Adaptive EKF tracking method, system and terminal based on variational Bayes and Sage-Husa

Through the adaptive extended Kalman filtering method of variational Bayes and Sage-Husa, the problems of target tracking robustness and model error correction in noise environments are solved, and better tracking effect is achieved.

CN118897940APending Publication Date: 2024-11-05HARBIN ENG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410932975.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

The existing target tracking methods are not robust enough in noisy environments, making it difficult to correct model errors during target movement, resulting in poor tracking results.

Method used

Adaptive extended Kalman filtering method based on variational Bayes and Sage-Husa is adopted to select appropriate motion models for position prediction, correct the observation error covariance matrix and prediction error covariance matrix, and realize iterative tracking filtering.

Benefits of technology

It improves the robustness and model error correction ability in noise environments, improves the tracking effect, and is more accurate than the traditional Kalman filtering results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118897940A_ABST
    Figure CN118897940A_ABST
Patent Text Reader

Abstract

The invention discloses an adaptive EKF tracking method and system based on variational Bayes and Sage-Husa, a medium and a terminal, relates to the technical field of target tracking, and aims to solve the problem that the error correction capability of a model needs to be improved in a target motion process. The method comprises the steps of selecting a motion model according to an actual related environment, selecting a preset value based on the selected motion model, and performing one-step prediction on position estimation to obtain a position at a current moment and a target position variance; performing extended Kalman filtering on the position at the current moment and the target position variance, and solving an azimuth solution and a variance solution of the extended Kalman filtering at the current moment; correcting the observation error covariance matrix based on variational Bayesian inference; based on the obtained observation error covariance matrix, using a Sage-Husa filtering algorithm to obtain a corrected prediction error covariance matrix; and performing iteration based on the corrected prediction error covariance matrix and the observation error covariance matrix to perform tracking filtering at the next moment. And effective tracking of the target in the same environment is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of target tracking, and in particular to an adaptive EKF tracking method, system and program based on variational Bayes and Sage-Husa. Background Art

[0002] Environmental noise has significant randomness and uncertainty, and target tracking and detection are often disturbed by frequent noise fluctuations. Conventional extended Kalman filtering often requires presetting a more appropriate noise error model and does not modify it during the process. Therefore, if the preset error model does not match the actual environment well enough, performance may degrade, and in severe cases, tracking failure may occur. Therefore, it is necessary to study a filter tracking algorithm with adaptive model matching capabilities that is suitable for different environmental conditions.

[0003] This application is dedicated to studying a filter tracking algorithm with adaptive model matching capability that is applicable to different environmental conditions. Specifically, the following problems need to be solved: 1) Tracking in noisy environments requires robustness; 2) During the target movement process, the ability to correct model errors is required; 3) The overall tracking effect needs to be good. Summary of the invention

[0004] The purpose of the present invention is to propose an adaptive tracking method based on variational Bayes and Sage-Husa in view of the technical status quo that the robustness of existing target tracking methods in noisy environments, the model error correction capability during target motion and the overall tracking effect need to be improved.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: An adaptive EKF tracking method based on variational Bayes and Sage-Husa is provided, comprising the following steps: S10: Select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; S20: Perform extended Kalman filtering on the current position and the target position variance, and solve the orientation solution and variance solution of the extended Kalman filter at the current moment; S30: Modify the observation error covariance matrix based on variational Bayesian inference; S40: Based on the obtained observation error covariance matrix, a Sage-Husa filtering algorithm is used to obtain a modified prediction error covariance matrix; S50: Iterating and performing tracking filtering at the next moment based on the modified prediction error covariance matrix and observation error covariance matrix.

[0006] Compared with the prior art, the adaptive EKF tracking method based on variational Bayes and Sage-Husa provided by the present invention has the following technical effects: 1) Unlike the existing methods that treat the observation error covariance matrix as a parameter to be estimated and optimize it through maximum a posteriori probability estimation or maximum likelihood estimation, this method can effectively solve the robustness problem of the tracking algorithm under environmental noise with relatively stable first- and second-order characteristics; 2) It can effectively correct the fluctuations of the motion model; 3) With the ability to simultaneously correct the observation error covariance and the prediction error covariance matrix, the tracking results are better than those of traditional Kalman filtering.

