A radar online real-time maneuvering target tracking method for coping with plot interruption

Abstract

This invention specifically relates to a method for tracking maneuvering targets in real-time online radar response to track interruptions. The method includes: calculating the time difference between adjacent sampling points to determine if a track interruption has occurred; distinguishing between short-term and long-term interruptions based on the interruption count; if it is a long-term interruption, predicting the current position based on the previous state and maintaining the state parameters unchanged; if it is a short-term interruption, further determining whether it is a long-term interruption recovery; if so, calculating the distance between the predicted position and the actual measured position, comparing it with a threshold to determine if the target has been lost, and accordingly performing state preservation or multi-model filtering update; if not, based on the re-tracking flag and the interruption flag, performing state re-initialization and filtering, state preservation, or directly performing multi-model filtering update. This invention significantly improves the tracking continuity, state estimation accuracy, and robustness of radar in the event of track interruptions by adaptively managing the filtering process during and after the interruption, effectively reducing the probability of target loss.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and specifically to a method for tracking moving targets in real-time online in response to point interruptions. Background Technology

[0002] As a primary active detection device, one of the core functions of radar is to generate stable and continuous tracks for targets of interest over multiple scan cycles. However, in practical applications, the target track sequence acquired by radar often experiences interruptions, meaning that effective measurements of the target are not successfully obtained within a scan cycle. This is mainly due to the following reasons: ① Environmental and target characteristics: The target is obscured by terrain and clutter, and fluctuations in the target's cross-sectional area cause the echo to fall below the detection threshold; ② Electromagnetic environment factors: Active jamming or dense passive jamming set up by the enemy can obscure the true target track, and severe weather conditions can attenuate electromagnetic waves, causing a decrease in the target's signal-to-noise ratio; ③ The radar system itself: Insufficient sampling rate due to the setting of the scan interval or beam dwell time, and the target entering the radar blind zone or beam spacing, etc.

[0003] Track interruptions can severely disrupt the normal operation of tracking filters. Traditional radar maneuvering target tracking algorithms, such as α-β or α-β-γ based filtering algorithms, Kalman filters (KF) and their extensions in nonlinear systems, including Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF), as well as more advanced Interacting Multiple Model (IMM) algorithms, heavily rely on continuous and periodic measurement inputs for state updates. When track interruptions occur, the following two problems may arise: ① Accumulated state prediction error: During the interruption, the filter can only rely on the motion model for extrapolation prediction. For maneuvering targets, model mismatch will quickly cause the predicted position to deviate from the true trajectory; ② Filter covariance matrix out of control: Without measurement information updates, the error of the state estimation covariance matrix will continue to increase with the number of prediction steps. If the covariance matrix is ​​used directly for updating after the interruption ends, it may lead to filter divergence.

[0004] Existing methods for dealing with track interruptions mostly adopt a passive "interruption first, repair later" approach. This treats the track interruption as a given fact, and in the backend of data processing, uses historical observation data for interpolation, smoothing, and prediction to generate a visually continuous track. This type of approach, utilizing historical observation data or statistical fitting methods, is a "post-hoc" approach and is unsuitable for the real-time online processing requirements of modern radar systems. There is an urgent need for a processing method that can dynamically adjust the tracking strategy based on the number of interrupted frames, rationally manage filter states during the interruption, and perform intelligent discrimination and rapid correction when the track is restored.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the issue of radar tracking interruptions caused by temporary target obstruction, signal fluctuations, or radar scanning cycles in complex detection environments, this invention provides a real-time online radar tracking method for tracking moving targets, enabling continuous and stable tracking of moving targets by the radar system.

