Unmanned aerial vehicle sortie identification and positioning method and system

Abstract

The embodiment of the invention provides an unmanned aerial vehicle sortie identification and positioning method and system based on multi-positioning-source external detection data. According to the positioning method, detection data are clustered according to space-time density to generate a candidate sortie cluster, then the candidate sortie cluster is associated with a historical sortie cluster recorded by a past historical time window, and an optimal matching result of each sortie of the candidate sortie cluster is calculated according to sortie associated data. And determining a sortie identification result of the current time window according to the optimal matching result, thereby avoiding a positioning error caused by a fusion sortie object error in a multi-source data fusion stage. Furthermore, the clustering association process is not limited by the number of the unmanned aerial vehicles, and the unmanned aerial vehicles can be newly added at any time, so that the positioning method and system have better adaptability to different scenes.

Description

Technical Field

[0001] This invention belongs to the technical field of target positioning and tracking, specifically relating to a method and system for UAV sortie identification and positioning based on external detection data from multiple positioning sources. Background Technology

[0002] In recent years, drones have increasingly come into view. Besides their widespread military applications, drones are also rapidly expanding in the civilian market. To achieve accurate drone positioning, positioning systems typically employ multiple positioning sources, then fuse the positioning information from these sources to obtain the final positioning data, thus avoiding the inherent positioning defects of a single source.

[0003] In some situations, multiple drones need to fly simultaneously, using multiple positioning sources to acquire positioning data. The positioning data from different positioning sources are matched one by one by manual marking. When the number of drones changes, the newly added drones need to be marked. This sortie identification method relies on manual operation. Each expansion of drone sorties requires marking, which is a large workload and not conducive to the expansion of drone sorties.

[0004] Therefore, it is necessary to improve the existing technology to overcome the aforementioned defects. Summary of the Invention

[0005] Therefore, the present invention aims to solve the technical problems in the prior art where multiple positioning sources and multiple drones are prone to positioning data object matching errors and inability to adapt to changes in the number of targets during the positioning process.

[0006] To address the aforementioned technical problems, this invention provides a method for UAV sortie identification and positioning based on external detection data from multiple positioning sources. The method includes acquiring detection data from multiple positioning sources for multiple UAVs; wherein the detection data includes time signals and spatial signals; preprocessing the time signals and spatial signals, aligning the time signal data from multiple positioning sources to a time reference, and aligning the spatial signals to a spatial reference; extracting spatiotemporal features and motion features based on the time signals and spatial signals; for the current time window, clustering candidate sortie clusters based on the spatiotemporal features and motion features according to spatiotemporal density; associating the candidate sortie clusters with historical sortie clusters recorded in past historical time windows, and calculating the optimal matching result for each sortie in the candidate sortie cluster based on the sortie association data; determining the sortie identification result for the current time window based on the optimal matching result; and fusing multiple positioning data from multiple positioning sources for each sortie based on the sortie identification result to determine the fused positioning data for each sortie.

[0007] In one preferred embodiment, the step of "associating the candidate flight cluster with historical flight clusters recorded in past historical time windows, and calculating the optimal matching result for each flight in the candidate flight cluster based on the flight association data" includes constructing a cost matrix based on the spatiotemporal and motion characteristics of the ends of the candidate flight cluster and the historical flight clusters recorded in past historical time windows, solving the minimum cost of the cost matrix using the Hungarian optimal matching algorithm, and determining the optimal one-to-one matching result between the candidate flight cluster and the historical flight clusters recorded in past historical time windows based on the minimum cost.

[0008] In one preferred embodiment, the step of "determining the optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window based on the minimum cost" further includes determining that the matching fails if the minimum cost exceeds a preset cost threshold, and placing the historical flights that failed to match into a disappearing buffer; in subsequent time windows, using the Hungarian optimal matching algorithm to perform minimum cost matching calculations on the historical flights in the disappearing buffer; if no match is found after K consecutive time windows, the historical flights in the disappearing buffer are deleted; where K is a natural number greater than or equal to 1.

[0009] In one preferred embodiment, the step of "determining the optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window based on the minimum cost" includes: if the minimum cost exceeds a preset cost threshold, determining that the matching has failed, adding the flight of the candidate flight cluster in the current time window that failed the matching to the new flight candidate area; and using the Hungarian optimal matching algorithm to perform minimum cost matching calculation on the flights in the new flight candidate area in the subsequent K consecutive time windows. If the matching is successful, it is identified as a new flight; if the matching fails, the disappearance determination process is entered; where K is a natural number greater than or equal to 1.

