A method and system for cooperative detection and three-dimensional positioning of unmanned aerial vehicles based on a dynamic tetrahedral microphone array

Abstract

The application discloses a kind of unmanned aerial vehicle cooperative detection and three-dimensional positioning system and method based on dynamic tetrahedron microphone array, belong to sound source positioning and target detection technical field.The system includes multiple detection nodes and data processing unit, and detection node includes unmanned aerial vehicle airborne microphone node.Data processing unit selects four nodes to construct tetrahedron array according to the evaluation index of the relationship between positioning error and space geometry configuration, calculates the three-dimensional coordinates of sound source according to the unmanned aerial vehicle rotor sound signal collected by each node, and corrects the positioning result using sensor data, and outputs sound source position information in combination with geographic space data.The application maintains optimal positioning geometry configuration by dynamic array construction and reconstruction, realizes the two-way cooperation of target and detection unit using unmanned aerial vehicle airborne node, suppresses urban multipath interference in combination with three-dimensional geographic information, and has the advantages of high positioning accuracy, strong cooperative ability and good environmental adaptability.

Description

Technical Field

[0001] This invention relates to the field of sound source localization and target detection technology, and in particular to a method and system for UAV cooperative detection and three-dimensional localization based on a dynamic tetrahedral microphone array. Background Technology

[0002] Microphone array-based sound source localization technology is an important research direction in the field of UAV detection. Currently, tetrahedral microphone arrays, due to their symmetrical structure and mature localization algorithms, are already used in acoustic detection, calculating the three-dimensional coordinates of the sound source by collecting the time difference of arrival (TDR) from four microphones. Multi-node collaborative detection schemes expand the monitoring range by deploying multiple fixed microphone base stations, and some systems incorporate GPS synchronization and wireless transmission to achieve distributed acquisition. Furthermore, detection systems that fuse acoustic localization with optical imagery have also been studied, utilizing cameras to assist in verifying the existence of targets. In UAV acoustic detection, existing technologies mainly focus on analyzing and identifying the time-frequency characteristics of rotor noise.

[0003] Existing technologies have several drawbacks: 1. Traditional tetrahedral arrays are mostly fixed deployments with fixed array geometry. When the spatial distribution of detection nodes changes, the array configuration cannot be dynamically adjusted to maintain the optimal positioning geometry, and the positioning accuracy is constrained by the fixed configuration. 2. UAVs are only used as passive detection targets, and their onboard microphone sensors do not participate in the network as detection nodes. There is a lack of two-way collaboration mechanism between nodes and targets, and it is impossible to actively optimize the array spatial configuration using movable nodes. 3. Pure acoustic positioning is significantly affected by the multipath effect caused by building reflections in complex urban environments, resulting in decreased positioning reliability. Existing solutions lack occlusion analysis and multipath suppression methods assisted by three-dimensional geographic information, and also lack multi-sensor cross-verification mechanisms, making it difficult to meet the high-precision detection and positioning requirements of urban low-altitude UAVs.

[0004] Existing technologies suffer from problems such as fixed array configurations that cannot be dynamically optimized, UAVs being merely passive targets lacking two-way coordination, and poor reliability of purely acoustic positioning under urban multipath interference, resulting in insufficient positioning accuracy, coordination capabilities, and environmental adaptability. Summary of the Invention

[0005] (a) Technical problems to be solved This invention provides a UAV collaborative detection and 3D positioning system and method based on a dynamic tetrahedral microphone array, comprising multiple detection nodes including UAV-borne microphone nodes and a data processing unit. The data processing unit selects four nodes to construct a tetrahedral array according to evaluation indicators characterizing the relationship between positioning error and spatial geometric configuration, collects UAV rotor acoustic signals to calculate 3D coordinates, corrects the positioning results based on sensor data, and outputs sound source location information by combining geospatial data, thereby achieving UAV collaborative detection and high-precision 3D positioning.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a method and system for UAV cooperative detection and three-dimensional positioning based on a dynamic tetrahedral microphone array, comprising detection nodes and a data processing unit. The detection nodes include UAV onboard microphone nodes. The data processing unit selects four nodes from the plurality of detection nodes to construct a tetrahedral detection array according to the spatial position of each detection node and an evaluation index characterizing the relationship between positioning error and spatial geometric configuration. The data processing unit calculates the three-dimensional coordinates of the sound source based on the acoustic signals received by each detection node in the tetrahedral detection array. The acoustic signals are acoustic signals generated by the rotor of the UAV. The data processing unit also corrects the calculation results based on sensor data and outputs sound source location information based on geospatial data.

[0007] Preferably, the evaluation index includes at least one of geometric accuracy factor, positioning error ellipsoid volume, and root mean square error of sound source positioning; The data processing unit calculates the value of the evaluation index for any combination of four nodes among the currently available nodes, and selects the combination that makes the evaluation index optimal and the volume of the tetrahedron formed by the four nodes greater than a preset threshold to construct the tetrahedron detection array.

[0008] Preferably, when the spatial distribution of each detection node changes, causing the geometric accuracy factor of the constructed tetrahedral detection array to deteriorate beyond a preset threshold, or when a preset time interval is reached, the data processing unit reselects four nodes to reconstruct the tetrahedral detection array. The data processing unit sends a position adjustment command to the UAV onboard node, and the UAV onboard node adjusts its own spatial position according to the command to optimize the reconstructed tetrahedral array spatial configuration.

[0009] Preferably, the detection node further includes a fixedly installed microphone base station and a mobile microphone node, wherein the mobile microphone node is a soldier-carried node or a vehicle-mounted node. Each detection node is equipped with a microphone array unit consisting of four microphones, a positioning module, an inertial measurement unit, a wireless synchronization module, and an edge computing module. The four microphones are located at the four vertices of a regular tetrahedron.

[0010] Preferably, the data processing unit includes a time delay estimation module, a spatial intersection module, and an error compensation module; The time delay estimation module estimates the time delay of the sound signal received by the microphone in each detection node and calculates the azimuth and elevation angles of the sound source relative to each detection node. The spatial intersection module calculates the three-dimensional coordinates of the sound source through spatial geometric intersection based on the spatial position of the four detection nodes and the azimuth and elevation angles measured by each node. The error compensation module corrects the calculation results using an error compensation model.

