Target tracking method, target tracking system and storage medium

Abstract

The present application relates to a target tracking method, a target tracking system and a storage medium, wherein the target tracking method includes: determining the predicted state value of the tracking target and updating the predicted covariance matrix; performing measurement traversal to obtain the measurement points associated with the tracking target; determining the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate by calculating the new information vector obtained by the observation conversion value of the measurement points associated with the tracking target and the current predicted state value of the tracking target; when it is determined that the measurement covariance matrix of the tracking target needs to be updated; the target covariance matrix of the tracking target is obtained by comparing the first covariance matrix of the new information vector with the second covariance matrix of each type of target in the offline acquisition module, and updating the measurement covariance matrix of the tracking target; wherein the second covariance matrix is ​​obtained by pre-collecting the data statistics of each type of target. Through the present application, the calculation amount of the algorithm is reduced and the accuracy of the tracking algorithm is improved.

Description

Technical Field

[0001] The present application relates to the field of radar target tracking, and in particular to a target tracking method, a target tracking system and a storage medium. Background Art

[0002] At present, the millimeter-wave radar multi-target tracking method mainly obtains the target observation value of the millimeter-wave radar, then constructs a tracker model based on the interactive multi-model IMM structure, pre-tracks the target based on the adaptive Kalman filter method AKF, constructs a training sample set and a test sample set, constructs a classifier model and iteratively trains it, and then obtains the target category information in each range Doppler map (hereinafter referred to as RD map), and obtains the target tracking result of the millimeter-wave radar based on the target category information feedback. In this method, the target type is trained using the relevant information of the RD map detected by the radar, and the target type is used as a reference for the selection of the tracking model. Since there are many points of the same distance and speed of different targets in a multi-target environment, these points are in a block in the RD map and cannot be distinguished, which will affect the training of the classifier; in addition, this method also depends on the quality of the RD map, and it is necessary to save and analyze multiple frames of RD maps, which requires saving a large amount of data, and is relatively complex in calculation, which is not suitable for real-time systems.

[0003] Alternatively, a feature-driven multi-target tracking method is used. By obtaining the spatiotemporal spectral features of the target, a tree-like feature trace is constructed using a multi-hypothesis method. By calculating the credibility of each branch, the branch with the highest credibility is retained, and the branches with credibility less than the threshold are pruned. This method mainly uses the spatiotemporal spectral features of the target. Compared with radar targets, this feature is easier to extract, while the shape features of the target are difficult to extract under general radars. In addition, the multi-hypothesis tracking (MHT) method designed to improve the continuity and effectiveness of the trajectory also has the disadvantages of large calculation and complex construction.

[0004] There is no effective solution to the problems of inaccurate tracking and large amount of calculation in the millimeter wave radar multi-target tracking algorithm in related technologies. Summary of the invention

[0005] In this embodiment, a target tracking method, a target tracking system and a storage medium are provided to solve the problems of inaccurate tracking and large amount of calculation in the millimeter wave radar multi-target tracking algorithm in the related art.

[0006] In a first aspect, a target tracking method is provided in this embodiment, including:

[0007] Determine the predicted state value of the tracking target and update the predicted covariance matrix;

[0008] Conduct measurement traversal on the measurement data to obtain the measurement points associated with the tracking target;

[0009] Calculate the difference between the observed conversion value of the measurement point associated with the tracking target and the current predicted state value of the tracking target to obtain the new information vector;

[0010] Determine the number of target measurement points and the distribution probability of the measurement points that fall into the tracking target wave gate according to the new information vector;

[0011] Determine whether the measurement covariance matrix of the tracking target needs to be updated based on the number of target measurement points and the distribution probability of the measurement points;

[0012] If yes, then the target covariance matrix of the tracked target is obtained according to the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module; wherein the second covariance matrix is ​​obtained by pre-collecting data statistics of each type of target;

[0013] The measurement covariance matrix of the tracked target is updated through the target covariance matrix.

[0014] In some of the embodiments, the second covariance matrix is ​​obtained by pre-collecting statistics of various types of target data, including:

[0015] Collecting first motion data of various types of targets through radar, and obtaining a set of associated points of the first motion data;

[0016] Second motion data of each type of target obtained by fitting the associated point set of the first motion data;

[0017] Calculate the deviation between the second motion data and the first motion data to obtain a deviation value matrix for each type of target;

[0018] The mean of the deviation value matrix of each type of target is calculated, and the second covariance matrix of each type of target is obtained according to the mean of the deviation value matrix of each type of target, and the second covariance matrix of each type of target is saved in the offline acquisition module.

[0019] In some of the embodiments, the first motion data of each type of target is position information of each type of target.

[0020] In some of the embodiments, measuring data is traversed to obtain measurement points associated with the tracking target, including:

[0021] Conduct measurement traversal on the measurement data;

[0022] Based on the measurement traversal, filter the measurement points to be processed that are associated with multiple tracking targets;

[0023] Count the association probabilities between the measurement points to be processed and multiple tracking targets;

[0024] The tracking target with the highest association probability is associated with the measurement point to be processed to obtain the measurement point associated with the tracking target.

