Comprehensive photoelectric platform method and system for unmanned aerial vehicle detection, electronic equipment and storage medium

Abstract

The invention provides a comprehensive photoelectric platform method and system for unmanned aerial vehicle detection, electronic equipment and a storage medium, and relates to the technical field of unmanned aerial vehicle detection and localization. Synchronously triggering a photoelectric platform to capture a suspicious airspace high-definition video stream and establishing a space-time corresponding database of the two; synchronizing each radio frequency receiving node by taking a high-precision atomic clock as a time reference, and processing multiple paths of signal data to obtain a time-aligned basic signal; basic signal radio frequency features are extracted and associated with equipment identification, a mapping relation between the basic signal radio frequency features and remote controller models is established, and a hardware feature library is constructed; and finally, in combination with the library identification result, the signal time difference and the optical data, correcting a multipath error by using particle filtering, predicting and updating the position of an operator, and realizing comprehensive detection and positioning of the unmanned aerial vehicle after iterative convergence, thereby realizing comprehensive detection and positioning of the unmanned aerial vehicle in combination with radio frequency detection and optical observation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle detection and positioning, and particularly relates to a comprehensive optoelectronic platform method and system for unmanned aerial vehicle detection, an electronic device and a storage medium. BACKGROUND

[0002] In cities with high-rise buildings and complex electromagnetic environments, there are frequent illegal flights of unmanned aerial vehicles. It is necessary to solve the signal problems caused by non-line-of-sight propagation and building multipath reflection. At the same time, it is necessary to accurately identify unmanned aerial vehicle remote control signals to exclude interference, and to realize real-time tracing of the operator and positioning error meeting the requirements of the city, thereby providing support for airspace supervision and public safety.

[0003] At present, the mainstream scheme is to use a single radio frequency node monitoring system to collect unmanned aerial vehicle remote control signals through fixed radio frequency equipment, analyze the source direction according to the signal strength, identify the signal type in combination with a preset frequency band database, and estimate the approximate area of the operator by a single point azimuth angle to assist in completing detection and preliminary tracing.

[0004] However, this system has three defects: first, the anti-interference is weak, and the signal fluctuation caused by city multipath reflection is easy to misjudge the interference signal; second, the positioning accuracy is low, the single point azimuth angle cannot solve the non-line-of-sight deviation, the tracing error is large, and it is difficult to meet the requirements of accurate management; third, there is a lack of data verification, and the signal is not verified in combination with optical observation, which is easy to cause false positives and false negatives, and there is no complete detection closed loop. SUMMARY

[0005] The present application aims to provide a comprehensive optoelectronic platform method and system for unmanned aerial vehicle detection, an electronic device and a storage medium to solve the problems of signal misjudgment and low positioning accuracy in the prior art.

[0006] To solve the above technical problems, in a first aspect, the present application provides a comprehensive optoelectronic platform method for unmanned aerial vehicle detection, comprising: Collecting original signals of unmanned aerial vehicle remote control signals, generating multi-channel signal data with accurate time stamps based on the original signals, synchronously triggering an optoelectronic platform to capture high-definition video streams of suspicious airspace, and establishing a space-time correspondence database of the multi-channel signal data and the high-definition video streams; Configuring a high-precision atomic clock as a unified time reference, achieving time synchronization of each radio frequency receiving node through a precise time synchronization protocol, performing frequency conversion and signal demodulation processing on the signals based on the frequency characteristics and modulation characteristics in the multi-channel signal data, and obtaining time-aligned basic signals; Radio frequency (RF) features are extracted from the basic signal, and the RF features are associated with the device identification information in the multi-channel signal data. Through feature digitization processing and classification analysis, a mapping relationship between RF features and UAV remote controller model is established. Based on the mapping relationship, a hardware feature library containing feature templates, threshold parameters, and matching algorithms is constructed. Based on the identification results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, the multipath error is corrected by particle filtering technology and the operator's position at the next moment is predicted. The predicted position is then updated using the corrected multipath error compensation results. This process is iterated until convergence, thereby realizing the integrated detection and positioning of UAVs that combines radio frequency detection and optical observation.