[0010] In one preferred embodiment, the step of "generating new flight results for flights that failed to match" includes the steps of storing the trajectory information of the flights that failed to match into a disappearance buffer to confirm whether the flights will disappear in a future time window; and defining the detection data of the flights that failed to match as new candidate flights.

[0011] In one preferred embodiment, when the matching fails, a backtracking mechanism is triggered; the backtracking mechanism includes retaining the nearest neighbor to the current time window. The detection and trajectory data of each time window are processed, and the Hungarian algorithm is re-executed for optimal matching to correct possible mismatches, where k is an integer greater than 2; the detection data that cannot be associated through the backtracking mechanism will be used as new candidate flights, and the trajectory information that cannot be successfully backtracked will enter the disappearance determination process.

[0012] In one preferred embodiment, the step of "generating candidate frame sub-clusters by spatiotemporal density clustering based on the spatiotemporal features and the motion features for the current time window" includes the following steps: S1 Define the neighborhood: For a detection point p, its neighborhood is defined as follows: ; in, For spatial distance, For time distance, The spatial neighborhood threshold, The time neighborhood threshold; S2 Core Point Judgment: Based on the neighborhood definition above. The specific formula for determining whether the detection point p is a core point is as follows: ; in, The threshold representing the minimum number of points within the neighborhood; S3 Clustering Expansion: If the detection point p is the core point, the neighborhood... Points within the current cluster are added to the current cluster. This step is repeated for each core point in the neighborhood. During the expansion process, the clusters continue to grow until no more new points can be added. Points that are not core points and do not belong to any neighborhood are considered noise points. S4 Each spatiotemporal cluster represents belonging to the same candidate frame sub-cluster.

[0013] In one preferred embodiment, the step of "acquiring detection data of multiple drones from multiple positioning sources" includes: if a drone lacks external detection data from a positioning source within the current time window, retaining the missing detection data record and using the predicted detection data value based on past historical time window records to fill the detection data of the current time window; in the step of "fusing positioning data of multiple positioning sources for each sortie based on the sortie identification result to determine the fused positioning data of each sortie", the weighting weight of the drone is set to 0 or attenuated according to the missing detection duration.

[0014] In one preferred embodiment, the step of "fusing and positioning multiple positioning data from multiple positioning sources for each sortie based on the sortie identification result, and determining the fused positioning data for each sortie" includes the steps of determining a detection data model and a motion model for fusion calculation; collecting external detection data from multiple positioning sources within the current time window for each sortie that has been matched; calculating the weight of the external detection data of each positioning source based on the type of positioning source and real-time quality parameters, and using the weights to fuse the data to obtain the positioning data for each sortie.

[0015] In one preferred embodiment, the detection data model is [x, y, z, time, confidence], and the motion model includes at least one of the following models: uniform velocity model, uniform acceleration model, and coordinated turning model.

[0016] In one preferred embodiment, the step of "fusing and positioning multiple positioning data from multiple positioning sources for each sortie based on the sortie identification result, and determining the fused positioning data for each sortie" includes the steps of adjusting the detection covariance using Kalman filtering or factor graph; and obtaining at least one of the three-dimensional trajectory, velocity, and acceleration for each sortie.

[0017] This invention also provides a UAV sortie identification and positioning system based on external detection data from multiple positioning sources. The positioning system includes multiple positioning sources; multiple drones; a detection data acquisition module for acquiring detection data of the drones from the multiple positioning sources, wherein the detection data includes time signals and spatial signals; a data preprocessing module for preprocessing the time signals and spatial signals, aligning the time signal data from the multiple positioning sources with a time reference, aligning the spatial signals with a spatial reference, and extracting spatiotemporal features and motion features based on the time signals and spatial signals; a sortie cluster generation module for generating candidate sortie clusters based on the spatiotemporal features and motion features according to spatiotemporal density for the current time window; a sortie identification and association module for associating the candidate sortie clusters with historical sortie clusters recorded in past historical time windows, calculating the optimal matching result for each sortie in the candidate sortie cluster based on the sortie association data, and determining the sortie identification result for the current time window based on the optimal matching result; and a weighted fusion positioning module for fusing positioning data from multiple positioning sources for each sortie based on the sortie identification result, and determining the fused positioning data for each sortie.

