A multi-scene configuration-oriented multi-sensor cooperative tracking fusion system

Abstract

A multi-scene configuration-oriented multi-sensor cooperative tracking fusion system belongs to the technical field of target tracking and information fusion. In order to solve the problem of insufficient configuration flexibility of the multi-sensor tracking fusion system, the shipborne and airborne platform cooperative tracking fusion subsystem based on file configuration constructs an integrated processing flow around scene parameter configuration, local tracking, track management, track association and feedback fusion; the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraint combines the real motion characteristics of the target with tracking fusion by establishing the motion models of multiple targets such as high-speed aircraft, accompanying body target and split target; the active and passive radar tracking fusion subsystem oriented to intelligent configuration of data flow carries out cooperative processing of multiple source data streams around multi-station passive radar positioning, active radar tracking, identity discrimination, time alignment and state completion; the three subsystems generate system-level track results, which are merged into a general system situation map generation module for display.

Description

Technical Field

[0001] This invention belongs to the field of target tracking and information fusion technology, specifically relating to a multi-sensor collaborative tracking and fusion system for multi-scenario configuration. Background Technology

[0002] In complex environments, a single platform or sensor can no longer meet the requirements for large-scale, continuous, and high-precision target detection. With the development of multi-platform collaboration, multi-domain linkage, and multi-source information fusion, how to uniformly organize, continuously track, and effectively fuse target information from different platforms, systems, and arrival methods has become a key issue in multi-sensor information processing.

[0003] Existing multi-sensor tracking fusion systems typically suffer from the following shortcomings: First, the system deployment is tightly coupled with the mission scenario. Platform layout, sensor parameters, target attributes, and environmental conditions often rely on manual hard-coding settings, lacking a unified and flexible mission configuration mechanism, resulting in insufficient system scalability and scenario adaptability. Second, in shipborne and airborne platform collaborative detection scenarios, the spatial distribution of different platforms is discrete, and the observation conditions vary significantly. Local tracking results require further trajectory correlation and fusion, making the overall system process quite complex. Third, in high-speed aircraft target detection scenarios, the target's flight speed is high, its motion phases are distinct, and it may be accompanied by multiple types of targets such as accompanying targets and split targets. If only simplified models are used, it is difficult to accurately reflect the real motion patterns, thus affecting subsequent state estimation and fusion effects. Fourth, in active and passive radar collaborative scenarios, there are significant differences between sensors in terms of measurement methods, data arrival order, and time consistency, which can easily lead to problems such as interruptions, out-of-order data, and local missing data, thereby affecting the fusion center's ability to identify the identity of targets and the collaborative tracking fusion performance.

[0004] While existing research has developed numerous methods for target tracking, trajectory association, and information fusion, most methods still focus on a single, independent processing step. There is a lack of a unified system architecture capable of coordinating different scenario subsystems within a single system framework to complete target modeling and tracking, trajectory association, trajectory fusion, and intelligent processing of multi-source data streams. Due to this lack of a unified system architecture, it is difficult to uniformly access, organize, and output system-level target trajectory results generated in different scenarios, and it is also difficult to form a comprehensive situational awareness result within the same framework. Summary of the Invention

[0005] The problem this invention aims to solve is to overcome the lack of flexibility in the configuration of existing multi-sensor tracking fusion systems for various mission scenarios, and to propose a multi-sensor collaborative tracking fusion system for multi-scenario configuration.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A multi-sensor collaborative tracking and fusion system for multi-scenario configuration includes a file-based shipborne and airborne platform collaborative tracking and fusion subsystem, a high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and an active and passive radar tracking fusion subsystem for intelligent configuration based on data flow.

[0008] The file-based shipborne and airborne platform collaborative tracking fusion subsystem, the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and the active and passive radar tracking fusion subsystem oriented to intelligent configuration of data flow are respectively connected to the overall system situation map generation module.

[0009] The file-based shipborne and airborne platform collaborative tracking fusion subsystem builds an integrated processing flow around scene parameter configuration, local tracking, track management, track association and feedback fusion;

[0010] The high-speed vehicle target modeling and tracking fusion subsystem under real parameter constraints combines the real motion characteristics of targets with tracking fusion by establishing multiple target motion models such as high-speed vehicles, companion targets, and split targets.

[0011] The active and passive radar tracking fusion subsystem for intelligent configuration of data streams carries out multi-source data stream collaborative processing around multi-station passive radar positioning, active radar tracking, identity discrimination, time alignment and state completion.

[0012] The three subsystems generate system-level target trajectory results for their respective scenarios and input them into the overall system situation map generation module to achieve comprehensive organization and unified display of target information across multiple scenarios.

[0013] Furthermore, the method for establishing the file-configured shipborne and airborne platform collaborative tracking fusion subsystem includes the following steps:

[0014] S1.1. By configuring the shipborne platform, airborne platform, aerial targets and environmental conditions in a unified manner through configuration files, a collaborative detection mission scenario between shipborne and airborne platforms is constructed.