[0007] The above scheme combines the variational inference static approximation method to improve the robustness of the algorithm. At the same time, combined with Sage-Husa's approximation of the prediction error covariance matrix, it has a flexible detection range that changes with the observed data, better performs azimuth tracking, and thus achieves better applicable effective target tracking in the same environment.

[0008] The present invention also provides an adaptive EKF tracking system based on variational Bayes and Sage-Husa, comprising: A data processing unit, used to select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; Perform extended Kalman filtering on the current position and target position variance, and solve the azimuth solution and variance solution of the extended Kalman filter at the current moment; The observation error covariance matrix is ​​modified based on variational Bayesian inference; It is also used to obtain the corrected prediction error covariance matrix based on the obtained observation error covariance matrix using Sage-Husa filtering; The tracking filter at the next moment is iterated based on the modified prediction error covariance matrix and observation error covariance matrix.

[0009] The present invention also provides a storage medium, wherein the computer storage medium stores instructions, and when the instructions are executed, an adaptive EKF tracking method based on variational Bayes and Sage-Husa is implemented.

[0010] The present invention also provides a terminal, comprising a processor and a communication interface coupled to the processor; the processor is used to run a computer program or instruction to implement the adaptive EKF tracking method based on variational Bayes and Sage-Husa described in the first aspect.

[0011] Compared with the prior art, the beneficial effects of the adaptive EKF tracking system based on variational Bayes and Sage-Husa provided by the present invention are the same as the beneficial effects of the adaptive EKF tracking method based on variational Bayes and Sage-Husa described in the above technical scheme, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 is a schematic flow chart of an adaptive EKF tracking method based on variational Bayes and Sage-Husa in an embodiment of the present invention; Figure 2 It is a flow chart of an adaptive extended Kalman filter tracking algorithm based on variational Bayes and Sage-Husa noise online filtering in an embodiment of the present invention; Figure 3 The Q tracking coefficient is selected as 0.0002 (i.e., the Q value is modified by 0.02% each time the adaptive data is used), and the R tracking coefficient is selected as 0.6 (i.e., 60% of the historical weights are retained each time). The tracking result graphs of the algorithms VBSA-EKF (patented algorithm), VB-EKF, EKF, and SA-EKF; Figure 4 is the Euclidean distance error of the algorithm. DETAILED DESCRIPTION

[0013] In order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, words such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and effects. For example, the first threshold and the second threshold are only used to distinguish different thresholds, and their order is not limited. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different.

[0014] It should be noted that, in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0015] The algorithms and names involved in the present invention are explained, where: VBSE-EKF: the algorithm proposed in this paper; VB-EKF: Variational Bayesian Extended Kalman Filter algorithm; SH-EKF: Sage-Husa extended Kalman filter algorithm; EKF: Extended Kalman filter.

[0016] like Figures 1 to 4 As shown, the adaptive EKF tracking method based on variational Bayes and Sage-Husa provided by the present invention comprises the following steps: S10: Select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; S20: Perform extended Kalman filtering on the current position and the target position variance, and solve the orientation solution and variance solution of the extended Kalman filter at the current moment; S30: Modify the observation error covariance matrix based on variational Bayesian inference; S40: Based on the obtained observation error covariance matrix, a Sage-Husa filtering algorithm is used to obtain a modified prediction error covariance matrix; S50: Iterating and performing tracking filtering at the next moment based on the modified prediction error covariance matrix and observation error covariance matrix.