[0007] Optionally, based on the identification results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, multipath errors are corrected using particle filtering technology to predict the operator's position at the next moment. The predicted position is then updated using the corrected multipath error compensation results. This process is iterated until convergence, achieving integrated UAV detection and positioning combining radio frequency detection and optical observation, including: The collected basic signals are input into the hardware feature library, and the identification result of the UAV remote controller model is output through the matching algorithm. The arrival time difference of the basic signals between different radio frequency receiving nodes is calculated, and the initial position estimation area is generated by combining the signal propagation speed. Optical observation data corresponding to the recognition result is extracted from the spatiotemporal correspondence database. The optical observation data includes the pixel coordinates of the UAV in the video frame. The initial position estimation area and the optical observation data are input into the position prediction unit to generate the operator's initial position prediction point. Based on the particle filter model, multipath propagation paths are simulated, and the theoretical signal propagation delay between the initial position prediction point and each radio frequency receiving node is calculated. The theoretical signal propagation delay is compared with the actual arrival time difference to generate a multipath error compensation amount. Based on the multipath error compensation amount, the initial position prediction point is corrected, and the updated prediction position is output. Repeat the error compensation and position update operations until the deviation of the predicted position is less than the convergence threshold for two consecutive times, and obtain the integrated detection and positioning result of the UAV combining radio frequency detection and optical observation.

[0008] Optionally, based on a particle filter model to simulate multipath propagation paths, the theoretical signal propagation delay between the initial position prediction point and each radio frequency receiving node is calculated, including: Based on the initial position prediction point and the position of the radio frequency receiving node, a set of particles representing different signal reflection paths is generated, and each particle contains a path length parameter and a reflection point position parameter. Calculate the matching degree between the path length parameter of each particle in the particle set and the initial position prediction point, and retain a preset number of particles with the highest matching degree to form a particle subset; The particles in the particle subset are randomly paired to form multiple particle pairs. The reflection point position parameters of each particle pair are exchanged. A new reflection path is generated based on the exchanged reflection point position parameters. The new particles corresponding to the new reflection path are combined into a new particle set. Randomly select some particles from the new particle set, randomly perturb the reflection point position parameters of the selected particles to form an optimized particle set, and calculate the path length of each propagation path from the initial position prediction point to each radio frequency receiving node based on the optimized particle set. Based on the propagation speed of electromagnetic waves, the path length of each propagation path is converted into a single-path propagation time. The single-path propagation times are then weighted and combined according to particle weights to generate the theoretical signal propagation delay.

[0009] Optionally, radio frequency (RF) features are extracted from the basic signal, and these RF features are associated with device identification information in the multi-channel signal data. Through feature digitization processing and classification analysis, a mapping relationship between the RF features and the drone remote controller model is established. Based on this mapping relationship, a hardware feature library containing feature templates, threshold parameters, and matching algorithms is constructed, including: The signal amplitude, frequency offset, and modulation period are separated from the basic signal as radio frequency features, which are stored in the form of numerical vectors. The preset device encoding field in the multi-channel signal data is associated with the numerical vector of the radio frequency feature to form a device-feature binding dataset; The bound dataset is subjected to feature digitization processing, the numerical vector of the radio frequency feature is quantized into a fixed-dimensional digital matrix, and the digital matrix is ​​clustered to generate feature clusters corresponding to different drone remote controller models. Establish a mapping relationship between the feature clusters and the drone remote controller models, and extract the center vector of each model corresponding to the feature cluster based on the mapping relationship to generate feature templates; Based on the mapping relationship and the cluster boundary distance, a threshold parameter is set, and a matching algorithm based on similarity calculation is designed. The feature template, threshold parameter and matching algorithm are integrated to build a dynamically updatable hardware feature library.

