A multi-target tracking method and device for networked radars in a complex electromagnetic environment based on joint fusion

Abstract

The application provides a networking radar multi-target tracking method and device in a complex electromagnetic environment based on joint fusion, and relates to the technical field of radar information processing. The method fuses active ranging / angulation data and passive radiating source parameters, and constructs a time-labeled fusion measurement sequence; a joint track initiation and multi-threshold fast data association model is used, combined with a nonlinear, non-Gaussian robust filter and a track quality criterion, to realize stable tracking and accurate state estimation of multi-targets, and to improve the anti-interference capability and tracking reliability of the system.

Description

Technical Field

[0001] This invention relates to the field of radar information processing technology, and in particular to a method and apparatus for multi-target tracking of networked radar in complex electromagnetic environments based on joint fusion. Background Technology

[0002] Networked radar—a distributed network of multiple radars or passive reconnaissance devices with different systems, frequency bands, polarizations, and operating modes connected by communication links—has become a core means of modern regional space countermeasures to suppress jamming, counter radiation attacks, and counter stealth and low-altitude penetration threats. Multi-sensor data fusion can provide more accurate and complete trajectory information for situation assessment than any single radar; therefore, "multi-target tracking data fusion" is a key technological direction for current radar information processing systems.

[0003] As the regional electromagnetic environment becomes increasingly complex, target echoes exhibit non-Gaussian, nonlinear, and even intermittently distorted statistical characteristics. Simultaneously, networked radars are increasingly incorporating joint systems of active detection and passive radiation source reconnaissance to leverage spectral dispersion and geometric multi-baseline advantages to counter strong suppression interference. Future development focuses on two main lines: "active-passive collaborative observation and distributed real-time fusion" and "robust filtering against strong interference," requiring tracking algorithms to balance high accuracy, low latency, and robustness.

[0004] Most existing studies rely on clutter background assumptions and employ linear Gaussian models based on Kalman families or JPDA. When the observation noise is colored interference, false target deception, or range gate loss, these methods are prone to track divergence. The lack of a unified data association and filtering mechanism that can simultaneously utilize active and passive redundant information and operate robustly in a nonlinear, non-Gaussian framework has become a bottleneck restricting the reliable tracking of networked radars in extreme electromagnetic environments. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and device for multi-target tracking of networked radar in complex electromagnetic environments based on joint fusion. This method overcomes the shortcomings of existing methods in nonlinear and non-Gaussian interference environments, which make it difficult to achieve robust correlation and accurate tracking. It realizes deep fusion of active and passive observation information and highly robust multi-target tracking.

[0006] To achieve the above objectives, the present invention provides the following solution: A multi-target tracking method for networked radars in complex electromagnetic environments based on joint fusion includes: Within the same detection cycle, the active ranging / angle measurement data and passive radiation source parameter measurement data of each radar node are received, and the measurement data are compressed and merged to obtain a time-stamped fused measurement sequence. For the fused measurement sequence, a joint active-passive track initiation algorithm is used to determine candidate initial tracks within the set confirmation area; For each candidate initial trajectory, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and the radiation source attributes, as well as the directional consistency constraints of the target's predicted azimuth and pitch angles, in order to complete the matching of measurement data and trajectories. The matched measurement data is input into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state. When the track quality meets the terminal criteria, the target identifier, six-dimensional state vector, and covariance matrix are output to complete the multi-target tracking data fusion.

[0007] Preferably, within the same detection period, active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node are received. The measurement data are then compressed and merged to obtain a time-stamped fused measurement sequence, including: The raw point data reported by each radar node within the current detection period are synchronized and aligned according to the local timestamp to obtain a node-level time series; the detection period is determined by the radar scanning period. In each node-level time series, for multiple points whose spatial location deviation and Doppler / radiation source attribute differences are both below a preset threshold, point clustering is performed and a single compressed measurement vector is formed by weighted averaging to reduce the number of redundant points. The compressed single measurement vectors of each radar node are concatenated and sorted according to a unified network time index, and a "node identifier-time stamp" composite mark is added to each single compressed measurement vector to finally generate the fused measurement sequence with time stamp.