[0025] In some embodiments, determining the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate according to the innovation vector includes:

[0026] The target measurement point quantity includes a first value and a second value, and the measurement point distribution probability includes a first distribution probability and a second distribution probability; wherein,

[0027] The probability data interconnection method is used to determine whether the new information vector is less than the preset first threshold.

[0028] If yes, the measurement point corresponding to the innovation vector is counted into the number of measurement points associated with the tracking target;

[0029] After counting all the new information vectors, the first value is obtained;

[0030] Obtain the first distribution probability according to the first value and the chi-square distribution probability density table;

[0031] The probability data interconnection method is used to determine whether the new information vector is less than the preset second threshold.

[0032] If yes, the measurement point corresponding to the innovation vector is counted into the number of measurement points associated with the tracking target;

[0033] After counting all the new information vectors, we get the second value;

[0034] The second distribution probability is obtained based on the second value and the chi-square distribution probability density table.

[0035] In some of the embodiments, judging whether the measurement acoustic covariance matrix of the tracking target needs to be updated according to the number of target measurement points and the distribution probability of the measurement points includes:

[0036] Use the following formula to determine whether the measurement covariance matrix of the tracking target needs to be updated:

[0037] ;

[0038] On the basis that the calculated result is less than the probability tolerance, the measurement covariance matrix of the tracked target is updated.

[0039] In some embodiments, a target covariance matrix of the tracked target is obtained according to the first covariance matrix of the innovation vector and the second covariance matrix of each type of target in the offline acquisition module, including:

[0040] Calculate the difference between the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module to obtain a covariance matrix difference set;

[0041] The covariance matrix of the target type corresponding to the smallest covariance matrix difference in the covariance matrix difference set is used as the target covariance matrix of the tracking target.

[0042] In some of the embodiments, measurement data is traversed to obtain measurement points without associated targets;

[0043] The measurement points of unrelated targets are clustered to generate new initial tracking targets.

[0044] In a second aspect, a target tracking system is provided in this embodiment, and the target tracking method described in the first aspect is applied. The target tracking system includes: a target tracking module and an offline acquisition module, wherein:

[0045] The target tracking module includes: target prediction module, target measurement matching module, each dimension matching distance statistics storage module, target matrix type update module, target state update module and new target generation module, among which,

[0046] The target prediction module is used to determine the predicted state value and predicted covariance of the tracked target according to the current state of the tracked target and the state transfer matrix;

[0047] The target measurement matching module is used to traverse all measurements and obtain the associated points of the tracking target;

[0048] The module for storing statistics of matching distances in each dimension is used to record the new information vectors that meet the preset requirements and determine whether the measurement covariance matrix of the tracked target needs to be updated;

[0049] The target matrix type update module is used to update the measurement covariance matrix of the tracked target according to the detection results of the matching distance statistics storage module in each dimension;

[0050] A target state updating module is used to update the tracking target state according to the update result of the target matrix type updating module;

[0051] A new target generation module is used to cluster the measurement points of unrelated targets and generate new starting tracking targets;

[0052] The offline acquisition module is used to collect the covariance matrix of each type of target to update the measurement covariance matrix of the tracked target in the target state update module.

[0053] According to a third aspect, a storage medium is provided in this embodiment, on which a computer program is stored, and when the program is executed by a processor, the target tracking method described in the first aspect is implemented.

[0054] Compared with the related art, the target tracking method provided in this embodiment determines the predicted state value of the tracked target and updates the predicted covariance matrix; performs measurement traversal on the measurement data to obtain the measurement points associated with the tracked target; calculates the difference between the measurement points associated with the tracked target and the observed conversion value of the current predicted state value of the tracked target to obtain the new information vector; determines the number of target measurement points and the distribution probability of the measurement points that fall into the tracked target wave gate according to the new information vector; determines whether the measurement covariance matrix of the tracked target needs to be updated according to the number of target measurement points and the distribution probability of the measurement points; if so, obtains the target covariance matrix of the tracked target according to the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module; wherein the second covariance matrix is ​​obtained by pre-collecting the data statistics of each type of target; and updates the measurement covariance matrix of the tracked target through the target covariance matrix. Wherein, the measurement covariance matrix in the tracked target is updated in a timely manner through the offline acquisition module, which improves the tracking accuracy of the target tracking algorithm, reduces data storage, and reduces the calculation amount of the algorithm.

[0055] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0057] Figure 1 It is a hardware structure block diagram of the terminal of the target tracking method of this embodiment.

[0058] Figure 2 is a flow chart of the target tracking method of this embodiment.

[0059] Figure 3 It is a flow chart of obtaining the covariance matrix of various types of targets in the offline acquisition module of this embodiment.

[0060] Figure 4 Schematic diagram of the target tracking system structure of this embodiment. DETAILED DESCRIPTION

[0061] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0062] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by people with general skills in the technical field to which this application belongs. The words "one", "a", "the", "these" and the like in this application do not indicate a quantitative limitation, and they may be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in this application are intended to cover non-exclusive inclusions; for example, a process, method and system, product or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone. Generally, the character " / " indicates that the objects associated with each other are in an "or" relationship. The terms "first", "second", "third", etc. in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0063] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 FIG. 1 is a hardware structure diagram of a terminal of the target tracking method of this embodiment. Figure 1 As shown, the terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 and a memory 104 for storing data, wherein the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0064] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the target tracking method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0065] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by a communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0066] In this embodiment, a target tracking method is provided. Figure 2 is a flow chart of the target tracking method of this embodiment. Figure 2 As shown, the process includes the following steps:

[0067] Step S201, determining the predicted state value of the tracking target and updating the predicted covariance matrix.

