0080] After M observations, the observation data is first smoothed by the SG filtering algorithm. The target state vector In the absence of prior information, it is necessary to use the PLE algorithm to select the observation data of the previous moments to find the t The initial estimate X at time 0 ple , the algorithm needs to pre-set the value of matrix R for calculation. Then the state vector and observation model are substituted into the Unscented Kalman Filter (UKF) algorithm, which can be calculated based on the system at t k Estimated value X at time k｜k , calculate t k+1 The estimated value of the system state at time X k+1｜k+1 It should be noted that at the last observation moment, the observation model only contains DOA and TOA. At this time, the UKF algorithm should be adjusted accordingly, and only DOA and TOA observation information should be used for state update and state prediction correction. t Starting from time 0, the UKF algorithm is used to gradually recurse to t k+1 At this moment, the target is located and tracked based on the estimated value of the system state at that moment.

[0081] Step 5: Design a collaborative interactive target positioning and tracking method based on the IMM model

[0082] Using the multiple target kinematic models in step 1, combined with the observation model in step 1 and the target positioning and tracking model in step 4, IMM collaborative interaction positioning and tracking targets are performed. The specific algorithm implementation process of the multi-model collaborative interaction algorithm (IMM) is as follows: Figure 3 Figure 1 shows the IMM algorithm flow for two models within one IMM algorithm cycle. The algorithm includes an interactor, multiple filters, a model probability updater, and an estimation mixer. The IMM algorithm achieves adaptive filtering of dynamic systems with uncertain models by performing weighted interactions and probability updates between models to determine the optimal estimated state at the current moment.

[0083] As a real-time update algorithm, the basic concept of the IMM algorithm is to use multiple different motion models to match the different motion patterns of a maneuvering target. The transition probabilities between different models are a Markov matrix. The target state estimation and model probability update use Kalman filtering. An IMM cycle includes four steps: model input interaction, sub-model filtering and prediction, model probability update, and fusion output:

[0084] (1) Model input interaction: At any tracking moment, set model filters corresponding to the number of possible target motion models, and each filter corresponds to a possible system dynamic model.

[0085] (2) Sub-model filtering and prediction: Perform state estimation on each filter to obtain the state estimation value and covariance matrix of each model.

[0086] (3) Model probability update: Update the weight coefficient of each model and the probability of model update based on the observed data.

[0087] (4) Fusion output: The estimation results of all models are weighted and fused according to the weight coefficient to obtain the final state estimation.

[0088] In order to verify the effectiveness of the method of the present invention, a detailed description is given below with reference to specific examples and accompanying drawings.

[0089] In the simulation conditions, the observation platform is assumed to be stationary, and the observation system has a total of 4 receiving array elements. The specific array structure is as follows: Figure 8 As shown, the first array element is set as the reference array element, and the reference array element is the coordinate origin. The sound source signal is a linear frequency modulation signal with a frequency band range of 30kHz to 60kHz and a signal sampling frequency of f sis 0.005 s, and the pulse period T is 1 s. The target is equipped with a depth sensor that can provide real-time depth information, so z is always known during the motion process. Therefore, only the xy-plane trajectory of the target is studied, and the subsequent direction and speed are the azimuth of the xy-plane and the xy-plane speed by default. The initial position of the target is assumed to be (300 m, 350 m, 30 m). It is assumed that the target emits an LFM signal at the initial time, and then continues to emit periodic pulse signals at intervals of pulse period T during the motion process. The underwater sound speed is 1500 m / s. The signal-to-noise ratio is set to 0 dB, and the observation station observes the target for 60 s of motion, 1 s for a set of DOA-TOA-TDOA data, a total of 60 sets of data. The standard deviation of the angle DOA observation noise is 0.02°, the standard deviation of the signal arrival time TOA noise is 0.002 ms, and the standard deviation of the adjacent pulse arrival time difference TDOA noise is 0.02 ms.

[0090] The specific implementation process is as follows:

[0091] In an underwater acoustic target tracking system, the motion model and the observation model are the core components of the target state estimation. The fusion process of these models in the unscented Kalman filter is crucial to the tracking accuracy.

[0092] The motion model receives the state estimation value at the current time and the sampling period as input. In the underwater tracking system, three motion models are commonly used: constant velocity (CV) model, constant turn (CT) model, and constant acceleration (CA) model.

[0093] For the CV model, the input is a four-dimensional state vector containing the target position (x, y) and velocity (v x ,v y ), assuming that the target moves at a constant speed. The CV model linearly extrapolates the position by the speed through the state transition matrix, and the speed remains unchanged.

[0094] For the CT model, the state vector increases by one dimension, the angular velocity ω, assuming that the target moves at a constant angular velocity. The CT model considers the rotational motion of the target, updates the speed direction by the angular velocity, and makes the position evolve along the curved trajectory.