[0008] Preferably, for the fused measurement sequence, a joint active-passive track initiation algorithm is used within a set confirmation region to determine candidate initial tracks, including: Within the current detection cycle, a ring-shaped confirmation area is formed with each of the aforementioned fusion measurement sequences as the center, according to the outer radius and the inner radius; Active and passive measurement data that simultaneously satisfy the conditions of spatial position residual being less than a first threshold and radiation source main frequency difference being less than a second threshold are retrieved within the confirmed area, and the active and passive measurement data are combined to form a measurement pair set. For each measurement pair, using the two-dimensional position given by the active measurement and the radiation source parameters given by the passive measurement, the initial six-dimensional state vector and its covariance of the target are calculated using a constant velocity motion model and recorded as a temporary track record. During the subsequent detection period, a prediction-gating-association process is performed on each of the temporary track records; when the same temporary track record is successfully associated with a fusion measurement at least once within L consecutive detection periods, and the observation residual of each association is lower than a preset threshold, the temporary track record is promoted to the candidate initial track; where L≥2.

[0009] Preferably, for each candidate initial trajectory, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, to complete the matching of measurement data and trajectories, including: Using the state estimate of the candidate initial track that has been updated in the previous detection cycle, the predicted position and velocity of the candidate initial track in the current detection cycle are calculated, and an ellipsoidal first gate region is delineated based on the predicted position and velocity to limit the measurement entries in the fused measurement sequence to be compared. For each measurement item that falls into the first gated area, the kinematic parameter difference and radiation source attribute difference of the measurement item are compared respectively, and then combined into a single similarity score according to a preset ratio. The measurement item is retained only when the single similarity score does not exceed the first threshold, resulting in a coarsely screened and retained measurement item; For the measurement items retained after coarse screening, the azimuth and pitch differences are further examined. Only when both the azimuth and pitch differences are within the corresponding angle thresholds are their directions considered to be consistent with the predicted directions; otherwise, they are discarded. For measurement items that pass the direction consistency check, a second threshold is applied for fine screening. If multiple measurement items still meet the second threshold, the one with the smallest similarity score is selected as the associated measurement with the candidate initial track in the current detection period. If no measurement item passes the fine screening, it is determined that the candidate initial track lacks effective measurement input in this period.

[0010] Preferably, the matched measurement data is input into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state, including: At the beginning of the current detection cycle, based on the trajectory state estimate of the previous cycle, a nonlinear state transition model is performed on each candidate trajectory to obtain the set of prior states at the beginning of the current cycle. The matched measurement data are sequentially injected into a robust filter to perform likelihood evaluation and weight correction on the prior state set, forming a weighted posterior state set. Calculate the effective representation of the posterior state set, and trigger a resampling operation when the effective representation is lower than a preset sampling threshold; During the resampling operation, the posterior state set is weighted and resampled according to the weight to generate a new equal-weight state set in order to suppress particle degradation. After the resampling operation is completed, the predicted state for the next detection period is calculated based on the new equal-weighted state set, and the predicted state and dispersion of the next detection period are used as the gating reference for the next period.

[0011] Preferably, when the trajectory quality meets the terminal criteria, the target identifier, a six-dimensional state vector, and a covariance matrix are output to complete multi-target tracking data fusion, including: At the end of each detection cycle, for the updated track, the observation residuals, number of consistent valid measurements, state covariance convergence and duration of continuous existence of the most recent cycles are statistically analyzed and combined to form a track quality score. The track quality score is compared with a pre-set terminal criterion; when the track quality score meets the criterion requirements, the updated track is marked as "fusion confirmation status", otherwise it remains in "pending confirmation status". For tracks marked as "fusion confirmed state", a unique target identifier is assigned, and the three-dimensional position component and three-dimensional velocity component in the latest state estimate are extracted and combined into a six-dimensional state vector, while retaining the corresponding state covariance matrix. The unique target identifier, the six-dimensional state vector, and the covariance matrix are output to the result interface and simultaneously registered in the track management table; The output results are pushed to the upper command-control or display terminal using the aforementioned result interface to complete the final output of multi-target tracking data fusion.