[0010] Optionally, a mapping relationship is established between the feature clusters and the drone remote controller models, and the center vector of each model corresponding to the feature cluster is extracted based on the mapping relationship to generate a feature template, including: Read the device encoding field from the bound dataset, and map each feature cluster to a unique drone remote controller model based on the model information contained in the device encoding field, thus forming a mapping relationship between model and feature cluster. Based on the mapping relationship, the feature clusters associated with each drone remote controller model are located, and the arithmetic mean of all numerical vectors within the feature clusters is calculated to obtain the center vector representing the features of each model. The center vector of each drone remote controller model is standardized, the numerical precision of all vector elements is adjusted to a uniform standard, and the standardized center vector is encapsulated into a data structure with a fixed format to generate a feature template with uniform dimensions and encoding format.

[0011] Optionally, a high-precision atomic clock is configured as a unified time reference, and a precise time synchronization protocol is used to synchronize the time of each radio frequency receiving node. Based on the frequency and modulation characteristics of the multi-channel signal data, frequency conversion and demodulation processing are performed on the signals to obtain a time-aligned base signal, including: A highly stable atomic clock is deployed as a unified time reference. The atomic clock outputs a standard clock signal to all radio frequency receiving nodes and distributes the standard clock signal through a wired transmission link, so that the deviation between the local clock of each radio frequency receiving node and the unified time reference is less than a microsecond threshold, thereby completing the time synchronization between nodes. Frequency characteristic parameters and modulation waveform characteristic parameters are separated from the multi-channel signal data, and the original signals of different bands are converted into a unified intermediate frequency signal using a frequency conversion unit; Demodulation is performed on the unified intermediate frequency signal to remove the carrier component and extract the baseband signal, resulting in a basic signal containing timing information; The basic signal is time-calibrated based on the time synchronization results between nodes to eliminate phase offset caused by transmission delay between nodes and generate a time-aligned basic signal.

[0012] Optionally, the raw signals of the UAV remote control signal are collected, and multi-channel signal data with precise timestamps are generated based on the raw signals. Simultaneously, an optoelectronic platform is triggered to capture a high-definition video stream of the suspected airspace, and a spatiotemporal correspondence database of the multi-channel signal data and the high-definition video stream is established, including: The system captures the raw signal emitted by the drone remote controller at multiple radio frequency receiving nodes. The raw signal contains frequency bands and modulation waveforms, and the initial time information of the raw signal is recorded by the signal acquisition unit. The initial time information is input into the time stamping unit, and a unique time identifier is added to each of the original signals in combination with the global clock reference to generate multi-channel signal data with precise timestamps. Based on the time identifier sequence in the multi-channel signal data, the photoelectric sensor is controlled to perform directional scanning of the suspicious airspace and capture a high-definition video stream containing UAV images; The time identifiers of the multi-channel signal data are matched and mapped with the frame timecodes of the high-definition video stream to form a spatiotemporal correspondence database containing signal feature fields and video frame coordinates.

[0013] Secondly, this application provides an integrated optoelectronic platform system for UAV detection, comprising: The acquisition module is used to acquire the raw signals of the UAV remote control signal, generate multi-channel signal data with precise timestamps based on the raw signals, synchronously trigger the photoelectric platform to capture high-definition video streams in suspicious airspace, and establish a spatiotemporal correspondence database of the multi-channel signal data and the high-definition video streams. The conversion module is used to configure a high-precision atomic clock as a unified time reference, and to achieve time synchronization of each radio frequency receiving node through a precision time synchronization protocol. Based on the frequency characteristics and modulation characteristics in the multi-channel signal data, the module performs frequency conversion and signal demodulation processing to obtain a time-aligned base signal. The module is used to extract radio frequency features from the basic signal, associate the radio frequency features with the device identification information in the multi-channel signal data, establish a mapping relationship between radio frequency features and UAV remote controller model through feature digitization processing and classification analysis, and construct a hardware feature library containing feature templates, threshold parameters and matching algorithms based on the mapping relationship. The update module is used to correct multipath errors and predict the operator's position at the next moment by using particle filtering technology, based on the recognition results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, and then using the corrected multipath error compensation results to update the predicted position. This process is iterated until convergence, thereby realizing the integrated detection and positioning of UAVs that combines radio frequency detection and optical observation.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor, configured to execute the computer program, implements the steps of an integrated optoelectronic platform method for unmanned aerial vehicle (UAV) detection as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the integrated optoelectronic platform method for UAV detection as described in the first aspect above.