[0018] The UAV sortie identification and positioning method and system based on external detection data from multiple positioning sources provided in this invention generates candidate sortie clusters by clustering the detection data according to spatiotemporal density. These candidate sortie clusters are then associated with historical sortie clusters recorded in past time windows. The optimal matching result for each sortie in the candidate sortie cluster is calculated based on the associated data, and the sortie identification result for the current time window is determined according to the optimal matching result. This avoids positioning errors caused by incorrect sortie objects during the multi-source data fusion stage. Furthermore, the clustering and association process is not limited by the number of UAVs; UAVs can be added at any time, thus making this positioning method and system more adaptable to different scenarios. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0020] Figure 1 This is a flowchart illustrating the UAV sortie identification and positioning method based on external detection data from multiple positioning sources provided in an embodiment of the present invention. Figure 2 To execute Figure 1 The schematic diagram of the system structure corresponding to the embodiment shown is shown. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0023] In this invention, unless otherwise stated, directional terms such as "upper," "lower," "top," and "bottom" are generally used in relation to the direction shown in the accompanying drawings, or in relation to the vertical, perpendicular, or gravitational direction of the component itself; similarly, for ease of understanding and description, "inner" and "outer" refer to the inner and outer contours of each component itself, but the above directional terms are not intended to limit this invention.

[0024] This invention provides a method for identifying and locating drone flights based on external detection data from multiple positioning sources. The multiple positioning sources specifically refer to devices that acquire the time / space information of drones using different measurement / positioning principles. Examples include radar devices, which can detect the distance, orientation, and speed of drones by emitting electromagnetic waves and receiving reflected echoes; 5G-A devices, which leverage the high bandwidth and low latency of 5G-Advanced technology to provide high-speed data transmission support for airspace management, assisting in the real-time interaction and sharing of drone-related monitoring data; RemoteID devices (remote identification devices), which can identify and remotely transmit drone identity information, facilitating the positioning system to obtain key identification information such as the drone's identity and location; and TDOA (Time Difference of Arrival) devices, which measure the time difference of signals arriving at different receiving stations and combine this with multi-station location information for positioning calculations, thereby achieving accurate determination of the drone's location.

[0025] Different positioning sources have their own positioning advantages. In this embodiment of the invention, using multiple positioning sources simultaneously for UAV positioning and monitoring can effectively combine the positioning advantages of different sources while avoiding the inherent positioning defects of a single source, thereby improving the positioning accuracy of the UAV. Specifically, in this embodiment of the invention, the multiple positioning sources include radar equipment, 5G-A equipment, RemoteID equipment (remote identification equipment), and TDOA equipment (time difference of arrival equipment). The number of the same positioning device can be one or more, and there is no limitation here.

[0026] like Figure 1 As shown, the UAV sortie identification and positioning method based on external detection data from multiple positioning sources includes the following seven steps.

[0027] Step S11: Acquire detection data of multiple drones from multiple positioning sources; wherein the detection data includes time signals and spatial signals.

[0028] Step S12: Preprocess the time signal and the spatial signal, align the time signal data from multiple positioning sources with a time reference, and align the spatial signal with a spatial reference.

[0029] Step S13: Extract the spatiotemporal features and motion features based on the time signal and the spatial signal.

[0030] Step S14: For the current time window, generate candidate frame sub-clusters by spatiotemporal density clustering based on the spatiotemporal features and the motion features.

[0031] Step S15: Associate the candidate flight cluster with the historical flight cluster recorded in the past historical time window, and calculate the optimal matching result for each flight in the candidate flight cluster based on the flight association data.

[0032] Step S16: Determine the flight identification result for the current time window based on the optimal matching result.

[0033] Step S17: Based on the flight identification result, perform fusion positioning on multiple positioning data from multiple positioning sources for each flight to determine the fused positioning data for each flight.

[0034] Step S15, "For the current time window, generate candidate secondary clusters by spatiotemporal density clustering based on the spatiotemporal features and motion features," is an algorithm that combines the time dimension (such as the timestamp of data generation / detection, time interval) and the spatial dimension (such as geographical coordinates, network topology location), using "density" as the core criterion (i.e., dividing clusters by whether the number of data points in the neighborhood meets a threshold). The core advantage of spatiotemporal density-based clustering algorithms is that they can effectively identify irregularly shaped and unevenly dense clusters in the spatiotemporal dimension and filter out spatiotemporal noise points. The clustering algorithms defined in this step include ST-DBSCAN (Spatio-Temporal DBSCAN), STS-DBSCAN (Spatio-Temporal StreamDBSCAN), OPTICS-ST (Spatio-Temporal OPTICS), CT-DBSCAN (Continuous TrajectoryDBSCAN), or DBSCAN-ST+ (Enhanced Spatiotemporal DBSCAN).