[0017] When implementing: S1. Select a motion model and a preset value according to the actual relevant environment, make a one-step prediction on the position estimation, and obtain the current position and target position variance; Select the preset value, specifically: select the corresponding initial value of the target prediction error variance matrix and the initial value of the target observation error covariance matrix; The initial value of the target prediction error covariance matrix is ​​denoted as Q 0 ; The initial value of the target observation error covariance matrix is ​​denoted as R 0 ; Selecting a motion model means setting the motion trajectory; One-step prediction is achieved through the following process: x k-1 For a given k -1 time target estimated initial value, F k|k-1 for k- 1 moment target motion transfer matrix, Q k for k The target prediction error covariance matrix at time ; (1) x k|k-1 To pass k -1 The target estimate at time one step is obtained by k Time location, To pass k- 1 moment target estimate is obtained by one-step prediction k The variance of the moment position; S2, perform extended Kalman filtering on the current position and target position variance, and solve the azimuth solution and variance solution of the extended Kalman filter at the current moment; The following sub-steps are included: S21, through one-step prediction, based on k The observation transfer matrix at time H k And the observation error covariance matrix R k ,get k Time location z k and the covariance matrix at that location C k ; (2) S22, calculate the Kalman coefficient K and convert it into k Time location z k and observational data y k Input the Kalman filter to get the predicted k Azimuth solution and variance solution of the time-extended Kalman filter; (3) (4) (5) for k The observation transfer matrix at time H k The transpose of xk 、P k for k -1 time extended Kalman filter azimuth solution, variance solution; S3: Correcting the observation error covariance matrix based on variational Bayesian inference; Combined with variational Bayesian inference, according to k -1 moment solution, historical information of the observation error covariance matrix, and correction of the observation error covariance matrix; The observation error covariance matrix is ​​denoted as R k ; The essence of variational inference is to find the extreme value of the functional. KL divergence is the target functional. q Function, which minimizes the KL divergence and approaches the required posterior distribution; Combined with the posterior distribution of underwater target tracking, it is usually assumed that the target and noise are independent of each other, and the joint probability distribution is assumed to be in the form of a product, satisfying the selected approximate function framework, namely: (6) set up is the posterior distribution of the target motion orientation, is the target noise posterior distribution; The observation error covariance matrix is ​​derived R k : (7) (8) (9) in, V k is the sum of matrix samples, v k is the freedom degree of matrix samples. Each time a matrix sample is added, the freedom degree increases by 1. By correcting the sum of the freedom degree and the matrix sample, the influence of historical data can be effectively controlled. Adaptively generate the prediction error covariance matrix and perform adaptive mean processing based on the environment.

[0018] Step 4: Obtain the corrected prediction error covariance matrix based on the Sage-Husa filter through the obtained observation error covariance matrix; S4: Obtain the corrected prediction error covariance matrix based on the obtained observation error covariance matrix and Sage-Husa filtering; The Sage-Husa filter in step 4 is implemented using the windowed average of the observations instead of the true random model; The modified forecast error covariance matrix is ​​obtained by introducing historical data and historical weights; The predicted error covariance matrix is: (10) By introducing historical weights, (10) is modified to (11): (11) in, d k is the historical weight correction coefficient; is the modified forecast error covariance matrix, j for 0 Time has come k-1 time.

[0019] S5: Iterate and perform tracking filtering at the next moment based on the modified prediction error covariance matrix and observation error covariance matrix.

[0020] Compared with the prior art, the adaptive EKF tracking method based on variational Bayes and Sage-Husa provided by the present invention has the following technical effects: 1) Unlike the existing methods that treat the observation error covariance matrix as a parameter to be estimated and optimize it through maximum a posteriori probability estimation or maximum likelihood estimation, this method can effectively solve the robustness problem of the tracking algorithm under environmental noise with relatively stable first- and second-order characteristics; 2) It can effectively correct the fluctuations of the motion model; 3) With the ability to simultaneously correct the observation error covariance and the prediction error covariance matrix, the tracking results are better than those of traditional Kalman filtering.

[0021] The above scheme combines the variational inference static approximation method to improve the robustness of the algorithm. At the same time, combined with Sage-Husa's approximation of the prediction error covariance matrix, it has a flexible detection range that changes with the observed data, better performs azimuth tracking, and thus achieves better applicable effective target tracking in the same environment.