[0011] Preferably, the data processing unit is provided with a cross-validation module. The cross-validation module acquires the image captured by the camera, projects the sound source coordinates onto the image plane according to the three-dimensional coordinates of the sound source and the spatial pose of the camera, verifies the presence of a drone target at the projected position using a target detection algorithm, and corrects the three-dimensional coordinates of the sound source by minimizing the reprojection error when the deviation between the acoustic positioning result and the image detection result exceeds a threshold.

[0012] Preferably, the data processing unit includes a map fusion module, which maps the three-dimensional coordinates of the sound source to the geospatial data. The geospatial data includes a three-dimensional city model, which at least contains building outlines and elevation information. The map fusion module combines the building outlines and elevation information to perform sound propagation path occlusion analysis, identify multipath effects, and generate three-dimensional visual markers of the sound source in the scene. By dynamically reconstructing a tetrahedral array, the multipath interference caused by urban buildings is reduced, thereby improving the three-dimensional positioning accuracy of low-altitude UAVs.

[0013] Preferably, the wireless synchronization module uses the UWB protocol or PTP protocol to achieve time synchronization between the detection nodes, and each detection node collects acoustic signals with a unified time reference and the collected signals have a unified timestamp; the edge computing module completes the acquisition and preprocessing of acoustic signals at each detection node; the positioning module is a GPS receiver module or a Beidou receiver module; the spatial distance between the four microphones is a fixed value; the sensor data includes at least one of image data collected by the camera, radar detection data, and infrared sensor data.

[0014] Preferably, a method for UAV cooperative detection and 3D localization based on a dynamic tetrahedral microphone array includes the following steps: S1, Deploy multiple detection nodes, including UAV onboard microphone nodes, each detection node acquires its own spatial position and attitude and completes time synchronization; S2, each detection node reports its status information to the data processing unit, and the data processing unit establishes and maintains a global node status database. S3, the data processing unit selects four nodes from multiple detection nodes to construct a tetrahedral detection array based on the spatial distribution of each detection node and according to the evaluation index that characterizes the relationship between positioning error and spatial geometric configuration. S4, each detection node in the constructed tetrahedral array collects acoustic signals, which are acoustic signals generated by the rotor of the UAV; S5, perform time delay estimation on the acoustic signals collected by each detection node, and calculate the azimuth and elevation angles of the sound source relative to each detection node; S6. Based on the spatial positions of the four detection nodes and the azimuth and elevation angles measured by each node, the three-dimensional coordinates of the sound source are calculated through spatial geometric intersection. S7, acquire sensor data, and correct the calculation results based on the sensor data; S8 maps the corrected three-dimensional coordinates of the sound source to geospatial data and outputs the sound source location information; Steps S2 to S8 are executed repeatedly to continuously detect and perform three-dimensional positioning and tracking of the UAV target.

[0015] Preferably, the evaluation index in S3 includes at least one of geometric accuracy factor, positioning error ellipsoid volume, and root mean square error of sound source positioning. When four nodes are selected, the value of the evaluation index for any combination of four nodes is calculated, and the combination that makes the evaluation index optimal and the tetrahedral volume greater than the threshold is selected. The data processing unit issues a position adjustment command to the UAV onboard node, and the UAV onboard node adjusts its own spatial position according to the command to optimize the reconstructed array configuration. When the displacement of any node in the selected nodes causes the geometric accuracy factor of the constructed array to deteriorate beyond the threshold or reaches the preset time interval, S3 is re-executed. The time delay estimation in S5 adopts a generalized cross-correlation and phase transformation algorithm. The sensor data in S7 includes at least one of image data acquired by the camera, radar detection data, and infrared sensing data. The correction is to correct the three-dimensional coordinates of the sound source by minimizing the reprojection error.

[0016] Compared with existing technologies, this invention provides a method and system for UAV cooperative detection and three-dimensional localization based on a dynamic tetrahedral microphone array, which has the following beneficial effects: The present invention provides a UAV cooperative detection and 3D positioning system and method based on a dynamic tetrahedral microphone array, which has the following beneficial effects: (1) The present invention uses a data processing unit to dynamically select four nodes from multiple detection nodes to construct a tetrahedral array using an evaluation index that characterizes the relationship between positioning error and spatial geometric configuration. This overcomes the defect of fixed array geometric configuration and maintains the optimal geometric configuration when the spatial distribution of nodes changes through dynamic reconstruction, thereby improving positioning accuracy and adaptability.

[0017] (2) By setting up an onboard microphone node for the UAV, the UAV can participate in array construction as both a detection target and a network detection unit. The data processing unit sends position adjustment instructions to the onboard node of the UAV to optimize the array spatial configuration, thereby realizing bidirectional collaboration between the detection node and the target and expanding the collaborative detection coverage.

[0018] (3) The present invention obtains data from cameras, radar or infrared sensors through a multi-view image cross-validation module to correct the acoustic positioning results, and combines a three-dimensional geographic information module to perform sound propagation path occlusion analysis and multipath effect identification, which effectively suppresses sound multipath interference in complex urban environments and improves the reliability and accuracy of three-dimensional positioning of low-altitude UAVs. Attached Figure Description

[0019] Figure 1 This is a diagram showing the overall system architecture of the method of the present invention; Figure 2 This is a flowchart of the UAV collaborative detection and 3D positioning method of the present invention; Figure 3 This is a flowchart illustrating the construction process of the dynamic tetrahedral array of the method system of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Urban low-altitude UAV detection and positioning based on fixed base stations and individual soldier nodes.

[0022] Purpose of implementation: The purpose of this embodiment is to verify the effectiveness of the system of the present invention in a typical urban low-altitude defense scenario. Specifically, it verifies whether the data processing unit can dynamically select four nodes to construct a tetrahedral detection array based on the spatial position of each detection node and an evaluation index characterizing the relationship between positioning error and spatial geometric configuration when the fixed microphone base station and the mobile microphone node (carried by a single soldier) work together in the hybrid network detection layer. Furthermore, it verifies the accuracy of calculating the three-dimensional coordinates of the UAV target based on this array, and the effect of three-dimensional geographic information-assisted sound propagation obstruction analysis and image cross-validation on improving positioning accuracy. This embodiment focuses on testing the comprehensive performance of the system's dynamic tetrahedral array construction function, sound source localization calculation function, and multi-view image cross-validation function in a real urban environment.