[0095] For the CA model, the state vector contains six components of position (x, y), velocity (v x ,v y ), and acceleration (a x ,a y ), assuming that the target moves at a constant acceleration. The CA model considers the influence of acceleration, making the speed change linearly with time, and the position evolves along the quadratic curve.​​​

[0096] The state transfer equations of the three motion models are:

[0097] CV model: Position updates are calculated by adding the position and velocity times the time interval. Velocity remains constant, indicating uniform linear motion. CT model: Both position and velocity are updated using a rotation matrix, taking into account the influence of angular velocity. CA model: Position updates include a first-order term (velocity) and a second-order term (half the acceleration times the square of the time interval). Velocity updates are calculated by adding the acceleration and times the time interval. Acceleration remains constant.

[0098] The output of the motion model is the predicted state vector at the next moment, and each model takes into account the corresponding process noise covariance to reflect the uncertainty in the motion process.

[0099] The observation model converts the predicted state vector into observable physical quantities. In an underwater acoustic tracking system, observable quantities include time of arrival (TOA), angle of arrival (DOA), and time difference of arrival (TDOA). TOA represents the time it takes for an acoustic wave to travel from the target to each hydrophone array element. DOA represents the directional angle at which the acoustic wave arrives at the array. TDOA represents the time difference between the arrivals of adjacent pulse signals relative to a reference array element.

[0100] During the UKF fusion process, the model mixing probability is first calculated using the Interacting Multi-Model (IMM) algorithm. The states and covariances of the different models are then weighted together to produce the initial state. For each mixed model, a set of Sigma points is generated to capture the statistical properties of the state distribution. These Sigma points are generated through the Cholesky decomposition of the current state estimate and covariance matrix and are assigned different weights for subsequent calculations.

[0101] In the prediction step, the system substitutes each Sigma point into the corresponding motion model (CV, CT, or CA) for evolution to obtain the set of Sigma points of the predicted state. The mean of the predicted state is then calculated by weighted averaging, and the covariance of the predicted state is calculated by the weighted outer product, and the process noise covariance is added.

[0102] During the update step, the system substitutes the predicted state Sigma points into the observation model and calculates the corresponding observation Sigma points. The mean of the predicted observations is obtained through weighted averaging, and the covariance of the predicted observations and the cross-covariance between the state and the observations are calculated. During this process, the angular observations are normalized to ensure they are within the range of -π to π.

[0103] The system updates its state using the Kalman gain, which determines the influence of the actual observation on the state estimate. The system calculates the observation residual (the difference between the actual observation and the predicted observation), normalizes the angle component, and then applies the Kalman gain to correct the predicted state and covariance. The system also calculates the model likelihood, which is the probability density of the observation residual under the predicted observation distribution, and uses it to update the model probability.

[0104] Finally, the system uses the updated model probabilities to perform a weighted combination of the filtering results from different models to obtain the final state estimate. The system may adjust model weights for different motion phases, for example, favoring the CV model during constant speed, the CT model during cornering, and the CA model during acceleration, thereby accurately tracking targets with complex motion characteristics.

[0105] Through this iterative fusion process, the Unscented Kalman Filter continuously combines predicted and observed information, improving the accuracy of state estimation over time and effectively tracking underwater moving targets. In practice, the system receives acoustic signals through a hydrophone array, extracts observations such as TOA, DOA, and TDOA through digital signal processing, and then reconstructs the target's trajectory using the aforementioned state estimation algorithm.

[0106] Assume that the target's actual trajectory and parameter changes during motion are as follows:

[0107] From 1 to 21 seconds, the motion model is CV; the uniform velocity is 5 m / s; the azimuth angle is 100°;

[0108] From 21 to 60 seconds, the motion model is CA; the uniform acceleration is 0.05 m / s 2 ; Azimuth 110°;

[0109] Figure 4 The model demonstrates target trajectory tracking performance. In this scenario, the dotted line represents the target's true trajectory, the dashed line represents the algorithm's estimated trajectory, the starting and ending points are marked with dots and asterisks, respectively, and the circles represent the model switching points. Comparing the three-dimensional trajectories shows a high degree of agreement between the algorithm's estimated and true trajectories, validating the tracking accuracy of the proposed method.

[0110] Figure 5 A time-varying curve of the normalized root mean square error (RMSE) of target tracking using the proposed algorithm under the CV-CA model is presented. The overall error remains low, with a maximum value not exceeding 1.2 meters, and fluctuates between 0.6 and 0.8 meters for most of the time period, demonstrating the proposed algorithm's good stability and reliability.

[0111] Assume that the target's actual trajectory and parameter changes during motion are as follows:

[0112] From 1 to 21 seconds, the motion model is CV; the uniform velocity is 5 m / s; the azimuth angle is 70°;

[0113] From 21 to 60 seconds, the motion model is CT; the uniform angular velocity is -0.0873 rad / s; the speed is 5 m / s; the azimuth angle is constantly changing;

[0114] Figure 6 The underwater motion tracking effect of this model is demonstrated. In this scenario, a negative turning angular velocity indicates a clockwise turn. The dotted line represents the target's actual trajectory, while the dashed line represents the algorithm's estimated trajectory. The starting and ending points are marked with a circle and an asterisk, respectively, and the circle represents the model switching point. The three-dimensional trajectory curve shows that the algorithm's estimated trajectory and the actual trajectory are nearly identical, demonstrating that the proposed method can accurately track the motion of underwater targets.