[0012] A networked radar multi-target tracking device based on joint fusion in complex electromagnetic environments includes: The data receiver and fusion unit is used to receive active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node within the same detection cycle, and to compress and merge the measurement data to obtain a time-stamped fused measurement sequence. An initial track candidate determiner is used to determine candidate initial tracks within a set confirmation region for the fused measurement sequence using a joint active-passive track initiation algorithm. A data association model builder is used to construct an improved multi-threshold fast data association model for each candidate initial track, based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, in order to complete the matching of measurement data and tracks. Nonlinear robust filters are used to input matched measurement data into nonlinear, non-Gaussian robust filters to recursively update and extrapolate the track state. The multi-target tracking output unit is used to output target identifiers, six-dimensional state vectors, and covariance matrices when the track quality meets the terminal criteria, so as to complete the fusion of multi-target tracking data.

[0013] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention, under conditions of strong interference, high noise, and uncertain target characteristics, fully leverages the complementary advantages of active ranging / angle measurement information and passive radiation source parameters. Through the construction of a joint confirmation region, an adaptive initiation mechanism, a multi-threshold rapid association model, and a nonlinear robust filtering process, it achieves stable identification and accurate tracking of multiple targets. Compared to traditional methods, this invention offers significant advantages in improving data association accuracy, suppressing false track generation, and enhancing system anti-interference capabilities. It is suitable for the practical deployment and intelligent expansion of next-generation networked radar systems in complex regional spatial environments. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the technical route provided for an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] The purpose of this invention is to provide a method and device for multi-target tracking of networked radar in complex electromagnetic environments based on joint fusion. This method overcomes the shortcomings of existing methods in nonlinear and non-Gaussian interference environments, which make it difficult to achieve robust correlation and accurate tracking. It realizes deep fusion of active and passive observation information and highly robust multi-target tracking.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this invention provides a multi-target tracking method for networked radars in complex electromagnetic environments based on joint fusion, comprising: Step 100: Receive active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node within the same detection period, compress and merge the measurement data to obtain a time-stamped fused measurement sequence. Step 200: For the fused measurement sequence, a joint active-passive track initiation algorithm is used within the defined confirmation area to determine candidate initial tracks; Step 300: For each candidate initial track, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and the radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, in order to complete the matching of measurement data and tracks. Step 400: Input the matched measurement data into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state; Step 500: When the track quality meets the terminal criteria, output the target identifier, six-dimensional state vector and covariance matrix to complete the multi-target tracking data fusion.

[0020] Preferably, within the same detection period, active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node are received. The measurement data are then compressed and merged to obtain a time-stamped fused measurement sequence, including: The raw point data reported by each radar node within the current detection period are synchronized and aligned according to the local timestamp to obtain a node-level time series; the detection period is determined by the radar scanning period. In each node-level time series, for multiple points whose spatial location deviation and Doppler / radiation source attribute differences are both below a preset threshold, point clustering is performed and a single compressed measurement vector is formed by weighted averaging to reduce the number of redundant points. The compressed single measurement vectors of each radar node are concatenated and sorted according to a unified network time index, and a "node identifier-time stamp" composite mark is added to each single compressed measurement vector to finally generate the fused measurement sequence with time stamp.