[0007] Other features and advantages of the invention will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0008] According to a first aspect of the present invention, a method for tracking a maneuvering target in real-time online response to point track interruptions is provided, the method comprising: Parameter initialization: Set the sampling point number, the number of consecutive interruptions of the tracking point, and the re-tracking flag; initialize the observation matrix, target state vector, target multi-model state transition matrix set, probability vector of each model, probability transition matrix of each model, process noise covariance matrix of each model, observation noise covariance matrix of each model, and target state covariance matrix of each model. The interruption of the sampling point is determined based on the time difference between two consecutive sampling points. Once a point interruption is determined, it is determined whether it is a long-term or short-term interruption based on the cumulative number of interruptions at the current sampling point. When there is a long interruption, the target state vector is updated and the current position is predicted based on the model states and probabilities of the previous sampling point; the multi-model parameters, target state vector and covariance matrix are kept the same as those of the previous sampling point. For short-term interruptions, interrupt recovery is determined based on the cumulative number of interrupt frames from the previous sampling point. When it is an interruption recovery, the recovery association judgment is performed: calculate the distance between the predicted position during the interruption and the current measurement position, and compare it with the dynamic threshold; if it exceeds the threshold, it is determined that the target has been lost, and the multi-model parameters, target state vector and covariance matrix are kept the same as the previous sampling point; otherwise, it is determined that the target has not been lost, and multi-model filtering is updated. If no interrupt recovery is performed, regular or short-term interrupt handling is executed: based on the re-tracking flag and interrupt count, the appropriate options are selected: perform state re-initialization and filtering, maintain state, or directly perform multi-model filtering update.

[0009] In some exemplary embodiments, the multiple models include a uniform motion model, a coordinated turning model with multiple different turning angular velocities, and a stationary model; The vector composed of the probabilities of each model is

[0010] in, Indicates the number of models. express A dimensional vector of all 1s; The probability transition matrix of each model is

[0011] in, Indicates the number of models. express A 1-dimensional matrix of all ones. express A diagonal matrix; The process noise covariance matrix of each model is

[0012] in, Indicates the root mean square of the measurement noise. Indicates the sampling time interval; The measurement noise covariance matrix of each model is as follows:

[0013] in, Indicates the root mean square of the measurement noise; The target state covariance matrix of each model is initialized as follows: .

[0014] In some exemplary embodiments, the target state vector is initialized to

[0015] in, and Indicates the target is in the first place. sampling points direction and Direction coordinates and Indicates the target is in the first place. sampling points direction and Direction coordinates and They represent the first The sampling point and the first The time for each sampling point.

[0016] In some exemplary embodiments, determining whether an interrupt is a long-term or short-term interrupt based on the cumulative number of interrupt frames at the current sampling point specifically involves: If the cumulative number of interrupt frames at the current sampling point is greater than or equal to 2, it is a long interrupt; otherwise, it is a short interrupt.

[0017] In some exemplary embodiments, the interruption recovery determination based on the cumulative number of interrupted frames from the previous sampling point specifically includes: If the cumulative number of interrupted frames at the previous sampling point is greater than or equal to 2, then it is considered an interrupt recovery; otherwise, it is not considered an interrupt recovery.

[0018] In some exemplary embodiments, updating the target state vector and predicting the current position based on the model states and probabilities of the previous sampling point specifically involves:

[0019] in, Indicates the first The target state vector of the sampling point is the first... One element, The following formula is obtained

[0020] in, Indicates the first The sampling point The probability of each model, The following formula is obtained

[0021] in, Indicates the first The sampling point of the nth sampling point The state of each model.

[0022] In some exemplary embodiments, the dynamic threshold is obtained by the following formula:

[0023] in, This indicates finding the median of an array. Indicates the extended range factor. This indicates the number of consecutive breaks in the dot pattern. Indicates the first k The distance between the predicted location and the current measurement location during the sampling point interruption period.

[0024] In some exemplary implementations, based on the re-tracking flag and the interruption count, the execution state is selected to be reinitialized and filtered, the state is maintained, or a multi-model filtering update is performed directly. Specifically: If the retracking flag is 1 and the interruption count is 0, then after state initialization, multi-model filtering is performed to obtain the multi-model parameters of the current sampling point, the target state vector, and the covariance matrix, and the retracking flag is set to 0. If the retracking flag is 1 and the interruption count is 1, then the multi-model parameters, target state vector, and covariance matrix remain the same as the previous sampling point. If the re-tracking flag is 0, the multi-model parameters, target state vector, and covariance matrix of the current sampling point are obtained through multi-model filtering.