[0016] This application provides a comprehensive optoelectronic platform method for UAV detection. By acquiring the raw signals of the UAV's remote control signals, it generates multi-channel signal data with precise timestamps, synchronously triggers the optoelectronic platform to capture high-definition video streams from suspicious airspace, and establishes a spatiotemporal correspondence database between the two. This enables spatiotemporal correlation between radio frequency signals and optical data, laying the foundation for subsequent multi-source data fusion analysis. By configuring a high-precision atomic clock as a unified time reference, the time of each radio frequency receiving node is synchronized. Processing the multi-channel signal data yields a time-aligned base signal, eliminating time deviations between multiple nodes and ensuring time consistency in signal analysis, thus laying the foundation for subsequent... Differential positioning provides reliable data; by extracting the radio frequency characteristics of basic signals and associating them with device identification information, a mapping relationship is established between them and the drone remote controller model, and a hardware feature library is constructed. This enables accurate identification and model matching of drone remote control signals, effectively distinguishes interference signals, and improves the accuracy of signal identification. By combining the identification results of the hardware feature library, multi-node signal time difference, and optical observation data, particle filtering technology is used to correct multipath errors and iteratively update the operator's position. This integrates the advantages of radio frequency detection and optical observation, effectively solving the positioning deviation caused by non-line-of-sight and multipath reflection, and achieving high-precision comprehensive detection and positioning of drones and operators. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a comprehensive optoelectronic platform method for UAV detection provided in this application embodiment; Figure 2 A flowchart illustrating a specific implementation of an integrated optoelectronic platform method for UAV detection provided in this application embodiment; Figure 3 A scene diagram illustrating a comprehensive optoelectronic platform method for unmanned aerial vehicle (UAV) detection provided in an embodiment of this application; Figure 4 This is a schematic diagram of a comprehensive optoelectronic platform system for UAV detection provided in an embodiment of this application. Detailed Implementation

[0019] In urban drone monitoring scenarios, existing monitoring solutions relying on a single radio frequency node face significant bottlenecks: on the one hand, multipath reflections from buildings in urban areas can cause signal instability, making it easy for the system to mistake other interference signals for drone remote control signals, resulting in inaccurate identification; on the other hand, relying solely on a single node to analyze signal direction to estimate the operator's position has a large error in non-line-of-sight environments, making it difficult to meet the precise positioning needs of urban management; at the same time, the system only looks at radio frequency signals without combining them with optical observation data, making it impossible to verify whether the signal truly corresponds to the drone, which can easily lead to false alarms or missed alarms, and prevents the formation of a complete monitoring process.

[0020] To address these issues, this application proposes a comprehensive optoelectronic platform method for UAV detection. This method first acquires UAV remote control signals and generates multi-channel data with precise time stamps, while simultaneously triggering the device to capture high-definition video of suspicious airspace, establishing a correlation database between the two. Next, a high-precision clock synchronizes the time of multiple signal receiving nodes, processing the signals to obtain basic data with a unified time reference. Then, signal features are extracted and correlated with device information to establish a feature library capable of accurately identifying remote control models. Finally, combining the feature library identification results, multi-node signal time differences, and video data, technical means are used to correct signal errors, repeatedly optimizing operator position estimation until accuracy is achieved. This solution not only solves the signal misjudgment problem through feature library and video data, but also improves positioning accuracy through multi-node synchronization and error correction, forming a complete "signal acquisition-identification-positioning-verification" process, fundamentally overcoming the shortcomings of existing single-radio frequency node systems.