[0015] S1.2. Each platform sensor processes the local observation information and uses an IMM-GLMB filter to complete target tracking and local track generation; let the measurement set received during the update phase be... Calculate the local multi-target posterior density, and then extract the local track state set from the local multi-target posterior density. The calculation formula is as follows:

[0016]

[0017] in, For the first State extraction operator for each track. This represents the local multi-object posterior density. Represents a labeled set of multi-objective states. Indicates the first The first platform The flight path at time State estimation, No. The first platform The flight path at time The probability of its existence. The threshold for track confirmation;

[0018] S1.3. When a local track is interrupted, the position change during the interruption time interval and the prior maximum speed are used to construct a correlation criterion. The two-dimensional allocation cost function is combined to complete the interruption correlation, thereby achieving unified interruption track numbering and track completion.

[0019] S1.4. For local tracks generated by different platforms, a track association method based on shared Transformer coding is used to perform cross-platform track association. The association decision variables are solved by using the association probability of candidate track pairs, and the associated track groups are extracted.

[0020] S1.5. For the local tracks that have been associated, the generalized covariance cross criterion is used for fusion to obtain the fusion center. A system-level flight path at time State estimation Covariance ;

[0021] S1.6. In the feedback fusion mode, the system trajectory obtained in step S1.5 is fed back to the local platform to predict the local trajectory at the next moment;

[0022] S1.7. The OSPA distance is used to evaluate the quality of the local tracks of each platform and the fused system tracks, resulting in a set of system-level tracks output by the file-configured shipborne and airborne platform collaborative tracking fusion subsystem.

[0023] Furthermore, the method for establishing the high-speed vehicle target modeling and tracking fusion subsystem under the constraints of real parameters includes the following steps:

[0024] S2.1. Establish a target motion model based on the true parameters of the high-speed aircraft target, including constructing the target state vector from the position vector and velocity vector, and then constructing a motion model constrained by the true parameters, the expression of which is:

[0025]

[0026]

[0027] in, For position vectors, For velocity vectors, Let be the target state vector. Indicates the first A motion model constrained by real parameters This represents the set of true parameters corresponding to the model. Let T be the first derivative of the target state vector, and let T denote the transpose.

[0028] S2.2. In the target tracking phase, the UKF-JIPDA method based on interactive multi-models is adopted for any local track. Calculate the first The state estimates under each model are then combined with the IMM fusion method to obtain the local track state estimates and covariance.

[0029] S2.3. The local tracks obtained in step S2.2 are associated using a track association method based on shared Transformer coding. Then, in the multi-sensor track fusion stage, the local estimates of multiple platforms are gradually fused using a sequential inverse covariance cross method until the sequential fusion of multiple sensors is completed, resulting in a system-level track set output by the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints.

[0030] Furthermore, the method for establishing the active and passive radar tracking fusion subsystem for intelligent configuration of data streams includes the following steps:

[0031] S3.1. A passive radar positioning method based on chi-square distribution statistics is used to screen multi-station direction finding line combinations; a chi-square statistic and corresponding correlation likelihood function are constructed based on the candidate direction finding line combinations, assuming the three observation stations are at time... The correlation likelihood function corresponding to the candidate direction finding line combination is: Introducing binary assignment variables Then the total likelihood function Represented as:

[0032] ;

[0033] The final association result was obtained by solving the following optimization problem:

[0034]

[0035]

[0036] in, , , These represent the times of the three observation stations. The number of candidate direction finding lines formed Indicates the first , , Do the direction finding lines form a valid combination?

[0037] The set of local tracks for passive radar is then represented as:

[0038]

[0039] in, Indicates time The set of local tracks of passive radar. Indicates the source of the passive radar track. Indicates the first Local track status of passive radar;

[0040] S3.2. To address the issue of track association in interruptions, an identity discrimination method based on RWKV time-series modeling is adopted, combined with the local track set obtained by IMM-GLMB tracking using active radar. Constructing local track sets for active and passive radar :

[0041]

[0042] in, Indicates the active radar at time The resulting set of local tracks;

[0043] when When there is an interruption in the flight path, boundary-aware features are extracted for both the old and new flight path segments and mapped to a unified metric space. Let their normalized representations be respectively... and Then its associated score Represented as:

[0044] ;

[0045] A score matrix is ​​constructed based on the correlation scores of all candidate pairs. Under one-to-one constraints, the Hungarian algorithm is used to perform global matching on candidate pairs that meet the threshold conditions, resulting in a set of local tracks after the interruption of continuity.