[0115] Figure 7 The time-varying trend of the normalized root mean square error (RMSE) of target tracking for the proposed algorithm under the CV-CT model is presented. Except for a peak error of approximately 3 meters in the 30-40 second interval, the error generally remains below 1 during other time periods, fully verifying the strong tracking robustness and reliability of the proposed algorithm. It should be noted that the algorithms under both combined motion models show an increasing trend in the tracking end-point error, likely due to increased measurement noise and impacted tracking performance as the target's motion distance increases.

[0116] Compiling the above-mentioned method steps into a program and then storing it on a hard disk or other non-transitory storage medium constitutes an embodiment of the present invention's "a non-transitory readable recording medium"; and electrically connecting the storage medium to a computer processor and completing target tracking based on periodic pulse signals through data processing constitutes an embodiment of the present invention's "a target tracking system based on periodic pulse signals".

[0117] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computers containing computer-usable program code, or on a usable storage medium (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.).

[0118] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0119] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0121] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are 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 will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A target tracking method based on periodic pulse signals, characterized in that: The following steps are involved: S1. Construct multiple kinematic models of the target and observation models of periodic pulse signals based on the target's "azimuth angle - pulse signal propagation time - next period pulse arrival time difference". The observation models are used to describe the correspondence between "azimuth angle - pulse signal propagation time - next period pulse arrival time difference" and state data observations, which include but are not limited to the target's three-dimensional coordinates, motion state, and measurement noise; S2. Searching for the peak of the multiple signal classification spectrum function to estimate the direction of arrival of the signal, and calculating the azimuth of the target based on the direction of arrival; The pulse signal propagation time is determined by the spatial position relationship of different observation points and the time difference of receiving the same period pulse signal; Perform cross-correlation calculation on adjacent periodic pulse signals at the same observation point to find the peak position of the correlation function, and use the delay corresponding to this position as the arrival time difference of the next periodic pulse; S3. Substitute the pulse signal azimuth, pulse signal propagation time, and next cycle pulse arrival time difference data from the start to the previous moment, as well as the pulse signal azimuth and pulse signal propagation time observed at the last moment, into the observation model to calculate the state data observation value of the pulse signal emission point at each moment; S4. Smoothe the observed data using the SG filtering algorithm, then extract the state data from the previous few moments and use the pseudo-linear initial value estimation algorithm to obtain the estimated state data value at the starting time; S5. Starting from the starting moment, multiple kinematic models are integrated, the estimated state data value and the observed state data value at the current moment are substituted, and the estimated state data value at the next moment is gradually derived through the unscented Kalman filter algorithm to complete the tracking of the target.

2. A target tracking method based on a periodic pulse signal according to claim 1, characterized in that: The various kinematic models include uniform linear motion model, uniform acceleration motion model, uniform angular velocity motion model, and segmented combinations of the above motion models.

3. The target tracking method based on periodic pulse signals according to claim 2, characterized in that: The method of determining the propagation time of a pulse signal by the spatial position relationship of different observation points and the time difference of receiving the same periodic pulse signal is called the triangulation positioning algorithm.

4. The target tracking method based on periodic pulse signals according to claim 3, characterized in that: When the target is equipped with a depth sensor, the observation model of the target is constructed by the following steps: Where, θ k is the azimuth of the signal at the kth moment, (x k ,y k ,z k ) is the three-dimensional coordinate of the target at the kth moment; t k,i is the propagation time of the pulse signal when the signal reaches the observation point i at the kth moment, where the coordinates of the observation point i are expressed as (x i ,y i ,z i ), i=1,2…; △T k is the arrival time difference of the next cycle pulse at the kth moment, and the observation model at the last observation moment only includes θ k and t k,i ; T r is the period of the target transmitting pulse signal, d x 、d y is the displacement of the target along the x and y axes within the Tr time; n k is the observation noise at time k.

5. The target tracking method based on periodic pulse signals according to claim 4, characterized in that: The steps of fusing multiple kinematic models in S5 include: setting the initial state and initial probability for each kinematic model based on prior knowledge or historical data; predicting the mixed state and covariance at the next moment based on the probability and state of each model at the previous moment; updating the predicted state of the model using the state data observation value, calculating the updated mixed state and covariance and the likelihood function value of each model and updating the model probability, and performing weighted aggregation based on the updated model probability to obtain the motion state model of the target after fusion at each moment.

6. A non-transitory readable recording medium for storing one or more programs comprising a plurality of instructions, characterized in that: When the instruction is executed, the processing circuit is caused to execute the target tracking method based on periodic pulse signals according to any one of claims 1 to 5.

7. A target tracking system based on a periodic pulse signal, comprising a processing circuit and a memory electrically coupled thereto, characterized in that: The memory configuration stores at least one program, the program including a plurality of instructions, and the processing circuit runs the program to execute a target tracking method based on a periodic pulse signal according to any one of claims 1 to 5.

Images