[0021] Preferably, for the fused measurement sequence, a joint active-passive track initiation algorithm is used within a set confirmation region to determine candidate initial tracks, including: Within the current detection cycle, a ring-shaped confirmation area is formed with each of the aforementioned fusion measurement sequences as the center, according to the outer radius and the inner radius; Active and passive measurement data that simultaneously satisfy the conditions of spatial position residual being less than a first threshold and radiation source main frequency difference being less than a second threshold are retrieved within the confirmed area, and the active and passive measurement data are combined to form a measurement pair set. For each measurement pair, using the two-dimensional position given by the active measurement and the radiation source parameters given by the passive measurement, the initial six-dimensional state vector and its covariance of the target are calculated using a constant velocity motion model and recorded as a temporary track record. During the subsequent detection period, a prediction-gating-association process is performed on each of the temporary track records; when the same temporary track record is successfully associated with a fusion measurement at least once within L consecutive detection periods, and the observation residual of each association is lower than a preset threshold, the temporary track record is promoted to the candidate initial track; where L≥2.

[0022] Specifically, such as Figure 2 As shown, the purpose of this invention is to provide a data fusion method for multi-target tracking by networked radar in complex electromagnetic environments, including a networked radar track initiation algorithm, a networked radar data association algorithm, and a networked radar tracking filtering algorithm under complex electromagnetic environments. This method can generate target track information at the networked radar fusion center based on the "residual" observation information from each radar station under complex electromagnetic environments.

[0023] The specific process of the network radar track initiation algorithm under complex electromagnetic environments in this embodiment is as follows: make This indicates that the network radar members, after point compression and merging, are in The first time received One measurement, among which , express The total number of measurements after compression and merging at any given moment. Record the network radar in... All measurements received at any given time are If measurement Originating from the goal Then the following measurement equation is satisfied: (1) in For the measurement matrix, For the goal State vector, measurement noise It is an independent Gaussian random process with zero mean and has a covariance matrix. ,like Clutter, on the other hand, satisfies a uniform distribution across the entire space. The track initiation process involves starting from the entire measurement dataset. The process extracts one or more measurement sequences originating from the target, and simultaneously provides an initial state estimate of the target. The target trajectory is included. The corresponding measurement sequence is used To indicate, here express Time and target trajectory The corresponding measurement number.

[0024] (1) Establishing the initial track Initial time Each received measurement establishes the initial trajectory starting point. If there is a rough estimate of the target's velocity, let its maximum value be... The minimum value is Assume the sensor uses a constant-period scanning method, with a scanning period of... Then, taking the starting point as the center, and respectively... and For radius formation The area of ​​time is confirmed, thereby establishing several initial tracks.

[0025] (2) Two-step extrapolation for Candidate target tracks at any given time Perform first-order polynomial extrapolation to form The time corresponds to the confirmed area of ​​this initial track, where To predict the measurement location, , Measurements were taken in the confirmed area; using... , Perform second-order polynomial extrapolation to obtain Measurement confirmation area at time (predicted measurement locations are respectively) and ).

[0026] (3) Calculate the cumulative interest For each measurement combination in the measurement confirmation area .

[0027] definition: (2) in: express Time and measurement combination Corresponding measurement number; The combination of measurement obtained through polynomial fitting The corresponding target state estimate.

[0028] make (3) Pick The measurement combination corresponding to the minimum value is used as Time Track Extended measurements, from Candidate target tracks at any time The departure track received a unique extension. (If it is confirmed that there is no measurement in the area, the candidate target track is deleted.) Then, the track is removed in the first step of the track establishment loop until the track start requirements are met.

[0029] Preferably, for each candidate initial trajectory, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, to complete the matching of measurement data and trajectories, including: Using the state estimate of the candidate initial track that has been updated in the previous detection cycle, the predicted position and velocity of the candidate initial track in the current detection cycle are calculated, and an ellipsoidal first gate region is delineated based on the predicted position and velocity to limit the measurement entries in the fused measurement sequence to be compared. For each measurement item that falls into the first gated area, the kinematic parameter difference and radiation source attribute difference of the measurement item are compared respectively, and then combined into a single similarity score according to a preset ratio. The measurement item is retained only when the single similarity score does not exceed the first threshold, resulting in a coarsely screened and retained measurement item; For the measurement items retained after coarse screening, the azimuth and pitch differences are further examined. Only when both the azimuth and pitch differences are within the corresponding angle thresholds are their directions considered to be consistent with the predicted directions; otherwise, they are discarded. For measurement items that pass the direction consistency check, a second threshold is applied for fine screening. If multiple measurement items still meet the second threshold, the one with the smallest similarity score is selected as the associated measurement with the candidate initial track in the current detection period. If no measurement item passes the fine screening, it is determined that the candidate initial track lacks effective measurement input in this period.

