Interrupted track association method and system based on combination of dynamic model and pseudo measurement

Abstract

The invention belongs to the field of ballistic target tracking, and discloses an interrupted track association method and system based on combination of a dynamic model and pseudo measurement, which can more accurately estimate the target state even under the condition of data missing or batch skipping by combining the dynamic model and the pseudo measurement. According to the method, the abnormal condition in the measurement data can be effectively processed, the track interruption risk caused by data missing is reduced, and the stability of the system is improved. According to the method, pseudo measurement of polynomial function fitting is utilized, the original track can be updated more effectively, errors are reduced, and continuity and consistency of the track are improved. The method is suitable for track association in various targets and complex environments, and the association strategy can be dynamically adjusted according to actual conditions. According to the method, the track fuzzy similarity is calculated and the association cost matrix is constructed, so that a more scientific basis can be provided for track matching, and an association result is optimized.

Description

Technical Field

[0001] The invention belongs to the field of ballistic target tracking, and in particular relates to an interrupted track association method and system based on a combination of a dynamic model and pseudo-measurement. Background Art

[0002] In the process of mid-course tracking of ballistic missile group targets, due to various reasons such as the fast target flight speed, dense target composition, echo interference and low radar data rate, the measurement data detected by each channel of the phased array radar has various abnormal conditions such as data missing and data skipping. The occurrence of these abnormal conditions often leads to the interruption of the target track. From the data level, the occurrence of this phenomenon will lead to inconsistency of the batch number of the same target and the lack of continuity of the target track; from the command level, this phenomenon will directly affect the commander's on-the-spot decision-making and decision-making scheduling. The above problems can be uniformly summarized as the problem of interrupted track association.

[0003] The research on the problem of interrupted track association is currently divided into two directions: based on traditional tracking algorithms and based on neural network methods. In the study of traditional tracking algorithms, SW Yeom et al. adopted the Interacting Multiple Model (IMM) method to reversely predict the new starting track to the last update time of the old track, calculate the allocation cost of the new and old tracks based on the position information, and solve the allocation relationship between the new and old tracks with two-dimensional allocation. Du Jian, Mao Yifan et al. determined the fuzzy factor weight set, membership function and track fuzzy similarity based on position, velocity and acceleration information, and introduced fuzzy mathematics theory to complete the association of interrupted track segments. On this basis, Qi Lin et al. considered the various maneuvering possibilities of the target in the interruption interval, set a variety of possible motion models based on the idea of ​​multiple assumptions, and defined the final similarity as the maximum value of all similarities between the current track and the predicted state based on the old track according to different dynamic models. Long Weiyu et al. introduced uncertain variables to consider the influence of interference in actual measurement, and used uncertain variables to correct process noise and measurement noise in track association to ensure a more accurate reflection of the actual measurement situation. Wang Yuan, Han Wei and others considered the Doppler blind zone information in the radar tracking problem and incorporated it into the cost matrix of the two-dimensional allocation problem. Liu Lijun and others used the dynamic model of space targets and track extrapolation technology to effectively solve the track interruption problem caused by repeated target changes in radar single-channel measurement.

[0004] During the mid-course tracking of ballistic targets, due to various reasons such as the fast target flight speed, dense target composition, echo interference and low radar data rate, the measurement data detected by each channel of the phased array radar have various abnormal conditions such as data missing and data skipping. The occurrence of these abnormal conditions often leads to the interruption of the target track. Summary of the invention

[0005] The purpose of the present invention is to overcome the problem of target track interruption caused by abnormal conditions such as missing data and data skipping in measurement data, and to provide an interrupted track association method and system based on a combination of dynamic model and pseudo-measurement.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides an interrupted track association method based on a combination of a dynamic model and pseudo-measurement, comprising: Obtain the original track and current track of the target, perform filtering estimation on the current track, obtain the motion characteristics of the current track, generate the state estimation vector sequence of the original track according to the original track of the target, and generate the state estimation vector sequence of the current track according to the current track and motion characteristics of the target; Associating the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs; Acquire status data of the original track, and identify the track interruption interval of the original track according to the status data of the original track; Fitting the state data of the original track to obtain a polynomial function, fitting the polynomial function to a set of candidate track matching pairs to obtain a pseudo measurement of the interruption interval; Based on the pseudo-measurement, the track interruption interval of the original track is updated, the track fuzzy similarity between the original track and the current track is calculated, and the associated cost matrix is ​​obtained; A two-dimensional allocation model is established according to the association cost matrix, and the association result between the original track and the current track is obtained according to the two-dimensional allocation model.