[0068] Specifically, the radar data processor and radar signal are two important components in modern radar systems. The radar receives the signal and first processes it in the signal processor to suppress clutter, interference signals and detect target signals. Then it needs to be processed in the data processor to extract the target coordinate information to the maximum extent possible, so as to estimate the motion trajectory of the target in the control area and give its position shift at the next moment, so as to achieve the purpose of high-precision real-time tracking of the target. This implementation mainly tracks and optimizes the targets of millimeter-wave radars. In the process of millimeter-wave radar target tracking, Kalman filtering is used to predict the state and predict the covariance. Among them, the predicted state of the tracking target is based on the current state of the target and the state transition model to predict the state at the next moment, mainly including the state transfer matrix Φ, process noise ω and predicted state quantity x(k+1|k). Among them, the state transfer matrix describes the evolution of the system state over time. For example, in a uniform speed model, the state transfer matrix will consider position and velocity, while in more complex models, it may include acceleration or other dynamic characteristics; process noise simulates random dynamics that cannot be captured by the system model, and process noise is usually assumed to be Gaussian white noise; the predicted state quantity is the state at the next moment predicted from the current state x(k) through the state equation. The state of the tracking target is predicted by the following formula:

[0069] (1)

[0070] Where G is the system noise coefficient matrix, Φ is the state transfer matrix, and ω is the process noise.

[0071] After predicting the state, update the prediction covariance matrix to reflect the uncertainty of the predicted state. The prediction covariance matrix formula is as follows:

[0072] (2)

[0073] Where Φ is the state transfer matrix, T is the transposed sign, and Q is the variance of the process noise ω.

[0074] By predicting the state value of the tracked target and updating the predicted covariance matrix, the uncertainty of the target state can be updated to achieve effective target tracking and state estimation.

[0075] Step S202: perform measurement traversal on the measurement data to obtain measurement points associated with the tracking target.

[0076] When the target is detected, the detected target signal is sent to the data recorder to record the spatial position, amplitude value, radial velocity and other target characteristic parameters of the target. The data recorder is generally implemented by a computer. The point track data output by the data recorder is called measurement data. Measurement refers to the noise-contaminated observation value related to the target state. It is not the original data point of the radar, but the point track output by the data recorder after signal processing. The point track can be divided into free point track and related point track according to whether it has been interconnected with the established target track. Among them, the point track related to the known target is called related point track, and the point track that is not interconnected with the established target track is called free point track. The measurement can be obtained by correct measurement of the target by other sensors or by fitting the time dimension based on the point set of the target trajectory. It may also be an erroneous measurement from false targets, clutter, and interference targets.

[0077] The measurement data output by the data recorder is traversed, and the measurement points are associated with the tracked target through a gate. The gate is divided into an initial gate and a related gate. The initial gate is a region with the free point trace as the center, which is used to determine the possible range of the observation value of the target. In the initial stage of the track, in order to better capture the target, the initial gate is generally slightly larger. The related gate refers to a region with the predicted position of the tracked target as the center, which is used to determine the possible range of the observation value of the target. The gate size is related to the radar measurement error size, the probability of correctly receiving the echo, etc. It is necessary to make the real measurement fall into the gate with a high probability, and at the same time, the number of irrelevant points in the related gate is not too large. In this embodiment, the gate size is set according to the actual situation. After all the measurement data are traversed, the measurement points falling into the set gate are associated with the tracked target to obtain the measurement points associated with the tracked target.

[0078] Step S203, calculating the difference between the observed conversion value of the measurement point associated with the tracking target and the current predicted state value of the tracking target to obtain an innovation vector.

[0079] Specifically, the new information vector represents the difference between the measurement point associated with the tracking target and the current predicted state value of the tracking target. This difference provides new information about the current state of the tracking target. First, the current predicted state value of the tracking target is converted into the current observation value. In the Kalman filter, the observation model formula is as follows:

[0080] (3)

[0081] Among them, z k is the current observation value, H is the observation matrix, x k is the predicted state value, v k For observation noise, it is usually assumed to be Gaussian noise with zero mean.

[0082] The state transition model is used to predict the state of the current observation value at the next moment, and then the predicted state is mapped to the observation space through the observation matrix H to obtain the predicted observation value. The specific formula is as follows:

[0083] (4)

[0084] in, is the predicted state at time K-1, is the predicted observation value corresponding to time k.

[0085] The difference between the current observation value and the predicted observation value is calculated to obtain the new information vector. The specific formula is as follows:

[0086] (5)

[0087] Step S204, determining the number of target measurement points falling into the tracking target gate and the distribution probability of the measurement points according to the innovation vector.