[0035] In one specific embodiment, step S15 specifically includes the following steps: Step T11: Construct a cost matrix based on the spatiotemporal and motion characteristics of the candidate flight clusters and the ends of historical flight clusters recorded in the past historical time windows, and use the Hungarian optimal matching algorithm to solve for the minimum cost of the cost matrix.

[0036] Step T12: Determine the optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window based on the minimum cost.

[0037] The specific calculation process of step T11 is as follows, S1 to S4: S1 Define the neighborhood: For a detection point p, its neighborhood is defined as shown in Formula 1: ; (Formula 1) in, For spatial distance, For time distance, The spatial neighborhood threshold, This is the time neighborhood threshold.

[0038] Spatial distance is typically calculated using Euclidean distance or ENU coordinate distance. In this embodiment, Euclidean coordinate distance is used, and the specific calculation method is shown in Formula 11: ;(Formula 11) The time distance specifically refers to the difference in timestamps, and the specific calculation method is shown in Formula 12: ;(Formula 12) Based on the above neighborhood definition process, only detection points that are close in both space and time are considered neighborhood points, thereby avoiding the incorrect clustering of targets in different time periods into the same sortie.

[0039] S2 Core Point Judgment: Based on the neighborhood definition above. To determine whether the detection point p is a core point, the specific determination is shown in Formula 2: ; (Formula 2) in, This is the minimum number of points in the neighborhood threshold. If the number of points in the neighborhood reaches the threshold, then p is the core point. The core point is the starting point for cluster expansion, ensuring that clustering only generates a set of points with sufficiently high density. Noisy points or isolated points will not form clusters, thus further avoiding the phenomenon of incorrectly matching location data with drones.

[0040] S3 Clustering Expansion: If the detection point p is the core point, the neighborhood... Points within a given cluster are added to the current cluster. This step is repeated for each core point within its neighborhood. During the expansion process, the clusters grow continuously until no more new points can be added. Points that are not core points and do not belong to any neighborhood are considered noise points. This cluster expansion step can identify drone sub-clusters within the same time window.

[0041] S4 outputs that each spatiotemporal cluster belongs to the same candidate sub-cluster.

[0042] After completing S1 to S4, step T11 continues with the following steps to associate the candidate flight with the current flight.

[0043] T1: Calculate the similarity between candidate flight clusters and historical flight clusters based on the dimensions of distance, velocity, direction, and signal.

[0044] T2: Construct the cost matrix and use the Hungarian algorithm for optimal matching. The specific calculation formula is shown in Formula 3: ;(Formula 3) Among them, the M candidate flights detected in the current time window are C1, C2...C M The N historically identified flights are H1, H2, ..., H. N C ij Let represent the similarity cost between the i-th historical sortie and the j-th candidate sortie.

[0045] The spatial distance error is calculated as shown in Formula 4: ;(Formula 4) in, Indicates historical sorties H i The predicted location, Indicates candidate sortie C j The central position.

[0046] The speed error is calculated as shown in Formula 5: ;(Formula 5) in, Indicates historical sorties H i Historical estimated speed Indicates candidate sortie C j The average speed.

[0047] The calculation of signal or detection error is shown in Formula 6: ;(Formula 6) in, Indicates historical sorties H i The feature value detected by device K (referring to the Kth positioning source) This represents the feature value detected by device K (referring to the Kth location source) for candidate flight Cj. This indicates the number of devices (location sources) participating in the feature calculation.

[0048] In Formula 3, α, β, and γ can be adaptively adjusted according to the requirements of the positioning system, as long as their sum is 1. Through the adaptive weighting mechanism, combined with the characteristics of the positioning source itself and the quality of real-time detection data, the weight ratio of each positioning data in the fusion process is dynamically adjusted, which significantly enhances the robustness of the data fusion process, avoids interference from a single positioning source or low-quality detection data on the fusion result, and improves the accuracy of the fusion result.