[0022] As an implementable method, the steps of selecting the preset value are as follows: Select the initial values ​​of the prediction error covariance matrix and the observation error covariance matrix corresponding to the target.

[0023] By selecting the initial values ​​of the target prediction error covariance matrix and the initial values ​​of the observation error covariance matrix as input parameters, data support is provided for subsequent process steps.

[0024] As an implementable method, an extended Kalman filter is performed on the current position and the target position variance to solve the orientation solution and variance solution of the extended Kalman filter at the current moment, including the following steps: S21, through one-step prediction, based on the observation transfer matrix at the current moment and the observation error covariance matrix of the target, obtain the observation value corresponding to the predicted position at the current moment and the variance of the corresponding observation value; S22, calculate the Kalman coefficient and input the real observation data of the current position into the Kalman filter to obtain the azimuth solution and variance solution of the extended Kalman filter at the current moment.

[0025] Through the one-step prediction, basic data can be provided for the correction of the observation error covariance matrix, ensuring the correction effect of the observation error covariance matrix.

[0026] As an implementable method, the observation error covariance matrix is ​​corrected based on variational Bayesian inference, including the following steps: Combined with variational Bayesian inference, the observation error covariance matrix at the current moment is corrected according to the orientation solution, variance solution, historical information and observation error covariance matrix at the previous moment.

[0027] The accuracy of the tracking results is ensured by correcting the observation error covariance matrix.

[0028] As an implementable method, the prediction error covariance matrix is ​​obtained as: By introducing historical weights and iteratively correcting the prediction error covariance matrix, the corrected prediction error covariance matrix is: in, Q k for k The prediction error covariance matrix of the target at time instant, is the modified forecast error covariance matrix, F k|k-1 for k-1 The target motion transfer matrix at the moment, P k for k The variance solution of the extended Kalman filter at the moment, the observation value corresponding to the predicted position at the moment z k and observational data y k , d k is the historical weight correction coefficient, j 0 time to k-1 time.

[0029] By correcting the prediction error covariance matrix, the accuracy of tracking results is guaranteed.

[0030] The present invention also provides an adaptive EKF tracking system based on variational Bayes and Sage-Husa, comprising: A data processing unit, used to select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; Perform extended Kalman filtering on the current position and target position variance, and solve the azimuth solution and variance solution of the extended Kalman filter at the current moment; The observation error covariance matrix is ​​modified based on variational Bayesian inference; It is also used to obtain the corrected prediction error covariance matrix based on the obtained observation error covariance matrix using Sage-Husa filtering; The tracking filter at the next moment is iterated based on the modified prediction error covariance matrix and observation error covariance matrix.

[0031] As an implementable method, the data processing unit is further used to obtain the observation value corresponding to the predicted position at the current moment and the variance of the corresponding observation value based on the observation transfer matrix at the current moment and the observation error covariance matrix of the target through one-step prediction; Calculate the Kalman coefficient and input the real observation data of the current position into the Kalman filter to obtain the azimuth solution and variance solution of the extended Kalman filter at the current moment.

[0032] As an implementable method, the data processing unit is also used to combine variational Bayesian inference to correct the error covariance matrix at the current moment according to the orientation solution, variance solution, historical information and observation error covariance matrix at the previous moment.

[0033] The present invention also discloses a terminal, comprising a processor and a communication interface coupled to the processor; the processor is used to run a computer program or instruction to implement the adaptive EKF tracking method based on variational Bayes and Sage-Husa of the first aspect.

[0034] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on the computer, the process or function of the embodiment of the present invention is executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a terminal, a user device or other programmable device. The computer program or instruction can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instruction can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless means. The computer-readable storage medium can be any available medium that can be accessed by the computer or a data storage device such as a server, data center, etc. that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a tape; it can also be an optical medium, such as a digital video disc (DVD); it can also be a semiconductor medium, such as a solid state drive (SSD).