[0088] The innovation vector is calculated by observing the difference of the transformed values, and the covariance matrix of the innovation vector is calculated by the following formula:

[0089] (6)

[0090] Among them, V k is the innovation vector, is the transpose of the new information vector, H is the observation matrix, and H ’ is the transpose of the observation matrix, R is the measurement noise covariance matrix, which represents the degree of inaccuracy of the actual sensor observation, and contains the covariance relationship between multiple observations, P is the target state covariance matrix, and E is the expectation. Among them, under the probabilistic data interconnection (PDA) method, the determination of whether a measurement belongs to a certain target and the probability of belonging to this target is affected by the target state covariance matrix P and the measurement noise covariance matrix R. The smaller P is, the more accurate the system's estimation of the target state is, and the smaller R is, the more accurate the sensor's detection of the target is.

[0091] The result calculated by the following formula is compared with the preset value to determine the number of target measurement points and the distribution probability of the measurement points that fall into the tracking target wave gate:

[0092] (7)

[0093] pass Measure whether the measurement points meet the preset rules, where the vector V contains state values ​​of different dimensions. It is V squared and then divided by the new information vector covariance matrix S k, in order to eliminate the influence of inconsistent variances of state values ​​in different dimensions.

[0094] Since V contains two independent Gaussian distribution variables, That is, X with 2 degrees of freedom 2 Distribution variables. In general, PDA algorithms The size of the alternative measurement value is judged, when the difference between the alternative measurement value and the target prediction value observation conversion value When the value is less than a certain threshold value γ, it is considered that the measurement falls into the wave gate of the target. According to this method, the number of target points that fall into the corrected target wave gate among all measurement points is counted, and the value is calculated according to X 2 The distribution probability density table determines the distribution probability of the measurement points.

[0095] Step S205, judging whether the measurement covariance matrix of the tracked target needs to be updated through the number of target measurement points and the distribution probability of the measurement points, and if so, obtaining the target covariance matrix of the tracked target based on the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module; wherein the second covariance matrix is ​​obtained by pre-collecting data statistics of each type of target; and updating the measurement covariance matrix of the tracked target through the target covariance matrix.

[0096] Among them, when counting the number of target measurement points and the distribution probability of measurement points, two thresholds are set and To count the number of target measurement points within the two threshold ranges ( and ) and the distribution probability of the measurement points ( and ), first detect whether there is a tracking target Reaching the preset threshold N, if yes, then entering the R measurement noise covariance matrix in the measurement covariance matrix to determine whether to update, the judgment is made by the following formula:

[0097] (8)

[0098] in, Represents the probability tolerance. When the above formula conditions are not met, the R measurement noise covariance matrix corresponding to the tracked target is updated.

[0099] Before updating, first collect various types of target data and count the distribution of each dimension through the offline acquisition module. The target types include but are not limited to people, non-motor vehicles, medium and large and small motor vehicles, etc. The target motion data is generally location information. Perform covariance matrix statistics on the data of various types of targets in advance, and save the statistical results in the offline acquisition module. The specific covariance matrix set is:

[0100] (9)

[0101] The covariance matrix S of the target's innovation vector is tracked k The covariance matrix Cov of each type of target in the offline acquisition module is calculated as follows:

[0102] (10)

[0103] The calculation results are sorted, and the minimum value is used as the target covariance matrix of the tracking target. The measurement covariance matrix of the tracking target is updated using the R measurement noise covariance matrix corresponding to the target covariance matrix of the tracking target.

[0104] Through the above steps S201 to S205, the predicted state value of the tracked target is determined and the predicted covariance matrix is ​​updated; the measurement data is measured and traversed to obtain the measurement points associated with the tracked target; the difference between the measurement points associated with the tracked target and the observed conversion value of the current predicted state value of the tracked target is calculated to obtain the new information vector; according to the new information vector, the number of target measurement points and the distribution probability of the measurement points falling into the tracked target wave gate are determined; through the number of target measurement points and the distribution probability of the measurement points, it is judged whether the measurement covariance matrix of the tracked target needs to be updated; if so, the target covariance matrix of the tracked target is obtained according to the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module; wherein the second covariance matrix is ​​obtained by pre-collecting data statistics of each type of target; the measurement covariance matrix of the tracked target is updated through the target covariance matrix. Compared with the prior art of obtaining target tracking results by combining Kalman filtering and range Doppler mapping, and constructing a tree-like feature trace by adopting a multi-hypothesis method through spatiotemporal spectral features to achieve target tracking, this embodiment first collects various types of target data and statistics the distribution of various dimensions through an offline acquisition module to obtain the covariance matrix of each type of target. After tracking and predicting the tracking target, when it is found that the measurement covariance matrix of the tracking target needs to be updated, it is directly matched with the measurement covariance matrix detected in advance in the offline acquisition module for update, thereby improving the accuracy of the target tracking algorithm. When the tracking target matrix type is updated, it is updated through the offline acquisition module, without the need to store a large amount of data, reducing the amount of calculation, and improving the calculation efficiency, thereby further improving the real-time performance of target tracking.