[0025] According to a second aspect of the present invention, a radar system is provided, characterized in that it includes a memory and a processor, the memory storing a computer program, the processor executing the program to implement the above-described maneuvering target tracking method.

[0026] The radar online real-time tracking method for maneuvering targets in response to track interruptions, provided by embodiments of the present invention, can maintain the tracking gate during track interruptions and quickly perform correlation and status correction after track recovery. It can be used in radar systems for monitoring and early warning of maneuvering targets on the sea, ground, and in the air, aiming to improve the real-time online tracking capability of radar under incomplete observation conditions.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Significantly improves tracking robustness and continuity during point interruptions: By introducing an adaptive decision process based on interruption duration, this invention can dynamically distinguish between short interruptions (0 or 1 frame) and long interruptions (2 frames or more). For different interruption situations, it intelligently switches between three strategies: filter update, state preservation, and model prediction, effectively suppressing error accumulation and covariance matrix divergence caused by blind extrapolation during periods without observation.

[0028] 2. Intelligent reassociation and rapid state correction after interruption recovery are achieved. A special discrimination mechanism is designed for recovery from long interruptions. By calculating and comparing the difference between the predicted position during the interruption and the new position at the time of recovery, the state is adaptively discriminated, avoiding the forced association of the recovered point with the erroneously predicted trajectory.

[0029] 3. Possesses significant value for online real-time engineering applications: The decision-making conditions of this invention are clear, and the computational load is controllable. All decisions rely solely on parameters that can be directly obtained or easily calculated, such as the number of interrupted frames, time difference, and location distance, without requiring complex iterative optimization, thus meeting the high real-time requirements of radar systems. Furthermore, this invention incorporates a pause mechanism for model probability updates when the point is interrupted, performing only state transfer and prediction. This allows the resource consumption of multi-model filtering to focus on effective observation moments, improving both overall filtering efficiency and filtering accuracy.

[0030] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0031] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0032] Figure 1 This is a flowchart of a radar online real-time tracking method for maneuvering targets that responds to point interruptions, according to the present invention; Figure 2 This is a graph showing the change of distance of data 1 with the number of frames in an embodiment of the present invention; Figure 3 This is a graph showing the change of the azimuth angle of data 1 with the number of frames in an embodiment of the present invention; Figure 4 This is a graph showing the change of the time interval between two frames of data 1 in this embodiment of the invention as a function of the number of frames; Figure 5 This is a diagram of the measured trajectory and the actual trajectory of data 1 in this embodiment of the invention; Figures 6(a)-(g) are measurement point traces at the time of frame breakage in data 1 in the embodiments of the present invention; Figure 7 This is the target trajectory diagram after data 1 in this embodiment of the invention has been processed by the method of the present invention and the traditional IMM method; Figure 8 This is the curve showing the change of the root mean square error of data position 1 with the number of frames in this embodiment of the invention; Figure 9 This is a graph showing the change in distance between data 2 and the number of frames in an embodiment of the present invention; Figure 10 This is a graph showing the change of the azimuth angle of data 2 with the number of frames in an embodiment of the present invention; Figure 11This is a graph showing the change of the time interval between two frames of data 2 in this embodiment of the invention as a function of the number of frames; Figure 12 This is a diagram of the measured trajectory and the actual trajectory of data 2 in this embodiment of the invention; Figure 13 This is a measurement point trace map at the moment of data frame breakage in the embodiment of the present invention; Figure 14 This is the target trajectory diagram after data 2 in this embodiment of the invention has been processed by the method of the present invention and the traditional IMM method; Figure 15 This is the curve showing the change of the root mean square error of data position 2 with the number of frames in this embodiment of the invention. Detailed Implementation

[0033] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the invention will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0034] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0035] To address the shortcomings and deficiencies of existing technologies, this example implementation provides a method for tracking maneuvering targets in real-time online radar response to point track interruptions. This method adaptively identifies the duration of point track interruptions and adjusts the tracking strategy accordingly. Specifically, it updates the filtering state normally during short interruptions and maintains state prediction during long interruptions. Upon recovery, it uses a distance threshold to determine whether to re-associate. Furthermore, it incorporates a multi-model filtering framework to manage model probabilities during interruptions. Ultimately, while ensuring real-time performance, this significantly improves the continuity and accuracy of radar tracking maneuvering targets under intermittent observation.