[0049] During this matching process, matching failures may occur. Even if the Hungarian algorithm outputs an optimal match, some "matches" are unreliable, requiring the definition of rejection criteria. A preset cost threshold is set; if the matching cost exceeds this threshold, the match is considered invalid. The specific definition of exceeding the cost threshold is shown in Formula 7: ;(Formula 7) If the minimum cost C ij Exceeding the preset cost threshold θ assoc If a match fails, the historical flights that failed to match are placed in the disappearance buffer. In subsequent time windows, the Hungarian optimal matching algorithm is used to perform minimum cost matching calculations on the historical flights in the disappearance buffer. If no match is found after K consecutive time windows, the historical flights in the disappearance buffer are deleted. Here, K is a natural number greater than or equal to 1.

[0050] If the minimum cost C ij Exceeding the preset cost threshold θ assoc If the match fails, the candidate flight clusters in the current time window that failed to match are added to the new flight candidate area. In the subsequent K consecutive time windows, the Hungarian optimal matching algorithm is used to perform minimum cost matching calculations on the flights in the new flight candidate area. If the match is successful, it is identified as a new flight. If the match fails, the disappearance determination process is entered. Here, K is a natural number greater than or equal to 1.

[0051] If detection data is missing, the matching cost exceeds a threshold, or there are abnormalities in trajectory prediction residuals, a backtracking step is triggered. The backtracking step specifically includes: the system retains the detection and trajectory data from the most recent k time windows and re-executes the Hungarian algorithm for optimal matching to correct possible mismatches. Detections that fail to be backtracked will be considered as new flight candidates, and trajectories that fail to be backtracked will enter the disappearance determination process. When there are partial missing data in multi-source positioning data, the backtracking step ensures that the positioning method can still maintain the continuity of the UAV trajectory and ensure that the UAV's positioning accuracy is within a preset reasonable range. This avoids trajectory interruption or a significant drop in positioning accuracy due to partial missing data, improving the adaptability of the positioning system in scenarios with incomplete data.

[0052] The positioning method provided in this invention can use missing data records to assess the health status of various positioning sources. At the same time, it can perform backtracking correction operations based on the missing data records, promptly detect potential faults in positioning sources and correct them, reduce the impact of positioning source anomalies on the operation of the positioning system, and thus improve the long-term stability and reliability of the positioning system.

[0053] After "acquiring detection data from multiple positioning sources for multiple drones; wherein the detection data includes time signals and spatial signals" (i.e., step S11), the process further includes the step of: if a drone lacks external detection data from a positioning source within the current time window, then the missing detection data record is retained and the detection data prediction value based on past historical time window records is used to fill the detection data for the current time window. At this point, in the subsequent "fusion positioning of multiple positioning data from multiple positioning sources for each sortie based on the sortie identification result, and determination of the fused positioning data for each sortie" (i.e., step S17), the weighting weight of the drone is set to 0 or attenuated according to the duration of the missing detection.

[0054] In this embodiment, step S22 can ensure that the time reference of external detection data from multiple positioning sources is aligned, the spatial base station is aligned, the detection data format is aligned, and obviously abnormal data is removed during the positioning method calculation process.

[0055] To ensure time consistency among devices or nodes within the system and achieve time synchronization during data acquisition, transmission, and processing, a unified time reference needs to be constructed using multi-source time synchronization methods to complete time reference alignment. In this embodiment, multi-source time synchronization methods specifically include, but are not limited to, the following types: GPS (Global Positioning System) time synchronization, which provides a high-precision absolute time reference for the target device by receiving time signals transmitted by GPS satellites, ensuring that the device's time is consistent with the globally unified time reference; BeiDou (BeiDou Navigation Satellite System) time synchronization, which relies on time information transmitted by the BeiDou satellite constellation to provide a reliable absolute time reference for the target device, achieving precise alignment between the device's time and the BeiDou system time reference; and PTP (Precision Time Protocol) time synchronization, which transmits high-precision time signals through network links and, based on the time synchronization mechanisms specified in the PTP protocol (such as master-slave clock negotiation, delay measurement and correction), achieves relative time reference alignment among multiple devices within a local area network or wide area network, meeting the short-distance or long-distance time synchronization requirements between devices.