[0105] In some of the embodiments, the second covariance matrix is ​​obtained by collecting statistics of various types of target data in advance. Figure 3 is a flow chart of obtaining the covariance matrix of each type of target in the offline acquisition module of this embodiment, such as Figure 3 As shown, the process includes the following steps:

[0106] Step S301, collecting first motion data of each type of target through radar, and obtaining a set of associated points of the first motion data; the first motion data of each type of target is the position information of each type of target.

[0107] Specifically, in the common application scenarios of millimeter wave radar, the motion data of various types of targets detected by the radar are collected, wherein the target types include but are not limited to people, non-motor vehicles, medium, large and small motor vehicles, etc. The first motion data can be the position information, amplitude value, radial velocity and other characteristic parameters of each type of target. In this embodiment, the first motion data adopts the position information X of each type of target. t and Y t , represents the x and y coordinate positions of the target at the corresponding time t. For example, a target moves during the time 0~T, and the radar detects it at a frame rate F, with a total number of detection frames of F×T. The associated point set A of the first motion data is obtained by the same method as in step S202 above:

[0108] (11)

[0109] Step S302, second motion data of each type of target obtained by fitting the associated point set of the first motion data;

[0110] Specifically, the second motion data is obtained by fitting the time dimension through other sensors or using the point set associated with the point set A, and then the point set B of the second motion data is obtained:

[0111] (12)

[0112] Among them, it is necessary to ensure that at any time t, there is only one pair (X ref , Y ref ) corresponds to the points in the associated point set A.

[0113] Step S303, calculating the deviation between the second motion data and the first motion data to obtain a deviation value matrix of each type of target;

[0114] Specifically, after obtaining two point sets, the ability of the radar to detect the x, y coordinate position of this type of target is evaluated. t , Y t ), find the reference point (X ref , Y ref ), calculate the deviation:

[0115] (13)

[0116] Calculate the deviation operation between all points in A and the corresponding points in B to obtain the deviation value matrix:

[0117] (14)

[0118] Where M is the number of elements in the associated point set A, P diff is an M×2 matrix.

[0119] Based on the above steps S301 to S303, the same operation is performed on each type of target to obtain the deviation value matrix of each type of target.

[0120] Step S304, calculating the mean of the deviation value matrix of each type of target, calculating the second covariance matrix of each type of target according to the mean of the deviation value matrix of each type of target, and saving the second covariance matrix of each type of target to the offline acquisition module.

[0121] Specifically, the mean value of the deviation value matrix of each type of target is calculated respectively, and the calculation formula is as follows:

[0122] (15)

[0123] The covariance matrix of each type of target, i.e., the second covariance matrix, is calculated based on the mean of the deviation value matrix of each type of target. The calculation formula is as follows:

[0124] (16)

[0125] Get the covariance matrix ;in, is the variance of ΔX, is the variance of ΔY, is the covariance between ΔX and ΔY.

[0126] The covariance matrices of all types of targets are combined to obtain the covariance matrix set of each type of target in the offline acquisition module:

[0127] (17)

[0128] The covariance matrix of each type of target is saved in the offline acquisition module, which is used to timely update the measurement covariance matrix of different types of targets during target tracking, so as to improve the accuracy of target tracking, reduce the storage of data during tracking, reduce the amount of data calculation, and improve the real-time performance of target tracking.

[0129] In another embodiment, performing measurement traversal on the measurement data to obtain measurement points associated with the tracking target includes:

[0130] The measurement data is traversed; based on the measurement traversal, the measurement points to be processed that are associated with multiple tracking targets are screened; the association probabilities between the measurement points to be processed and the multiple tracking targets are counted; the tracking target with the highest association probability is associated with the measurement points to be processed to obtain the measurement points associated with the tracking targets.

[0131] Specifically, the measurement nurse section is traversed, and after the traversal, the target measurement matching is performed to correctly assign the observed measurement points to the corresponding targets. In the target measurement matching process, it is first necessary to define a wave gate for each tracking target. The wave gate is an area based on the target prediction state and uncertainty. If the measurement point falls within this wave gate, the measurement point is identified as a candidate measurement point associated with the tracking target. For each tracking target, the algorithm traverses all the measurement points synchronously acquired at the current time and checks whether they fall within the wave gate of the corresponding target. Once it is determined which measurement points fall within the wave gate of the target, these measurement points are marked as associated with the target. This process is repeated for all targets to ensure that all measurement points have the opportunity to be correctly associated. In actual operation, a measurement point may fall into the wave gates of multiple targets at the same time, that is, the measurement point is associated with multiple targets. In this case, further probability calculation is required to determine the target to which the measurement point is most likely to belong. For each measurement point that falls into multiple target wave gates, the algorithm calculates the probability that the measurement point comes from each target. This usually involves the application of Bayesian theory, which takes into account factors such as the probability of target detection, the probability of the measurement point falling into the wave gate, and the probability of clutter. Based on the above probability calculations, the measurement point will be assigned to the target with the highest association probability. This means that the measurement point is more likely to have originated from this target and should be used to update the state estimate of this target.