[0036] refer to Figure 1 As shown, the method of the present invention may specifically include the following steps: Initialize the tracking parameters; Calculate the time difference between adjacent sampling points to determine whether a point interruption has occurred and update the interruption count; Based on whether the interrupt count exceeds the threshold, distinguish between short-time interrupts and long-time interrupts; If the interruption is long-term, the current position is predicted based on the state of the previous moment and the state parameters are kept unchanged; If it is a short-term interruption or a normal situation, further determine whether it has just recovered from a long-term interruption; If it is a long-term interruption recovery, the distance between the predicted position and the actual measured position is calculated, and the distance is compared with the threshold to determine whether the target has been lost, and state preservation or multi-model filtering update is performed accordingly. If not, then based on the re-tracking flag and the interruption flag, perform state re-initialization and filtering, state preservation, or directly perform multi-model filtering update.

[0037] This invention significantly improves the tracking continuity, state estimation accuracy, and robustness of radar in the event of a point interruption by adaptively managing the filtering process during and after the interruption, and effectively reduces the probability of target loss.

[0038] The steps in this exemplary embodiment will now be described in more detail with reference to the accompanying drawings and embodiments.

[0039] Step 1: Parameter initialization and quantitative setting.

[0040] Let the sampling point number Number of consecutive interruptions in the dot pattern Re-tracking flags Set the observation matrix:

[0041] Initialization values ​​for the target state vector:

[0042] in, and Indicates the target is in the first place. sampling points direction and Direction coordinates and Indicates the target is in the first place. sampling points direction and Direction coordinates and They represent the first The sampling point and the first The time for each sampling point.

[0043] The set of multi-model state transition matrices of the target To enhance the robustness of tracking maneuvering targets, the complete set of the target's multi-model state transition matrix consists of a CV model, a ±20° CT model, a ±45° CT model, a ±70° CT model, and a stationary model. The CV model is mainly used to adapt to the target's linear motion, the multiple CT models are used to describe maneuvering motions with different degrees of turning, and the stationary model is used as a memory reset mechanism to maintain numerical stability.

[0044] Initialization values ​​for the vector composed of probabilities from each model:

[0045] in, Indicates the number of models. express A vector of all 1s in dimension.

[0046] Initial values ​​of the probability transition matrix for each model:

[0047] in, express A 1-dimensional matrix of all ones. express A diagonal matrix.

[0048] Measurement noise covariance matrix of each model :

[0049] in, Indicates the root mean square of the measurement noise. Indicates the sampling time interval.

[0050] Observation noise covariance matrix of each model :

[0051] in, This represents the root mean square of the measurement noise.

[0052] Initial values ​​of the target state covariance matrix for each model :

[0053] Step 2: Perform a dot interruption determination.

[0054] Calculate the first The sampling point and the first Time difference between sampling points:

[0055] like If it is zero, then determine the first... The trace at each sampling point is interrupted, causing ; Step 3: Check whether the current sampling point has been interrupted for two or more consecutive frames.

[0056] If the first sampling points If the current sampling point is interrupted for two or more consecutive frames, proceed to step 4 for long-term interruption processing; otherwise, if the current sampling point is not interrupted or only one frame is interrupted, proceed to step 5 for interruption recovery determination.

[0057] Step 4: Long-term interrupt handling, perform state prediction or maintenance.

[0058] Step 4.1: State Prediction According to the The model states at each sampling point With probability Update the target state vector :

[0059] in, Indicates the first The sampling point The probability of each model, , Indicates the first The sampling point of the nth sampling point The state of each model.

[0060] The predicted number Location of each sampling point

[0061]

[0062] in, Indicates the first The target state vector of the sampling point is the first... Each element.

