CN116520311A - A Method of Adaptive Track Initiation Based on GLMB - Google Patents

Abstract

The invention belongs to the technical field of multi-target tracking, and specifically relates to a GLMB-based adaptive track initiation method. The invention traverses all possible tracks according to the measurement data at two consecutive moments, and passes the speed screening rules and Doppel The Le information screening rule preliminarily screens the measurement data to remove most of the clutter, and calculates the target measurement probability according to the posterior probability to distinguish the measurement source. The target measurement probability can be used to distinguish whether the measurement originates from a newborn target or a living target. The speed information implicit in the Doppler measurement calculates the initial operating state of the new target and enters the next step of the prediction process to complete the track initiation function.

Description

An adaptive track initiation method based on GLMB

Technical Field

The invention belongs to the technical field of multi-target tracking, and in particular relates to an adaptive track initiation method based on GLMB (Generalized Labeled Multi-Bernoulli).

Background Art

After decades of development, multi-target tracking (MTT) technology has formed two sets of algorithm systems. One is the traditional MTT algorithm, which decomposes the multi-target tracking problem into multiple single-target tracking problems through the data association (DA) algorithm. Due to the combinatorial explosion characteristic of the DA algorithm, as the number of targets increases, the matching combinations required to be enumerated will increase exponentially, making the algorithm lose real-time performance; the other is a multi-target tracking algorithm based on random finite modeling. By modeling multi-target states and multi-target measurements as random finite sets (RFS), it can naturally describe the mechanisms of track start and stop, and can completely avoid the measurement-track association process. Due to the systematic and scientific nature of the RFS theory, it has attracted widespread attention and research from scholars.

In the process of multi-target tracking, measurement is the only source of information, but in the MTT algorithm based on RFS, the density of new targets is generally given by prior information. Usually, new targets are fixed at several positions, and the transmission parameters of new targets are given by prior information. In real detection scenarios, the prior information required for new targets is unknown, and they may appear at any position in the detection area. The standard GLMB filter will not be able to correctly start the track in this situation.

In view of the above problems, the present invention proposes a track adaptive initiation algorithm suitable for GLMB filter, which can complete the track initiation process only based on measurement information when the position of the new target is uncertain.

Summary of the invention

In order to overcome the problems in the prior art, the present invention proposes an adaptive track initiation method based on GLMB.

The technical solution of the present invention to solve the above technical problems is as follows:

The present invention provides an adaptive track initiation method based on GLMB, comprising the following steps:

Convert the polar coordinate system measurement data received by the radar into the rectangular coordinate system measurement data;

According to the prior information, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density;

The clutter in the measurement data is filtered out by using the velocity screening rule and the Doppler information screening rule;

Adopting sequential filtering method to update multi-target posterior probability density;

The probability of target measurement is calculated through the posterior probability density in the updating process, and the surviving targets and new targets are distinguished through the probability of target measurement, and the measurements related to the new targets are retained for the new track at the next moment.

Furthermore, according to the prior information, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density, which specifically includes the following steps:

According to the new component calculated at the previous moment, the single target density of the new target is calculated;

According to the posterior information transmitted at the previous moment, the single target density of the surviving target is calculated;

The single target density of the new target and the single target density of the surviving target are connected in parallel to obtain the multi-target prediction probability density.

Furthermore, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density, which also includes: and calculating the initial operation state of the new target according to the speed information hidden in the Doppler information.

Furthermore, the multi-objective posterior probability density is:

In the formula, represents the mean, variance, and weight of the missing measurement; represents the mean, variance and weight of the position measurement; θ(l) represents the track association mapping with label l; δ(θ(l)) is the De Kotta function. When θ(l) = 0, δ(θ(l)) = 1, which means that the measurement is not associated with the track, indicating that the track is missed; conversely, θ(l) ≠ 0, δ(θ(l)) = 0, which means that the track and measurement information are updated, indicating that the track is normal.