[0046]

[0047] in, Indicates by The set of local tracks obtained after determining the identity of interrupted tracks. express Local flight paths in This represents the threshold for identity discrimination. This represents the associated confirmation variable after a one-to-one match;

[0048] S3.3. To address the issues of out-of-order routing and local missing data, a trajectory completion method based on neural controlled differential equations is adopted, combining timestamps and state availability information to construct a continuous control path:

[0049]

[0050] in, Indicates by The constructed continuous control path This represents the constructor for a continuous control path.

[0051] The hidden state propagates in the continuous time domain driven by a continuous control path, and the continuous time modeling form is as follows:

[0052]

[0053] in, Indicates time The hidden state, This represents the initial hidden state. This represents a continuous-time vector field function parameterized by a neural network;

[0054] Next, the hidden states at each time step are decoded to obtain the completed state output:

[0055]

[0056] in, Represents the state decoding function. Indicates the length of the local track window. Indicates the first Output the completed state corresponding to each moment;

[0057] S3.4. For the local tracks generated by different sensors after processing in steps S3.1-S3.3, a multi-sensor track association method based on shared Transformer coding is used to associate the tracks, extract the associated track groups, and perform track fusion using the generalized covariance cross criterion. Then, the OSPA distance is used to evaluate the track quality, resulting in a system-level track set output by the active and passive radar tracking fusion subsystem for intelligent configuration of data stream.

[0058] Furthermore, the overall system situation map generation module receives the system-level track set generated by the three subsystems, and organizes, aggregates and displays target information from multiple scenarios in a unified manner. Based on unified time, unified coordinates and unified attribute descriptions, it comprehensively manages the system-level tracks to form a unified situation map for displaying target information from multiple scenarios.

[0059] The beneficial effects of this invention are:

[0060] This invention discloses a multi-sensor collaborative tracking and fusion system for multiple scenarios, which organically combines system configuration, target modeling and tracking, trajectory association, trajectory fusion, feedback correction, and intelligent data stream configuration to form a unified processing framework suitable for different scenario requirements. Compared with existing multi-sensor tracking and fusion methods, this invention not only achieves stable tracking and effective fusion under collaborative detection conditions of shipborne and airborne platforms, but also extends to high-speed aircraft target tracking and fusion scenarios under real parameter constraints and collaborative processing scenarios of heterogeneous data streams from active and passive radars. The system-level target trajectory results generated by each scenario subsystem are further uniformly input into the overall system situation map generation module, realizing the comprehensive organization and display of target information across multiple scenarios, and possessing good scenario adaptability, module scalability, and engineering application value. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of the structure of a multi-sensor collaborative tracking and fusion system for multi-scenario configuration according to the present invention;

[0062] Figure 2 Figure (a) shows the GCI fusion result of the shipborne and airborne platform collaborative tracking based on file configuration as described in this invention, and Figure (b) shows the OSPA distance curves corresponding to each local platform and the fusion result.

[0063] Figure 3 Figure (a) shows the SICI track fusion result of high-speed aircraft target tracking under real parameter constraints as described in this invention, and Figure (b) shows the OSPA distance curves corresponding to each local sensor and the fusion result.

[0064] Figure 4 Figure (a) shows the GCI fusion result of active and passive radar tracking in the intelligent configuration of data flow described in this invention, and Figure (b) shows the OSPA range curves corresponding to active radar, passive radar and fusion result. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0066] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0067] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 - Appendix Figure 4 Detailed explanation is as follows:

[0068] Example 1:

[0069] A multi-sensor collaborative tracking and fusion system for multi-scenario configuration includes a file-based shipborne and airborne platform collaborative tracking and fusion subsystem, a high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and an active and passive radar tracking fusion subsystem for intelligent configuration based on data flow.

[0070] The file-based shipborne and airborne platform collaborative tracking fusion subsystem, the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and the active and passive radar tracking fusion subsystem oriented to intelligent configuration of data flow are respectively connected to the overall system situation map generation module.

[0071] The file-based shipborne and airborne platform collaborative tracking fusion subsystem builds an integrated processing flow around scene parameter configuration, local tracking, track management, track association and feedback fusion;

[0072] The high-speed vehicle target modeling and tracking fusion subsystem under real parameter constraints combines the real motion characteristics of targets with tracking fusion by establishing multiple target motion models such as high-speed vehicles, companion targets, and split targets.

[0073] The active and passive radar tracking fusion subsystem for intelligent configuration of data streams carries out multi-source data stream collaborative processing around multi-station passive radar positioning, active radar tracking, identity discrimination, time alignment and state completion.

[0074] The three subsystems generate system-level target trajectory results for their respective scenarios and input them into the overall system situation map generation module to achieve comprehensive organization and unified display of target information across multiple scenarios.