[0030] Specifically, the process of the network radar data association algorithm under interference conditions in this embodiment is as follows: This algorithm combines the joint statistical distance of the target state and attributes with the motion direction characteristics to reasonably limit the number of effective echoes in the association gate, and then uses the Fast Data Association (FAFDA) algorithm to process the echoes falling into the tracking gate.

[0031] (1) Determine the joint statistical distance tracking gate The tracking gate is determined using the statistical distance between spatial state parameters and radiation source attribute parameters. .

[0032] (4) In the formula, The threshold is determined by a pre-set prior probability of receiving a correct echo. It has been determined.

[0033] (5) In the formula, Indicates new information, express covariance, Indicates the predicted location of the target With effective echo The distance between them.

[0034] (6) In the formula, Let k be a vector containing the attribute information of the target at time k. This is a vector composed of the target attribute information corresponding to the tracking gate at time k-1; The covariance matrix of the measurement error of the target attribute information. For carrier frequency; For repetition frequency; Pulse width; This is the antenna scanning period.

[0035] (2) Restriction on direction of movement set up for Always on the target The filtered value; for Always on the target The predicted value, target The predicted direction of motion is and .

[0036] Limit the azimuth and pitch angles of the direction of motion. (7) (8) In the formula: It is a coefficient, representing the prior probability that a valid echo will fall into the tracking gate, as selected in advance. To determine; and This represents the standard deviation of the predicted direction.

[0037] (9) (10) (11) (12) (3) FAFDA algorithm Assuming in At any given moment, the radar must track within its detection area. One target, and received One measurement, namely ; .

[0038] Step 1: Determine the measurement points With known targets Interconnected simple events, determine and identify common point tracks—track pairs—within each target tracking gate, and calculate measurements. Falling into the target The probability density PDF within the tracking gate.

[0039] (13) (14) in: It is clutter density; It is the probability of radar detection; With zero mean and covariance matrix as The normal distribution density function.

[0040] Step 2: Track common points within the gate for each target. probability density value To attenuate, that is: (15) in, This represents the probability density value after decay.

[0041] Step 3: Calculate the interconnection probability (16) (17) Step 4: Attenuate the interconnection probability of common points falling within the tracking gates of each target, i.e. (18) Step 5: Calculate the new interconnection probability and ,Right now: (19) (20) The FAFDA algorithm considers the correlation between measurement points and target tracks, as well as the correlation between targets. It overweights measurement points that appear in only one track association domain, and underweights common measurement points that appear in several track association domains.

[0042] Preferably, the matched measurement data is input into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state, including: At the beginning of the current detection cycle, based on the trajectory state estimate of the previous cycle, a nonlinear state transition model is performed on each candidate trajectory to obtain the set of prior states at the beginning of the current cycle. The matched measurement data are sequentially injected into a robust filter to perform likelihood evaluation and weight correction on the prior state set, forming a weighted posterior state set. Calculate the effective representation of the posterior state set, and trigger a resampling operation when the effective representation is lower than a preset sampling threshold; During the resampling operation, the posterior state set is weighted and resampled according to the weight to generate a new equal-weight state set in order to suppress particle degradation. After the resampling operation is completed, the predicted state for the next detection period is calculated based on the new equal-weighted state set, and the predicted state and dispersion of the next detection period are used as the gating reference for the next period.

[0043] Optionally, the process of the network radar tracking filtering algorithm under interference conditions in this embodiment is as follows: The measurement data after measurement synthesis is filtered using the UPF algorithm.

[0044] (1) Initialization .

[0045] Uniformly sampled based on the initial mean and covariance. One particle. , , .