Furthermore, the probability ρ(z) of the target measurement is:

in, represents the association probability between the surviving target i and the measurement in the measurement set Z _k at time k. Conversely, 1- _pi represents the association probability between the new target and the clutter in the measurement set Z _k at time k. p _D,k is the sensor detection probability. z _k is the measurement point at time k. represents the Gaussian density with mean m and variance P, _Hk is the observation matrix, R is the position measurement noise covariance, and h(·) is defined as follows:

In the above formula, (x, y) represents the target position, and ( _xs , _ys ) represents the sensor position.

Compared with the prior art, the present invention has the following technical effects:

The present invention traverses all possible tracks according to the measurement data of two consecutive moments, and preliminarily screens the measurement data to remove most of the clutter through speed screening rules and Doppler information screening rules, and distinguishes the measurement source by calculating the target measurement probability according to the posterior probability. The target measurement probability can distinguish whether the measurement originates from a new target or a surviving target, and the speed information implicit in the Doppler measurement is used to calculate the initial operating state of the new target and enter the next prediction process to complete the track initiation function.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a flowchart of the adaptive track initiation GLMB algorithm of the present invention;

FIG2 is a diagram showing the actual motion trajectory of the target of the invention;

FIG3 is a schematic diagram of the invented sensor measurement;

FIG4 is a GLMB tracking effect diagram of the invention track adaptive initiation;

FIG5 is a diagram showing the tracking effect of the GLMB invention;

FIG. 6 is a schematic diagram of the OSPA distance of the present invention.

DETAILED DESCRIPTION

In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. The specific features, structures or characteristics in one or more embodiments may be combined in any suitable form. Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by technicians in the technical field of the present invention.

1-6 , the present invention aims at the problems existing in the prior art and proposes an adaptive track initiation method based on GLMB, which can estimate the motion state of the target with only the measurement information of the previous moment, and incorporate it into the prediction process of the filtering algorithm, and remove unreasonable new components in the process of branch pruning, which can effectively reduce the probability of false short tracks and improve the tracker performance.

A GLMB-based adaptive track initiation method specifically comprises the following steps:

Step 1: Convert the radar measurement data into coordinates.

The polar coordinate measurement data of the radar is [ _rz , _θz ], _rz is the measured distance in polar coordinates, and _θz is the azimuth in polar coordinates; the measured distance _rz =r+ _vr , and the azimuth is _θz =θ+ _vθ , where r represents the actual distance between the sensor and the tracked target, _vr represents the distance noise, θ represents the actual azimuth, and _vθ represents the angle noise.

The classic measurement conversion is:

Where _xz represents the x-axis coordinate of the converted rectangular coordinate system; _yz represents the y-axis coordinate of the converted rectangular coordinate system.

Due to the existence of noise, an erroneous estimation result will be given after nonlinear transformation. The precise compensation procedure of these errors depends on the angle noise v _θ . Assuming that the angle noise v _θ can be explained by a symmetric probability density function, since the symmetric probability density function of the angle noise is symmetric, we can get the expectation:

E(sin v _θ )＝0

E stands for expectation, the Sin function is an odd function, and theorem: the expected value of an odd function with a probability density is 0. This formula describes the above theorem.

Through the expectation of the above formula, we can get:

Where ε _θ is the orientation compensation factor. When ε _θ ≠ 1, a bias is expected, and when ε _θ ≠ 0, an unbiased transformation can be given:

Among them, x _m represents the x-axis coordinate in the rectangular coordinate system, and y _m represents the y-axis coordinate in the rectangular coordinate system.

The position component covariance corresponding to the measurement point is as follows:

Rm is the covariance of the position component. Rm is a 2*2 matrix. In order to conveniently represent the calculation process of the four position values, the specific parameters are as follows:

Among them, the compensation factor ε _θ =E(cosv _θ ), ε′ _θ =E(cos2v _θ ).

Step 2. According to the prior information, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density.

Since the value of the new component is calculated from the measurement value at the previous moment, the new component cannot be directly connected in parallel with the surviving component, and additional state prediction of the new component is required.

