CN115792895A - Millimeter wave radar target tracking method combining UKF and random matrix - Google Patents
Millimeter wave radar target tracking method combining UKF and random matrix Download PDFInfo
- Publication number
- CN115792895A CN115792895A CN202211622703.9A CN202211622703A CN115792895A CN 115792895 A CN115792895 A CN 115792895A CN 202211622703 A CN202211622703 A CN 202211622703A CN 115792895 A CN115792895 A CN 115792895A
- Authority
- CN
- China
- Prior art keywords
- state
- track
- matrix
- ukf
- updating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a millimeter wave radar target tracking method combining a UKF and a random matrix, which comprises the following steps: step S1: defining a track state, wherein the track state comprises a kinematic state and an extended state, and modeling the kinematic state and the extended state respectively; step S2: predicting a flight path state, wherein the predicting of the flight path state comprises predicting a flight path state and a state covariance matrix at the moment k, and filtering the kinematic state by adopting Unscented Kalman (UKF); and step S3: matching and associating the measured point clouds; and step S4: updating the track kinematics state; step S5: and updating the track expansion state. The invention provides a new track and measurement likelihood calculation method based on a random matrix, which improves the association accuracy rate under the condition that a plurality of tracks are close to each other, adopts unscented Kalman to predict and update the track state, further improves the estimation precision of multi-target tracking, and is particularly suitable for high maneuvering target tracking scenes.
Description
Technical Field
The invention belongs to the technical field of target detection and tracking, and particularly relates to a millimeter wave radar target tracking method combining a UKF and a random matrix.
Background
The target tracking technology is a process of estimating the number and the state (such as position, speed and the like) of targets according to measurement received by a sensor, and the technology is widely applied to the fields of battlefield monitoring, air defense and reverse guidance, automatic driving and the like, and the traditional millimeter wave radar target tracking usually assumes that one target generates at most one sensor measurement at one moment and is called as point target tracking; however, with the improvement of the range resolution of the millimeter wave radar, the individual contour features of the target can be distinguished, the target occupies a plurality of range distinguishing units, each target can generate a plurality of measurements at the moment, and the traditional point target tracking is not applicable any more, so that the extended target tracking is proposed; the traditional point target tracking method can only estimate the motion states of the target such as the position, the speed and the like, and the extended target tracking method fully utilizes the target information contained in measurement and estimates the complex shape information of the target such as the size, the direction, the structure and the like through the spatial distribution of a plurality of measurements; an extended target tracking algorithm based on a random matrix is a classical algorithm, a set of complete Bayesian theory framework is established, and the motion state and the extended state of a target can be estimated simultaneously; wherein the extended state is modeled as an ellipse (including size, direction) and described by a random matrix variable, and the variable obeys an inverse weishat distribution.
The method in the existing multi-sensor fusion positioning initialization time generally has the following problems:
on one hand, the extended target tracking algorithm based on the random matrix only considers the likelihood of the measurement position information and the target extension state during measurement association, ignores the Doppler information specific to the millimeter wave radar, and therefore, the error association is generated under the condition that the distances of a plurality of targets are close, and the tracking performance is reduced;
on the other hand, as the road environment is increasingly complex, the variety of road traffic participants is great, the traditional kalman filter assumes that the target is in a linear motion state and performs tracking filtering on the target, and a large number of non-linear motion targets exist in the actual road, so that the performance of the traditional kalman filter is poor when the traditional kalman filter processes such targets.
Disclosure of Invention
The invention provides a millimeter wave radar target tracking method combining a UKF and a random matrix, which aims to solve the technical problems in the background technology.
In order to achieve the purpose, the invention adopts the technical scheme that: a millimeter wave radar target tracking method combining a UKF and a random matrix specifically comprises the following steps.
The beneficial effect of adopting above technical scheme is:
1. a new track and measurement likelihood calculation method is provided based on a random matrix, the association accuracy rate under the condition that a plurality of tracks are close to each other is improved, unscented kalman is adopted to predict and update the track state, the estimation precision of multi-target tracking is further improved, and the method is particularly suitable for high maneuvering target tracking scenes;
2. the method introduces the measurement likelihood value to calculate the track survival probability in the track management stage, can effectively reduce the generation of false targets, and improves the overall tracking effect.
