Alternating iteration robust track association method for realizing error association decoupling

Abstract

The invention discloses an alternate iteration robust track correlation method for realizing error correlation decoupling, which relates to the technical field of signal processing, and comprises the following steps of: firstly, in a multi-target flight scene, aiming at a multi-false alarm condition existing in a sensor A and a sensor B, determining an error correlation between the sensor A and the sensor B; constructing a GNN track correlation model for the same target by using the local detection tracks of the two sensors; secondly, under the condition that influence of system errors and random noise of a sensor A and a sensor B is considered, joint estimation is carried out on system errors and track correlation, and the system errors comprise translation system errors or angle and distance system errors; and finally, optimizing the translation system error or the angle and distance system error based on a maximum consistent correlation number optimization algorithm to complete decoupling of error correlation. The method provided by the invention can solve the problem of track association which is relatively complex and relatively high in difficulty, has relatively good universality and application range, and can improve the optimization performance of track association.

Description

Technical Field

[0001] The invention relates to the technical field of signal processing, and in particular to an alternating iterative anti-error track association method for realizing error association decoupling. Background Art

[0002] Track association is mainly aimed at the problem of multi-sensor tracking of multi-target tracks. It determines to which target the observed track of each sensor belongs, and then associates and maps the tracks belonging to the same source targets to improve the accuracy of track information.

[0003] In the process of multi-sensor tracking of multiple targets, the internal structure of the system is relatively complex. During the tracking process, the sensors will continuously produce distance deviation and angle deviation, causing the system error and track association to couple with each other, and the error will continue to accumulate and spread. The traditional robust track association algorithm mainly estimates the system error and track association jointly, and realizes error estimation compensation in the alternating iterative process.

[0004] Currently, researchers have proposed a variety of methods for alternating iterative robust track association estimation, such as interpolation compensation method, heuristic algorithm, expectation maximization (EM) algorithm framework, etc.

[0005] As shown in Example 1, in the 1994 Proceedings of SPIE journal article "Track association using correction for bias and missing data", the angle system error estimation is achieved by taking the median of the difference to solve the problem of alternating iterations of track association and system error estimation.

[0006] As shown in Example 2, in the journal paper “Simultaneous Track-to-Track Association and Bias Removal Using MultistartLocal Search” published in the 2008 Proceedings of Aerospace Conference, a multi-start alternating iterative heuristic algorithm is used to achieve joint estimation of system error and track association.

[0007] As shown in Example 3, in the 2010 journal article “Joint Data Association, Registration, and Fusion using EM-KF” in Aerospace and Electronic Systems, the joint estimation of track association and system error is achieved in an alternating iterative process through the E-step and M-step of the expectation-maximization (EM) algorithm framework.

[0008] However, the above track association algorithm only considers the influence of translation system error. When the sensor generates system error structures such as distance deviation and angle deviation, there is a coupling relationship between system error estimation and track association. In the process of alternating iteration of track association and system error estimation, the error association error will continue to increase, thereby reducing the correctness of the association algorithm. Therefore, the present invention proposes an alternating iterative robust track association method that realizes error association decoupling, so as to solve the shortcomings of the above prior art. Summary of the invention

[0009] The purpose of the present invention is to provide an alternating iterative robust track association method for realizing error correlation decoupling, so as to solve the problem of correlation coupling caused by lack of consideration of distance deviation, angle deviation, etc. in the alternating iterative robust track correlation estimation in the prior art, and the problem that the system error estimator is sensitive to error correlation, thereby greatly reducing the stability and reliability of the correlation estimation.

[0010] To achieve the above object, the present invention provides an alternating iterative robust track association method for realizing error association decoupling, comprising the following steps: Step 1: In the scenario of multi-target flight, for the multiple false alarms of sensor A and sensor B, the local detection tracks of the two sensors are used to build a GNN track association model for the same target; Step 2: Under the condition of considering the influence of the system error and random noise of sensor A and sensor B, jointly estimate the system error and track association, wherein the system error includes translation system error or angle and distance system error; Step 3: Based on the maximum consistent association number (MaxCAN) optimization algorithm, the translation system error or the angle and distance system error in step 2 are optimized to complete the decoupling of the error association.

[0011] Preferably, in step 1, the local detection tracks of sensor A and sensor B are and , under the condition of multiple targets and multiple false alarms on both sides, the accuracy of the two sensors in associating the same target track The calculation expression is as follows: (1) In the formula, m is the observation space dimension, is a fixed constant, is the target airspace density, is the multiplication factor, 、 represents the false alarm target density observed by sensors A and B respectively, is the standard deviation of the average correlation noise.