[0021] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The core of this application is to provide a comprehensive optoelectronic platform method for UAV detection, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes: S101. Collect the raw signal of the UAV remote control signal, generate multi-channel signal data with precise timestamps based on the raw signal, synchronously trigger the photoelectric platform to capture high-definition video streams in suspicious airspace, and establish a spatiotemporal correspondence database of the multi-channel signal data and the high-definition video streams. Optionally, step S101 may specifically include the following steps: S1011. The original signal emitted by the drone remote controller is captured at multiple radio frequency receiving nodes. The original signal includes a frequency band and a modulation waveform. The initial time information of the original signal is recorded by the signal capture unit. S1012. Input the initial time information into the time stamping unit, and add a unique time identifier to each of the original signals in combination with the global clock reference to generate multi-channel signal data with precise timestamps. S1013. Based on the time identifier sequence in the multi-channel signal data, control the photoelectric sensor to perform directional scanning of the suspicious airspace and capture a high-definition video stream containing UAV images; S1014. Match and map the time identifiers of the multi-channel signal data with the frame timecodes of the high-definition video stream to form a spatiotemporal correspondence database containing signal feature fields and video frame coordinates.

[0023] In the above steps, the radio frequency receiving node is a device used to receive signals transmitted by the drone remote controller. It can be deployed in different locations to cover a wider monitoring range. The raw signal refers to the unprocessed electrical signal directly transmitted by the drone remote controller, which includes a frequency band and a modulated waveform. The frequency band is the frequency range of signal transmission, and the modulated waveform is the waveform form that carries the information in the signal. The signal acquisition unit is a module used to acquire the raw signal and record the initial moment of signal reception. The initial time information is the time data of the raw signal being received, recorded by the signal acquisition unit. The time stamping unit is a module that adds time stamps to the raw signal. The global clock reference is a reference source that provides a unified time standard for all monitoring devices, ensuring that the time of each device is consistent. The time identifier is a unique time code generated by the time stamp unit to identify the time information of each signal data stream; the multi-channel signal data is a collection of signals acquired by multiple radio frequency receiving nodes and for which time identifiers have been added; the photoelectric sensor is a device used to capture airspace images and videos, enabling directional scanning; the suspicious airspace is the airspace range where drone activity may be detected; the high-definition video stream is a high-definition continuous image sequence captured by the photoelectric sensor, and the frame time code is the time code corresponding to each image in the high-definition video stream; the spatiotemporal correspondence database is a database that associates multi-channel signal data with the high-definition video stream, where the signal feature field is information describing the signal attributes, and the video frame coordinates are the position information of the target within each frame of the high-definition video stream.

[0024] In this embodiment, step S1011 first captures the raw signals emitted by the drone remote controller at multiple radio frequency receiving nodes. These raw signals include the frequency band of the signal transmission and the modulation waveform carrying the information. Simultaneously, the signal acquisition unit records the initial time information when each raw signal is received. For example, in a drone monitoring scenario in area A, multiple radio frequency receiving nodes are deployed at different locations in the area. When the remote controller of a drone emits a signal, multiple radio frequency receiving nodes simultaneously receive this raw signal. The signal acquisition unit of each node records the initial time information of its own reception of the signal. For instance, one node receives the signal at 9:00:01:00 AM, and another node receives the same signal at 9:00:01:02 AM.

[0025] Secondly, in step S1012, the initial time information recorded by each radio frequency receiving node in step S1011 is input into the time stamping unit. The time stamping unit, in conjunction with a pre-set global clock reference, generates a unique time identifier for each raw signal, thereby integrating multiple sets of raw signals with time identifiers into multi-channel signal data with precise timestamps. For example, in the scenario of area A mentioned above, the initial time information recorded by each radio frequency receiving node is input into the time stamping unit. This unit uses the global clock reference of the monitoring system in area A as a reference to generate a time identifier T001 for the raw signal received at 9:00:01:00 AM and a time identifier T002 for the raw signal received at 9:00:01:02 AM. Then, the signal data with time identifiers such as T001 and T002 are organized into multi-channel signal data to ensure that each data channel has a precise timestamp.