[0023] Implementation System: The detection nodes used in this embodiment include three fixed microphone base stations and two man-portable portable microphone nodes. The three fixed base stations are installed on the rooftops of three high-rise buildings in the core urban area, with coordinates A(100,200,85), B(450,150,92), and C(380,480,78), respectively, in meters. The two man-portable nodes are carried by patrol personnel, with initial positions at D(200,350,1.5) and E(500,300,1.5), respectively. Each detection node integrates a microphone array unit consisting of four microphones with identical characteristics, located at the four vertices of a regular tetrahedron with an edge length of 0.15 meters. Each node also integrates a GPS / BeiDou receiver module for obtaining position coordinates, an IMU inertial measurement unit for obtaining attitude angles, a UWB wireless synchronization module for time synchronization between detection nodes, and an edge computing module for acquiring and preprocessing acoustic signals at the node. The data processing unit is deployed in the command center and connected to each detection node via a wireless network. The unit includes an array construction module, a latency estimation module, a spatial intersection module, an error compensation module, a cross-validation module, and a map fusion module. The system also deploys a multi-view image acquisition module, which provides real-time image sequences from a camera installed on the rooftop of the building where node A is located. The 3D geographic information module pre-stores a 3D city model of the area, including building outlines and elevation information.

[0024] Implementation process: After deployment, each detection node acquires its own three-dimensional spatial coordinates via GPS / BeiDou receiver modules, obtains its attitude angles via IMU inertial measurement units, and achieves high-precision time synchronization between the detection nodes via UWB wireless synchronization modules, with a synchronization accuracy better than 1 microsecond. Each node begins acquiring acoustic signals using a unified time reference, and the acquired signals are stamped with a unified timestamp.

[0025] After completing self-localization and time synchronization, each detection node reports its status information to the data processing unit via a wireless network, including node ID, 3D spatial coordinates, attitude angle, synchronization status, battery level, and network connection status. Upon receiving the reported information, the data processing unit establishes and maintains a global node status database. The database uses the node ID as the primary key and updates the spatial location and status parameters of each node in real time.

[0026] Based on the spatial distribution of currently available nodes, the data processing unit selects four nodes from five detection nodes to construct the optimal tetrahedral detection array. Specifically, the selection method involves calculating the volume of the tetrahedron formed by any combination of four nodes and its corresponding geometric precision factor (GMP). Calculations show that among the five possible combinations (A, B, D, E), the combination with the smallest GMP value (1.5) corresponds to a tetrahedron volume of 2.31 × 10^6 cubic meters, satisfying the condition that the tetrahedron volume exceeds a preset threshold of 500 cubic meters. The data processing unit selects this combination to construct the tetrahedral detection array.

[0027] In the constructed tetrahedral array, four microphones at each detection node synchronously acquire ambient acoustic signals at a sampling rate of 48 kHz. The edge computing module of each node preprocesses the acquired acoustic signals, filtering out low-frequency ambient noise and high-frequency interference through a bandpass filter (passband frequency from 200 Hz to 8000 Hz), performing a short-time Fourier transform on the filtered signal using a Hamming window with a window length of 1024 points and a 50% overlap rate, and then performing dereverberation processing on the transformed signal, employing spectral subtraction to suppress the impact of room reverberation on time delay estimation.

[0028] Within each detection node, the preprocessed signals from the four microphones are evaluated using a generalized cross-correlation-phase transform algorithm for cross-correlation time delay estimation. Taking node A as an example, the four microphones are located at A1, A2, A3, and A4. The cross-power spectrum of the signals received by any two microphones A1 and A2 is calculated. After applying a phase transform weighting function, the generalized cross-correlation function is calculated. Peak detection is performed on the generalized cross-correlation function, and the time delay value corresponding to the peak position is the time difference of arrival between the two microphones. Using three independent time differences of arrival, combined with the spatial coordinates of the four microphones, the azimuth and elevation angles of the sound source relative to node A are calculated. Similarly, the azimuth and elevation angles measured by nodes B, D, and E are calculated respectively. In this embodiment, node A measures an azimuth of 135.2 degrees and an elevation of 28.7 degrees, node B measures an azimuth of 112.5 degrees and an elevation of 32.1 degrees, node D measures an azimuth of 156.8 degrees and an elevation of 25.3 degrees, and node E measures an azimuth of 98.4 degrees and an elevation of 35.6 degrees.

[0029] The data processing unit calculates the three-dimensional coordinates of the sound source using the spatial geometric intersection method based on the spatial coordinates of the four detection nodes A, B, D, and E, and the azimuth and elevation angles measured by each node. The sound source position is denoted as S(x, y, z), and the detection node position is denoted as Pᵢ(xᵢ, yᵢ, zᵢ). The unit direction vector of the sound source observed from node Pᵢ is calculated from the azimuth and elevation angles. According to spatial geometric relationships, there are constraint equations between the sound source position, node positions, and the unit direction vector. By simultaneously solving the observation equations of the four nodes, the optimal sound source position is solved using the nonlinear least squares method. Through iterative solution, the three-dimensional coordinates of the sound source are obtained as S = (320.5, 260.8, 52.3) meters.

[0030] Since the sound source is located in a far-field environment (the distance between the sound source and the array is greater than ten times the array aperture), an error compensation model is used to correct the solution results. In this embodiment, the error compensation model uses a radial basis function neural network with a three-layer structure. The input layer is the initial three-dimensional coordinates of the spatial intersection solution (3 neurons), the hidden layer uses a Gaussian kernel function (12 neurons), and the output layer is the compensated three-dimensional coordinates (3 neurons). The network training data comes from 5000 sets of samples generated by simulation. The samples cover the range of sound source distance from 50 meters to 500 meters, azimuth angle from 0 degrees to 360 degrees, and elevation angle from 0 degrees to 90 degrees. The input is the initial coordinates of the time difference data solution containing noise, and the label is the true coordinates. After correction by the error compensation model, the sound source coordinates in this embodiment are corrected to S' = (318.7, 262.3, 53.1) meters.