[0046] (2) Then, at each sampling time, UKF updates the Sigma point of each particle. In the UKF algorithm, due to the presence of noise terms, the state needs to be expanded. .

[0047] 1) For each particle, Sigma point sampling is performed. This embodiment uses a proportionally symmetric sampling method: (twenty one) In the formula: , .

[0048] (twenty two) (twenty three) 2) Prediction equation: (twenty four) (25) (26) (27) (28) (29) (30) 3) Update equation: (31) (32) (33) 4) Use the mean and variance estimated by UKF as the proposed distribution for calculating the weights of PF: ,in, The mean is covariance is The overall distribution.

[0049] (34) In interference environments, the observation noise is non-normally distributed, and the corresponding probability density function (i.e., the probability density function of the measurement noise) must be substituted into it for calculation.

[0050] Calculate the weights of all particles and normalize them.

[0051] 5) Output State estimation at time: State estimation: (35) Variance estimation: (36) (3) Resampling Evaluate particle weights to replicate high-weight particles and reduce low-weight particles, thus reproducing the old discrete particle swarm. With the new particle swarm Replace them and assign the same weights to each particle. The replacement must satisfy the following conditions: (37) (4) Recursion Return to step 2 and let .

[0052] Preferably, when the trajectory quality meets the terminal criteria, the target identifier, a six-dimensional state vector, and a covariance matrix are output to complete multi-target tracking data fusion, including: At the end of each detection cycle, for the updated track, the observation residuals, number of consistent valid measurements, state covariance convergence and duration of continuous existence of the most recent cycles are statistically analyzed and combined to form a track quality score. The track quality score is compared with a pre-set terminal criterion; when the track quality score meets the criterion requirements, the updated track is marked as "fusion confirmation status", otherwise it remains in "pending confirmation status". For tracks marked as "fusion confirmed state", a unique target identifier is assigned, and the three-dimensional position component and three-dimensional velocity component in the latest state estimate are extracted and combined into a six-dimensional state vector, while retaining the corresponding state covariance matrix. The unique target identifier, the six-dimensional state vector, and the covariance matrix are output to the result interface and simultaneously registered in the track management table; The output results are pushed to the upper command-control or display terminal using the aforementioned result interface to complete the final output of multi-target tracking data fusion.

[0053] Specifically, in this embodiment, during the target track initiation stage, the initial target track is established using observation data within the radar network and based on the network radar track initiation algorithm under complex electromagnetic environments. During the target track maintenance stage, the network radar data association algorithm under complex electromagnetic environments performs "measurement-track" association to determine new observation data for track correction. During the target track update stage, the associated observation data is used to filter and predict track information based on the network radar tracking filtering algorithm under complex electromagnetic environments.

[0054] Corresponding to the above method, this embodiment also discloses a networked radar multi-target tracking device based on joint fusion in complex electromagnetic environments, including: The data receiver and fusion unit is used to receive active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node within the same detection cycle, and to compress and merge the measurement data to obtain a time-stamped fused measurement sequence. An initial track candidate determiner is used to determine candidate initial tracks within a set confirmation region for the fused measurement sequence using a joint active-passive track initiation algorithm. A data association model builder is used to construct an improved multi-threshold fast data association model for each candidate initial track, based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, in order to complete the matching of measurement data and tracks. Nonlinear robust filters are used to input matched measurement data into nonlinear, non-Gaussian robust filters to recursively update and extrapolate the track state. The multi-target tracking output unit is used to output target identifiers, six-dimensional state vectors, and covariance matrices when the track quality meets the terminal criteria, so as to complete the fusion of multi-target tracking data.

[0055] The beneficial effects of this invention are as follows: (1) This invention constructs a multi-target tracking method that integrates active and passive measurement information in a complex electromagnetic environment, which effectively improves the target initiation and identification capabilities of radar systems under conditions of strong interference, deception signals and low signal-to-noise ratio. By combining active and passive track initiation mechanisms, the initiation sensitivity to low-observable targets is enhanced, overcoming the problem of strong dependence on initial observation in traditional methods.