[0075] Furthermore, the file-configured shipborne and airborne platform collaborative tracking fusion subsystem is the first type of scenario subsystem in the overall system. This subsystem takes a unified scenario configuration file as input, completes the unified configuration of parameters such as platform, target, and environment, and sequentially performs local measurement generation, local target tracking, interrupted track completion, cross-platform track association, system-level track fusion, feedback correction, and track quality assessment under the same scenario constraints. This subsystem ultimately outputs system-level target track results under the shipborne and airborne platform collaborative detection scenario, and inputs these results into the overall system situation map generation module for the overall system's comprehensive organization and unified display of target information from multiple scenarios.

[0076] Furthermore, the method for establishing the file-configured shipborne and airborne platform collaborative tracking fusion subsystem includes the following steps:

[0077] S1.1. By configuring the shipborne platform, airborne platform, aerial targets and environmental conditions in a unified manner through configuration files, a collaborative detection mission scenario between shipborne and airborne platforms is constructed.

[0078] S1.2. Each platform sensor processes the local observation information and uses an IMM-GLMB filter to complete target tracking and local track generation; let the measurement set received during the update phase be... Calculate the local multi-object posterior density;

[0079] Furthermore, the local multi-objective posterior density is expressed as:

[0080]

[0081] in, This represents the updated labeled multi-objective state set. Indicates the set of measurements used for the current update. This indicates a uniqueness constraint for the label. To predict the label space, For the reason The set of finite labeled subsets constituted For the updated set of labels, and These are the historical association hypothesis sets and their indices. and These are the current measurement association mapping set and its mappings. For the corresponding assumed weights, The set of labels corresponding to the set of multi-objective states. For the consistency constraint of the label set, This represents the single-objective state probability density under the corresponding assumption;

[0082] Then, the local track state set is obtained by extracting the local multi-object posterior density, and the calculation formula is as follows:

[0083]

[0084] in, For the first State extraction operator for each track. This represents the local multi-object posterior density. Represents a labeled set of multi-objective states. Indicates the first The first platform The flight path at time State estimation, No. The first platform The flight path at time The probability of its existence. The threshold for track confirmation;

[0085] S1.3. When a local track is interrupted, the position change during the interruption time interval and the prior maximum speed are used to construct a correlation criterion. The two-dimensional allocation cost function is combined to complete the interruption correlation, thereby achieving unified interruption track numbering and track completion.

[0086] Furthermore, a correlation criterion is constructed using the position change during the interruption time interval and the prior maximum velocity, namely:

[0087]

[0088] in, and They represent the first Old flight path segments of each platform At the end and new flight path segments on the same platform At the start time The state estimation, where the distance calculation takes the position components of both, is performed in the formula. and These represent the end time of the old track segment and the start time of the new track segment, respectively. Indicates the target's prior maximum speed. This represents a constant. Finally, the two-dimensional allocation cost function is used to complete the interrupt association, achieving unified interrupt track numbers and track completion.

[0089] S1.4. For local tracks generated by different platforms, a track association method based on shared Transformer coding is used to perform cross-platform track association. The association decision variables are solved by using the association probability of candidate track pairs, and the associated track groups are extracted.

[0090] Furthermore, let's set the first platform as the first... The track state sequence is The second platform's first The track state sequence is Then the association probability of the candidate track pair is:

[0091]

[0092] in, For the first track and the first The correlation probability of the flight paths To share Transformer track association functions, This represents the Sigmoid activation function. According to... Construct the correlation matrix and solve for the correlation decision variables under one-to-one constraints. :

[0093]

[0094] in, and These represent the number of tracks formed by the two sensors, respectively. When... When, it indicates the first track and the first The two tracks were determined to be tracks for the same target. This constitutes a group of tracks associated with the same target:

[0095]

[0096] in, Indicates the first A group of tracks associated with the same target.

[0097] S1.5. For the local tracks that have been associated, the generalized covariance cross criterion is used for fusion to obtain the fusion center. A system-level flight path at time State estimation Covariance ;

[0098] Furthermore, let's assume Indicates the first The first platform The trajectory has a state estimate and covariance of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ... and The GCI fusion result is:

[0099]

[0100]

[0101] in, Represents the fusion weights, satisfying and . and They represent the first fusion center. A system-level flight path at time State estimation and covariance;

[0102] S1.6. In the feedback fusion mode, the system trajectory obtained in step S1.5 is fed back to the local platform to predict the local trajectory at the next moment;

[0103] Furthermore, the formula for predicting the local trajectory at the next moment is:

[0104]

[0105]

[0106] in, and They represent the first The first platform Next-moment state prediction and covariance of a local flight path. Represents the state transition matrix. Represents the process noise input matrix. This represents the process noise covariance matrix.

[0107] S1.7. The OSPA distance is used to evaluate the quality of the local tracks of each platform and the fused system tracks, resulting in a set of system-level tracks output by the file-configured shipborne and airborne platform collaborative tracking fusion subsystem.