[0007] A further improvement of the present invention is that the original track and the current track of the target are obtained, and the current track is filtered and estimated. The specific method for obtaining the motion characteristics of the current track is as follows: The current track is processed by the transfer equation and the measurement equation to obtain the motion characteristics of the current track, where the transfer equation is:

[0008] The measurement equation is:

[0009] in, for k The state vector at the moment, is a zero-mean Gaussian white noise, for k The nonlinear state function of the system at time for The Jacobi matrix of for k The moment measurement vector, is the nonlinear measurement function of the system at time k, is Gaussian white noise with zero mean.

[0010] A further improvement of the present invention is that, according to the original track of the target, a method for generating a state estimation vector sequence of the original track is as follows:

[0011] According to the current track and motion characteristics of the target, the method for generating the state estimation vector sequence of the current track is as follows:

[0012] in, is the state estimation vector set of the original track, with a total of The original track, For the i The starting time of the original track, For the i The end time of the original track, is the state estimation vector set of the current track, with a total of Current tracks, For the i The start time of the current track, For the i The end time of the current track, .

[0013] A further improvement of the present invention is that the specific method of associating the state estimation vector sequence of the current track with the state estimation vector sequence of the original track is as follows:

[0014] in, is the state estimation vector set of the original track exist The state estimation vector at time , is the state estimation vector set of the original track exist The state estimation vector at time The corresponding covariance matrix is, is the state estimation vector set of the current track exist The state estimation vector at time , is the state estimation vector set of the current track exist The state estimation vector at time The corresponding covariance matrix, state vector , is the upper quantile of the chi-square test with 3 degrees of freedom, To find the trace of the covariance matrix, is the maximum speed of the target; If the above formula is satisfied, the association is completed. Otherwise, the state estimation vector set of the original track is and the state estimation vector set of the current track Does not belong to the set of potential associations.

[0015] A further improvement of the present invention is that a least squares algorithm is used when fitting the state data of the original track to obtain a polynomial function.

[0016] A further improvement of the present invention is that the state data of the original track is fitted to obtain a polynomial function. The specific method is as follows: The state estimation vector set for the original track , extracted from Time to end time The state estimation vector is m Group, along the radar ENU coordinate system In the three-axis direction, the state estimation vector set of the original track The state estimation vector is split into three sets of position data that change in time sequence, as shown below:

[0017] For the data set given by the above formula, in the radar ENU coordinate system x For example, assuming that there is n Polynomial ; In the least squares fitting, the polynomial coefficients obtained according to time k are required to make the residual sum of squares between the time-varying position prediction data calculated by the polynomial and the actual data Minimum, among which n Polynomial and residual sum of squares The mathematical expression is as follows:

[0018]

[0019] According to the necessary conditions for finding the extreme value of a multivariate function, Find the partial derivatives and set each equation equal to 0, as shown below:

[0020] Put the radar ENU coordinate system xSubstitute the data in the axial direction into the above formula and solve for the parameters ; Get the original track in the radar ENU coordinate system x Axis direction corresponding to n Polynomial ; Similarly, yes y Axis and z The above operations are performed on the axial direction data to obtain the corresponding n Order polynomial.

[0021] A further improvement of the present invention is that a specific method of fitting a polynomial function to a set of candidate track matching pairs to obtain pseudo-measurements of the interruption interval is as follows: Traverse the candidate track matching pair set The state estimation vector set of the original track that meets the association requirements and the state estimation vector set of the current track ; Estimate the vector set from the state of the current track Extract the time stamps in sequence , combined with the state estimation vector set of the original track The polynomial fitted by historical information , and predict the state estimation vector set of the original track in turn from Time has come The three-axis position information that changes in the radar ENU coordinate system in time series is combined with the measurement model to convert and generate the state estimation vector set of the original track The state estimation vector set to the current track Pseudo-measurement of track interruption interval .

[0022] A further improvement of the present invention is that the track interruption interval of the original track is updated based on the pseudo measurement, and the track fuzzy similarity between the original track and the current track is calculated. The specific method of obtaining the association cost matrix is ​​as follows: According to the pseudo-measurement combined with the motion characteristics of the interruption interval, the state estimation vector set of the original track combined with the pseudo-measurement correction is obtained. The state sequence in the interruption interval, thus obtaining the state estimation vector set of the original track exist The time is updated based on pseudo measurement The state estimation vector at time , and the corresponding covariance matrix ; Determine the fuzzy factor set, fuzzy factor weight set and membership function:

[0023]

[0024]

[0025] in, is the fuzzy factor set, for Middle i A fuzzy factor, is the position ambiguity factor, is the speed ambiguity factor; is the fuzzy factor weight set, For the i Fuzzy Factor The corresponding weight size, For the i The membership function of the fuzzy factors is For the i Fuzzy Factor The corresponding spread, For the i Fuzzy Factor The corresponding adjustment degree; Based on the Kalman filter theory, i Fuzzy Factor The calculation method is as follows:

[0026] Use the corresponding elements of the filtered covariance matrix as the first i Fuzzy Factor The extent of i Fuzzy Factor The corresponding spread The calculation is as follows:

[0027] The calculated i Fuzzy Factor and i Fuzzy Factor The corresponding spread Substituting into the above formula, we can get the membership degree, and then calculate the track fuzzy similarity between the original track and the current track as follows:

[0028] Match the candidate track pairs The above calculation is performed on the original track and the current track in the image to obtain the fuzzy similarity matrix between the original track and the current track, and then the fuzzy similarity matrix is ​​obtained. The reciprocal of , forming the association cost matrix as follows: .

[0029] A further improvement of the present invention is that a two-dimensional allocation model is established according to the association cost matrix. The specific method of obtaining the association result of the original track and the current track according to the two-dimensional allocation model is as follows: The two-dimensional allocation model is as follows:

[0030] in, for The corresponding binary variable, 0 means no choice, 1 means choice.

[0031] The optimal solution of the two-dimensional allocation model is solved by the auction algorithm, and the track association result between the original track and the current track is obtained.

[0032] In a second aspect, the present invention is based on an interrupted track association system combining a dynamic model and pseudo-measurement, comprising: The estimation vector sequence generation module obtains the original track and current track of the target, performs filtering estimation on the current track, obtains the motion characteristics of the current track, generates the state estimation vector sequence of the original track according to the original track of the target, and generates the state estimation vector sequence of the current track according to the current track and motion characteristics of the target; A vector sequence association module is used to associate the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs; An interruption interval identification module is used to obtain the status data of the original track, and identify the track interruption interval of the original track according to the status data of the original track; A pseudo-measurement acquisition module is used to fit the state data of the original track to obtain a polynomial function, and fit the polynomial function to a set of candidate track matching pairs to obtain a pseudo-measurement of the interruption interval; The associated cost matrix acquisition module is used to update the track interruption interval of the original track based on the pseudo measurement, calculate the track fuzzy similarity between the original track and the current track, and obtain the associated cost matrix; The association result acquisition module is used to establish a two-dimensional allocation model according to the association cost matrix, and obtain the association result of the original track and the current track according to the two-dimensional allocation model.

[0033] Compared with the prior art, the present invention has the following beneficial effects: By combining dynamic models and pseudo-measurements, the present invention can estimate the target state more accurately, even in the presence of missing data or batch skipping. The method can effectively handle abnormal conditions in the measurement data, reduce the risk of track interruption due to missing data, and improve the stability of the system. The present invention utilizes pseudo-measurements fitted by polynomial functions to more effectively update the original track, reduce errors, and improve the continuity and consistency of the track. The present invention is suitable for track association in a variety of targets and complex environments, and can dynamically adjust the association strategy according to actual conditions. By calculating the fuzzy similarity of the track and constructing an association cost matrix, the present invention can provide a more scientific basis for track matching, thereby optimizing the association results. In summary, the method can achieve rapid processing in a real-time system, adapt to dynamically changing target tracks, and facilitate real-time monitoring and decision-making. It has better estimation performance in the track interruption interval and higher track association accuracy and robustness in terms of track association performance. It also has great application potential and value in many military fields, such as ballistic target tracking and interception. It can provide effective track correlation information for commanders' on-the-spot decision-making and decision-making scheduling in response to problems such as track interruption and track skipping in the measured data of phased array radar. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of the present invention; Figure 2 This is a schematic diagram of track interruption; Figure 3 and Figure 4 To simulate the experimental scenario 1, the track interruption interval estimation performance under the same interruption start time and different process noises; Figure 5 and Figure 6 To simulate the experimental scenario 2, the track interruption interval estimation performance under different interruption start times and different process noises; Figure 7 and Figure 8 To simulate the experimental scenario 3, track correlation performance under different interruption durations and different process noises; Fig. 9 This is a schematic diagram of track splicing results; Fig.10 It is a system schematic diagram of the present invention. DETAILED DESCRIPTION

[0035] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0036] Embodiment 1: See also Figure 1 , based on the interrupted track association method combining dynamic model and pseudo measurement, including: S1, obtain the original track and current track of the target, filter and estimate the current track to obtain the motion characteristics of the current track, generate the state estimation vector sequence of the original track according to the original track of the target, and generate the state estimation vector sequence of the current track according to the current track and motion characteristics of the target.

[0037] S2, associate the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs.

[0038] S3, obtaining status data of the original track, and identifying the track interruption interval of the original track according to the status data of the original track.

[0039] S4, fitting the state data of the original track to obtain a polynomial function, fitting the polynomial function to a set of candidate track matching pairs, and obtaining a pseudo measurement of the interruption interval.

[0040] S5, based on the pseudo measurement, the track interruption interval of the original track is updated, the track fuzzy similarity between the original track and the current track is calculated, and the associated cost matrix is ​​obtained.

[0041] S6, establishing a two-dimensional allocation model according to the association cost matrix, and obtaining the association result of the original track and the current track according to the two-dimensional allocation model.

[0042] Embodiment 2: See also Fig.10 , based on the interrupted track correlation system combining dynamic model and pseudo measurement, including: The estimation vector sequence generation module obtains the original track and current track of the target, performs filtering estimation on the current track, obtains the motion characteristics of the current track, generates the state estimation vector sequence of the original track according to the original track of the target, and generates the state estimation vector sequence of the current track according to the current track and motion characteristics of the target; A vector sequence association module is used to associate the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs; An interruption interval identification module is used to obtain the status data of the original track, and identify the track interruption interval of the original track according to the status data of the original track; A pseudo-measurement acquisition module is used to fit the state data of the original track to obtain a polynomial function, and fit the polynomial function to a set of candidate track matching pairs to obtain a pseudo-measurement of the interruption interval; The associated cost matrix acquisition module is used to update the track interruption interval of the original track based on the pseudo measurement, calculate the track fuzzy similarity between the original track and the current track, and obtain the associated cost matrix; The association result acquisition module is used to establish a two-dimensional allocation model according to the association cost matrix, and obtain the association result of the original track and the current track according to the two-dimensional allocation model.

[0043] Embodiment 3: The specific method of obtaining the original track and current track of the target, filtering and estimating the current track, and obtaining the motion characteristics of the current track is as follows: The current track is processed by the transfer equation and the measurement equation to obtain the motion characteristics of the current track, where the transfer equation is: Formula 1 The measurement equation is: Formula 2 in, for k The state vector at the moment, is a zero-mean Gaussian white noise, for k The nonlinear state function of the system at time for The Jacobi matrix of for k The moment measurement vector, is the nonlinear measurement function of the system at time k, is Gaussian white noise with zero mean.

[0044] According to the original track of the target, the method of generating the state estimation vector sequence of the original track is as follows: Formula 3 According to the current track and motion characteristics of the target, the method for generating the state estimation vector sequence of the current track is as follows: Formula 4 in, is the state estimation vector set of the original track, with a total of The original track, For the i The starting time of the original track, For the i The end time of the original track, is the state estimation vector set of the current track, with a total of Current tracks, For the i The start time of the current track, For the i The end time of the current track, .

[0045] Embodiment 4: The specific method of associating the state estimation vector sequence of the current track with the state estimation vector sequence of the original track is as follows: Formula 5 in, is the state estimation vector set of the original track exist The state estimation vector at time , is the state estimation vector set of the original track exist The state estimation vector at time The corresponding covariance matrix is, is the state estimation vector set of the current track exist The state estimation vector at time , is the state estimation vector set of the current track exist The state estimation vector at time The corresponding covariance matrix, state vector , is the upper quantile of the chi-square test with 3 degrees of freedom, To find the trace of the covariance matrix, is the maximum speed of the target; If the above formula is satisfied, the association is completed. Otherwise, the state estimation vector set of the original track is and the state estimation vector set of the current track It does not belong to the potential association set. After the rough association is completed, a candidate matching track pair set is formed. .

[0046] Embodiment 5: When fitting the state data of the original track to obtain a polynomial function, the least squares algorithm is used. The specific method is as follows: The state estimation vector set for the original track , extracted from Time to end time The state estimation vector is m Group, along the radar ENU coordinate system In the three-axis direction, the state estimation vector set of the original track The state estimation vector is split into three sets of position data that change in time sequence, as shown below: Formula 6 For the data set given by formula 6, in the radar ENU coordinate system x For example, assuming that there is n Polynomial In the least squares fitting, the polynomial coefficients obtained according to time k are required to make the residual sum of squares between the time-varying position prediction data calculated by the polynomial and the actual data Minimum. n Polynomial and residual sum of squares The mathematical expression is as follows: Formula 7 Formula 8 According to the necessary conditions for finding the extreme value of a multivariate function, Find the partial derivatives and set each equation equal to 0, as shown below: Formula 9 Put the above radar ENU coordinate system x Substitute the data in the axial direction into formula 9 to solve the parameters . Thus, the original track can be obtained in the radar ENU coordinate system x Axis direction corresponding to n Polynomial Similarly, y Axis and z Perform the above operations on the axial direction data to get the corresponding n Order polynomial.

[0047] Embodiment 6: The specific method of fitting the polynomial function to the candidate track matching pair set and obtaining the pseudo measurement of the interruption interval is as follows: Traverse the candidate track matching pair set The state estimation vector set of the original track that meets the association requirements and the state estimation vector set of the current track ; Estimate the vector set from the state of the current track Extract the time stamps in sequence , combined with the state estimation vector set of the original track The polynomial fitted by historical information , and predict the state estimation vector set of the original track in turn from Time has come The three-axis position information that changes in the radar ENU coordinate system in time series is combined with the measurement model to convert and generate the state estimation vector set of the original track The state estimation vector set to the current track Pseudo-measurement of track interruption interval .

[0048] Embodiment 7: The specific method of updating the track interruption interval of the original track based on pseudo measurement, calculating the track fuzzy similarity between the original track and the current track, and obtaining the association cost matrix is ​​as follows: (1) Target status update during interruption According to the interruption interval pseudo-measurement obtained in step 3, and the motion equation and measurement equation shown in formula 1 and formula 2, the pseudo-measurement correction can be estimated. The state sequence during the interruption period. for exist The time is updated based on pseudo measurement The state estimation vector at time , is the corresponding covariance matrix.

[0049] The problem of ballistic target interruption track association is a typical uncertainty data combing problem. The data processing method based on fuzzy theory is an effective method to deal with this type of problem. Before calculating the track fuzzy similarity, it is necessary to first determine the fuzzy factor set, fuzzy factor weight set and membership function.

[0050] Define the fuzzy factor set , , for Middle i A fuzzy factor, is the position ambiguity factor, is the speed fuzzy factor; define the fuzzy factor weight set , , for The corresponding weight size, the larger the weight, the greater the influence of the fuzzy factor corresponding to the weight on the track association, and the sum of the weight values ​​is 1. Definition For the i The membership function of the fuzzy factors is a normal membership function. The specific expression is as follows: Formula 10 In the formula, for The corresponding spread, for The corresponding adjustment degree.

[0051] Based on the Kalman filter theory, The calculation method is as follows: Formula 11 Use the corresponding elements of the filtered covariance matrix as The extension of The calculation is as follows: Formula 12 The calculated and Substituting into formula 10, we can get the membership degree, and then calculate the track fuzzy similarity between the original track and the current track. as follows: Formula 13 right The above calculation is performed on the original track and the current track in the image, and the fuzzy similarity matrix between the original track and the current track can be obtained. The reciprocal of , forming the association cost matrix as follows: Formula 14 Embodiment 8: A two-dimensional allocation model is established according to the association cost matrix. The specific method of obtaining the association result between the original track and the current track according to the two-dimensional allocation model is as follows: The two-dimensional allocation model is as follows: Formula 15 In the formula, for The corresponding binary variable, 0 means no choice, 1 means choice.

[0052] By solving the optimal solution of the two-dimensional assignment problem through the auction algorithm, the track association result between the original track and the current track can be obtained.

[0053] Embodiment 9: In the simulation experiment, the average root mean square error ARMSE and the probability of correct association (The Probability of Correct Association, P CA ) are used to evaluate the estimation performance of track interruption interval and track association performance.

[0054] Solution k The ARMSE at this moment needs to be carried out in two steps. First, determine each group of associated and The RMSE of the matched pairs is then averaged over the total number of new raw track association pairs at that moment. k The ARMSE at the moment is calculated as follows: (16) In the formula, For the nThe target filter state in the group association pair, For the n The target true state in the group matching pair, M is the number of Monte Carlo simulations, for k The number of associated logarithms of the new original track at time.

[0055] Track segment association accuracy P CA Original track With current track The ratio of the total number of correct associations to the product of the number of targets and the number of simulations is calculated as follows: (17) In the formula, is the target number, For M The total number of correct associations in the simulation experiment.

[0056] In order to verify the estimation performance and track association accuracy of the interrupted track segment association method based on the combination of dynamic model and pseudo-measurement proposed in the present invention in the track interruption interval, this section distinguishes different variables such as the interruption duration and whether the interruption start time is synchronized, verifies the performance of the method, and compares it with two algorithms. Algorithm 1 is the traditional track segment association algorithm (TTSA) that only relies on prediction information, and algorithm 2 is a track association algorithm that predicts the targetstate based on fitting historical data but does not combine the dynamic model (PBHA). Two different process noises are set in the simulation scenario, the process noise ω a The standard deviation is (5 m ,5 m ,5 m ,5 m / s ,5 m / s ,5 m / s ), process noise ω b The standard deviation is ω a The number of Monte Carlo simulations for the algorithm's estimated performance in the track interruption interval is 100, and the number of Monte Carlo simulations for the associated performance in the track interruption interval is 1000. Taking track synchronization interruption as an example, the track interruption diagram can be found in Figure 2 .

[0057] See also Figure 3 and Figure 4 As the interruption occurs and continues, the ARMSE estimated by TTSA increases sharply. After the interruption ends, due to the introduction of measurement, the ARMSE gradually decreases. However, the method proposed in this invention can still maintain stable tracking performance in the track interruption interval, and the estimated ARMSE is basically consistent with that before the tracking interruption. PBHA completely relies on fitting historical data to generate the track interruption interval state, and the algorithm performance is between the two. See Figure 5 and Figure 6 , once a track is interrupted, the ARMSE estimated by TTSA will begin to increase. As the interruption ends and new measurements are introduced, the ARMSE begins to gradually decrease, while the estimation performance of the method proposed in this invention can continue to remain stable. As the process noise increases, the estimation errors of different algorithms will increase. This proves that when the track is interrupted asynchronously, the algorithm estimation performance can also obtain the same conclusion as when the track is interrupted synchronously. See Figure 7 and Figure 8 The results show that the proposed method is always better than TTSA and PBHA, and has a higher track association accuracy. Taking track synchronization interruption as an example, after the interrupted track is correctly associated, the track interruption interval is spliced. Fig. 9 In summary, compared with the traditional track segment association method which only relies on prediction information to estimate the current track interruption interval status, the present invention has better estimation performance in the track interruption interval and higher track association accuracy and robustness in terms of track association performance.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. The interrupted track association method based on the combination of dynamic model and pseudo measurement is characterized by: include: Obtain the original track and current track of the target, perform filtering estimation on the current track, obtain the motion characteristics of the current track, generate the state estimation vector sequence of the original track according to the original track of the target, and generate the state estimation vector sequence of the current track according to the current track and motion characteristics of the target; Associating the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs; Acquire status data of the original track, and identify the track interruption interval of the original track according to the status data of the original track; Fitting the state data of the original track to obtain a polynomial function, fitting the polynomial function to a set of candidate track matching pairs to obtain a pseudo measurement of the interruption interval; Based on the pseudo-measurement, the track interruption interval of the original track is updated, the track fuzzy similarity between the original track and the current track is calculated, and the associated cost matrix is ​​obtained; A two-dimensional allocation model is established according to the association cost matrix, and the association result between the original track and the current track is obtained according to the two-dimensional allocation model.

2. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: The specific method of obtaining the original track and current track of the target, filtering and estimating the current track, and obtaining the motion characteristics of the current track is as follows: The current track is processed by the transfer equation and the measurement equation to obtain the motion characteristics of the current track, where the transfer equation is: The measurement equation is: in, for k The state vector at the moment, is a zero-mean Gaussian white noise, for k The nonlinear state function of the system at time for The Jacobi matrix of for k The moment measurement vector, is the nonlinear measurement function of the system at time k, is Gaussian white noise with zero mean.

3. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 2 is characterized in that: According to the original track of the target, the method of generating the state estimation vector sequence of the original track is as follows: According to the current track and motion characteristics of the target, the method for generating the state estimation vector sequence of the current track is as follows: in, is the state estimation vector set of the original track, with a total of The original track, For the i The starting time of the original track, For the i The end time of the original track, is the state estimation vector set of the current track, with a total of Current tracks, For the i The start time of the current track, For the i The end time of the current track, .

4. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: The specific method of associating the state estimation vector sequence of the current track with the state estimation vector sequence of the original track is as follows: in, is the state estimation vector set of the original track exist The state estimation vector at time , is the state estimation vector set of the original track exist The state estimation vector at time The corresponding covariance matrix is, is the state estimation vector set of the current track exist The state estimation vector at time , is the state estimation vector set of the current track exist The state estimation vector at time The corresponding covariance matrix, state vector , is the upper quantile of the chi-square test with 3 degrees of freedom, To find the trace of the covariance matrix, is the maximum speed of the target; If the above formula is satisfied, the association is completed. Otherwise, the state estimation vector set of the original track is and the state estimation vector set of the current track Does not belong to the set of potential associations.

5. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: The least squares algorithm is used to fit the state data of the original track to obtain the polynomial function.

6. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 or 5, characterized in that: The specific method of fitting the state data of the original track to obtain the polynomial function is as follows: The state estimation vector set for the original track , extracted from Time to end time The state estimation vector is m Group, along the radar ENU coordinate system In the three-axis direction, the state estimation vector set of the original track The state estimation vector is split into three sets of position data that change in time sequence, as shown below: For the data set given by the above formula, in the radar ENU coordinate system x For example, assuming that there is data in the axis direction n Polynomial ; In the least squares fitting, the polynomial coefficients obtained according to time k are required to make the residual sum of squares between the time-varying position prediction data calculated by the polynomial and the actual data Minimum, among which n Polynomial and residual sum of squares The mathematical expression is as follows: According to the necessary conditions for finding the extreme value of a multivariate function, Find the partial derivatives and set each equation equal to 0, as shown below: Put the radar ENU coordinate system x Substitute the data in the axial direction into the above formula and solve for the parameters ; Get the original track in the radar ENU coordinate system x Axis direction corresponding to n Polynomial ; Similarly, yes y Axis and z The above operations are performed on the axial direction data to obtain the corresponding n Order polynomial.

7. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: The specific method of fitting the polynomial function to the candidate track matching pair set and obtaining the pseudo measurement of the interruption interval is as follows: Traverse the candidate track matching pair set The state estimation vector set of the original track that meets the association requirements and the state estimation vector set of the current track ; Estimate the vector set from the state of the current track Extract the time stamps in sequence , combined with the state estimation vector set of the original track The polynomial fitted by historical information , and predict the state estimation vector set of the original track in turn from Time has come The three-axis position information that changes in the radar ENU coordinate system in time series is combined with the measurement model to convert and generate the state estimation vector set of the original track The state estimation vector set to the current track Pseudo-measurement of track interruption interval .

8. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: The specific method of updating the track interruption interval of the original track based on pseudo measurement, calculating the track fuzzy similarity between the original track and the current track, and obtaining the association cost matrix is ​​as follows: According to the pseudo-measurement of the interruption interval combined with the motion characteristics, the state estimation vector set of the original track combined with the pseudo-measurement correction is obtained. The state sequence in the interruption interval, thus obtaining the state estimation vector set of the original track exist The time is updated based on pseudo measurement The state estimation vector at time , and the corresponding covariance matrix ; Determine the fuzzy factor set, fuzzy factor weight set and membership function: in, is the fuzzy factor set, for Middle i A fuzzy factor, is the position ambiguity factor, is the speed ambiguity factor; is the fuzzy factor weight set, For the i Fuzzy Factor The corresponding weight size, For the i The membership function of the fuzzy factors is For the i Fuzzy Factor The corresponding spread, For the i Fuzzy Factor The corresponding adjustment degree; Based on the Kalman filter theory, i Fuzzy Factor The calculation method is as follows: Use the corresponding elements of the filtered covariance matrix as the first i Fuzzy Factor The extent of i Fuzzy Factor The corresponding spread The calculation is as follows: The calculated i Fuzzy Factor and i Fuzzy Factor The corresponding spread Substituting into the above formula, we can get the membership degree, and then calculate the track fuzzy similarity between the original track and the current track as follows: Matching pairs of candidate tracks The above calculation is performed on the original track and the current track in the image to obtain the fuzzy similarity matrix between the original track and the current track, and then the fuzzy similarity matrix is ​​obtained. The reciprocal of , forming the association cost matrix as follows: 。 9. The interrupted track association method based on the combination of dynamic model and pseudo measurement according to claim 1 is characterized in that: A two-dimensional allocation model is established according to the association cost matrix. The specific method of obtaining the association result between the original track and the current track according to the two-dimensional allocation model is as follows: The two-dimensional allocation model is as follows: in, for The corresponding binary variable, 0 means no choice, 1 means choice; The optimal solution of the two-dimensional allocation model is solved by the auction algorithm, and the track association result between the original track and the current track is obtained.

10. The interrupted track association system based on the combination of dynamic model and pseudo measurement is characterized by: include: The estimation vector sequence generation module obtains the original track and current track of the target, performs filtering estimation on the current track, obtains the motion characteristics of the current track, generates the state estimation vector sequence of the original track according to the original track of the target, and generates the state estimation vector sequence of the current track according to the current track and motion characteristics of the target; A vector sequence association module is used to associate the state estimation vector sequence of the current track with the state estimation vector sequence of the original track to obtain a set of candidate track matching pairs; An interruption interval identification module is used to obtain the status data of the original track, and identify the track interruption interval of the original track according to the status data of the original track; A pseudo-measurement acquisition module is used to fit the state data of the original track to obtain a polynomial function, and fit the polynomial function to a set of candidate track matching pairs to obtain a pseudo-measurement of the interruption interval; The associated cost matrix acquisition module is used to update the track interruption interval of the original track based on the pseudo measurement, calculate the track fuzzy similarity between the original track and the current track, and obtain the associated cost matrix; The association result acquisition module is used to establish a two-dimensional allocation model according to the association cost matrix, and obtain the association result of the original track and the current track according to the two-dimensional allocation model.

Images