[0056] (2) This invention introduces a multi-threshold fast data association model based on the joint statistical features of motion state and radiation source attributes. Combined with directional consistency constraints, it effectively reduces the risk of false association and missed association, and can still maintain track continuity and resolution in multi-target dense interference scenarios.

[0057] (3) To address nonlinear and non-Gaussian observation noise, this invention employs a robust filtering mechanism to support particle degradation detection and resampling operations, thereby improving the stability and adaptability of state estimation. In particular, it can maintain reliable tracking performance even when the target is highly maneuverable and measurements are intermittent or mismatched.

[0058] (4) Through track quality assessment and terminal criterion control, the present invention can dynamically output highly reliable target identification and state estimation results, ensuring that the final fused data has engineering usability and decision support value, and is suitable for the actual application deployment and application expansion of various radar networking systems.

[0059] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0060] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion, characterized in that, include: Within the same detection cycle, the active ranging / angle measurement data and passive radiation source parameter measurement data of each radar node are received, and the measurement data are compressed and merged to obtain a time-stamped fused measurement sequence. For the fused measurement sequence, a joint active-passive track initiation algorithm is used to determine candidate initial tracks within the set confirmation area; For each candidate initial trajectory, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and the radiation source attributes, as well as the directional consistency constraints of the target's predicted azimuth and pitch angles, in order to complete the matching of measurement data and trajectories. The matched measurement data is input into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state. When the track quality meets the terminal criteria, the target identifier, six-dimensional state vector, and covariance matrix are output to complete the multi-target tracking data fusion.

2. The multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion as described in claim 1, characterized in that, Within the same detection period, active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node are received. These measurement data are then compressed and merged to obtain a time-stamped fused measurement sequence, including: The raw point data reported by each radar node within the current detection period are synchronized and aligned according to the local timestamp to obtain a node-level time series; the detection period is determined by the radar scanning period. In each node-level time series, for multiple points whose spatial location deviation and Doppler / radiation source attribute differences are both below a preset threshold, point clustering is performed and a single compressed measurement vector is formed by weighted averaging to reduce the number of redundant points. The compressed single measurement vectors of each radar node are concatenated and sorted according to a unified network time index, and a "node identifier-time stamp" composite mark is added to each single compressed measurement vector to finally generate the fused measurement sequence with time stamp.

3. The multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion as described in claim 1, characterized in that, For the fused measurement sequence, a joint active-passive track initiation algorithm is used within the defined confirmation region to determine candidate initial tracks, including: Within the current detection cycle, a ring-shaped confirmation area is formed with each of the aforementioned fusion measurement sequences as the center, according to the outer radius and the inner radius; Active and passive measurement data that simultaneously satisfy the conditions of spatial position residual being less than a first threshold and radiation source main frequency difference being less than a second threshold are retrieved within the confirmed area, and the active and passive measurement data are combined to form a measurement pair set. For each measurement pair, using the two-dimensional position given by the active measurement and the radiation source parameters given by the passive measurement, the initial six-dimensional state vector and its covariance of the target are calculated using a constant velocity motion model and recorded as a temporary track record. During the subsequent detection period, a prediction-gating-association process is performed on each of the temporary track records; when the same temporary track record is successfully associated with a fusion measurement at least once within L consecutive detection periods, and the observation residual of each association is lower than a preset threshold, the temporary track record is promoted to the candidate initial track; where L≥2.