[0012] Preferably, the expression for jointly estimating the system error and the track association in step 2 is as follows: (2) In the formula, is the system error, satisfying ; is the correlation coefficient matrix between the tracks of sensors A and B; is the association cost matrix, which satisfies the following formula: (3) In the formula, , is the detection probability of sensors A and B, is the likelihood function of two homologous tracks, , It means that there is no track element observed by the other party in the track set of sensors A and B, which satisfies the following formula: (4) (5).

[0013] Preferably, the specific process of optimizing the translation system error or the angle and distance system error in step 2 based on the maximum consistent association number optimization algorithm in step 3 is as follows: S31, MaxCAN algorithm initialization; input sensor A, sensor B local detection track , ; S32, through the GNN track association model formula (1) and the joint estimation algorithm formula (2), the track association coefficient matrix is ​​calculated and the global association cost matrix ; S33, introduce the LMedS system error estimator, and solve the error estimation result of the translation system only ; S34. Calculate the improved consistent correlation number ,like , then there is no mismatched element, execute the MinAC (minimum association cost) algorithm, and jump to S36; if , then there are elements with incorrect pairing, and proceed to the next step; S35. Judgment Is it reduced? If so, execute S36; if not, then , Perform error compensation and return to S32; S36, the algorithm ends and the global minimum track correlation coefficient matrix is ​​output .

[0014] Preferably, in S33, only the translation system error estimation result is solved The calculation expression is as follows: (6) In the formula, It is the error vector formed by subtracting the corresponding track elements of the two track sets of sensor A and sensor B, and med is the median operation.

[0015] Preferably, the improved consistent correlation number in S34 is The calculation process is as follows: S341A, calculate the correlation coefficient matrix as Total number of track element pairs in , the calculation expression is as follows: (7) S342A, let the correlation coefficient matrix The number of successful pairing of track elements under the condition of pairing errors is , then the consistent correlation number Satisfies the following formula: (8) From equation (8), we can conclude that when the threshold T of the false association detector is set to When , all track associations are correct, indicating that there are no mismatched elements. On the contrary, when the threshold T is set When , all track relationships are wrong. ; S343A, based on the conclusions drawn from S342A, make optimization adjustments and consistent correlation numbers Improved to the following expression: (9).

[0016] Preferably, in S35, the detection track , The expression for error compensation is as follows: (10) (11) (12).

[0017] Preferably, the specific execution process of the MinAC algorithm in S34 is as follows: S341B, MinAC algorithm initialization; input sensor A, B local detection track information , , set the minimum association cost matrix The initial value is +∞, and the maximum number of iterations is N; S342B. Judge the relationship between the current iteration number n and the maximum iteration number N. If the current iteration number n > N, execute S344B; if the current iteration number n < N, calculate the track association coefficient matrix through the GNN track association model formula (1) and the joint estimation formula (2) and the global association cost matrix , and judge the global association cost matrix and the minimum association cost matrix relationship. If , then , , execute S343B; if , then execute S344; S343B. Let the current iteration number n = n + 1. Given the calculated , calculate the estimation results with only translational system errors and the estimation results with only angular and distance system errors respectively; realize the error compensation for the detection tracks , of sensors A and B, and return to S342; S344B. The algorithm ends and outputs .

[0018] Preferably, the calculation expression of the estimation result with only angular and distance system errors in S343B is as follows: (13) In the formula, represents the estimated residual vector generated by the association detection of sensors A and B.

[0019] Therefore, the present invention adopts the above-mentioned alternating iterative robust track association method for realizing error association decoupling, and has the following beneficial effects: (1) The present invention is a robust track association method proposed under the conditions of multiple targets and two-party multiple false alarms, which can solve relatively complex and difficult track association problems, and has good generality and application scope; (2) The optimization algorithms based on the minimum association cost (MinAC) and the maximum consistent association number (MaxCAN) proposed by the present invention can reduce the sensitivity of the alternating iteration process to incorrect associations through relatively robust system error estimators in both system error and residual environments, and further improve the optimization performance of track association.

[0020] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The present invention is an overall flow chart of an alternating iterative robust track association method for realizing error association decoupling; Figure 2 A flow chart of the MinAC algorithm according to an embodiment of the present invention; Figure 3 This is a flow chart of the MaxCAN algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1-Figure 3 , an alternating iterative robust track association method for realizing error association decoupling, comprising the following steps: Step 1: In the scenario of multi-target flight, for the multiple false alarms of sensor A and sensor B, the local detection tracks of the two sensors are used to construct a GNN track association model for the same target; the local detection tracks of sensor A and sensor B are respectively and , under the condition of multiple targets and multiple false alarms on both sides, the accuracy of the two sensors in associating the same target track The calculation expression is as follows: (1) In the formula, m is the observation space dimension, is a fixed constant, is the target airspace density, is the multiplication factor, 、 represents the false alarm target density observed by sensors A and B respectively, is the standard deviation of the average correlation noise.