[0108] Furthermore, let the true target set be... The set of tracks to be evaluated is uniformly represented as , If we represent a specific platform or convergence center, then the OSPA distance is defined as:

[0109]

[0110] in, Indicates the first A real goal at any moment state, Represents the set of tracks to be evaluated Central arrangement The corresponding estimated state, and They are respectively and The momentum, Let be the set of permutations of the track set numbers to be evaluated. Distance calculation uses the position component from the state. To cut off the distance, denoted as the distance order.

[0111] System-level track output and situation map integration. After track fusion, feedback correction, and quality assessment, the system-level track results obtained from the fusion center are input into the overall system situation map generation module for unified organization and display of target information across multiple scenarios. Assume the shipborne and airborne platform collaborative tracking fusion subsystem is at time... The output system-level track set is as follows:

[0112]

[0113] in, and The aforementioned GCI fusion centers are respectively the fusion centers obtained by fusion. System-level track state estimation and covariance, This refers to the number of system-level tracks.

[0114] The high-speed vehicle target modeling and tracking fusion subsystem under realistic parameter constraints is the second type of scenario subsystem in the overall system. This subsystem is designed for high-speed vehicle target detection scenarios, addressing the characteristics of high target flight speed, distinct motion phases, complex motion patterns, and the simultaneous appearance of multiple target types such as accompanying targets and split targets. Supported by a unified tracking and fusion processing flow, it introduces a target modeling method constrained by realistic parameters. This is combined with the UKF-JIPDA target tracking algorithm based on interactive multi-models and the sequential inverse covariance cross-track fusion algorithm to achieve target state estimation and multi-sensor track fusion in this scenario. The key to this subsystem is establishing motion models for different types of targets based on realistic parameters, and then generating system-level target track results for the high-speed vehicle target detection scenario, which are ultimately input into the overall system situation map generation module.

[0115] Furthermore, the method for establishing the high-speed vehicle target modeling and tracking fusion subsystem under the constraints of real parameters includes the following steps:

[0116] S2.1. Establish a target motion model based on the true parameters of the high-speed aircraft target, including constructing the target state vector from the position vector and velocity vector, and then constructing a motion model constrained by the true parameters, the expression of which is:

[0117]

[0118]

[0119] in, For position vectors, For velocity vectors, Let be the target state vector. Indicates the first A motion model constrained by real parameters This represents the set of true parameters corresponding to the model. Let T be the first derivative of the target state vector, and let T denote the transpose.

[0120] Furthermore, let the unified motion model of the high-speed aircraft in each flight phase be:

[0121]

[0122]

[0123] in,

[0124]

[0125]

[0126]

[0127] in, and Indicates stage indicator parameters; the active segment has... Free flight segment has Re-entry stage has ; Represents a position vector; Represents the velocity vector. ; Indicates engine thrust; Indicates the time of high-speed aircraft The quality; Represents the acceleration due to gravity. Represents the Earth's gravitational constant; This represents the aerodynamic drag acceleration term. Indicates the drag coefficient. Indicates the reference area of ​​a high-speed aircraft. Indicates atmospheric density; This represents the harmonic perturbation acceleration of the Earth's second zone. This represents the harmonic coefficient of the Earth's second zone. It represents the radius of the Earth's equator.

[0128] Based on this, we further model the release process of the accompanying target and the generation process of the split target to describe the release and flight motion characteristics of the accompanying target, as well as the multi-batch generation, random velocity impulse and destruction characteristics of the split target.

[0129] S2.2. In the target tracking phase, the UKF-JIPDA method based on interactive multi-models is adopted for any local track. Calculate the first The state estimates under each model are then combined with the IMM fusion method to obtain the local track state estimates and covariance.

[0130] Furthermore, for any local track In its first The predicted state under each model is:

[0131]

[0132] in Indicates the first The track is in The initial state estimates obtained through interactive mixing under each model Indicates the first The parameter set of a true parameter-constrained model. Indicates the first The state prediction function corresponding to a real parameter constrained model.

[0133] After the UKF-JIPDA update, the The state estimates under each model are:

[0134]

[0135] in, Represents a set of related events. Indicates the current measurement set, Indicates a joint related event. Indicates the first A motion model, Indicates the first Under the model, in the joint related events The updated state estimate obtained under the given conditions.

[0136] Further IMM fusion based on the model's posterior probability yields the local track state estimate and covariance:

[0137]

[0138]

[0139] in, Indicates the first The updated state estimation covariance under each model Indicates the number of models. Indicates the first The track is in Posterior probabilities under each model;

[0140] S2.3. The local tracks obtained in step S2.2 are associated using a track association method based on shared Transformer coding. Then, in the multi-sensor track fusion stage, the local estimates of multiple platforms are gradually fused using a sequential inverse covariance cross method until the sequential fusion of multiple sensors is completed, resulting in a system-level track set output by the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints.