Drawings
FIG. 1 is a logic block diagram of the millimeter wave radar target tracking method of the invention combining a UKF and a random matrix;
Detailed Description
The invention provides a millimeter wave radar target tracking method combining a UKF and a random matrix, which solves the problems in the background technology.
In order to achieve the purpose, the invention adopts the technical scheme that: the method comprises the following steps:
step S1: and (3) defining a track state, wherein the track state is divided into two parts: a kinematic state and an extended state, which are modeled respectively;
step S2: predicting a flight path state, predicting a flight path state and a state covariance matrix at the moment k, and filtering a kinematic state by adopting Unscented Kalman (UKF);
and step S3: measuring point cloud matching correlation;
and step S4: updating the track kinematics state;
step S5: updating a track expansion state;
step S6: and maintaining and deleting the flight path.
S1, defining a track state, and specifically comprising the following steps:
step S1.1: firstly, modeling a track kinematic state into a CTRV model, wherein the CTRV model comprises position information, speed information, yaw angle information and angular speed information of a track, and the CTRV model is defined as follows:
in the formula, p x Denotes the position in the x direction, p y Indicating the position in the y direction, v the track speed magnitude, phi the yaw angle,
represents an angular velocity;
step S1.2: secondly, modeling the extended state information X into an inverse Weissett distribution model, which is defined as follows:
IW(X；v,V)
wherein V is a degree of freedom estimated value, V is a parameter matrix, X is an extended state matrix, and the mean value estimation is as follows:
step S1.3: in the method, the track expansion state based on the inverse Weisset model has the following physical characteristics:
X＝AΛA T
in the formula I 1 The major axis of the ellipse,/ 2 Representing length of minor axis, theta being ellipticalOrientation angle in the range of [ - π, + π]。
S2, predicting the track state, which comprises the following specific steps:
step S2.1: first, the scatter particles are calculated:
x k|k-1,0 ＝x k-1|k-1
wherein λ = α 2 (L + k) -L is a scale parameter, alpha and k are two constant number factors, and L is the number of state dimensions;
i column element, x, representing the square root of the matrix k-1|k-1 For a posteriori estimation of the state at the time k-1, P k-1|k-1 A posteriori estimation of the state covariance matrix at the time k-1;
step S2.2: performing one-step prediction on all the scattered particles, and calculating to obtain a priori estimated mean value X k|k-1 And a prior covariance matrix P k|k-1 :
χ i ＝f(x k|k-1,i ),i＝0,1,..,2L
In the formula, f is a state transition equation, Q is a process noise matrix, and w is a weight factor:
wherein, the definitions of the lambda, the alpha and the L are consistent with the former, and the beta is a constant factor;
step S2.3: further calculating a measurement predicted value y through the state predicted value k|k-1 :
y i ＝Hχ i
H is a state observation matrix;
step S2.4: after the kinematics state prediction is finished, continuously predicting the track expansion state V at the moment K k|k-1 ：
In the formula, X k|k-1 Expanding the state matrix for time k, V k|k-1 For prediction of the k-time parameter matrix, V k-1 For the estimation of the parameter matrix at the time k-1, v k|k-1 Extended State freedom prediction for time k, v k-1 And (4) estimating the degree of freedom of the expansion state at the moment of k-1, wherein T is the working period of the millimeter wave radar, and tau represents a time constant variable.