[0063] Step 4.2: State Maintenance Since no new measurements are available for updating, the multi-model parameters are preserved. and ), target state vector and covariance matrix With the All sampling points are the same:

[0064]

[0065]

[0066]

[0067] Step 4.3: End the current step and proceed to step 8.

[0068] Step 5: Interruption Recovery Determination If the first sampling points If the current sampling point has just recovered from a state of two or more frames being interrupted, proceed to step 6 to restore the association judgment and processing; otherwise, if the previous frame was in a normal state, proceed to step 7 for normal or short-term interruption processing.

[0069] Step 6: Restore Association Judgment and Processing Step 6.1: Based on the predicted number when the dot pattern is interrupted... Each sampling point location , and the measurement of the first Each sampling point location Calculate the distance between two locations. :

[0070] Step 6.2: Calculate the distance threshold:

[0071] in, This indicates finding the median of an array. This represents the extended range factor, which can be 1.5.

[0072] Step 6.3: Threshold Comparison If distance Proceed to step 6.3.1; if the distance... Proceed to step 6.3.2.

[0073] Step 6.3.1: Determine that the target has been lost; re-tracking is required at the next sampling point. Maintain multiple model parameters ( and ), target state vector and covariance matrix With the All sampling points are the same.

[0074]

[0075]

[0076]

[0077]

[0078] End the current step and proceed to step 8.

[0079] Step 6.3.2: Determine that the target has not been lost, and update the multi-model parameters through multi-model filtering. and ), target state vector and covariance matrix End the current step and proceed to step 8.

[0080] Step 7: Regular or short-term interruption handling like =1 and If the value is 0, proceed to step 7.1; if =1 and If the value is 1, proceed to step 7.2; if If the value is 0, proceed to step 7.3.

[0081] Step 7.1: This situation occurs after the target was lost in step 6. A new measurement is received in the next frame, and the measurement is continuous with the previous time.

[0082] Step 7.1.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] Initialize as follows

[0083] Step 7.1.2: Update the multi-model parameters through multi-model filtering ( and ), target state vector and covariance matrix .

[0084] Step 7.1.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] Set to 0, end the current step, and proceed to step 8.

[0085] Step 7.2: This situation occurs after the target was determined to be lost in step 6, and the process is interrupted again in the next frame. At this time, the system still has not obtained reliable measurements that can be used for re-initialization. Maintain multiple model parameters ( and ), target state vector and covariance matrix With the All sampling points are the same.

[0086]

[0087]

[0088]

[0089]

[0090] End the current step and proceed to step 8.

[0091] Step 7.3: This is a normal tracking state, including continuous normal tracking and immediate recovery after only one frame interruption. The multi-model parameters are updated through multi-model filtering. and ), target state vector and covariance matrix End the current step and proceed to step 8.

[0092] Step 8: Loop control and termination determination.

[0093] If sampling is complete, then the process ends; if sampling is not complete, then... Repeat step 2.

[0094] The effectiveness of the proposed method for tracking maneuvering targets in real-time online radar response to point track interruptions can be illustrated by the following simulation experiments: (1) Experimental conditions: The experiment was conducted using the digital array generalized radar low-speed-small detection dataset "LSS-DAUR-1.0" published in the Journal of Radar. The planar tracking accuracy of two sets of measured data from speedboats was used to verify the invented method; the data names were "20231014102639_DAUR_TR_Speedboat_03_1" and "20230801153955_DAUR_TR_Speedboat_10_9". The selected multi-models included the CV model, the ±20° CT model, the ±45° CT model, the ±70° CT model, and the stationary model. , to Both are 1. to All are 1.

[0095] (2) Experimental content and results analysis: Experiment 1: In this experiment, the measured data of the speedboat "20231014102639_DAUR_TR_Speedboat_03_1" was processed using a radar online real-time response method for tracking maneuvering targets with interrupted tracking points, as described in this invention.