[0056] To eliminate delay deviations caused by factors such as transmission distance, link bandwidth, and network congestion in the transmission link, and to ensure the timeliness and accuracy of data transmission, link delay measurement must be performed, followed by delay compensation based on the measurement results. This includes link delay measurement and link delay compensation. Specifically, a preset delay measurement mechanism (such as round-trip time measurement or one-way delay measurement) is used to obtain the total delay time incurred during data transmission from the sending end to the receiving end via the transmission link, as well as the delay components of each segment in the link (such as sending end buffer delay, physical link transmission delay, and receiving end buffer delay), forming complete link delay measurement data. Based on the above link delay measurement data, combined with the system's preset compensation strategy (such as hardware-based real-time delay cancellation, software-based timestamp correction, and protocol-based dynamic delay compensation), the data transmission time at the sending end or the data reception time at the receiving end is adjusted, or the timestamp information of the transmitted data is corrected, so that the deviation between the actual data acquisition time at the receiving end and the theoretically expected time is controlled within a preset threshold range, thereby achieving effective compensation for link delay.

[0057] To eliminate the differences in measurement coordinates between different positioning sources and to achieve collaborative association and fusion processing of positioning data from various sources, the measurement coordinates of different positioning sources need to be transformed to a unified regional coordinate system, thus achieving spatial benchmark consistency.

[0058] To achieve unified reception, parsing, and fusion processing of positioning data from different sources, the raw data structures output by each source need to be standardized and transformed into a unified detection vector format. This unified detection vector format must include at least x, y, and z parameters representing target spatial information, as well as type parameters representing target attributes, signal characteristic parameters, and confidence level parameters representing data reliability. This format conversion ensures that positioning data from different sources have a consistent data structure, guaranteeing the compatibility and effectiveness of subsequent data processing.

[0059] To improve the reliability of positioning data, quality labeling and preliminary filtering operations need to be performed. During preliminary filtering, obvious outliers are removed, including but not limited to points where distance jumps exceed preset thresholds, time information is incorrect, and coordinates deviate from a reasonable range. At the same time, the filtered valid data is labeled with an initial confidence level to provide a quality basis for subsequent data processing and ensure the effectiveness of data application.

[0060] Specifically, step S17 includes the following steps.

[0061] Step S171: Determine the detection data model and motion model used for fusion computation. The detection data model is specifically [x, y, z, time, confidence], consisting of x, y, and z parameters representing the target's spatial information, as well as type parameters representing target attributes, signal characteristic parameters, and confidence parameters representing data reliability. The motion model includes at least one of the following models: uniform velocity model, uniform acceleration model, and coordinated turning model. An interactive multi-model adaptive switching mechanism is used to switch between these three different motion models. Through interactive fusion between models (including mixing state estimates and updating model probabilities), the system can quickly adapt to sudden or gradual changes in the target's motion pattern, improving the accuracy and robustness of state estimation in complex dynamic scenarios.

[0062] Step S172: For each matched flight, collect detection data from multiple location sources within the current time window, specifically expressed as follows: .

[0063] Step S173: Calculate the weight of the detection data for each positioning source based on the type of positioning source and real-time quality parameters, and fuse the data using the weights to obtain the positioning data for each sortie. The weight calculation is shown in Formula 8: ;(Formula 8) Where, q i The device health level (0~1) is expressed as follows: ; Among them, A i S represents the availability rate (1 - the percentage of tests not received). i For stability (detection fluctuation normalization), C i For consistency (normalization of deviation from fusion results), α1, α2, and α3 can be adjusted according to the requirements of the positioning system, as long as the sum of the three is 1.

[0064] σ1 is the real-time error estimate, and its specific calculation is shown in Formula 9: ;(Formula 9) Among them, z i,k Let be the detection value of the i-th device over the past k time intervals; This is the average detected value. The smaller the value, the more stable the detection and the higher the accuracy.

[0065] Preferably, the method further includes step S174: adjusting the detection covariance using Kalman filtering or factor graph; obtaining at least one of the three-dimensional trajectory, velocity, and acceleration for each sortie. The calculation of the detection covariance adjusted by Kalman filtering or factor graph is shown in Formula 10: ; (Formula 10) The positioning method provided in this invention can automatically identify drone sorties without requiring manual intervention in sortie identification and matching operations. This effectively reduces manual intervention costs, improves sortie identification efficiency, avoids errors that may be introduced by manual operation, and ensures the accuracy and automation level of sortie identification. Even when drone trajectories intersect (including new data backtracking scenarios), are obstructed, or have lost detection signals, it can still assign continuous sortie identifiers to the drone, ensuring the continuity of the identifiers and preventing them from breaking or becoming disordered due to the aforementioned complex scenarios. This ensures continuous tracking and identity association of the drone.