According to the prior information, the multi-target prediction probability density is obtained:

Where X represents the set of target states, and Δ(X) represents the label difference indicator, which is defined as

That is, when the labels of the elements in the set X are different, Δ(X) = 1; otherwise, Δ(X) = 0. δ(·) represents the Dirac delta function, which is defined as: L(X ₊ ) represents the label projection function, Indicates survival tag The weight of Indicates new label The weight of . and p _{+, B} (·, l) represent the survival target density and the new target density respectively. For a given label set L, p _S represents the survival probability of the target at the next moment.

Among them, the single target density of the surviving target is:

in,

In the above formula, represents a Gaussian density with mean m and variance P, is the Gaussian component weight of the surviving target; F is the state transfer matrix, Q is the covariance of the noise; is the scaling factor.

The single target density of the new target is:

In the above formula, represents a Gaussian density with mean m and variance P, is the Gaussian component weight of the new target. F is the state transfer matrix, Q is the covariance of the noise; Assign existence probabilities to new targets.

The new target needs to use the new target parameters calculated in step 3 at the previous moment to estimate the new target motion state.

Step 3. Filter out the clutter in the measurement data of two consecutive moments by using the speed screening rule and the Doppler information screening rule; and calculate the initial running state of the new target based on the speed information hidden in the Doppler information.

Assume that the target estimated state at time k is m is the target estimated state mean, and the corresponding measurement set at the time is Z _k = {Y _k , D _k }, where the position measurement set Doppler measurement collection represents the Doppler measurement. Since the position measurement set is in polar coordinate form, it needs to be converted into rectangular coordinates, and and are the radial distance and azimuth measured by the radar, and They are the x-axis and y-axis coordinates after coordinate conversion.

Here we take one measurement value from each of the conversion measurement sets at two consecutive moments k-1 and k The velocity information and Doppler information are used to filter the measurements respectively. The filtering rules are as follows:

The speed information screening rules are:

The Doppler information screening rules are:

Where ||·|| ₂ represents the 2-norm of the vector, Δt represents the time interval between moments k-1 and k, v ₁ and v ₂ represent the minimum and maximum speeds of the detected target, respectively; σ _d is the measurement noise of the Doppler information.

The motion speed during the Doppler information screening process can be calculated by the following formula:

After measurement and screening, the state vector and covariance of the new trajectory at time k are calculated through the subsequent algorithm. The calculation process is as follows:

The required mean for the new target and covariance They are:

Among them, the position component of the mean is directly measured as:

The corresponding position component covariance is:

The velocity component of the mean is The corresponding covariance is:

The Doppler measurement equation is:

By observation, the above formula can be restructured into:

in,

In the above formula, n _d,k represents the Doppler observation noise,

Estimation of velocity components using linear minimum mean square error criterion and velocity component covariance You can get:

The velocity component covariance is:

in, is the location of the Doppler radar.

The required weights of the new components at the prediction step are set as:

Step 4: Introduce Doppler measurement and use sequential filtering to update the posterior probability density.

Since Doppler radar has unique Doppler measurement (radial velocity) compared to conventional radar measurement data, a sequential filtering approach is used to introduce Doppler measurement into the δ-GLMB update process.

The specific implementation steps are as follows:

The updated posterior probability density of δ-GLMB is:

Among them, θ(l) represents the track association mapping with label l; δ(θ(l)) is the De Kotta function, which is used to distinguish between missed detection targets and regular detection situations. When θ(l) = 0, δ(θ(l)) = 1, which means that the measurement is not associated with the track, indicating that the track is missed; on the contrary, θ(l) ≠ 0, δ(θ(l)) = 0, which means that the track and measurement information are updated, indicating that the track is normal. represents a Gaussian density with mean m and variance P, is the Gaussian component weight of the target, l represents the label information corresponding to the component, Indicates the measurement point corresponding to the component.