[0026] Next, in step S1013, based on the time identifier sequence in the multi-channel signal data generated in step S1012, the system sends a control command to the photoelectric sensor, controlling the photoelectric sensor to perform a directional scan of suspicious airspace where drone activity may exist. During the scan, a high-definition video stream containing drone images is captured. For example, in the scenario of area A, the system reads the time identifier sequence in the multi-channel signal data and finds that the signals corresponding to time identifiers such as T001 and T002 all come from the airspace northeast of the area. This airspace is determined to be suspicious. Subsequently, the system controls the photoelectric sensor deployed in the area to turn northeast for a directional scan. During the scan, a high-definition video stream of the drone in the airspace is successfully captured, and the drone's flight status can be clearly seen in the video stream.

[0027] Finally, in step S1014, the time identifiers of the multi-channel signal data generated in step S1012 are matched and mapped one-to-one with the frame timecodes of the high-definition video stream captured in step S1013. The signal features corresponding to each time identifier are organized into signal feature fields, and the position information of the UAV in the video image of the corresponding time frame is recorded as video frame coordinates. Finally, a spatiotemporal correspondence database containing signal feature fields and video frame coordinates is formed. For example, in the scene of area A, the time identifier T001 is matched with the frame timecode of the video frame corresponding to 9:00:01:00 AM in the high-definition video stream. The signal frequency, modulation method, etc. corresponding to T001 are organized into signal feature fields, and the horizontal and vertical coordinates of the UAV in the image of the video frame are recorded as video frame coordinates. Similarly, other time identifiers such as T002 are matched and processed to finally construct a complete spatiotemporal correspondence database, realizing the temporal and spatial association between signal data and video data.

[0028] In practical applications, to strengthen drone surveillance in key areas such as core business districts and transportation hubs within its jurisdiction, the airspace management department of City A deployed a three-tiered drone detection system around these key areas: The first tier consisted of eight radio frequency receiving nodes deployed within a 1-kilometer radius of the key areas, evenly distributed on rooftops of high-rise buildings and lampposts to ensure no signal monitoring blind spots; the second tier deployed a high-precision time stamp unit on the rooftop of the regional center office building, connected to the city's government network's global clock reference, which is synchronized with the National Time Service Center and provides microsecond-level time accuracy; the third tier deployed one high-definition photoelectric sensor on each of the four diagonally opposite high-rise buildings in the key areas. These sensors have 360-degree rotation scanning capabilities, support 1080P resolution video capture, and can respond to millisecond-level control commands. The system backend detected suspected drone remote control signal fluctuations in the airspace east of the core business district and immediately triggered the S101 procedure: First, three of the eight radio frequency receiving nodes simultaneously captured the raw signal of the remote control. The signal contained a 2.4GHz frequency band and an FSK modulation waveform. Each node's signal acquisition unit recorded the initial time information through its built-in time module—9:12:03:150 ms for the east node, 9:12:03:152 ms for the north node, and 9:12:03:151 ms for the south node. Next, the initial time information from these three nodes was transmitted to the central time stamping unit via a dedicated data link. The time stamping unit, using a global clock reference, generated unique time identifiers T20240828091203150, T20240828091203152, and T20240828091203151 for the three raw signals, and associated the time identifiers with the corresponding signal frequency and modulation waveform. After data integration, three signal data streams with precise timestamps are generated. Subsequently, the system backend reads the time identifier sequence of these three data streams, analyzes it to determine that the signal source is roughly in the eastern airspace, and immediately sends directional scanning commands to two photoelectric sensors on the east and south sides. Within one second, the sensors adjust to a 30-degree angle to the east and begin capturing high-definition video streams at a rate of 25 frames per second. Within 10 seconds, an image of a small drone is captured in the video. Finally, the system matches the time identifiers of the three signal data streams with the frame timecodes of the video stream. For example, T20240828091203150 is matched with the 375th frame corresponding to 9:12:03:150 in the video stream. The pixel coordinates of the drone in that frame (X:850, Y:420) are extracted as the video frame coordinates. At the same time, the frequency, modulation mode, etc. of the signal are recorded as signal feature fields. Finally, a spatiotemporal correspondence database containing 20 sets of matching data is constructed to prepare data for subsequent signal identification and operator positioning.