[0141] Furthermore, suppose the results of the two associated local tracks of the same target are as follows: and Then the consistent fusion estimator of the two is:

[0142]

[0143]

[0144]

[0145]

[0146] in, Indicates the fusion weights. and Let represent the state estimate and covariance after ICI fusion, respectively. Based on this, the current fusion result is recursively fused with the next local sensor estimate in the same way until multi-sensor sequential fusion is completed.

[0147] Furthermore, system-level track output and situation map integration are implemented. After completing target modeling and tracking, track association, SICI fusion, and OSPA track quality assessment, the fused system-level track results are input into the overall system situation map generation module for unified organization and display of target information across multiple scenarios. Assume the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints is at time... The output system-level track set is

[0148]

[0149] in, and They represent the first and second digits obtained by sequential fusion, respectively. System-level track state estimation and covariance, This indicates the number of system-level tracks.

[0150] The active / passive radar tracking fusion subsystem with intelligent data stream configuration is the third scenario subsystem in the overall system. This subsystem is designed for active / passive radar cooperative detection scenarios. Addressing the differences between active and passive sensors in observation methods, data arrival order, and time consistency, which can easily lead to track interruptions, data disorder, and partial missing data, this subsystem adds an intelligent data stream configuration step before track association and fusion. It integrates statistical association, passive positioning, intelligent identity discrimination, and continuous time state completion methods into a unified processing flow, achieving time alignment, identity discrimination, state completion, and cooperative tracking fusion processing under multi-source data stream conditions. This subsystem ultimately generates system-level target track results for active / passive radar cooperative detection scenarios and inputs them into the overall system situation map generation module.

[0151] Furthermore, the method for establishing the active and passive radar tracking fusion subsystem for intelligent configuration of data streams includes the following steps:

[0152] S3.1. A passive radar positioning method based on chi-square distribution statistics is used to screen multi-station direction finding line combinations; a chi-square statistic and corresponding correlation likelihood function are constructed based on the candidate direction finding line combinations, assuming the three observation stations are at time... The correlation likelihood function corresponding to the candidate direction finding line combination is: Introducing binary assignment variables Then the total likelihood function Represented as:

[0153] ;

[0154] The final association result was obtained by solving the following optimization problem:

[0155]

[0156]

[0157] in, , , These represent the times of the three observation stations. The number of candidate direction finding lines formed Indicates the first , , Do the direction finding lines form a valid combination?

[0158] The set of local tracks for passive radar is then represented as:

[0159]

[0160] in, Indicates time The set of local tracks of passive radar. Indicates the source of the passive radar track. Indicates the first Local track status of passive radar;

[0161] S3.2. To address the issue of track association in interruptions, an identity discrimination method based on RWKV time-series modeling is adopted, combined with the local track set obtained by IMM-GLMB tracking using active radar. Constructing local track sets for active and passive radar :

[0162]

[0163] in, Indicates the active radar at time The resulting set of local tracks;

[0164] when When there is an interruption in the flight path, boundary-aware features are extracted for both the old and new flight path segments and mapped to a unified metric space. Let their normalized representations be respectively... and Then its associated score Represented as:

[0165] ;

[0166] A score matrix is ​​constructed based on the correlation scores of all candidate pairs. Under one-to-one constraints, the Hungarian algorithm is used to perform global matching on candidate pairs that meet the threshold conditions, resulting in a set of local tracks after the interruption of continuity.

[0167]

[0168] in, Indicates by The set of local tracks obtained after determining the identity of interrupted tracks. express Local flight paths in This represents the threshold for identity discrimination. This represents the associated confirmation variable after a one-to-one match;

[0169] S3.3. To address the issues of out-of-order routing and local missing data, a trajectory completion method based on neural controlled differential equations is adopted, combining timestamps and state availability information to construct a continuous control path:

[0170]

[0171] in, Indicates by The constructed continuous control path This represents the constructor for a continuous control path.

[0172] The hidden state propagates in the continuous time domain driven by a continuous control path, and the continuous time modeling form is as follows:

[0173]

[0174] in, Indicates time The hidden state, This represents the initial hidden state. This represents a continuous-time vector field function parameterized by a neural network;

[0175] Next, the hidden states at each time step are decoded to obtain the completed state output:

[0176]

[0177] in, Represents the state decoding function. Indicates the length of the local track window. Indicates the first Output the completed state corresponding to each moment;

[0178] S3.4. For the local tracks generated by different sensors after processing in steps S3.1-S3.3, a track association method based on shared Transformer coding is used to associate multi-sensor tracks, extract associated track groups, and perform track fusion using the generalized covariance cross criterion. Then, the OSPA distance assessment is performed to evaluate track quality, resulting in a system-level track set output by the active and passive radar tracking fusion subsystem for intelligent configuration of data stream.

[0179] Furthermore, the system-level fusion trajectory results are as follows:

[0180]

[0181] in, This indicates that the active and passive radar tracking fusion subsystem is at time [time missing]. The output system-level track set and They represent the first fusion center. State estimation and covariance of system-level tracks, This indicates the number of system-level tracks. This set of system-level tracks is input into the overall system situation map generation module and is used for the unified organization and display of target information across multiple scenarios.