Update the branch weight to:

The single target normalization constant is:

When θ(l) = 0, the track is missed and the measurement is lost. The parameters of are given by the following formula:

When θ(l)≠0, position measurement The parameters are obtained by processing the position measurement and the Doppler measurement as follows:

First, use the position measurement to update the status:

Where _Hc represents the target observation matrix, represents the likelihood function, p _D is the detection probability of the radar, _Kyp and _Syp represent the gain and innovation covariance of the position measurement respectively, and the corresponding calculation process is as follows:

Compared with traditional filters, in order to effectively utilize Doppler information, a sequential filtering method is adopted, that is, the position information is first used to update the state, and then the Doppler measurement is used to further update the transition state. After obtaining a more accurate state estimate and likelihood function, the position and Doppler measurement information are finally used to calculate the weights.

Next, the Doppler information y _d is used for sequential updating. The specific steps are as follows:

Use Doppler information to sequentially update the target state:

The weight of the position component is:

in, They are expressed as the clutter intensity of the position component and the Doppler component respectively. Predicted Doppler measurement for:

The Doppler measurement gain and covariance are:

Where R _c and σ _d are the standard deviations of the position measurement and Doppler measurement noise, respectively. The Jacobian matrix of the Doppler measurement is:

H _d (m)＝[h ₁ , h ₂ , h ₃ , h ₄ ]

The parameters in the formula are:

Step 5. Calculate the probability of target measurement through the posterior probability density in the update process, distinguish surviving targets from new targets through the probability of target measurement, and retain the measurements related to the new targets for the new track at the next moment.

After the above calculations, a birth trajectory is established based on each measurement value. These measurements include clutter that is not eliminated by the measurement filtering rules, measurements of currently surviving targets, and measurements generated by the new targets we are interested in. Although most of the clutter is eliminated by measurement filtering before calculation, there will still be some clutter that cannot be eliminated by simple filtering rules, which can often be corrected by subsequent filtering algorithms. The very few clutter that are retained in the end will fall near the surviving target measurements, which will cause the filter to think that this is a measurement point generated by the current surviving target, which will lead to the divergence of the filtering results. In response to the above phenomenon, we also need to group the measurement set and divide the measurement set into surviving target-related points and new target-related points. The steps for measurement division are as follows:

At time k, the measurement set Z _k consists of the surviving target-related points Z _{B, k} and the newly generated target-related points Z _{S, k} . Here, the probability ρ(z) of the target measurement in the measurement set Z _k is defined as:

in, represents the association probability between the surviving target i and the measurement in Z _k at time k, and vice versa, 1- _pi represents the association probability between the new target i and the clutter in the measurement Z _k at time k. _{p D,k} is the sensor detection probability, z _k is the measurement point at time k, represents the Gaussian density with mean m and variance P, _Hk is the observation matrix, R is the position measurement noise covariance, and h(·) is defined as follows:

In the above formula, (x, y) represents the state vector of the target, and ( _xs , _ys ) represents the sensor position.

If ρ(z _k )>0.5, the measurement is considered to be from the current surviving target. Otherwise, when ρ(z _k )≤0.5, it indicates that the measurement is from a new target or clutter. According to the above rules, the measurement set is divided into two parts: Z _B,k and Z _S,k. Z _B,k is used to generate the new target component at the next moment, and Z _S,k is used in the subsequent GLMB update process.

The experimental conditions in this application are: new positions appear at different times and in different positions, respectively (-750, 20), (-750, -750), (750, -750), (750, 750), (-750, 750), the survival probability of the target is p _S = 0.99, the detection probability of the sensor is p _D = 0.98, the total number of targets is 9, each object moves with a constant speed and state vector, the state vector is in and represents the position component, and represents the velocity component, and the initial states and birth and death times of the nine tracking objects are shown in Table 1. Figure 2 shows the motion trajectory of each target, and Figure 3 shows the data measured by the sensor, which includes the measurement generated by the target and clutter.