[0031] The data processing unit maps the corrected 3D coordinates of the sound source, S' = (318.7, 262.3, 53.1) meters, onto a 3D city model. A query of the 3D geographic information module reveals that these coordinates correspond to the southern part of the top of a 62-meter-high office building, approximately 8 meters from the roof edge. The map fusion module combines building outlines and elevation information to perform sound propagation path obstruction analysis, determining whether the sound propagation path between the sound source and each detection node is obstructed by buildings. Analysis shows no obstruction between node A and the sound source, partial obstruction between node B and the sound source but with a diffraction path, and no obstruction between nodes D and E and the sound source. The system's overall location reliability is 92%.

[0032] In the multi-view image acquisition module, a camera installed on the rooftop of the building where node A is located synchronously acquired an image sequence, and the spatial pose of this camera is known. The cross-validation module of the data processing unit projects the three-dimensional coordinates S' of the sound source onto the image plane of the camera, with the projection position corresponding to the image coordinates (1240, 680). A target detection algorithm is used to detect this area, and a UAV target is detected near the projection position, with the center coordinates of the detection box at (1235, 675), deviating from the acoustic projection position by approximately 5 pixels. The deviation between the acoustic localization result and the image detection result is less than a preset threshold of 10 pixels, so the cross-validation module determines that the acoustic localization result is reliable. By minimizing the reprojection error, the sound source coordinates are fine-tuned to S'' = (318.5, 262.8, 52.9) meters. Finally, a visual marker of the UAV target on a 3D map is output, and the localization result is recorded.

[0033] The system executes the steps of node status reporting, tetrahedral array construction, acoustic signal acquisition, time delay estimation, spatial intersection calculation, image cross-validation, and coordinate output in a 1-second cycle. During continuous tracking, individual soldier nodes D and E move with the patrol personnel. When the cumulative displacement of node D exceeds 5 meters, the array construction module triggers array reconstruction, re-selecting the optimal four-node combination from the currently available nodes to construct the detection array, ensuring continuous high-precision positioning and tracking of the moving UAV target.

[0034] Implementation results: In this embodiment, under the aforementioned typical urban environment, a quadcopter UAV (flying at an altitude of approximately 50 meters and a horizontal distance of approximately 300 meters) was detected and located. After 10 repeated experiments, the average horizontal positioning error was 0.8 meters, and the average vertical error was 1.2 meters. The positioning success rate was 95% for positioning errors less than 3 meters. Compared with traditional fixed tetrahedral arrays (fixed four nodes, no dynamic reconstruction), the horizontal positioning accuracy was improved by approximately 40%, and the vertical positioning accuracy was improved by approximately 35%.

[0035] Example 2: Dynamic networking and multi-sensor cross-verification based on UAV onboard nodes.

[0036] Purpose of implementation: The purpose of this embodiment is to verify the core collaborative logic of the present invention, in which the UAV's onboard microphone node participates in collaborative detection, and the UAV is both a detection target and a network detection unit. Specifically, it verifies the feasibility of the UAV's onboard microphone node being selected by the data processing unit as one of the detection nodes and participating in the construction of a tetrahedral detection array; it verifies the effectiveness of the UAV adjusting its spatial position to optimize the spatial configuration of the tetrahedral array after the data processing unit issues a position adjustment command to the UAV's onboard node; and it verifies the reliability of cross-validation and correction of the sound source localization results using auxiliary sensor data (including image data collected by the camera and radar detection data). This embodiment focuses on verifying the system's dynamic array optimization function and multi-sensor fusion localization function.

[0037] Implementation System: The detection nodes used in this embodiment include three fixed microphone base stations, one vehicle-mounted mobile microphone node, and one UAV-borne microphone node. The positions of the three fixed base stations are F(0, 0, 20), G(200, 0, 18), and H(100, 180, 22), respectively, with coordinates in meters. The initial position of the vehicle-mounted node is I(-50, 100, 1.5). The UAV-borne microphone node is mounted on a quadcopter UAV, with an initial hovering position of J(120, 80, 40). This node integrates a microphone array unit consisting of four microphones with identical characteristics (the four microphones are located at the four vertices of a regular tetrahedron, with an edge length of 0.12 meters), a GPS / BeiDou receiver module, an IMU inertial measurement unit, a UWB wireless synchronization module, and an edge computing module. It is powered by an onboard power supply and communicates with the ground nodes via a wireless network. The data processing unit is deployed on an edge server, and each detection node connects to the data processing unit via a wireless network. The system deploys a multi-view image acquisition module, including a camera installed at node F and a millimeter-wave radar (operating frequency 77 GHz) deployed near node G, used to acquire image data and radar detection data, respectively. The 3D geographic information module has a pre-stored 3D city model of the area.

[0038] Implementation process: After each detection node completes self-localization and time synchronization, it reports its status information to the data processing unit. The data processing unit establishes a global node status database, currently showing five available nodes: F, G, H, I, and J.

[0039] Based on the spatial distribution of currently available nodes, the data processing unit selects four nodes from five to construct a tetrahedral detection array, using the minimum geometric precision factor as the criterion. Calculations show that among the five-to-four combinations, the combination F, G, H, and J has the smallest geometric precision factor of 1.8. Therefore, the data processing unit selects this combination to construct the tetrahedral detection array, meaning that the UAV-borne node J is chosen to participate in the network.

[0040] The data processing unit further evaluates whether the current array's geometric precision factor value meets the accuracy requirements. In this embodiment, the preset geometric precision factor threshold is 2.0, and the current value of 1.8 meets the requirements. However, considering the mobility of the airborne node J, the data processing unit issues a position adjustment command to the airborne node J, suggesting that it move from its current position (120, 80, 40) to (130, 90, 55) to further optimize the spatial configuration of the tetrahedral array. After receiving the command, the airborne node J is positioned by the UAV flight control system, taking approximately 8 seconds to reach the designated position. Upon arrival, the airborne node J re-reports its position coordinates, and the data processing unit recalculates the array's geometric precision factor value, reducing it to 1.5, confirming that the array configuration optimization is complete.