[0182] Furthermore, the overall system situation map generation module receives the system-level track set generated by the three subsystems, and organizes, aggregates and displays target information from multiple scenarios in a unified manner. Based on unified time, unified coordinates and unified attribute descriptions, it comprehensively manages the system-level tracks to form a unified situation map for displaying target information from multiple scenarios.

[0183] Furthermore, the working method of the overall system situation diagram generation module includes the following steps:

[0184] S4.1. System-level track access and unified organization. Receives data from the three subsystems at specific times. The output system-level track set forms the total system input track set:

[0185]

[0186] in, , and These represent the system-level track sets output by the three scenario subsystems, Indicates the total system at time [time]. The received system-level track set;

[0187] S4.2. Spatial location-based track consistency aggregation; for the total system input track set Spatial consistency is determined for the flight paths within the data. Let... For any two system-level tracks, their corresponding state estimates are as follows: and When the distance between the positional components of two tracks is less than a set spatial threshold, they are considered as candidate consistent tracks. With the assistance of target attribute, temporal consistency, and track quality assessment results, an aggregation judgment is made to form a unified situational target set. .

[0188] S4.3. Situation map generation; unifying the set of situational targets. The input situation map generation and display module displays information uniformly based on target location, track identifiers, track quality, and source scenario. This enables the comprehensive organization and visualization of the output results from the three scenario subsystems within the same situation map.

[0189] Table 1 is a comparison table of the average OSPA distances of each shipborne platform, airborne platform, and fused track in the file-based collaborative tracking fusion scenario of shipborne and airborne platforms described in this embodiment, under the two modes of fusion without feedback and fusion with feedback. It is used to illustrate the improvement of track quality of each local platform and system-level track under the condition of fusion with feedback.

[0190] Table 1

[0191]

[0192] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0193] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A multi-sensor collaborative tracking and fusion system for multi-scenario configuration, characterized in that, This includes a file-based configuration-based shipborne and airborne platform collaborative tracking fusion subsystem, a high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and an active and passive radar tracking fusion subsystem oriented towards intelligent configuration of data streams; The file-based shipborne and airborne platform collaborative tracking fusion subsystem, the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints, and the active and passive radar tracking fusion subsystem oriented to intelligent configuration of data flow are respectively connected to the overall system situation map generation module. The file-based shipborne and airborne platform collaborative tracking fusion subsystem builds an integrated processing flow around scene parameter configuration, local tracking, track management, track association and feedback fusion; The high-speed vehicle target modeling and tracking fusion subsystem under real parameter constraints combines the real motion characteristics of targets with tracking fusion by establishing multiple target motion models such as high-speed vehicles, companion targets, and split targets. The active and passive radar tracking fusion subsystem for intelligent configuration of data streams carries out multi-source data stream collaborative processing around multi-station passive radar positioning, active radar tracking, identity discrimination, time alignment and state completion. The three subsystems generate system-level target trajectory results for their respective scenarios and input them into the overall system situation map generation module to achieve comprehensive organization and unified display of target information across multiple scenarios.

2. The multi-sensor collaborative tracking and fusion system for multi-scenario configuration according to claim 1, characterized in that, The method for establishing the file-configured shipborne and airborne platform collaborative tracking fusion subsystem includes the following steps: S1.

1. By configuring the shipborne platform, airborne platform, aerial targets and environmental conditions in a unified manner through configuration files, a collaborative detection mission scenario between shipborne and airborne platforms is constructed. S1.

2. Each platform sensor processes the local observation information and uses an IMM-GLMB filter to complete target tracking and local track generation; let the measurement set received during the update phase be... Calculate the local multi-target posterior density, and then extract the local track state set from the local multi-target posterior density. The calculation formula is as follows: in, Indicates the first State extraction operator for each track. This represents the local multi-object posterior density. Represents a labeled set of multi-objective states. Indicates the first The first platform The flight path at time State estimation, No. The first platform The flight path at time The probability of its existence. The threshold for track confirmation; S1.

3. When a local track is interrupted, the position change during the interruption time interval and the prior maximum speed are used to construct a correlation criterion. The two-dimensional allocation cost function is combined to complete the interruption correlation, thereby achieving unified interruption track numbering and track completion. S1.

4. For local tracks generated by different platforms, a track association method based on shared Transformer coding is used to perform cross-platform track association. The association decision variables are solved by using the association probability of candidate track pairs, and the associated track groups are extracted. S1.

5. For the local tracks that have been associated, the generalized covariance cross criterion is used for fusion to obtain the fusion center. A system-level flight path at time State estimation Covariance ; S1.

6. In the feedback fusion mode, the system trajectory obtained in step S1.5 is fed back to the local platform to predict the local trajectory at the next moment; S1.