[0132] In some embodiments, determining the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate according to the innovation vector includes:

[0133] The target measurement point quantity includes a first value and a second value, and the measurement point distribution probability includes a first distribution probability and a second distribution probability; wherein,

[0134] By using the probability data interconnection method, it is determined whether the new information vector is less than a preset first threshold value. If so, the measurement point corresponding to the new information vector is counted into the number of measurement points associated with the tracking target; after counting all the new information vectors, a first value is obtained; and a first distribution probability is obtained according to the first value and the chi-square distribution probability density table;

[0135] The probability data interconnection method is used to determine whether the new information vector is less than the preset second threshold. If so, the measurement point corresponding to the new information vector is included in the number of measurement points associated with the tracking target; after counting all the new information vectors, a second value is obtained; and the second distribution probability is obtained based on the second value and the chi-square distribution probability density table.

[0136] Specifically, when counting the number of target measurement points and the distribution probability of measurement points, two thresholds are set: and To count the number of target measurement points within the two threshold ranges ( and ) and the distribution probability of the measurement points ( and ). Since V in the above formula (7) contains two independent Gaussian distribution variables, That is, X with 2 degrees of freedom 2 Distribution variables. In general, PDA algorithms The size of the alternative measurement value is judged, when the difference between the alternative measurement value and the target prediction value observation conversion value When it is less than a certain threshold γ, the measurement is considered to fall into the wave gate of the target.

[0137] By querying X 2 The distribution probability density table shows that when γ=9, the probability of the two-dimensional measurement point falling into the wave gate is 98.9%; when γ=4, the probability of the two-dimensional measurement point falling into the wave gate is 86.5%. 2 The probability distribution characteristics of the distribution are used to track the target and the corresponding associated measurement points. In the process of tracking the target associated measurement point in step S202, the threshold γ is set to the first threshold ( can be around 9 or larger), in the probabilistic data association method (PDA), the tracking target is associated with the corresponding measurement point With the first threshold Compare and select the ones that are smaller than the first threshold of The corresponding measurement points are counted in the number of measurement points associated with the tracking target. , get all the values ​​that are less than the first threshold Count the number of measurement points and get the first value . Based on the first value With X 2 The distribution probability density table is matched to obtain the corresponding first distribution probability By setting a larger threshold γ, it can be ensured that most of the valid measurements fall within the target gate.

[0138] Then set a second threshold , Compare Smaller (e.g. =4), a smaller threshold (such as 4) is set to count the ratio of the number of measurement points in the wavelet gate to the number of all matched measurement points. This ratio provides a metric to evaluate whether the current gate size is suitable for the current target. If the ratio of the number of measurement points in the wavelet gate to the number of matched measurement points exceeds a certain range of probability corresponding to the threshold, this indicates that the current gate size may be too large or too small and needs to be adjusted. For example, if the ratio of the number of measurement points in the wavelet gate is much higher than 86.5%, this may mean that the gate is too small, resulting in many measurement points being mistakenly excluded; conversely, if the ratio is much lower than 86.5%, the gate may be too large and needs to be reduced to improve tracking accuracy.

[0139] In the probabilistic data association method (PDA), the tracking target is associated with the corresponding measurement point. With the second threshold Compare and select those that are smaller than the second threshold of The corresponding measurement points are counted in the number of measurement points associated with the tracking target. , get all the values ​​that are less than the second threshold Count the number of measurement points and get the second value . According to the second value With X 2 The distribution probability density table is matched to obtain the corresponding second distribution probability .

[0140] In another embodiment, judging whether the measurement acoustic covariance matrix of the tracking target needs to be updated based on the number of target measurement points and the distribution probability of the measurement points includes:

[0141] Use the following formula to determine whether the measurement covariance matrix of the tracking target needs to be updated:

[0142] ;

[0143] On the basis that the calculated result is less than the probability tolerance, the measurement covariance matrix of the tracked target is updated.

[0144] Specifically, traverse all tracking targets, when the second value of a tracking target After reaching a certain threshold, the R measurement covariance matrix of the tracked target is updated and judged. The specific judgment formula is as follows:

[0145] (18)

[0146] in, Represents probability tolerance. Probability tolerance represents the tolerance between the statistically obtained gate landing probability ratio and the theoretical gate landing probability of the measurement noise covariance in use. If the value of is large enough, it means that there are more statistical samples, so that the statistical ratio is closer to the true value. In this case, the probability tolerance can be selected at a lower point, otherwise a higher point should be selected. The specific value can be selected based on experience or adjusted according to actual test conditions, and this embodiment does not make specific limitations on this.

[0147] If the calculation result satisfies the above inequality, the measurement covariance matrix of the tracked target does not need to be updated in this evaluation; if the calculation result does not satisfy the above inequality, the measurement covariance matrix of the tracked target needs to be updated. And by traversing the covariance matrices of various types of targets in the offline acquisition module, find the covariance matrix for the domestic target to update.

[0148] In some other embodiments, a target covariance matrix of the tracked target is obtained according to the first covariance matrix of the innovation vector and the second covariance matrix of each type of target in the offline acquisition module, including:

[0149] Calculate the difference between the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module to obtain a covariance matrix difference set;

[0150] The covariance matrix of the target type corresponding to the smallest covariance matrix difference in the covariance matrix difference set is used as the target covariance matrix of the tracking target.