[0041] The microphone subarrays of nodes F, G, H, and J in the array synchronously acquire acoustic signals. The acoustic signals include those generated by the rotor of the target UAV and those generated by the rotor of the airborne node J itself. Because the microphone array of the airborne node J is installed at a vibration-isolated location and is far from its own rotor, the acoustic signal from its own rotor is attenuated to below -20 dB at the microphone, and does not substantially interfere with the acquisition of the target acoustic signal.

[0042] After the edge computing modules of each node complete the acoustic signal preprocessing, they use a generalized cross-correlation-phase transform algorithm to estimate the time delay and calculate the azimuth and elevation angles of the target sound source relative to each node. In this embodiment, node F measures an azimuth of 45.3 degrees and an elevation of 15.2 degrees, node G measures an azimuth of 32.7 degrees and an elevation of 18.5 degrees, node H measures an azimuth of 58.1 degrees and an elevation of 12.8 degrees, and node J measures an azimuth of 40.5 degrees and an elevation of 22.3 degrees. The data processing unit calculates the three-dimensional coordinates of the target sound source as S = (165.2, 82.5, 28.6) meters using the spatial geometric intersection method.

[0043] The multi-view image acquisition module synchronously provides auxiliary sensing data. A camera installed at node F acquires real-time images. The cross-validation module of the data processing unit projects the sound source coordinates S = (165.2, 82.5, 28.6) meters onto the camera's image plane, with the projection position corresponding to image coordinates (980, 540). A target detection algorithm detects the area, finding a UAV target near the projection position. The detection box center coordinates are (975, 538), deviating from the acoustic projection position by approximately 5 pixels. This deviation is less than the preset threshold of 10 pixels, and the image cross-validation passes. Simultaneously, a millimeter-wave radar deployed near node G synchronously outputs the target distance as 175 meters, azimuth angle as 35.5 degrees, and elevation angle as 17.8 degrees. Comparing the radar data with the acoustic positioning results, the acoustic positioning results correspond to a distance of 178.3 meters, an azimuth angle of 35.2 degrees, and an elevation angle of 17.6 degrees. The acoustic and radar distance deviations were 3.3 meters, azimuth deviations were 0.3 degrees, and elevation deviations were 0.2 degrees. All deviations were less than the preset thresholds (distance threshold 5 meters, angle threshold 2 degrees), and the radar data verification was successful.

[0044] Because the deviation between the image detection results and the acoustic localization results is small and the radar data shows good consistency, the cross-validation module determines that the acoustic localization results have high reliability. By minimizing the reprojection error (combining the center coordinates of the image detection box as a constraint), the sound source coordinates are fine-tuned to S'=(164.8, 82.9, 28.4) meters. Finally, the corrected coordinates are mapped to the 3D city model provided by the 3D geographic information module, generating a 3D visualization marker for the UAV target at the corresponding location. The marker color is green (indicating a localization confidence level higher than 90%), and the localization result is output.

[0045] The system continuously performs detection and positioning with a cycle of 0.5 seconds. During the 15th cycle, airborne node J needs to leave the current area to perform other tasks and sends a withdrawal request to the data processing unit. After confirmation, the data processing unit removes the node from the available node set and immediately triggers array reconstruction. Currently, four nodes (F, G, H, and I) are available. A new tetrahedral detection array is constructed using these four nodes, with a geometric precision factor of 3.5, which still meets the system's positioning requirements (the maximum allowable geometric precision factor in this embodiment is 4.0), and detection and positioning continue normally. When airborne node J completes its task and rejoins the network, the data processing unit triggers reconstruction again, reselecting the optimal four-node combination.

[0046] Implementation results: This embodiment verifies the effectiveness of UAV-borne nodes participating in network deployment as collaborative detection units. The addition of the onboard nodes reduced the array geometric accuracy factor from 3.2 (without onboard nodes) to 1.8 (with onboard nodes), the horizontal positioning error from 1.5 meters to 0.6 meters, and the vertical error from 2.2 meters to 1.0 meter. Dual cross-validation using imagery and radar reduced the false alarm rate from 8% with acoustic positioning alone to below 0.5%. After the UAV-borne nodes actively adjusted their positions to optimize the array configuration, the geometric accuracy factor was further reduced from 1.8 to 1.5, improving positioning accuracy by approximately 20%, thus validating the effectiveness of the collaborative logic that "the UAV is both a detection target and a network detection unit."

[0047] Example 3: Large-scale node collaboration and multipath suppression in complex environments.

[0048] Purpose of implementation: The purpose of this embodiment is to verify the feasibility and effectiveness of the present invention in simultaneously conducting collaborative detection of multiple UAV targets under large-scale hybrid networking conditions (including fixed base stations, individual soldier nodes, vehicle-mounted nodes, and multiple UAV airborne nodes). Specifically, it verifies the optimal efficiency of the data processing unit in dynamically selecting four nodes to construct a tetrahedral detection array under large-scale node conditions based on minimizing the geometric accuracy factor; verifies the effectiveness of the 3D map fusion module in combining building outlines and elevation information to perform sound propagation path occlusion analysis and identify multipath effects; verifies the system's ability to simultaneously detect, locate, and track multiple sound source targets; and verifies the effect of dynamically reconstructing the tetrahedral array on reducing sound multipath interference caused by urban buildings and improving the 3D positioning accuracy of low-altitude UAVs. This embodiment focuses on testing the system's multi-target positioning function, 3D geographic information-assisted multipath suppression function, and large-scale node management function.