7. The OSPA distance is used to evaluate the quality of the local tracks of each platform and the fused system tracks, resulting in a set of system-level tracks output by the file-configured shipborne and airborne platform collaborative tracking fusion subsystem.

3. A multi-sensor collaborative tracking and fusion system for multi-scenario configuration according to claim 2, characterized in that, The method for establishing the high-speed vehicle target modeling and tracking fusion subsystem under real parameter constraints includes the following steps: S2.

1. Establish a target motion model based on the true parameters of the high-speed aircraft target, including constructing the target state vector from the position vector and velocity vector, and then constructing a motion model constrained by the true parameters, the expression of which is: in, For position vectors, For velocity vectors, Let be the target state vector. Indicates the first A motion model constrained by real parameters This represents the set of true parameters corresponding to the model. Let T be the first derivative of the target state vector, and let T denote the transpose. S2.

2. In the target tracking phase, the UKF-JIPDA method based on interactive multi-models is adopted for any local track. Calculate the first The state estimates under each model are then combined with the IMM fusion method to obtain the local track state estimates and covariance. S2.

3. The local tracks obtained in step S2.2 are associated using a track association method based on shared Transformer coding. Then, in the multi-sensor track fusion stage, the local estimates of multiple platforms are gradually fused using a sequential inverse covariance cross method until the sequential fusion of multiple sensors is completed, resulting in a system-level track set output by the high-speed aircraft target modeling and tracking fusion subsystem under real parameter constraints.

4. A multi-sensor collaborative tracking and fusion system for multi-scenario configuration according to claim 3, characterized in that, The method for establishing the active and passive radar tracking fusion subsystem for intelligent configuration based on data flow includes the following steps: S3.

1. A passive radar positioning method based on chi-square distribution statistics is used to screen multi-station direction finding line combinations; a chi-square statistic and corresponding correlation likelihood function are constructed based on the candidate direction finding line combinations, assuming the three observation stations are at time... The correlation likelihood function corresponding to the candidate direction finding line combination is: Introducing binary assignment variables Then the total likelihood function Represented as: ； The final association result was obtained by solving the following optimization problem: in, , , These represent the times of the three observation stations. The number of candidate direction finding lines formed Indicates the first , , Do the direction finding lines form a valid combination? The set of local tracks for passive radar is then represented as: in, Indicates time The set of local tracks of passive radar. Indicates the source of the passive radar track. Indicates the first Local track status of passive radar; S3.

2. To address the issue of track association in interruptions, an identity discrimination method based on RWKV time-series modeling is adopted, combined with the local track set obtained by IMM-GLMB tracking using active radar. Constructing local track sets for active and passive radar : in, Indicates the active radar at time The resulting set of local tracks; when When there is an interruption in the flight path, boundary-aware features are extracted for both the old and new flight path segments and mapped to a unified metric space. Let their normalized representations be respectively... and Then its associated score Represented as: ； A score matrix is ​​constructed based on the correlation scores of all candidate pairs. Under one-to-one constraints, the Hungarian algorithm is used to perform global matching on candidate pairs that meet the threshold conditions, resulting in a set of local tracks after the interruption of continuity. in, Indicates by The set of local tracks obtained after determining the identity of interrupted tracks. express Local flight paths in This represents the threshold for identity discrimination. This represents the associated confirmation variable after a one-to-one match; S3.

3. To address the issues of out-of-order routing and local missing data, a trajectory completion method based on neural controlled differential equations is adopted, combining timestamps and state availability information to construct a continuous control path: in, Indicates by The constructed continuous control path This represents the constructor for a continuous control path. The hidden state propagates in the continuous time domain driven by a continuous control path, and the continuous time modeling form is as follows: in, Indicates time The hidden state, This represents the initial hidden state. This represents a continuous-time vector field function parameterized by a neural network; Next, the hidden states at each time step are decoded to obtain the completed state output: in, Represents the state decoding function. Indicates the length of the local track window. Indicates the first Output the completed state corresponding to each moment; S3.

4. For the local tracks generated by different sensors after processing in steps S3.1-S3.3, a track association method based on shared Transformer coding is used to associate multi-sensor tracks, extract associated track groups, and perform track fusion using the generalized covariance cross criterion. Then, the OSPA distance assessment is performed to evaluate track quality, resulting in a system-level track set output by the active and passive radar tracking fusion subsystem for intelligent configuration of data stream.

5. A multi-sensor collaborative tracking and fusion system for multi-scenario configuration according to claim 4, characterized in that, The overall system situation map generation module receives the system-level track set generated by the three subsystems, and organizes, aggregates and displays target information from multiple scenarios in a unified manner. Based on unified time, unified coordinates and unified attribute descriptions, it comprehensively manages the system-level tracks to form a unified situation map for displaying target information from multiple scenarios.

Images