[0029] In the overall scheme of step S101 above, by capturing the original signal and generating multi-channel signal data with precise timestamps, simultaneously acquiring high-definition video streams and establishing a spatiotemporal correspondence database, precise correlation between UAV remote control signals and optical image data in the time and space dimensions is achieved. This correlation not only provides a data foundation for subsequently distinguishing interference signals and improving signal recognition accuracy, but also creates conditions for integrating radio frequency detection and optical observation results and improving operator positioning accuracy. Simultaneously, it makes the data of the entire detection process traceable, helping to form a complete detection process and providing reliable preliminary data support for UAV supervision in densely populated urban environments.

[0030] S102. Configure a high-precision atomic clock as a unified time reference, and use a precision time synchronization protocol to make each radio frequency receiving node time synchronized. Based on the frequency characteristics and modulation characteristics in the multi-channel signal data, perform frequency conversion and signal demodulation processing on the signal to obtain a time-aligned basic signal. Optionally, step S102 may specifically include the following steps: S1021. Deploy a highly stable atomic clock as a unified time reference. The atomic clock outputs a standard clock signal to all radio frequency receiving nodes and distributes the standard clock signal through a wired transmission link so that the deviation between the local clock of each radio frequency receiving node and the unified time reference is less than a microsecond threshold, thereby completing the time synchronization between nodes. S1022. Separate frequency characteristic parameters and modulation waveform characteristic parameters from the multi-channel signal data, and use a frequency conversion unit to convert the original signals of different bands into a unified intermediate frequency signal; S1023. Perform demodulation operation on the unified intermediate frequency signal, strip the carrier component and extract the baseband signal to obtain a basic signal containing timing information; S1024. Based on the time synchronization results between nodes, perform timing calibration on the basic signal to eliminate phase offset caused by transmission delay between nodes and generate a time-aligned basic signal.

[0031] In the above steps, a high-precision atomic clock is a clock device with high-stability time output capability, used to provide a unified and accurate time reference standard; a unified time base is a time standard followed by all RF receiving nodes, ensuring time consistency across nodes; a precise time synchronization protocol is the communication rule for aligning the time of each node with the unified time base; an RF receiving node is a device that receives drone remote control signals, and a local clock is a time recording module carried by the RF receiving node itself; a microsecond threshold is the maximum allowable deviation between the node's local clock and the unified time base, and a deviation less than this value is considered time synchronization; a wired transmission link is the physical line used to transmit the atomic clock standard clock signal, ensuring stable signal transmission; multi-channel signal data is a collection of signals with precise timestamps, frequency characteristic parameters are data describing the signal frequency attributes, and modulation waveform characteristic parameters are data describing the signal modulation method. The data consists of: a frequency conversion unit (FCU) and a time-aligned signal. The FCU is a module that converts signals from different frequency bands into fixed-frequency signals. The raw signal is the unprocessed signal emitted by the drone's remote controller. The unified intermediate frequency (IF) signal is a fixed-frequency signal obtained after converting raw signals from different bands, facilitating subsequent unified processing. Demodulation is the process of extracting raw information from modulated signals. The carrier component is the high-frequency signal carrying signal information. The baseband signal is the low-frequency raw information signal obtained after demodulation, containing the signal's timing information. The basic signal is the demodulated signal with timing information. Timing calibration is the operation of adjusting the signal's time order based on the time synchronization result. Transmission delay is the time delay generated during signal transmission between nodes. Phase offset is the signal phase inconsistency caused by transmission delay. The time-aligned basic signal is the basic signal that ensures the signals of each node remain consistent in the time dimension after timing calibration.