S3, measuring point cloud matching association, and specifically comprising the following steps:
step S3.1: firstly, all the measurements are traversed, and the likelihood value L1 of each measurement and track position is calculated j It is defined as follows:
in the formula, e j The position residual error information of the track and the measurement j is defined as follows:
e j ＝(z j -y k|k-1 )
in the formula, z j Is the location information of the jth measurement, z j ＝[rsinθ,rcosθ]R represents the distance information of the measured point cloud, and theta represents the azimuth information of the measured point cloud;
step S3.2: then calculating the track and measuring the likelihood value L2 of Doppler velocity j It is as follows:
wherein mu is the average value of the Doppler velocity of the track,
σ 2 is the Doppler velocity variance, v j Measuring the Doppler velocity of the point cloud;
step S3.3: combining the position likelihood value and the Doppler likelihood value to obtain a final likelihood value L of the flight path and the measurement j j It is defined as follows:
L j ＝w 1 L1 j +w 2 L2 j
in the formula, w 1 And w 2 Weighting factors of the distance likelihood value and the Doppler likelihood value respectively;
step S3.4: for any L j Determine its magnitude with a given threshold GIf the following conditions are met:
L j <G
considering the measurement j to fall into the track correlation threshold, and completing correlation matching;
and S4, updating the track kinematic state, and specifically comprising the following steps:
step S4.1: firstly, calculating the mean value of all measured point clouds falling into a track correlation threshold
n k Measuring the number of point clouds;
step S4.2: further calculating Kalman gain K k ：
K k|k-1 ＝T k S k|k-1 -1
In the formula S k|k-1 As an innovation covariance matrix:
in the above formula R k To measure a noise matrix;
step S4.3: updating the track state x at time k k|k ：
Sum state covariance P k|k :
P k|k ＝P k|k-1 -K k|k-1 P k|k-1 K k|k-1 T ；
And S5, updating the track expansion state, and specifically comprising the following steps:
step S5.1: firstly, updating an extended state parameter matrix V according to the current associated measurement information k|k ：
X in the above formula to ensure matrix positivity k|k-1 1/2 、S k|k-1 -1/2 、Y k -1/2 Matrix decomposition is carried out by using a square root method (cholesky decomposition method);
step S5.2: continuously updating the parameter v of the degree of freedom k|k-1 ：
ν k|k ＝ν k|k-1 +n k ；
Step S5.3: obtaining a parameter matrix V k|k And a degree of freedom parameter v k|k Then, updating the k time expansion state matrix X k|k :
S6, track maintenance and deletion, which comprises the following steps:
step S6.1: by calculating the Sum Sum _ L of the likelihood values of all the correlation measurements at the moment k ：
To update the track existence probability Prob k ＝α(Prob k-1 )+β(Sum_L k ) Wherein both α and β are scaling factors;
step S6.2: by setting an output threshold G output And a deletion threshold G delete Carrying out track management;
step S6.3: when existence probability Prob k >G output Outputting the track state; when existing probability Prob k <G delete If so, deleting the flight path; and otherwise, maintaining the track state and keeping updating.
The present invention has been described in connection with the accompanying drawings, and it is to be understood that the invention is not limited to the specific embodiments described above, but is intended to cover various insubstantial modifications of the invention based on the principles and technical solutions of the invention; the present invention is not limited to the above embodiments, and can be modified in various ways.
Claims (7)
1. A millimeter wave radar target tracking method combining UKF and a random matrix is characterized in that: the method comprises the following steps:
step S1: defining a track state, wherein the track state comprises a kinematic state and an extended state, and modeling the kinematic state and the extended state respectively;
step S2: predicting a flight path state, wherein the prediction of the flight path state comprises predicting a flight path state and a state covariance matrix at the moment k, and filtering a kinematic state by adopting Unscented Kalman (UKF);
and step S3: matching and associating the measured point clouds;
and step S4: updating the flight path kinematics state;
step S5: updating a track expansion state;
step S6: and maintaining and deleting the flight path.
2. The millimeter wave radar target tracking method combining the UKF and the random matrix according to claim 1, characterized in that: in the step S1, the track state is defined, and the specific steps are as follows:
step S1.1: firstly, modeling a track kinematic state into a CTRV model, wherein the CTRV model comprises position information, speed information, yaw angle information and angular speed information of a track, and the CTRV model is defined as follows:
in the formula, p x Denotes the position in the x direction, p y Indicating the position in the y direction, v the track speed magnitude, phi the yaw angle,
represents an angular velocity;