[0151] Specifically, when the measurement covariance matrix of the tracking target needs to be updated, the covariance matrix difference set is calculated by formula (10), the difference values ​​in the covariance matrix set are sorted, and the covariance matrix of the target type corresponding to the smallest covariance matrix difference is selected as the target covariance matrix of the tracking target. The measurement covariance matrix of the tracking target is updated with the target covariance matrix, and , Clear to zero and restart a new round of R matrix update evaluation for the tracking target.

[0152] In another embodiment, the measurement data is traversed to obtain measurement points without associated targets;

[0153] The measurement points of unrelated targets are clustered to generate new initial tracking targets.

[0154] Specifically, after traversing all the measurement data, there may be some measurement points that are not assigned to any known target. These measurement points are called unrelated measurement points. These measurement points may be because they do not match the predicted position of the existing target, or because they are newly appeared targets. In order to process these unrelated measurement points, a clustering algorithm can be used to group them. Among them, clustering algorithms include K-means, hierarchical clustering, DBSCAN, etc., which can be selected according to actual conditions. This embodiment does not make specific limitations on this. Through the clustering operation, unrelated measurement points that are close in space or similar in features can be clustered together to form new clusters. Each cluster can be regarded as a potential new target, thereby generating a new starting target. These new targets can then be further processed by the tracking algorithm, such as state estimation and prediction through Kalman filtering. Thereby improving the adaptability of the target tracking system, enabling it to handle newly appeared targets and enhancing the robustness of the system to complex scenes.

[0155] In this embodiment, a target tracking system is also provided, which applies the target tracking method described in any one of the above embodiments. Figure 4 Schematic diagram of the target tracking system structure of this embodiment. Figure 4 As shown, the target tracking system 40 includes: a target tracking module 41 and an offline acquisition module 42, wherein:

[0156] The target tracking module 41 includes: a target prediction module 411, a target measurement matching module 412, a dimension matching distance statistics storage module 413, a target matrix type update module 414, a target state update module 415 and a new target generation module 416, wherein:

[0157] The target prediction module 411 is used to determine the predicted state value and predicted covariance of the tracked target according to the current state of the tracked target and the state transfer matrix;

[0158] The target measurement matching module 412 is used to traverse all measurements and obtain the associated points of the tracking target;

[0159] Each dimension matching distance statistics storage module 413 is used to record the new information vector that meets the preset requirements and determine whether the measurement covariance matrix of the tracking target needs to be updated;

[0160] The target matrix type updating module 414 is used to update the measurement covariance matrix of the tracked target according to the detection results of the matching distance statistics storage module 413 in each dimension;

[0161] A target state updating module 415 is used to update the tracking target state according to the update result of the target matrix type updating module 414;

[0162] A new target generation module 416 is used to cluster the measurement points of unrelated targets to generate a new starting tracking target;

[0163] The offline acquisition module 42 is used to collect the covariance matrix of each type of target to update the measurement covariance matrix of the tracked target in the target state update module 414.

[0164] The system is used to implement the above-mentioned embodiments and preferred implementations, and those that have been described will not be repeated. The terms "module", "unit", "subunit", etc. used below can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0165] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0166] In this embodiment, an electronic device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0167] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0168] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0169] S1, determine the predicted state value of the tracking target and update the predicted covariance matrix;

[0170] S2, traverse the measurement data to obtain the measurement points associated with the tracking target;

[0171] S3, calculating the difference between the observed conversion value of the measurement point associated with the tracking target and the current predicted state value of the tracking target to obtain the new information vector;

[0172] S4, determining the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate according to the new information vector;

[0173] S5, judging whether the measurement covariance matrix of the tracked target needs to be updated based on the number of target measurement points and the distribution probability of the measurement points; if so, obtaining the target covariance matrix of the tracked target based on the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module; wherein the second covariance matrix is ​​obtained by pre-collecting data statistics of each type of target;

[0174] S6, updating the measurement covariance matrix of the tracked target through the target covariance matrix.

[0175] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and will not be repeated in this embodiment.

[0176] In addition, in combination with the target tracking method provided in the above embodiments, a storage medium may be provided in this embodiment to implement the target tracking method. The storage medium stores a computer program, and when the computer program is executed by a processor, any target tracking method in the above embodiments is implemented.

[0177] It should be understood that the specific embodiments described herein are only used to explain the application, rather than to limit it. Based on the embodiments provided in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of this application.

[0178] Obviously, the drawings are only some examples or embodiments of the present application. For ordinary technicians in the field, the present application can also be applied to other similar situations based on these drawings without creative work. In addition, it is understandable that although the work done in this development process may be complicated and lengthy, for ordinary technicians in the field, certain changes in design, manufacturing or production based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient content disclosed in this application.

[0179] The term "embodiment" in this application refers to a specific feature, structure or characteristic described in conjunction with the embodiment that can be included in at least one embodiment of the present application. The appearance of this phrase in various locations in the specification does not necessarily mean the same embodiment, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. It is clearly or implicitly understood by those of ordinary skill in the art that the embodiments described in this application can be combined with other embodiments without conflict.

[0180] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0181] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of patent protection. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the scope of protection of the present application. Therefore, the scope of protection of the present application shall be subject to the attached claims.