[0096] To plot the curves showing the changes in distance, azimuth, and time interval between two frames as a function of the number of frames, please refer to [references to be inserted here]. Figures 2-4 .from Figure 4It can be seen that there is a frame break at frame 9, frame 23, frame 25, frame 27, frame 34, three frames between frames 51 and 53, one frame between frame 65, a few frames between frames 78 and 80, one frame between frame 93, one frame between frame 95, and one frame between frame 97. Please refer to the diagram for drawing the measured trajectory and the actual trajectory. Figure 5 The measurement points at the moment of frame break are specifically drawn, see Figure 6.

[0097] The target trajectory is processed using the method of this invention and the traditional IMM method. Please refer to [link / reference needed]. Figure 7 It can be seen that the dots processed by the invented method are closer to the real track than the original measurement. The traditional IMM method deviates significantly from the real track after the 80th frame due to the interruption of dots.

[0098] For further statistical analysis of the root mean square error of the location, please refer to [link / reference needed]. Figure 8 The average root mean square error (RMSE) of the measured point position was 10.01 m. After processing with the proposed method, the average RMSE decreased to 8.65 m, demonstrating a significant improvement in target tracking performance with an improvement rate of 14%. In contrast, the traditional IMM method, after processing, exhibited an average RMSE of 19.94 m due to severe deviation from the true trajectory after the 80th frame. The total processing time required by the traditional IMM method and the proposed method were 0.056 and 0.058 seconds, respectively, with the proposed method only increasing the processing time by 3.5%.

[0099] Experiment 2: In this experiment, the measured data of the speedboat "20230801153955_DAUR_TR_Speedboat_10_9" was processed using a radar online real-time response method for tracking maneuvering targets with interrupted tracking points, as described in this invention.

[0100] To plot the curves showing the changes in distance, azimuth, and time interval between two frames as a function of the number of frames, please refer to [references to be inserted here]. Figures 9-11 .from Figure 11 It can be seen that there is a two-frame gap at frames 46 and 47, and a one-frame gap at frame 49. Please refer to [link to documentation] for plotting the measurement trajectory and the actual trajectory. Figure 12 The measurement points at the moment of frame breakage are specifically plotted; please refer to [link / reference]. Figure 13 .

[0101] The target trajectory is processed using the method of this invention and the traditional IMM method. Please refer to [link / reference needed]. Figure 14 It can be seen that the dots processed by the invented method are closer to the real track than the original measurement. The traditional IMM method deviates significantly from the real track after the 47th frame due to the interruption of dots.

[0102] For further statistical analysis of the root mean square error of the location, please refer to [link / reference needed]. Figure 15 The average root mean square error (RMSE) of the measured point position was 10.15 m. After processing with the proposed method, the average RMSE decreased to 7.44 m, demonstrating a significant improvement in target tracking performance with an improvement rate of 26%. In contrast, the traditional IMM method, after processing, exhibited an average RMSE of 31.94 m due to severe deviation from the true trajectory after the 47th frame. The total processing time required by the traditional IMM method and the proposed method were 0.032 and 0.036 seconds, respectively, with the proposed method only increasing the processing time by 11%.

[0103] In summary, this invention provides a real-time online radar target tracking method for handling point interruptions. By introducing an interruption frame count decision and an adaptive state management mechanism, it effectively solves the problem of severe performance degradation in traditional IMM methods after point interruptions. Without significantly increasing the computational burden, it significantly enhances the tracking continuity and robustness of the radar system under point interruption conditions, making it suitable for various radar target tracking scenarios with high real-time and reliability requirements.

[0104] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims.

[0105] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is defined only by the appended claims.