[0024] Step 2: Considering the influence of the system error and random noise of sensor A and sensor B, jointly estimate the system error and track association, where the system error includes translation system error or angle and distance system error; the expression for jointly estimating the system error and track association is as follows: (2) In the formula, is the system error, satisfying ; is the correlation coefficient matrix between the tracks of sensors A and B; is the association cost matrix, which satisfies the following formula: (3) In the formula, , is the detection probability of sensors A and B, is the likelihood function of two homologous tracks, , It means that there is no track element observed by the other party in the track set of sensors A and B, which satisfies the following formula: (4) (5).

[0025] Step 3: Optimize the translation system error or angle and distance system error in step 2 based on the maximum consistent association number (MaxCAN) optimization algorithm to complete the decoupling of the error association; wherein, the specific process of optimizing the translation system error or angle and distance system error in step 2 based on the maximum consistent association number optimization algorithm is as follows: S31, MaxCAN algorithm initialization; input sensor A, sensor B local detection track , ; S32, through the GNN track association model formula (1) and the joint estimation algorithm formula (2), the track association coefficient matrix is ​​calculated and the global association cost matrix ; S33, introduce the LMedS system error estimator, and solve the error estimation result of the translation system only ; Among them, the robust system error estimator LMedS is introduced to reduce the negative impact of erroneous association on alternating iterations and improve the reliability of track association; the specific calculation expression is as follows: (6) In the formula, It is the error vector formed by subtracting the corresponding track elements of the two track sets of sensor A and sensor B, and med is the median operation.

[0026] S34. Calculate the improved consistent correlation number ,like , then there is no mismatched element, execute the MinAC (minimum association cost) algorithm, and jump to S36; if , then there are elements with incorrect pairing, and S35 is executed; wherein the improved consistent association number The calculation process is as follows: S341A, calculate the correlation coefficient matrix as Total number of track element pairs in , the calculation expression is as follows: (7) S342A, let the correlation coefficient matrix The number of successful pairing of track elements under the condition of pairing errors is , then the consistent correlation number Satisfies the following formula: (8) From equation (8), we can conclude that when the threshold T of the false association detector is set to When , all track associations are correct, indicating that there are no mismatched elements. On the contrary, when the threshold T is set When , all track relationships are wrong. ; S343A, based on the conclusions drawn from S342A, make optimization adjustments and consistent correlation numbers Improved to the following expression: (9).

[0027] S35. Judgment Is it reduced? If so, execute S36; if not, then , Perform error compensation and return to S32; wherein, the detected track , The expression for error compensation is as follows: (10) (11) (12).

[0028] The specific implementation process of the MinAC algorithm is as follows: S341B, MinAC algorithm initialization; input sensor A, B local detection track information , , set the minimum association cost matrix The initial value is +∞, and the maximum number of iterations is N; S342B. Determine the relationship between the current iteration number n and the maximum iteration number N. If the current iteration number n > N, execute S344B; if the current iteration number n < N, calculate the track correlation coefficient matrix through the GNN track correlation model formula (1) and the joint estimation formula (2). and the global correlation cost matrix , and determine the relationship between the global correlation cost matrix and the minimum correlation cost matrix . If , then , , execute S343B; if , then execute S344; S343B. Let the current iteration number n = n + 1. Given the calculated , calculate the estimation results with only translational systematic errors and with only angular and distance systematic errors respectively; realize the error compensation for the detected tracks , of sensors A and B, and return to S342; among them, the calculation expression of the estimation result with only angular and distance systematic errors is as follows: (13) In the formula, represents the estimated residual vector generated by the associated detection of sensors A and B.

[0029] S344B. End the algorithm and output the global minimum track correlation coefficient matrix .

[0030] S36. End the algorithm and output the global minimum track correlation coefficient matrix .

[0031] Therefore, the present invention adopts the above-mentioned alternative iterative robust track correlation method for realizing error correlation decoupling, adopts the GNN track correlation modeling method, and realizes the minimization of the correlation cost function or the maximization of the consistent correlation number on the premise of considering systematic errors (translational systematic errors or angular and distance systematic errors) or systematic residuals, and improves the track correlation optimization performance during the alternating iteration process of track correlation and systematic error estimation.

[0032] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that: they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An alternating iterative robust track association method for achieving error association decoupling, characterized in that: The following steps are involved: Step 1: In the scenario of multi-target flight, for the multiple false alarms of sensor A and sensor B, the local detection tracks of the two sensors are used to build a GNN track association model for the same target; Step 2: Under the condition of considering the influence of the system error and random noise of sensor A and sensor B, jointly estimate the system error and track association, wherein the system error includes translation system error or angle and distance system error; Step 3: Based on the maximum consistent association number optimization algorithm, the translation system error or the angle and distance system error in step 2 is optimized to complete the decoupling of the error association.