Claims

1. A target tracking method, characterized in that: include: Determine the predicted state value of the tracking target and update the predicted covariance matrix; Performing measurement traversal on the measurement data to obtain measurement points associated with the tracking target; Calculate the difference between the observed conversion value of the measurement point associated with the tracking target and the current predicted state value of the tracking target to obtain an innovation vector; Determine the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate according to the new information vector; Determining whether the measurement covariance matrix of the tracking target needs to be updated based on the number of target measurement points and the distribution probability of the measurement points; If yes, then according to the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module, the target covariance matrix of the tracked target is obtained; wherein the second covariance matrix is ​​obtained by pre-collecting the data statistics of each type of target; The measurement covariance matrix of the tracking target is updated by using the target covariance matrix.

2. The target tracking method according to claim 1, characterized in that: The second covariance matrix is ​​obtained by collecting statistics of the target data of various types in advance, and includes: Collecting first motion data of various types of targets through radar, and obtaining a set of associated points of the first motion data; second motion data of each type of target obtained by fitting the associated point set of the first motion data; Calculating the deviation between the second motion data and the first motion data to obtain a deviation value matrix for each type of target; Calculate the mean of the deviation value matrix of each type of target, calculate the second covariance matrix of each type of target according to the mean of the deviation value matrix of each type of target, and save the second covariance matrix of each type of target in the offline acquisition module.

3. The target tracking method according to claim 2, characterized in that: The first motion data of each type of target is the position information of each type of target.

4. The target tracking method according to claim 1, characterized in that: The step of measuring and traversing the measurement data to obtain the measurement points associated with the tracking target includes: Performing measurement traversal on the measurement data; Based on the measurement traversal, screening measurement points to be processed that are associated with multiple tracking targets; Counting the association probabilities between the measurement point to be processed and the multiple tracking targets; The tracking target with the highest association probability is associated with the measurement point to be processed to obtain the measurement point associated with the tracking target.

5. The target tracking method according to claim 1, characterized in that: The step of determining the number of target measurement points and the distribution probability of the measurement points falling into the tracking target wave gate according to the innovation vector includes: The target measurement point quantity includes a first value and a second value, and the measurement point distribution probability includes a first distribution probability and a second distribution probability; wherein, Determine whether the new information vector is less than a preset first threshold value by using a probabilistic data interconnection method, If yes, the measurement point corresponding to the innovation vector is counted into the number of measurement points associated with the tracking target; After counting all the new information vectors, the first value is obtained; Obtaining the first distribution probability according to the first value and a chi-square distribution probability density table; Determine whether the new information vector is less than a preset second threshold value by using a probabilistic data interconnection method, If yes, the measurement point corresponding to the innovation vector is counted into the number of measurement points associated with the tracking target; After counting all the new information vectors, the second value is obtained; The second distribution probability is obtained according to the second value and a chi-square distribution probability density table.

6. The target tracking method according to claim 5, characterized in that: Judging whether the measurement sound covariance matrix of the tracking target needs to be updated according to the number of target measurement points and the distribution probability of the measurement points includes: The following formula is used to determine whether the measurement covariance matrix of the tracking target needs to be updated: ； On the basis that the calculated result is less than the probability tolerance, the measurement covariance matrix of the tracked target is updated.

7. The target tracking method according to claim 1, characterized in that: According to the first covariance matrix of the innovation vector and the second covariance matrix of each type of target in the offline acquisition module, a target covariance matrix of the tracked target is obtained, including: Calculate the difference between the first covariance matrix of the new information vector and the second covariance matrix of each type of target in the offline acquisition module to obtain a covariance matrix difference set; The covariance matrix of the target type corresponding to the smallest covariance matrix difference in the covariance matrix difference set is used as the target covariance matrix of the tracked target.

8. The target tracking method according to claim 1, characterized in that: The method further comprises: Conduct measurement traversal on the measurement data to obtain measurement points without related targets; The measurement points of the unrelated targets are clustered to generate new initial tracking targets.

9. A target tracking system, using the target tracking method according to any one of claims 1 to 8, characterized in that: The target tracking system includes: a target tracking module and an offline acquisition module, wherein: The target tracking module includes: a target prediction module, a target measurement matching module, a matching distance statistics storage module for each dimension, a target matrix type update module, a target state update module and a new target generation module, wherein: The target prediction module is used to determine the predicted state value and predicted covariance of the tracked target according to the current state of the tracked target and the state transfer matrix; The target measurement matching module is used to traverse all measurements to obtain the associated points of the tracking target; The module for storing statistics of matching distances in each dimension is used to record the new information vectors that meet the preset requirements and determine whether the measurement covariance matrix of the tracking target needs to be updated; The target matrix type updating module is used to update the measurement covariance matrix of the tracking target according to the detection results of the matching distance statistics storage module in each dimension; The target state updating module is used to update the tracking target state according to the update result of the target matrix type updating module; The new target generation module is used to cluster the measurement points of unrelated targets to generate new starting tracking targets; The offline acquisition module is used to collect the covariance matrix of each type of target to update the measurement covariance matrix of the tracked target in the target state update module.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the target tracking method according to any one of claims 1 to 8 are implemented.

Images