[0049] Implementation System: This embodiment deploys a large-scale detection node network in a city's central business district. The detection nodes include 12 fixed microphone base stations (distributed on the rooftops of high-rise buildings and at key street locations), 6 soldier-mounted nodes, 3 vehicle-mounted nodes, and 4 UAV-borne microphone nodes, totaling 25 detection nodes. Each detection node integrates a microphone array unit consisting of four identical microphones (each microphone is located at one of the four vertices of a regular tetrahedron, with edge lengths ranging from 0.12 meters to 0.18 meters), a GPS / BeiDou receiver module, an IMU inertial measurement unit, a wireless synchronization module (some nodes use the PTP protocol for sub-microsecond time synchronization), and an edge computing module. The data processing unit is deployed on a cloud server and connects to each detection node via a wireless network. The 3D geographic information module pre-stores a high-precision 3D city model of the area, including detailed building outlines and elevation information. The multi-view image acquisition module includes multiple fixed cameras distributed throughout the scene.

[0050] Implementation process: After each detection node completes self-localization and time synchronization, it reports its status information to the data processing unit. The data processing unit establishes a global node status database. There are currently 23 available nodes (2 of which are offline due to network failure).

[0051] The data processing unit selects four nodes to construct the optimal tetrahedral detection array based on the spatial distribution of the 23 available nodes, using the minimum geometric precision factor as the criterion. Due to the large number of nodes, exhaustively calculating the geometric precision factor values ​​for the 8855 combinations of 23-to-4 selection is computationally intensive (approximately 2.3 seconds per calculation). To improve computational efficiency, this embodiment employs a fast optimization algorithm based on spatial clustering: First, K-means clustering (forming 4 clusters) is performed based on the spatial distribution of the nodes, dividing the 23 nodes into 4 clusters according to their spatial location; then, the two nodes closest to the cluster center are selected from each cluster as candidates, resulting in 8 candidate nodes; finally, the geometric precision factor values ​​and tetrahedral volumes for 70 combinations of the 8 candidate nodes are calculated, and the combination with the minimum geometric precision factor and a tetrahedral volume greater than a preset threshold of 1000 cubic meters is selected. Calculations showed that the optimal combination of four nodes had coordinates of (150, 200, 95), (350, 180, 88), (220, 420, 82), and (280, 260, 52), with a geometric precision factor of 1.2 and a tetrahedral volume of approximately 3.5 × 10^6 cubic meters. The data processing unit selected this combination to construct the tetrahedral detection array.

[0052] Each node in the constructed tetrahedral array collects acoustic signals. Since there are three drones (i.e., three sound source targets) in the scene, the microphone arrays of each node receive a mixed signal from multiple sound sources. The edge computing module of each node first separates the mixed signal into three independent sound source components using a blind source separation algorithm (employing independent component analysis). Then, it performs generalized cross-correlation-phase transform time delay estimation on each sound source component to calculate the azimuth and elevation angles of each sound source relative to each node. In this embodiment, the observation results of node P1 (coordinates 150, 200, 95) for the three sound sources are as follows: sound source 1: azimuth 60.2 degrees, elevation 25.3 degrees; sound source 2: azimuth 120.5 degrees, elevation 18.7 degrees; sound source 3: azimuth 175.3 degrees, elevation 32.1 degrees. The observation results for node P2 (coordinates 350, 180, 88) are as follows: azimuth 45.8 degrees, elevation 22.1 degrees for sound source 1; azimuth 98.2 degrees, elevation 20.3 degrees for sound source 2; and azimuth 152.6 degrees, elevation 28.5 degrees for sound source 3. The observation results for node P3 (coordinates 220, 420, 82) are as follows: azimuth 72.5 degrees, elevation 20.8 degrees for sound source 1; azimuth 135.8 degrees, elevation 15.2 degrees for sound source 2; and azimuth 165.2 degrees, elevation 35.8 degrees for sound source 3. The observation results for node P4 (coordinates 280, 260, 52) are as follows: azimuth angle of sound source 1 is 55.3 degrees and elevation angle is 30.2 degrees; azimuth angle of sound source 2 is 110.7 degrees and elevation angle is 22.5 degrees; and azimuth angle of sound source 3 is 160.8 degrees and elevation angle is 40.3 degrees.

[0053] The data processing unit performs spatial geometric intersection calculations on each sound source to obtain the three-dimensional coordinates of the three sound sources: sound source 1 is S1=(215.3, 285.6, 42.8) meters, sound source 2 is S2=(285.7, 155.3, 28.5) meters, and sound source 3 is S3=(125.8, 320.4, 55.2) meters.

[0054] The densely built-up urban central business district experiences severe multipath effects during sound propagation (sound waves reflect off building walls and reach the microphone, forming multiple propagation paths and interfering with the accuracy of sound arrival time difference estimation). In this embodiment, the map fusion module combines the building outlines and elevation data of the 3D city model to analyze the sound propagation paths between each sound source and each detection node. Taking sound source 1S1 = (215.3, 285.6, 42.8) meters and node P1 = (150, 200, 95) meters as an example, the map fusion module calculates the straight-line path between the sound source and the node and detects whether the path intersects with the building model. The calculation shows that the straight-line path passes through a building (58 meters high) at a height of 62 meters, indicating an obstruction. The system then calculates the reflection path: the sound wave reflects off the south wall of the building (coordinates approximately (180, 230, 45)) and reaches node P1. The reflection path length is 1.3 times that of the straight-line path, corresponding to a time delay of approximately 1.2 milliseconds.

[0055] Based on the multipath analysis results, the data processing unit suppresses multipath components during the time delay estimation stage. For detector node-source pairs with obstructions, in the generalized cross-correlation-phase transform cross-correlation function, in addition to the main peak, there may be secondary peaks (corresponding to the reflection path). The system calculates the multipath time delay estimate based on the 3D geographic information, removes the secondary peaks corresponding to the time delay positions in the cross-correlation function, and retains only the main peak for sound arrival time difference estimation. After multipath suppression, the spatial geometric intersection solution is recalculated, and the coordinates of source 1 are corrected to S1'=(218.6, 282.3, 43.5) meters.

[0056] The system monitors the geometric precision factor (GMP) of the current array every 0.5 seconds. During the 30th cycle, due to changes in the positions of multiple moving nodes, the GMP value rose from 1.2 to 2.8, exceeding the preset threshold of 2.5, triggering array reconfiguration. The system then selects the optimal four-node combination from the current 23 available nodes; the new combination has a GMP of 1.5, meeting the requirements. The system continuously performs synchronous detection and 3D positioning tracking of the three UAV targets, with the positioning data of each target displayed in real-time on a 3D map in the form of trajectories.