[0066] like Figure 2 As shown, this embodiment of the invention also provides a UAV sortie identification and positioning system 100 based on external detection data from multiple positioning sources. The system 100 includes multiple positioning sources 10, multiple UAVs 20, a detection data acquisition module 30, a data preprocessing module 40, a sortie cluster generation module 50, a sortie identification and association module 60, and a weighted fusion positioning module 70.

[0067] Multiple positioning sources 10 include radar positioning device 11, 5G-A positioning device 12, RemoteID device 13, TDOA device 14, or other positioning devices capable of locating the position of UAVs. The total number of UAVs 20 can adaptively vary depending on different demand scenarios. The detection data acquisition module 30 is configured to acquire detection data of multiple UAVs 20 from multiple positioning sources 10, wherein the detection data includes time signals and spatial signals. The data preprocessing module 40 is configured to preprocess the time signals and spatial signals, aligning the time signal data from multiple positioning sources 10 to a time reference and aligning the spatial signals to a spatial reference. The sortie cluster generation module 50 is configured to generate candidate sortie clusters based on the spatiotemporal features and motion features according to spatiotemporal density for the current time window. The sortie identification and association module 60 is configured to associate the candidate sortie clusters with historical sortie clusters recorded in past historical time windows, calculate the optimal matching result for each sortie in the candidate sortie cluster based on the sortie association data, and determine the sortie identification result for the current time window based on the optimal matching result. The weighted fusion positioning module 70 is configured to perform fusion positioning on multiple positioning data from multiple positioning sources for each flight based on the flight identification result, and determine the fusion positioning data for each flight.

[0068] The UAV sortie identification and positioning method and system based on external detection data from multiple positioning sources provided by this invention generates candidate sortie clusters by clustering the detection data according to spatiotemporal density. These candidate sortie clusters are then associated with historical sortie clusters recorded in past time windows. The optimal matching result for each sortie in the candidate sortie cluster is calculated based on the associated data. The sortie identification result for the current time window is determined based on the optimal matching result. This avoids positioning errors caused by incorrect sortie objects during the multi-source data fusion stage. Furthermore, the clustering and association process is not limited by the number of UAVs; UAVs can be added at any time. Therefore, this positioning method and system are more adaptable to different scenarios.

[0069] Obviously, the embodiments described above are merely some, not all, embodiments of the present invention. Based on the embodiments of the present invention, those skilled in the art can make other variations or modifications without creative effort, and all such variations or modifications should fall within the scope of protection of the present invention.

Claims

1. A method for identifying and locating UAV sorties based on external detection data from multiple location sources, characterized in that, include: Acquire detection data of multiple drones from multiple location sources; The detected data includes time signals and spatial signals; The time signal and the spatial signal are preprocessed, and the time signal data from multiple positioning sources are aligned with a time reference, and the spatial signal is aligned with a spatial reference. Spatiotemporal features and motion features are extracted based on the time signal and the spatial signal; For the current time window, candidate frame sub-clusters are generated by clustering according to spatiotemporal density based on the spatiotemporal features and the motion features; The candidate flight clusters are associated with historical flight clusters recorded in past historical time windows, and the optimal matching result for each flight in the candidate flight cluster is calculated based on the flight association data. The flight identification result for the current time window is determined based on the optimal matching result; Based on the flight identification results, multiple positioning data from various positioning sources for each flight are fused for positioning to determine the fused positioning data for each flight.

2. The method as described in claim 1, characterized in that, The phrase "associating the candidate flight clusters with historical flight clusters recorded in past historical time windows, and calculating the optimal matching result for each flight in the candidate flight cluster based on the flight association data" includes: A cost matrix is ​​constructed based on the spatiotemporal and motion characteristics of the candidate flight clusters and the ends of historical flight clusters recorded in the past historical time windows. The minimum cost of the cost matrix is ​​solved using the Hungarian optimal matching algorithm. The optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window is determined based on the minimum cost.

3. The method as described in claim 2, characterized in that, The phrase "determining the optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window based on the minimum cost" also includes: If the minimum cost exceeds the preset cost threshold, the matching is determined to be a failure, and the historical flights that failed to match are placed in the disappearance buffer. In subsequent time windows, the Hungarian optimal matching algorithm is used to perform minimum cost matching calculations on the historical flights in the disappearing buffer. If no match is found after K consecutive time windows, the historical flights in the disappearing buffer are deleted; where K is a natural number greater than or equal to 1.