step S1.2: secondly, modeling the extended state information X into an inverse Weirsart distribution model, which is defined as follows:
IW(X；v,V)
wherein V is a degree of freedom estimated value, V is a parameter matrix, X is an extended state matrix, and the mean value estimation is as follows:
step S1.3: in the method, the track expansion state based on the inverse Weisset model has the following physical characteristics:
X＝AΛA T
in the formula I 1 The major axis of the ellipse,/ 2 Denotes the length of the minor axis, theta is the orientation angle of the ellipse, and ranges from [ -pi, + pi]。
3. The millimeter wave radar target tracking method combining the UKF and the random matrix according to claim 1, characterized in that: s2, predicting the track state, which comprises the following specific steps:
step S2.1: first, the scatter particles are calculated:
x k|k-1,0 ＝x k-1|k-1
wherein λ = α 2 (L + k) -L is a scale parameter, alpha and k are two constant number factors, and L is the number of state dimensions;
i column element, x, representing the square root of the matrix k-1|k-1 For state posterior estimation at time k-1, P k-1|k-1 A posteriori estimation of the state covariance matrix at the time k-1;
step S2.2: performing one-step prediction on all the scattered particles, and calculating to obtain a priori estimated mean value X k|k-1 And a priori covariance matrix P k|k-1 :
χ i ＝f(x k|k-1,i ),i＝0,1,..,2L
Wherein f is a state transition equation, Q is a process noise matrix, and w is a weight factor:
wherein λ = α 2 (L + k) -L is a scale parameter, alpha and k are two constant factors, L is the number of state dimensions, and beta is also a constant factor;
step S2.3: further calculating a measurement predicted value y through the state predicted value k|k-1 :
y i ＝Hχ i
H is a state observation matrix;
step S2.4: after the kinematics state prediction is finished, continuously predicting the track expansion state V at the moment K k|k-1 ：
In the formula, X k|k-1 Expanding the state matrix for time k, V k|k-1 For prediction of the k-time parameter matrix, V k-1 For the estimation of the parameter matrix at the time k-1, v k|k-1 Extended state freedom prediction for time k, v k-1 And estimating the degree of freedom of the expansion state at the moment of k-1, wherein T is the working period of the millimeter wave radar, and tau represents a time constant variable.
4. The millimeter wave radar target tracking method combining the UKF and the random matrix according to claim 1, characterized in that: s3, measuring point cloud matching association, and specifically comprising the following steps:
step S3.1: firstly, all the measurements are traversed, and the likelihood value L1 of each measurement and track position is calculated j It is defined as follows:
in the formula, e j The position residual error information of the track and the measurement j is defined as follows:
e j ＝(z j -y k|k-1 )
in the formula, z j Is the location information of the jth measurement, z j ＝[rsinθ,rcosθ]R represents the distance information of the measured point cloud, and theta represents the azimuth information of the measured point cloud;
step S3.2: secondly, calculating the likelihood value L2 of the track and the measured Doppler velocity j It is as follows:
wherein mu is the average value of the Doppler velocity of the track,
σ 2 is the Doppler velocity variance, v j Measuring the Doppler velocity of the point cloud;
step S3.3: combining the position likelihood value and the Doppler likelihood value to obtain a final likelihood value L of the flight path and the measurement j j It is defined as follows:
L j ＝w 1 L1 j +w 2 L2 j
in the formula, w 1 And w 2 Weighting factors of the distance likelihood value and the Doppler likelihood value respectively;
step S3.4: for any L j And judging the magnitude of the signal and a given threshold G, and if the magnitude satisfies:
L j <G
and considering the measurement j to fall into the track correlation threshold, and finishing correlation matching.
5. The method for tracking the millimeter wave radar target by combining the UKF and the random matrix according to claim 1, wherein: and S4, updating the track kinematic state, and specifically comprising the following steps:
step S4.1: firstly, calculating the mean value of all measured point clouds falling into a track correlation threshold
n k Measuring the number of point clouds;
step S4.2: further calculating the Kalman gain K k ：
K k|k-1 ＝T k S k|k-1 -1
In the formula S k|k-1 As an innovation covariance matrix:
in the above formula R k Measuring a noise matrix;
step S4.3: updating the track state x at time k k|k ：
Sum state covariance P k|k :
P k|k ＝P k|k-1 -K k|k-1 P k|k-1 K k|k-1 T 。
6. The millimeter wave radar target tracking method combining the UKF and the random matrix according to claim 1, characterized in that: and S5, updating the track expansion state, which comprises the following specific steps:
step S5.1: firstly, updating an extended state parameter matrix V according to the current associated measurement information k|k ：
X in the above formula to ensure positive matrix characterization k|k-1 1/2 、S k|k-1 -1/2 、Y k -1/2 Matrix decomposition is carried out by using a square root method;