[0057] Implementation results: Under large-scale hybrid networking (25 nodes), through dynamic tetrahedral array optimization, the system can select the optimal four-node combination from 23 available nodes in sub-second time. The geometric accuracy factor value is optimized from 3.8 in random selection to 1.2 after optimization, improving positioning accuracy by approximately 65%. Through multipath suppression processing assisted by 3D geographic information, the positioning error under building obstruction conditions is reduced from an average of 4.5 meters before suppression to an average of 1.8 meters after suppression, improving positioning accuracy by approximately 60%. This verifies the effectiveness of dynamically reconstructing the tetrahedral array in reducing acoustic multipath interference caused by urban buildings and improving the 3D positioning accuracy of low-altitude UAVs. The system can simultaneously detect, locate, and track multiple UAV targets; in this embodiment, three targets were successfully separated and located.

[0058] Comparative Example: A comparative experiment between a traditional fixed tetrahedral microphone array and the system of this invention.

[0059] Purpose of implementation: The purpose of this comparative experiment is to verify, through rigorous comparative experiments, whether the present invention represents a significant technological advancement compared to existing technologies (traditional fixed tetrahedral microphone array sound source localization systems). Specifically, under the same test scenario and environmental conditions, the system of the present invention and the traditional fixed array system were used to detect and locate the same UAV target. The positioning accuracy, positioning stability, positioning success rate, and adaptability to complex environments of the two systems were compared to quantitatively evaluate the actual technical effect of the present invention.

[0060] Implementation System: The system of this invention is executed according to the configuration of Embodiment 1, including 3 fixed microphone base stations and 2 individual soldier backpack nodes (a total of 5 detection nodes, of which 4 are dynamically selected to construct a tetrahedral array in actual work). Each node integrates a regular tetrahedral microphone subarray and adopts all technical means such as dynamic tetrahedral array construction and reconstruction, geometric accuracy factor optimization, error compensation model correction and multi-view image cross-validation.

[0061] Traditional fixed tetrahedral microphone array systems employ four microphone base stations in the same locations as the three fixed base stations in the system of this invention (selecting the same four fixed locations as the system of this invention). The four base stations are located at the four corners of the test site, with fixed positions that do not adjust with target movement, and do not include UAV-borne nodes. Each base station is configured with only a single microphone instead of a regular tetrahedral subarray, and uses a traditional four-station intersection localization algorithm based on time difference of arrival, lacking dynamic array reconstruction, geometric accuracy factor optimization, error compensation model, and cross-validation with auxiliary sensor data.

[0062] Implementation process: The test was conducted in an open area on the outskirts of a city, measuring 500 meters by 500 meters, with a few low-rise buildings (5 to 10 meters high). The target drone was a quadcopter, flying at an altitude of 30 to 50 meters, following a pre-programmed figure-eight path, at a speed of 5 to 10 meters per second. Environmental conditions included sunny skies, a light breeze (wind speed less than 3 meters per second), an ambient temperature of 25 degrees Celsius, and 60% humidity.

[0063] Two systems simultaneously detect and locate the same UAV target. The actual trajectory of the UAV measured by differential GPS is used as a reference (positioning accuracy better than 0.1 meters). The two systems output positioning results once per second, collecting a total of 600 sets of positioning data.

[0064] During the test, the positioning performance of the two systems was specifically recorded when the drone flew behind a building (i.e., under occlusion conditions). Under occlusion conditions, the traditional fixed array system showed a sharp increase in positioning error or even loss of positioning; the system of this invention, however, continued to maintain positioning output through dynamic array reconstruction and multipath suppression assisted by three-dimensional geographic information.

[0065] Based on statistical analysis of 600 sets of positioning data, the horizontal positioning error of the system of this invention is as follows: the average horizontal positioning error is 0.65 meters, while that of the traditional fixed array system is 2.80 meters, representing an improvement of 76.8%. The maximum horizontal positioning error of the system of this invention is 1.80 meters, while that of the traditional system is 6.50 meters, representing an improvement of 72.3%. The root mean square value of the horizontal positioning error of the system of this invention is 0.82 meters, while that of the traditional system is 3.20 meters, representing an improvement of 74.4%. Regarding vertical positioning, the average vertical positioning error of the system of this invention is 1.05 meters, while that of the traditional system is 4.20 meters, representing an improvement of 75.0%. The maximum vertical positioning error of the system of this invention is 2.50 meters, while that of the traditional system is 9.80 meters, representing an improvement of 74.5%. The root mean square value of the vertical positioning error of the system of this invention is 1.28 meters, while that of the traditional system is 5.10 meters, representing an improvement of 74.9%. In terms of positioning success rate (the percentage of positioning errors less than 3 meters), the system of this invention achieves 98.3%, while the traditional system achieves 52.7%, representing an improvement of 45.6%.

[0066] Under building obstruction conditions, the positioning error of traditional fixed array systems increases sharply to more than 10 meters, and even positioning loss occurs due to the inability to calculate coordinates; while the system of this invention, through multipath suppression assisted by three-dimensional geographic information and dynamic array reconstruction, can still maintain a positioning error of less than 3 meters under obstruction conditions.

[0067] In summary, this comparative example demonstrates through rigorous comparative experiments that, compared to traditional fixed tetrahedral microphone array sound source localization systems, this invention achieves significant improvements in positioning accuracy, positioning stability, and adaptability to complex environments through techniques such as dynamic tetrahedral array construction and reconstruction, geometric accuracy factor optimization, cross-validation of auxiliary sensor data, and multipath suppression assisted by three-dimensional geographic information. The positioning error is reduced by more than 70%, and the positioning success rate is increased by more than 45%, demonstrating significant technological advancements and practical value.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A UAV cooperative detection and 3D positioning system based on a dynamic tetrahedral microphone array, comprising detection nodes and a data processing unit, characterized in that: The detection nodes include UAV-borne microphone nodes. The data processing unit selects four nodes from the plurality of detection nodes to construct a tetrahedral detection array based on the spatial position of each detection node and according to the evaluation index characterizing the relationship between positioning error and spatial geometric configuration. The data processing unit calculates the three-dimensional coordinates of the sound source based on the acoustic signals received by each detection node in the tetrahedral detection array. The acoustic signals are acoustic signals generated by the rotor of the UAV. The data processing unit also corrects the calculation results based on sensor data and outputs sound source location information based on geospatial data.

2. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: The evaluation index includes at least one of the following: geometric accuracy factor, positioning error ellipsoid volume, and root mean square error of sound source positioning. The data processing unit calculates the value of the evaluation index for any combination of four nodes among the currently available nodes, and selects the combination that makes the evaluation index optimal and the volume of the tetrahedron formed by the four nodes greater than a preset threshold to construct the tetrahedron detection array.

3. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: When the spatial distribution of each detection node changes, causing the geometric accuracy factor of the constructed tetrahedral detection array to deteriorate beyond a preset threshold, or when a preset time interval is reached, the data processing unit reselects four nodes to reconstruct the tetrahedral detection array. The data processing unit sends a position adjustment command to the UAV onboard node, and the UAV onboard node adjusts its own spatial position according to the command to optimize the reconstructed tetrahedral array spatial configuration.

4. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: The detection node also includes a fixed microphone base station and a mobile microphone node, wherein the mobile microphone node is a soldier-carried node or a vehicle-mounted node. Each detection node is equipped with a microphone array unit consisting of four microphones, a positioning module, an inertial measurement unit, a wireless synchronization module, and an edge computing module. The four microphones are located at the four vertices of a regular tetrahedron.

5. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: The data processing unit includes a time delay estimation module, a spatial intersection module, and an error compensation module; The time delay estimation module estimates the time delay of the sound signal received by the microphone in each detection node and calculates the azimuth and elevation angles of the sound source relative to each detection node. The spatial intersection module calculates the three-dimensional coordinates of the sound source through spatial geometric intersection based on the spatial positions of the four detection nodes and the azimuth and elevation angles measured by each node; the error compensation module corrects the calculation results using an error compensation model.

6. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: The data processing unit is equipped with a cross-validation module. The cross-validation module acquires images captured by the camera, projects the sound source coordinates onto the image plane according to the three-dimensional coordinates of the sound source and the spatial pose of the camera, verifies the presence of a drone target at the projected position using a target detection algorithm, and corrects the three-dimensional coordinates of the sound source by minimizing the reprojection error when the deviation between the acoustic positioning result and the image detection result exceeds a threshold.

7. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 1, characterized in that: The data processing unit includes a map fusion module, which maps the three-dimensional coordinates of the sound source to the geospatial data. The geospatial data includes a three-dimensional city model, which at least contains building outlines and elevation information. The map fusion module combines the building outlines and elevation information to perform sound propagation path occlusion analysis, identify multipath effects, and generate three-dimensional visual markers of the sound source in the scene. By dynamically reconstructing a tetrahedral array, the multipath interference caused by urban buildings is reduced, thereby improving the three-dimensional positioning accuracy of low-altitude UAVs.

8. The UAV cooperative detection and three-dimensional positioning system based on a dynamic tetrahedral microphone array according to claim 4, characterized in that: The wireless synchronization module uses the UWB protocol or PTP protocol to achieve time synchronization between each detection node. Each detection node collects acoustic signals with a unified time reference and the collected signals have a unified timestamp. The edge computing module completes the acquisition and preprocessing of acoustic signals at each detection node; the positioning module is a GPS receiver module or a Beidou receiver module; the spatial distance between the four microphones is a fixed value; the sensor data includes at least one of image data acquired by the camera, radar detection data, and infrared sensor data.

9. A method for UAV cooperative detection and 3D localization based on a dynamic tetrahedral microphone array, applied to a UAV cooperative detection and 3D localization system based on a dynamic tetrahedral microphone array, characterized in that, Includes the following steps: S1, Deploy multiple detection nodes, including UAV onboard microphone nodes, each detection node acquires its own spatial position and attitude and completes time synchronization; S2, each detection node reports its status information to the data processing unit, and the data processing unit establishes and maintains a global node status database. S3, the data processing unit selects four nodes from multiple detection nodes to construct a tetrahedral detection array based on the spatial distribution of each detection node and according to the evaluation index that characterizes the relationship between positioning error and spatial geometric configuration. S4, each detection node in the constructed tetrahedral array collects acoustic signals, which are acoustic signals generated by the rotor of the UAV; S5, perform time delay estimation on the acoustic signals collected by each detection node, and calculate the azimuth and elevation angles of the sound source relative to each detection node; S6. Based on the spatial positions of the four detection nodes and the azimuth and elevation angles measured by each node, the three-dimensional coordinates of the sound source are calculated through spatial geometric intersection. S7, acquire sensor data, and correct the calculation results based on the sensor data; S8 maps the corrected three-dimensional coordinates of the sound source to geospatial data and outputs the sound source location information; Steps S2 to S8 are executed repeatedly to continuously detect and perform three-dimensional positioning and tracking of the UAV target.

10. A method for UAV cooperative detection and three-dimensional localization based on a dynamic tetrahedral microphone array according to claim 9, characterized in that: The evaluation index mentioned in S3 includes at least one of geometric accuracy factor, positioning error ellipsoid volume, and root mean square error of sound source positioning. When four nodes are selected, the value of the evaluation index of any combination of four nodes is calculated, and the combination that makes the evaluation index optimal and the tetrahedral volume is greater than the threshold is selected. The data processing unit sends a position adjustment command to the UAV airborne node, and the UAV airborne node adjusts its own spatial position according to the command to optimize the reconstructed array configuration. S3 is re-executed when the displacement of any node in the selected nodes causes the geometric accuracy factor of the constructed array to deteriorate beyond the threshold or reaches the preset time interval; the time delay estimation in S5 adopts a generalized cross-correlation and phase transformation algorithm; the sensor data in S7 includes at least one of image data acquired by the camera, radar detection data, and infrared sensing data, and the correction is to correct the three-dimensional coordinates of the sound source by minimizing the reprojection error.

Images