Table 1 Tracking target running status table

The standard state transfer matrix is set as:

The covariance matrix is:

Where v represents the process noise with covariance Q and mean 0, σ _v = 10 m/s. The standard deviation of the target's prior velocity is σ _s = 17 m/s, and the standard deviation of the Doppler observation noise is σ _d = 0.5 m/s. In the experiment, the clutter density follows a Poisson distribution, and the number of clutter points in each cycle follows a Poisson distribution with a mean of λ _c = 20. The position of each clutter point is uniformly distributed within the measurement range, and λ _c represents the average number of clutter points per unit volume. The Doppler radar position is set to The observation noise covariance matrix is R=diag([(π/180) ² ,100]). For comparison purposes, the pruning parameters of all filters are T=10 ⁻⁵ , the merging threshold is U=4, the maximum number of Gaussian components is J _max =100, and the multi-target extraction threshold is 0.5.

In order to verify the effectiveness of the algorithm, by comparing Figures 4 and 5, it can be seen that without the track adaptive initiation algorithm, GLMB cannot effectively track the target, and the target appearing in the unknown position cannot be detected at all, and even gives wrong tracking results when the target track intersects. Figure 6 shows the performance of the filtering algorithm. When there are fluctuations at 0, 10, 30, 40, and 50 moments, because there are new targets generated at this time, but it can be maintained within 2 seconds. Convergence shows the effectiveness of the track initiation algorithm.

In summary, the method of this embodiment can calculate the parameters required for the start of the new track in one step based on the Doppler measurement. Through measurement screening and grouping, the influence of clutter and the measurement points of surviving targets on the new track can be effectively reduced. The applied method can quickly complete the goal of track initiation by only measuring points at two consecutive moments, and the filter performance can converge quickly.

The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims (5)

1. A GLMB-based adaptive track initiation method, characterized by comprising the following steps: Convert the polar coordinate system measurement data received by the radar into the rectangular coordinate system measurement data; According to the prior information, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density; The clutter in the measurement data is filtered out by using the velocity screening rule and the Doppler information screening rule; Adopting sequential filtering method to update multi-target posterior probability density; The probability of target measurement is calculated through the posterior probability density in the updating process, and the surviving targets and new targets are distinguished through the probability of target measurement, and the measurements related to the new targets are retained for the new track at the next moment. 2. The method of adaptive track initiation based on GLMB according to claim 1 is characterized in that, according to the prior information, the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density, which specifically includes the following steps: According to the new component calculated at the previous moment, the single target density of the new target is calculated; According to the posterior information transmitted at the previous moment, the single target density of the surviving target is calculated; The single target density of the new target and the single target density of the surviving target are connected in parallel to obtain the multi-target prediction probability density. 3. According to claim 2, an adaptive track initiation method based on GLMB is characterized in that the single target density of the new target and the single target density of the surviving target are associated to obtain the multi-target prediction probability density, and it also includes: and calculating the initial operating state of the new target based on the speed information hidden in the Doppler information, and the initial operating state includes the mean, covariance and weight of the new target. 4. The GLMB-based adaptive track initiation method according to claim 3, characterized in that the multi-target posterior probability density is: In the formula, represents the mean, variance, and weight of the missing measurement; represents the mean, variance and weight of the position measurement; θ(l) represents the track association mapping with label l; δ(θ(l)) is the De Kotta function. When θ(l) = 0, δ(θ(l)) = 1, which means that the measurement is not associated with the track, indicating that the track is missed; conversely, θ(l) ≠ 0, δ(θ(l)) = 0, which means that the track and measurement information are updated, indicating that the track is normal. 5. The GLMB-based adaptive track initiation method according to claim 4, wherein the probability ρ(z) of the target measurement is: in, represents the association probability between the surviving target i and the measurement in the measurement set Z _k at time k. Conversely, 1- _pi represents the association probability between the new target and the clutter in the measurement set Z _k at time k. p _D,k is the sensor detection probability. z _k is the measurement point at time k. represents the Gaussian density with mean m and variance P, _Hk is the observation matrix, R is the position measurement noise covariance, and h(·) is defined as follows: In the above formula, (x, y) represents the target position, and ( _xs , _ys ) represents the sensor position.