Claims

1. A method for tracking maneuvering targets online in real time to address point-track interruptions, characterized in that, The method includes: Parameter initialization: Set the sampling point number, the number of consecutive interruptions of the tracking point, and the re-tracking flag; initialize the observation matrix, target state vector, target multi-model state transition matrix set, probability vector of each model, probability transition matrix of each model, process noise covariance matrix of each model, observation noise covariance matrix of each model, and target state covariance matrix of each model. The interruption of the sampling point is determined based on the time difference between two consecutive sampling points. Once a point interruption is determined, it is determined whether it is a long-term or short-term interruption based on the cumulative number of interruptions at the current sampling point. When there is a long interruption, the target state vector is updated and the current position is predicted based on the model states and probabilities of the previous sampling point; the multi-model parameters, target state vector and covariance matrix are kept the same as those of the previous sampling point. For short-term interruptions, interrupt recovery is determined based on the cumulative number of interrupt frames from the previous sampling point. When it is an interruption recovery, the recovery association judgment is performed: calculate the distance between the predicted position during the interruption and the current measurement position, and compare it with the dynamic threshold; if it exceeds the threshold, it is determined that the target has been lost, and the multi-model parameters, target state vector and covariance matrix are kept the same as the previous sampling point; otherwise, it is determined that the target has not been lost, and multi-model filtering is updated. If no interrupt recovery is performed, regular or short-term interrupt handling is executed: based on the re-tracking flag and interrupt count, the appropriate options are selected: perform state re-initialization and filtering, maintain state, or directly perform multi-model filtering update.

2. The method according to claim 1, characterized in that, The multiple models include a uniform motion model, a coordinated turning model with multiple different turning angular velocities, and a stationary model; The vector composed of the probabilities of each model is in, Indicates the number of models. express A dimensional vector of all 1s; The probability transition matrix of each model is in, Indicates the number of models. express A 1-dimensional matrix of all ones. express A diagonal matrix; The process noise covariance matrix of each model is in, Indicates the root mean square of the measurement noise. Indicates the sampling time interval; The measurement noise covariance matrix of each model is as follows: in, Indicates the root mean square of the measurement noise; The target state covariance matrix of each model is initialized as follows: 。 3. The method according to claim 1, characterized in that, The target state vector is initialized to in, and Indicates the target is in the first place. sampling points direction and Direction coordinates and Indicates the target is in the first place. sampling points direction and Direction coordinates and They represent the first The sampling point and the first The time for each sampling point.

4. The method according to claim 1, characterized in that, The determination of whether an interrupt is long-term or short-term based on the cumulative number of interrupt frames at the current sampling point is as follows: If the cumulative number of interrupt frames at the current sampling point is greater than or equal to 2, it is a long interrupt; otherwise, it is a short interrupt.

5. The method according to claim 1, characterized in that, The interrupt recovery determination based on the cumulative number of interrupted frames from the previous sampling point is as follows: If the cumulative number of interrupted frames at the previous sampling point is greater than or equal to 2, then it is considered an interrupt recovery; otherwise, it is not considered an interrupt recovery.

6. The method according to claim 2, characterized in that, The process of updating the target state vector and predicting the current position based on the model states and probabilities of the previous sampling point is as follows: in, Indicates the first The target state vector of the sampling point is the first... One element, The following formula is obtained in, Indicates the first The sampling point The probability of each model, The following formula is obtained in, Indicates the first The sampling point of the nth sampling point The state of each model.

7. The method according to claim 1, characterized in that, The dynamic threshold is obtained by the following formula: in, This indicates finding the median of an array. Indicates the extended range factor. This indicates the number of consecutive breaks in the dot pattern. Indicates the first k The distance between the predicted location and the current measurement location during the sampling point interruption period.

8. The method according to claim 1, characterized in that, Based on the re-tracking flag and interruption count, the following options are selected: re-initialize and filter the execution state, maintain the state, or directly perform multi-model filtering update. If the retracking flag is 1 and the interruption count is 0, then after state initialization, multi-model filtering is performed to obtain the multi-model parameters of the current sampling point, the target state vector, and the covariance matrix, and the retracking flag is set to 0. If the retracking flag is 1 and the interruption count is 1, then the multi-model parameters, target state vector, and covariance matrix remain the same as the previous sampling point. If the re-tracking flag is 0, the multi-model parameters, target state vector, and covariance matrix of the current sampling point are obtained through multi-model filtering.

9. A radar system, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the program to implement the maneuvering target tracking method as described in any one of claims 1 to 8.

Images