[0035] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit may implement several functions listed in a claim. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0036] Although the present invention has been described in conjunction with specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present invention. Accordingly, this specification and the accompanying drawings are merely exemplary illustrations of the present invention as defined by the appended claims and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present invention. Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, the present invention is intended to include such modifications and variations if they fall within the scope of the claims of the present invention and their equivalents.

Claims

1. An adaptive EKF tracking method based on variational Bayes and Sage-Husa, characterized in that: The following steps are involved: S10: Select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; S20: Perform extended Kalman filtering on the current position and the target position variance, and solve the orientation solution and variance solution of the extended Kalman filter at the current moment; S30: Modify the observation error covariance matrix based on variational Bayesian inference; S40: Based on the obtained observation error covariance matrix, a Sage-Husa filtering algorithm is used to obtain a modified prediction error covariance matrix; S50: Iterating and performing tracking filtering at the next moment based on the modified prediction error covariance matrix and observation error covariance matrix.

2. The method according to claim 1, characterized in that The steps of selecting the preset value are as follows: Select the initial values ​​of the prediction error covariance matrix and the observation error covariance matrix corresponding to the target.

3. The method according to claim 1, characterized in that The method of performing extended Kalman filtering on the current position and the target position variance to solve the orientation solution and variance solution of the extended Kalman filter at the current moment includes the following steps: S21, through one-step prediction, based on the observation transfer matrix at the current moment and the observation error covariance matrix of the target, obtain the observation value corresponding to the predicted position at the current moment and the variance of the corresponding observation value; S22, calculate the Kalman coefficient and input the real observation data of the current position into the Kalman filter to obtain the azimuth solution and variance solution of the extended Kalman filter at the current moment.

4. The method according to claim 1, characterized in that: The method of correcting the observation error covariance matrix based on variational Bayesian inference includes the following steps: Combined with variational Bayesian inference, the observation error covariance matrix at the current moment is corrected according to the historical information of the orientation solution, variance solution, and observation error covariance matrix at the previous moment.

5. The method according to claim 1, characterized in that The resulting prediction error covariance matrix is: By introducing historical weights and iteratively correcting the prediction error covariance matrix, the corrected prediction error covariance matrix is: Among them, Q k is the prediction error covariance matrix of the target at time k, is the modified forecast error covariance matrix, F k|k-1 is the target motion transfer matrix at time k-1, P k is the variance solution of the extended Kalman filter at time k, and the observation value z corresponding to the predicted position at time k k And the observed data y k , d k is the historical weight correction coefficient, j is from time 0 to time k-1.

6. An adaptive EKF tracking system based on variational Bayes and Sage-Husa, characterized in that: include: A data processing unit, used to select a motion model according to the actual relevant environment, select a preset value based on the selected motion model, perform a one-step prediction on the position estimation, and obtain the current position and the target position variance; Perform extended Kalman filtering on the current position and target position variance, and solve the azimuth solution and variance solution of the extended Kalman filter at the current moment; The observation error covariance matrix is ​​modified based on variational Bayesian inference; It is also used to obtain the corrected prediction error covariance matrix based on the obtained observation error covariance matrix using Sage-Husa filtering; The tracking filter at the next moment is iterated based on the modified prediction error covariance matrix and observation error covariance matrix.

7. The system according to claim 6, characterized in that: The data processing unit is further used to obtain the observation value corresponding to the predicted position at the current moment and the variance of the corresponding observation value through one-step prediction based on the observation transfer matrix at the current moment and the observation error covariance matrix of the target; The Kalman coefficient is calculated and the real observation data of the current position is input into the Kalman filter to obtain the azimuth solution and variance solution of the extended Kalman filter at the current moment.

8. The system according to claim 6, characterized in that: The data processing unit is also used to combine variational Bayesian inference to correct the observation error covariance matrix at the current moment according to the historical information of the orientation solution, variance solution, and observation error covariance matrix at the previous moment.

9. A terminal, characterized in that: It includes a processor and a communication interface coupled to the processor; the processor is used to run a computer program or instruction to implement the adaptive EKF tracking method based on variational Bayes and Sage-Husa described in the first aspect.