step S5.2: continuously updating the degree of freedom parameter v k|k-1 ：
ν k|k ＝ν k|k-1 +n k ；
Step S5.3: obtaining a parameter matrix V k|k And a degree of freedom parameter v k|k Then, updating the k-time expansion state matrix X k|k :
7. The millimeter wave radar target tracking method combining the UKF and the random matrix according to claim 1, characterized in that: s6, track maintenance and deletion, which comprises the following steps:
step S6.1: through calculating the Sum Sum _ L of the likelihood values of all the correlation measurements at the moment k ：
To update the track existence probability Prob k ＝α(Prob k-1 )+β(Sum_L k ) Wherein α and β are both scale factors;
step S6.2: by setting an output threshold G output And a deletion threshold G delete To perform track management(ii) a Step S6.3: when existence probability Prob k >G output Outputting the track state; when existence probability Prob k <G delete If so, deleting the flight path; and otherwise, maintaining the track state and keeping updating.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211622703.9A CN115792895A (en) | 2022-12-16 | 2022-12-16 | Millimeter wave radar target tracking method combining UKF and random matrix |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211622703.9A CN115792895A (en) | 2022-12-16 | 2022-12-16 | Millimeter wave radar target tracking method combining UKF and random matrix |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115792895A true CN115792895A (en) | 2023-03-14 |
Family
ID=85425901
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211622703.9A Pending CN115792895A (en) | 2022-12-16 | 2022-12-16 | Millimeter wave radar target tracking method combining UKF and random matrix |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115792895A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117075097A (en) * | 2023-10-12 | 2023-11-17 | 武汉理工大学三亚科教创新园 | Maritime radar target tracking method and system based on expanded target cluster division |
-
2022
- 2022-12-16 CN CN202211622703.9A patent/CN115792895A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117075097A (en) * | 2023-10-12 | 2023-11-17 | 武汉理工大学三亚科教创新园 | Maritime radar target tracking method and system based on expanded target cluster division |
| CN117075097B (en) * | 2023-10-12 | 2023-12-22 | 武汉理工大学三亚科教创新园 | Maritime radar target tracking method and system based on expanded target cluster division |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109633589A (en) | The Multi-target Data Associations assumed are optimized based on multi-model more in target following | |
| CN103729859B (en) | A kind of probability nearest neighbor domain multi-object tracking method based on fuzzy clustering | |
| CN108645413A (en) | The dynamic correcting method of positioning and map building while a kind of mobile robot | |
| CN108734725B (en) | Probability data correlation filtering extended target tracking method based on Gaussian process | |
| CN110503071B (en) | Multi-target tracking method based on variational Bayesian label multi-Bernoulli superposition model | |
| CN111722214B (en) | Method for realizing radar multi-target tracking PHD | |
| CN101221238B (en) | Dynamic deviation estimation method based on gauss average value mobile registration | |
| CN112613532B (en) | Moving target tracking method based on radar and cyclic neural network complement infrared fusion | |
| CN109633590A (en) | Extension method for tracking target based on GP-VSMM-JPDA | |
| CN104730537B (en) | Infrared/laser radar data fusion target tracking method based on multi-scale model | |
| CN107526070A (en) | The multipath fusion multiple target tracking algorithm of sky-wave OTH radar | |
| CN110689576A (en) | Automatic ware-based dynamic 3D point cloud normal distribution AGV positioning method | |
| CN113064155B (en) | Optimization method for track association under multi-target tracking of air radar | |
| CN107192995A (en) | A kind of Pure orientation underwater target tracking algorithm of multi-level information fusion | |
| CN112034445B (en) | Vehicle motion trail tracking method and system based on millimeter wave radar | |
| CN115792895A (en) | Millimeter wave radar target tracking method combining UKF and random matrix | |
| CN109214432B (en) | Multi-sensor multi-target joint detection, tracking and classification method | |
| CN117169818B (en) | Radar waveform design method for sea surface maneuvering target tracking | |
| CN108871365B (en) | State estimation method and system under course constraint | |
| CN112328959A (en) | Multi-target tracking method based on adaptive extended Kalman probability hypothesis density filter | |
| CN115905986A (en) | Steady Kalman filtering method based on joint strategy | |
| CN104777465B (en) | Random extended object shape and state estimation method based on B spline function | |
| CN111262556B (en) | Multi-target tracking method for simultaneously estimating unknown Gaussian measurement noise statistics | |
| CN111679270B (en) | Multipath fusion target detection algorithm under scene with uncertain reflection points | |
| CN102426357B (en) | Multi-target tracking method with function of image confirmation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |