A dynamic correction passive tracking method based on reference fragment tracks

Abstract

This invention provides a dynamic correction passive tracking method based on reference debris tracks, comprising: Step 1, fitting the reference debris tracks received by the passive tracking system to generate an interval track fitting model; Step 2, obtaining the measurement deviation based on the fitting model and the original sensor measurement sequence of the passive tracking system within the interval; Step 3, correcting the real-time input original measurements using the measurement deviation, and then substituting the corrected measurements into the unscented Kalman filter iterative process to complete forward filtering tracking and backward smoothing optimization, outputting a smoothed track. This invention effectively utilizes reference debris track information to dynamically correct the measurement deviation in the passive tracking system, resulting in stronger reliability; simultaneously, using the corrected measurements for target tracking significantly reduces the root mean square error, effectively improving track accuracy.

Description

Technical Field

[0001] This invention relates to the field of target tracking, and in particular to a dynamic correction passive tracking method based on reference debris tracks. Background Technology

[0002] Passive sensors are gaining increasing importance in target detection and tracking due to their advantages such as good concealment and strong anti-interference capabilities. Passive tracking technology utilizes passive sensors for long-term monitoring and continuous tracking of targets, improving the quality of target situational awareness. The tracking filtering process in this technology is characterized by strong nonlinearity and weak observability, making the tracking effect highly sensitive to sensor measurement errors. Therefore, eliminating or reducing measurement errors plays a crucial role in improving tracking performance. Measurement errors typically include two components: random error and fixed bias. Due to differences in the operating methods, ranges, and durations of various sensors, the target tracks observed by different sensors exhibit significant fragmentation characteristics, which are collectively referred to here as reference fragment tracks.

[0003] Random errors are typically handled using filtering algorithms. Different noise models, such as Gaussian white noise and colored noise, are used to mathematically model the errors. Then, filtering algorithms are used to reduce the impact of random errors on the tracking results. For example, for Gaussian noise models, there are extended Kalman filters and unscented Kalman filters, while for non-Gaussian noise, there are particle filters and Gaussian mixture filters.

[0004] For fixed biases, calibration tables are typically created after multiple experiments using simulated signal sources or cooperative targets. These tables are then used to eliminate or correct the fixed biases before tracking filtering, thereby improving tracking accuracy. However, several problems remain in implementation. First, due to the non-uniformity and instability of the signal transmission path, the fixed biases in actual signal measurements will vary. Second, direction finding correction based on a fixed signal source cannot cover all targets in the frequency and spatial domains. Third, as the sensor ages, component aging and drift can cause changes in the fixed bias, making periodic calibration insufficient to meet usage requirements.

[0005] In modern joint operations environments, continuous target tracking typically relies on joint detection using multiple sensors, including radar, television, infrared, laser, and electronic support measures (ESM). Due to operational requirements such as stealth, or differences in the operating methods, ranges, and durations of various sensors, target tracks observed by these sensors exhibit significant fragmentation characteristics; these are collectively referred to here as reference fragment tracks. In the field of passive tracking, reference fragment tracks are usually used for target guidance and multi-source verification; however, the technology for dynamic correction of passive sensors based on reference fragment tracks has not yet been discussed. Summary of the Invention

[0006] To address the problems existing in the prior art, a dynamic correction passive tracking method based on reference debris tracks is provided. This method aims to fully utilize the reference debris tracks of targets in the joint operations environment and system to dynamically correct the measurements in the passive tracking system. The debris tracks are fitted into continuous tracks and used as a reference to dynamically correct fixed sensor deviations. The corrected measurements are then used to complete the tracking filtering process to effectively improve track quality and solve the problem of poor tracking accuracy caused by sensor measurement deviations.

[0007] The technical solution adopted in this invention is as follows: A dynamic correction passive tracking method based on reference debris tracks, comprising:

[0008] Step 1: Fit the reference debris tracks received by the passive tracking system to generate an interval track fitting model;

[0009] Step 2: Based on the fitting model and the passive tracking system, obtain the measurement deviation from the original sensor measurement sequence within this interval;

[0010] Step 3: Correct the original measurements input in real time using measurement deviation, and then substitute the corrected measurements into the unscented Kalman filter iterative process to complete forward filtering tracking and backward smoothing optimization, and output the smoothed track.

[0011] Furthermore, in step 1, the adhesion fitting specifically involves: sorting the fragmented tracks according to time, adhering them together into a continuous track, and using polynomial fitting to obtain a fitting model for the continuous track.

[0012] Furthermore, step 2 includes the following specific sub-processes:

[0013] Step 2.1: Calculate the time range of the continuous tracks after adhesion in Step 2, and extract the original measurement sequence of the passive tracking system and the corresponding time point sequence within this time range;

[0014] Step 2.2: Use the fitting model to back-calculate the reference measurement sequence corresponding to the time point sequence;

[0015] Step 2.3: Calculate the measurement deviation based on the reference measurement sequence and the original measurement sequence.

[0016] Furthermore, in step 2.3, the elements in the reference measurement sequence and the original measurement sequence are subtracted one by one to obtain the error sequence, and the average value of all elements in the error sequence is used as the measurement deviation.

[0017] Furthermore, in the unscented Kalman filtering iteration process in step 3, the target state equation and the sensor measurement equation are established based on the components contained in the target state vector and the original sensor measurements.

[0018] Furthermore, in step 3, the forward filtering tracking process is as follows: the new original input measurement is corrected by the measurement deviation, and then the corrected measurement is used in the unscented Kalman filter iterative process to complete the forward filtering tracking of the target and output the tracking track.

[0019] Furthermore, the specific steps of the forward filtering tracking are as follows:

[0020] Step 3.1: Input the new original measurement, and use the measurement deviation to perform deviation correction to obtain the corrected measurement;

[0021] Step 3.2: Obtain Sigma sampling points and determine the weight of each Sigma sampling point;

[0022] Step 3.3: Substitute the Sigma sampling points into the target state equation to obtain the state prediction value, and then substitute the state prediction value into the measurement equation to obtain the observation prediction value.

[0023] Step 3.4: Weight the state prediction values ​​and observation prediction values ​​according to their weights to obtain the mean and covariance;

[0024] Step 3.5: Calculate the Kalman gain, and update the state based on the Kalman gain, mean, and covariance, then output the filtered tracking results.

[0025] Furthermore, in step 3, the specific process of backward smoothing optimization is as follows: the latest track output by forward filtering is used to perform backward smoothing on the track of the previous moment, and the smoothed track is output.

[0026] Compared with existing technologies, the beneficial effects of adopting the above technical solution are as follows:

[0027] (1) Unlike traditional deviation calibration techniques, this invention effectively utilizes reference debris track information to dynamically correct measurement deviations in passive tracking systems, resulting in greater reliability.

[0028] (2) Compared with the situation where the measurement deviation cannot be corrected, the present invention uses the corrected measurement for target tracking, and the root mean square error is significantly reduced, which effectively improves the accuracy of the track. Attached Figure Description

[0029] Figure 1 This is a flowchart of the passive tracking process for dynamic correction based on reference debris tracks in this invention.

[0030] Figure 2 This is a schematic diagram of a passive tracking system performing deviation calculation after receiving a reference track in one embodiment of the present invention.

[0031] Figure 3This is a comparison chart of the flight track tracked using uncorrected original measurements and corrected measurements in one embodiment of the present invention.

[0032] Figure 4 for Figure 3 A magnified view of a portion of the image.

[0033] Figure 5 This is a comparison chart of track errors when using uncorrected original measurements and corrected measurements in one embodiment of the present invention. Detailed Implementation

[0034] The embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar modules or modules having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. Rather, the embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0035] Before introducing the dynamic correction passive tracking method based on reference debris tracks proposed in this embodiment, it should be noted that, due to the differences in the operating mode, operating range and operating time of different sensors, the target tracks observed by various sensors exhibit obvious fragmentation characteristics, which are collectively referred to as reference debris tracks in this embodiment.

[0036] like Figure 1 As shown, this embodiment proposes a dynamic correction passive tracking method based on reference debris tracks. It effectively utilizes reference debris track information to dynamically correct measurement deviations in the passive tracking system, resulting in higher reliability. Simultaneously, using the corrected measurements for target tracking significantly reduces the root mean square error, effectively improving track accuracy. The specific steps are as follows:

[0037] Step 1: Fit the reference debris tracks received by the passive tracking system to generate an interval track fitting model;

[0038] Step 2: Based on the fitting model and the passive tracking system, obtain the measurement deviation from the original sensor measurement sequence within this interval;

[0039] Step 3: Correct the original measurements input in real time using measurement deviation, and then substitute the corrected measurements into the unscented Kalman filter iterative process to complete forward filtering tracking and backward smoothing optimization, and output the smoothed track.

[0040] Specifically, the above process can be divided into five parts: adhesion fitting, deviation calculation, tracking initialization, forward filtering tracking, and backward smoothing optimization.

[0041] Adhesion Fitting: The passive tracking system receives reference debris tracks, sorts them by time, and then adheses the debris tracks into a continuous track. A fitting model for the continuous track is obtained by using polynomial fitting.

[0042] Deviation calculation: Calculate the duration of the track after adhesion, extract the original measurement sequence and corresponding time point sequence of the passive tracking system within this time range, use the track fitting model to back-calculate the reference measurement sequence corresponding to the time point sequence, and finally use the reference measurement sequence and the original measurement sequence to calculate the measurement deviation.

[0043] Tracking initialization includes constructing the target motion state equation, sensor measurement equation, state noise model, and measurement noise model, etc.

[0044] Forward filtering tracking: Input new raw measurements and correct them using measurement bias. Then, use the corrected measurements for an unscented Kalman filter iterative process to complete the forward filtering tracking of the target and output the tracking track.

[0045] Backward smoothing optimization: The track from the previous time step is back-smoothed using the latest track output by forward filtering, and the smoothed track is output.

[0046] Furthermore, the above process is further explained using reference debris tracks received by a passive tracking system from two external auxiliary means, System A and System B:

[0047] The specific process of adhesion fitting is as follows:

[0048] S1.1, Set the multi-source trajectory of the reference debris received by the passive tracking system as... ,in Indicates the first Fragmented track sequence data provided by a reference source, This indicates the time corresponding to each waypoint. Indicates the first The total length of all reference tracks, i.e., the total number of trackpoints, is used to determine the total length of all reference tracks. The elements in the sequence are sorted and linked together to form a new continuous flight path. At this point, the reference debris tracks for systems A and B are respectively represented as follows: and Continuous flight path .

[0049] S1.2 The target motion model generally includes uniform motion, uniformly accelerated motion, and variable acceleration motion, as well as the alternation or synthesis of the above motion models, using time. To approximate the third-order polynomial of the independent variable, the position coordinates at a certain moment in the two-dimensional plane coordinate system are: Then the physical motion model of the target in each direction is:

[0050]

[0051] The track in S1.1 Substituting into the target's physical motion model, we get:

[0052]

[0053] The matrix on the left side of the equal sign is represented as The two matrices on the right are represented as and ,but Therefore, according to the least squares method, we have .

[0054] The specific process for deviation calculation is as follows:

[0055] S2.1 Obtain the flight path from S1.1 Duration of time interval and the raw measurement sequences of the sensors in the passive tracking system within this interval. ,in This represents the corresponding time point sequence, and the trajectory fitting model is used to inversely calculate... Corresponding reference measurement ;

[0056] S2.2, through Obtain the error sequences of the two measurement sequences, and calculate the average of all elements in the error sequences as the compensation value for the measurement bias. .

[0057] Considering the advantages of the unscented Kalman filter algorithm in passive tracking, such as high accuracy and good convergence, this algorithm is adopted in this embodiment. Therefore, the specific tracking initialization process is as follows:

[0058] set up express The effective correction time, i.e. Raw measurements taken after a certain time are not recommended. If correction is performed, then for The following moments ,in , This represents the target state vector, which includes components such as position, velocity, and acceleration. The original sensor measurements, including components such as distance and orientation, are represented as follows: The target state equation and the sensor measurement equation are respectively...

[0059]

[0060] in The state transition matrix has different matrix elements determined based on different target motion models. The noise transfer matrix, The measurement equation is a nonlinear function, and different functions are determined based on different sensor measurements. and Both are Gaussian white noise models, with covariance matrices of respectively and In this embodiment, the state transition matrix is: , For the sampling interval, take .

[0061] The specific process of forward filtering tracking is as follows:

[0062] S4.1. Set the parameters of the unscented transform algorithm (UT transform) in unscented Kalman filtering, assuming the state vector... The dimension is , Used to control the distribution of sampling points, with a value range of [value range missing]. , Used to describe the distribution of state variables, for Gaussian distribution, take... The values ​​of the distribution state of the sampling points. It is the scaling parameter, among which Typically, it is set to 0. In this embodiment, the state vector in the tracking system... The dimension is , , , , .

[0063] S4.2 Input new raw measurements Utilizing measurement deviation Perform deviation correction to obtain the corrected measurement. This is used in the filtering update process.

[0064] S4.3, State Prediction: Using express The Sigma sampling point at time t is calculated as follows:

[0065]

[0066] When the passive tracking system has received reference debris multi-source tracks, the initial state is... Take S1.1 The last element in the sequence is Otherwise, technicians will need to set it based on the target pattern library or usage experience; when hour, Set by technicians based on experience and patterns, when hour The iterative update value is given by step S4.4.

[0067] Determine the weight of each Sigma sampling point:

[0068]

[0069] Substituting the Sigma sampling points into the target state equation yields the predicted state value. Then, substituting the predicted state value into the measurement equation yields the predicted observation value.

[0070]

[0071] Predicted state values and observed predicted values We obtain the mean and covariance by weighting the values:

[0072]

[0073] S4.4, State Update: Calculating the Kalman Gain ;according to The state is updated using the predicted mean and covariance from S4.3, and the output filtered tracking result is as follows:

[0074] .

[0075] The specific process of backward smoothing optimization is as follows:

[0076] make The filtering result of the previous time step in S4.4 and Smoothing involves:

[0077]

[0078] Will and As respectively and The result after smoothing optimization.

[0079] After that, with new raw measurement input, the forward filtering tracking and backward smoothing optimization process can be repeated.

[0080] The process proposed in this embodiment involves 100 Monte Carlo simulations. In each simulation, the passive tracking system receives 10 reference track points from both system A and system B. The time step for a single simulation is 50 seconds. The process noise transfer matrix is... For unit moments, process noise covariance Measurement noise covariance The initial state covariance of the filter tracking is set to... .like Figure 2 This is a schematic diagram of the reference track and its deviation correction results from one of the Monte Carlo simulations. Figure 3 and Figure 4 This is a comparison chart of the target's passive tracking trajectory. Figure 5 This is a comparison chart of RMSE (Real-Time Sequence Error) calculated after 100 Monte Carlo simulations.

[0081] To verify the implementation effect of the present invention, the following three methods were compared: (1) tracking directly using the original measurement; (2) tracking with filtered measurement using the corrected measurement; and (3) tracking with smooth measurement using the corrected measurement. The third method is the technical implementation effect of the present invention.

[0082] Appendix Figure 3 The tracking results of three methods in a Monte Carlo simulation are shown. Method (3), which is the technical implementation of this invention, has the best effect, and its trajectory is closest to the actual trajectory. Method (2) is slightly better, and method (1) is the worst. (See attached diagram.) Figure 4 It can be seen that the advantages of the implementation effect of the present invention are relatively obvious, especially after the coordinates [5000, 1650].

[0083] Because of the appendix Figure 3 The comparison wasn't clear enough due to the results of a single Monte Carlo simulation. To further visualize the tracking performance of different methods, the root mean square error (RMSE) of position was used to compare the three methods. (See attached...) Figure 4 A comparison chart of RMSE calculated after 100 Monte Carlo simulations is given. It can be seen that the RMSE of tracking using method (3) and method (2) is significantly smaller than that of method (1). That is, tracking using the corrected measurement will inevitably result in better track accuracy, which is in line with expectations. By comparing the RMSE of method (3) and method (2), since method (3) adds a smoothing step, the overall track accuracy is slightly better than that of method (2). If the timeliness requirement is not considered, those skilled in the art can make a slight modification to change the one-step smoothing of this technology to multi-step smoothing, and the track accuracy can be further improved.

[0084] It should be noted that, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "set" and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances. The accompanying drawings in the embodiments are used to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0085] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A dynamic correction passive tracking method based on reference debris tracks, characterized in that, include: Step 1: Perform adhesion fitting on the reference debris track received by the passive tracking system to generate an interval track fitting model; The reference debris track includes target tracks generated by various sensor observations; Step 2: Based on the fitting model and the passive tracking system, obtain the measurement deviation from the original sensor measurement sequence within this interval; Step 3: Correct the original measurements input in real time using measurement deviation, and then substitute the corrected measurements into the unscented Kalman filter iterative process to complete forward filtering tracking and backward smoothing optimization, and output the smoothed track. In step 1, the adhesion fitting specifically involves: sorting the reference debris tracks by time and then adhering them into a continuous track, using polynomial fitting to obtain a fitting model for the continuous track; step 2 includes the following specific sub-processes: Step 2.1: Calculate the time range of the continuous tracks after adhesion in Step 1, and extract the original measurement sequence of the passive tracking system and the corresponding time point sequence within this time range; Step 2.2: Use the fitting model to back-calculate the reference measurement sequence corresponding to the time point sequence; Step 2.3: Subtract the elements in the reference measurement sequence from the original measurement sequence one by one to obtain the error sequence. Use the average value of all elements in the error sequence as the measurement deviation.

2. The dynamic correction passive tracking method based on reference debris tracks according to claim 1, characterized in that, In the unscented Kalman filtering iteration process in step 3, the target state equation and the sensor measurement equation are established based on the components contained in the target state vector and the original sensor measurements.

3. The dynamic correction passive tracking method based on reference debris tracks according to claim 1 or 2, characterized in that, In step 3, the forward filtering tracking process is as follows: the new original measurement is corrected by the measurement deviation, and then the corrected measurement is used in the unscented Kalman filter iterative process to complete the forward filtering tracking of the target and output the tracking track.

4. The dynamic correction passive tracking method based on reference debris tracks according to claim 2, characterized in that, The specific steps of the forward filtering tracking are as follows: Step 3.1: Input the new original measurement, and use the measurement deviation to perform deviation correction to obtain the corrected measurement; Step 3.2: Obtain Sigma sampling points and determine the weight of each Sigma sampling point; Step 3.3: Substitute the Sigma sampling points into the target state equation to obtain the state prediction value, and then substitute the state prediction value into the sensor measurement equation to obtain the observation prediction value. Step 3.4: Weight the state prediction values ​​and observation prediction values ​​according to their weights to obtain the mean and covariance; Step 3.5: Calculate the Kalman gain, and update the state based on the Kalman gain, mean, and covariance, then output the filtered tracking results.

5. The dynamic correction passive tracking method based on reference debris tracks according to claim 1, characterized in that, In step 3, the specific process of backward smoothing optimization is as follows: the latest track output by forward filtering is used to perform backward smoothing on the track of the previous moment, and the smoothed track is output.

Images