4. The multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion as described in claim 1, characterized in that, For each candidate initial trajectory, an improved multi-threshold fast data association model is constructed based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target's predicted azimuth and pitch angles, to complete the matching of measurement data and trajectories, including: Using the state estimate of the candidate initial track that has been updated in the previous detection cycle, the predicted position and velocity of the candidate initial track in the current detection cycle are calculated, and an ellipsoidal first gate region is delineated based on the predicted position and velocity to limit the measurement entries in the fused measurement sequence to be compared. For each measurement item that falls into the first gated area, the kinematic parameter difference and radiation source attribute difference of the measurement item are compared respectively, and then combined into a single similarity score according to a preset ratio. The measurement item is retained only when the single similarity score does not exceed the first threshold, resulting in a coarsely screened and retained measurement item; For the measurement items retained after coarse screening, the azimuth and pitch differences are further examined. Only when both the azimuth and pitch differences are within the corresponding angle thresholds are their directions considered to be consistent with the predicted directions; otherwise, they are discarded. For measurement items that pass the direction consistency check, a second threshold is applied for fine screening. If multiple measurement items still meet the second threshold, the one with the smallest similarity score is selected as the associated measurement with the candidate initial track in the current detection period. If no measurement item passes the fine screening, it is determined that the candidate initial track lacks effective measurement input in this period.

5. The multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion as described in claim 1, characterized in that, The matched measurement data is input into a nonlinear, non-Gaussian robust filter to recursively update and extrapolate the track state, including: At the beginning of the current detection cycle, based on the trajectory state estimate of the previous cycle, a nonlinear state transition model is performed on each candidate trajectory to obtain the set of prior states at the beginning of the current cycle. The matched measurement data are sequentially injected into a robust filter to perform likelihood evaluation and weight correction on the prior state set, forming a weighted posterior state set. Calculate the effective representation of the posterior state set, and trigger a resampling operation when the effective representation is lower than a preset sampling threshold; During the resampling operation, the posterior state set is weighted and resampled according to the weight size to generate a new equal-weight state set in order to suppress particle degradation. After the resampling operation is completed, the predicted state for the next detection period is calculated based on the new equal-weighted state set, and the predicted state and dispersion of the next detection period are used as the gating reference for the next period.

6. The multi-target tracking method for networked radar in complex electromagnetic environments based on joint fusion as described in claim 1, characterized in that, When the trajectory quality meets the terminal criteria, the target identifier, six-dimensional state vector, and covariance matrix are output to complete the multi-target tracking data fusion, including: At the end of each detection cycle, for the updated track, the observation residuals, number of consistent valid measurements, state covariance convergence and duration of continuous existence of the most recent cycles are statistically analyzed and combined to form a track quality score. The track quality score is compared with a pre-set terminal criterion; when the track quality score meets the criterion requirements, the updated track is marked as "fusion confirmation status", otherwise it remains in "pending confirmation status"; For tracks marked as "fusion confirmation state", a unique target identifier is assigned, and the three-dimensional position component and three-dimensional velocity component in the latest state estimate are extracted and combined into a six-dimensional state vector, while retaining the corresponding state covariance matrix. The unique target identifier, the six-dimensional state vector, and the covariance matrix are output to the result interface and simultaneously registered in the track management table; The output results are pushed to the upper command-control or display terminal using the aforementioned result interface to complete the final output of multi-target tracking data fusion.

7. A networked radar multi-target tracking device based on joint fusion in a complex electromagnetic environment, characterized in that, include: The data receiver and fusion unit is used to receive active ranging / angle measurement data and passive radiation source parameter measurement data from each radar node within the same detection cycle, and to compress and merge the measurement data to obtain a time-stamped fused measurement sequence. An initial track candidate determiner is used to determine candidate initial tracks within a set confirmation region for the fused measurement sequence using a joint active-passive track initiation algorithm. A data association model builder is used to construct an improved multi-threshold fast data association model for each candidate initial track, based on the joint statistical distance between the target motion state and radiation source attributes, as well as the directional consistency constraints of the target predicted azimuth and pitch angles, in order to complete the matching of measurement data and tracks. Nonlinear robust filters are used to input matched measurement data into nonlinear, non-Gaussian robust filters to recursively update and extrapolate the track state. The multi-target tracking output unit is used to output target identifiers, six-dimensional state vectors, and covariance matrices when the track quality meets the terminal criteria, so as to complete the fusion of multi-target tracking data.

Images