4. The method as described in claim 2, characterized in that, The phrase "determining the optimal one-to-one matching result between the candidate flight cluster and the historical flight cluster recorded in the past historical time window based on the minimum cost" includes: If the minimum cost exceeds the preset cost threshold, the matching is determined to be unsuccessful, and the flight numbers of the candidate flight clusters in the current time window that failed to match are added to the new flight candidate area. In the subsequent K consecutive time windows, the Hungarian optimal matching algorithm is used to perform minimum cost matching calculations on the flights in the candidate area of ​​the new flights. If the matching is successful, it is identified as a new flight; if the matching fails, it enters the disappearance determination process; where K is a natural number greater than or equal to 1.

5. The method as described in claim 1, characterized in that, The step of "generating candidate frame sub-clusters by clustering based on the spatiotemporal features and the motion features according to spatiotemporal density for the current time window" includes the following steps: S1 Define the neighborhood: For a detection point p, its neighborhood is defined as follows: ; in, For spatial distance, For time distance, The spatial neighborhood threshold, The time neighborhood threshold; S2 Core Point Judgment: Based on the neighborhood definition above. The specific formula for determining whether the detection point p is a core point is as follows: ; in, The threshold representing the minimum number of points within the neighborhood; S3 Clustering Expansion: If the detection point p is the core point, the neighborhood... Points within the current cluster are added to the current cluster. This step is repeated for each core point in the neighborhood. During the expansion process, the clusters continue to grow until no more new points can be added. Points that are not core points and do not belong to any neighborhood are considered noise points. S4 Each spatiotemporal cluster represents belonging to the same candidate frame sub-cluster.

6. The method as described in claim 1, characterized in that, The phrase "acquiring detection data of multiple drones from multiple location sources" includes: If a certain UAV lacks external detection data of the positioning source within the current time window, the record of missing detection data is retained and the detection data prediction value based on the historical time window record is used to fill the detection data of the current time window. In the step of "fusing and positioning multiple positioning data from multiple positioning sources for each sortie based on the sortie identification result, and determining the fused positioning data for each sortie", the weighting weight of a certain UAV is set to 0 or attenuated according to the missing measurement time.

7. The method as described in claim 1, characterized in that, The step of "fusing and locating multiple positioning data from multiple positioning sources for each flight based on the flight identification result, and determining the fused positioning data for each flight" includes the following steps: Determine the detection data model and motion model used for fusion computing; For each matched sortie, collect external detection data from multiple location sources within the current time window; The weight of the external detection data of each positioning source is calculated based on the type of positioning source and real-time quality parameters, and the weights are used to fuse the data to obtain the positioning data of each sortie.

8. The method as described in claim 7, characterized in that, The detection data model is [x, y, z, time, confidence level], and the motion model includes at least one of the following models: uniform velocity model, uniform acceleration model, and coordinated turning model.

9. The method as described in claim 7, characterized in that, The step of "fusing and locating multiple positioning data from multiple positioning sources for each flight based on the flight identification result, and determining the fused positioning data for each flight" includes the following steps: Kalman filtering or factor plotting can be used to adjust the detection covariance. Obtain at least one of the three-dimensional trajectory, velocity, and acceleration for each sortie.

10. A UAV sortie identification and positioning system based on external detection data from multiple positioning sources, characterized in that, The positioning system includes: Multiple location sources; Multiple drones; The detection data acquisition module is used to acquire detection data of the multiple positioning sources on the multiple drones; wherein the detection data includes time signals and spatial signals; The data preprocessing module is used to preprocess the time signal and the spatial signal, align the time signal data of the multiple positioning sources with a time reference, align the spatial signal with a spatial reference, and extract spatiotemporal features and motion features based on the time signal and the spatial signal. The flight cluster generation module is used to generate candidate flight clusters based on the spatiotemporal features and the motion features according to spatiotemporal density for the current time window; The flight identification and association module is used to associate the candidate flight cluster with the historical flight cluster recorded in the past historical time window, calculate the optimal matching result for each flight in the candidate flight cluster based on the flight association data, and determine the flight identification result for the current time window based on the optimal matching result; The weighted fusion positioning module is used to fuse multiple positioning data from multiple positioning sources for each flight based on the flight identification results, and to determine the fused positioning data for each flight.

Images