2. The alternating iterative robust track association method for achieving error association decoupling according to claim 1, characterized in that: The local detection tracks of sensor A and sensor B in step 1 are and , under the condition of multiple targets and multiple false alarms on both sides, the accuracy of the two sensors in associating the same target track The calculation expression is as follows: （1） In the formula, m is the observation space dimension, is a fixed constant, is the target airspace density, is the multiplication factor, 、 represents the false alarm target density observed by sensors A and B respectively, is the standard deviation of the average correlation noise.

3. The alternating iterative robust track association method for achieving error association decoupling according to claim 2, characterized in that: The expression for joint estimation of system error and track association in step 2 is as follows: （2） In the formula, is the system error, satisfying ; is the correlation coefficient matrix between the tracks of sensors A and B; is the association cost matrix, which satisfies the following formula: （3） In the formula, , is the detection probability of sensors A and B, is the likelihood function of two homologous tracks, , It means that there is no track element observed by the other party in the track set of sensors A and B, which satisfies the following formula: （4） （5）。 4. The alternating iterative robust track association method for achieving error association decoupling according to claim 3, characterized in that: The specific process of optimizing the translation system error or angle and distance system error in step 2 based on the maximum consistent association number optimization algorithm in step 3 is as follows: S31, MaxCAN algorithm initialization; input sensor A, sensor B local detection track , ; S32, through the GNN track association model formula (1) and the joint estimation algorithm formula (2), the track association coefficient matrix is ​​calculated and the global association cost matrix ; S33, introduce the LMedS system error estimator, and solve the error estimation result of the translation system only ; S34. Calculate the improved consistent correlation number ,like , then there is no mismatched element, execute the MinAC algorithm, and jump to S36; if , then there are elements with incorrect pairing, and proceed to the next step; S35. Judgment Is it reduced? If so, execute S36; if not, then , Perform error compensation and return to S32; S36, the algorithm ends and the global minimum track correlation coefficient matrix is ​​output .

5. The method for realizing error correlation decoupling by alternating iterative robust track correlation according to claim 4, characterized in that: Solve the error estimation results of the translation-only system in S33 The calculation expression is as follows: （6） In the formula, It is the error vector formed by subtracting the corresponding track elements of the two track sets of sensor A and sensor B, and med is the median operation.

6. The method for realizing error correlation decoupling by alternating iterative robust track correlation according to claim 5, characterized in that: Improved consistent correlation number in S34 The calculation process is as follows: S341A, calculate the correlation coefficient matrix as Total number of track element pairs in , the calculation expression is as follows: （7） S342A, let the correlation coefficient matrix The number of successful pairing of track elements under the condition of pairing errors is , then the consistent correlation number Satisfies the following formula: （8） From equation (8), we can conclude that when the threshold T of the false association detector is set to When , all track associations are correct, indicating that there are no mismatched elements. ; On the contrary, when the threshold T is set When , all track relationships are wrong. ; S343A, based on the conclusions drawn from S342A, make optimization adjustments and consistent correlation numbers Improved to the following expression: （9）。 7. The alternating iterative robust track association method for achieving error association decoupling according to claim 6, characterized in that: S35 detection track , The expression for error compensation is as follows: （10） （11） （12）。 8. The alternating iterative robust track association method for achieving error association decoupling according to claim 7, characterized in that: The specific implementation process of the MinAC algorithm in S34 is as follows: S341B, MinAC algorithm initialization; input sensor A, B local detection track information , , set the minimum association cost matrix The initial value is +∞, and the maximum number of iterations is N; S342B. Determine the relationship between the current iteration number n and the maximum iteration number N. If the current iteration number n > N, execute S344B; if the current iteration number n < N, calculate the track correlation coefficient matrix through the GNN track correlation model formula (1) and the joint estimation formula (2) and the global correlation cost matrix , and determine the relationship between the global correlation cost matrix and the minimum correlation cost matrix . If , then , , execute S343B; if , then execute S344; S343B, let the current number of iterations n = n + 1, it is known that the calculation , respectively calculate the estimated results of only the translation system error and the estimated results with only angle and distance systematic errors ; Realize the detection track of sensors A and B , Error compensation, return to S342; S344B, algorithm ends, output .

9. The alternating iterative robust track association method for achieving error association decoupling according to claim 8, characterized in that: Estimation results of only angle and distance systematic errors in S343B The calculation expression is as follows: （13） In the formula, Represents the estimated residual vector generated by the associated detection of sensors A and B.

Images