[0032] In this embodiment, firstly, in step S1021, a highly stable atomic clock is deployed as a unified time reference. This atomic clock continuously outputs a standard clock signal, which is distributed to all radio frequency receiving nodes via a wired transmission link. After receiving the signal, each node automatically calibrates its local clock, ensuring that the deviation between its local clock and the unified time reference is controlled within a microsecond threshold, thereby achieving time synchronization between nodes. For example, in a drone detection system in a key area of ​​City A, a rubidium atomic clock is deployed as a unified time reference. The clock's time stability can achieve a daily deviation of less than 1 microsecond. The standard clock signal is sent to 8 radio frequency receiving nodes via a shielded network cable (wired transmission link). Each node has a built-in time calibration module. After receiving the signal, it compares the difference between its local clock and the standard clock. If a node's local clock is 2 microseconds slower than the standard clock, the calibration module automatically advances its local clock by 2 microseconds. Ultimately, the deviation between all nodes and the unified time reference is controlled within 0.5 microseconds, achieving time synchronization.

[0033] Secondly, in step S1022, the frequency characteristic parameters and modulation waveform characteristic parameters of each signal are separated from the multi-channel signal data generated in S101 using a signal feature extraction algorithm. These parameters are then input into the frequency conversion unit. The frequency conversion unit converts the original signals of different bands into a unified intermediate frequency signal using a mixer based on a preset intermediate frequency value. For example, in the above scenario, the multi-channel signal data contains three original signals of different bands, namely 2.4GHz, 5.8GHz, and 915MHz. The frequency value (frequency characteristic parameter) and modulation type (modulation waveform characteristic parameter) of each signal are separated using a feature extraction algorithm. Subsequently, the frequency conversion unit sets the preset intermediate frequency value to 70MHz and mixes the 2.4GHz signal with the 2.33GHz local oscillator signal, the 5.8GHz signal with the 5.73GHz local oscillator signal, and the 915MHz signal with the 845MHz local oscillator signal using a mixer. Finally, the original signals of the three different bands are all converted into a unified intermediate frequency signal of 70MHz.

[0034] Next, in step S1023, the unified intermediate frequency (IF) signal obtained in step S1022 is demodulated. A coherent demodulation algorithm is used to remove the carrier component from the signal, extracting the baseband signal containing information such as UAV control commands. This baseband signal carries timing information of the signal changing over time, thus obtaining the basic signal containing timing information. For example, in the above scenario, for the 70MHz unified IF signal, a coherent demodulation algorithm is used. First, the carrier extraction module recovers the carrier signal with the same frequency and phase as the transmitter from the IF signal. Then, the IF signal is multiplied by the recovered carrier signal. After filtering the high-frequency components through a low-pass filter, the 70MHz carrier component is removed, and finally, the baseband signal with a frequency of 0-10MHz is extracted. This baseband signal contains the timing information of control commands such as "ascend" and "turn" sent by the UAV remote controller, forming the basic signal.

[0035] Finally, in step S1024, based on the inter-node time synchronization results obtained in step S1021, the timing of the basic signal obtained in step S1023 is adjusted using a timing calibration algorithm. The time difference caused by transmission delay in the basic signals of each node is calculated, and the phase of the signal is compensated according to the time difference to eliminate the phase shift caused by the transmission delay between nodes, ultimately generating a time-aligned basic signal. For example, in the above scenario, it is known that eight RF receiving nodes have achieved time synchronization. The timing calibration algorithm calculates that the basic signal of the north node arrives 0.3 microseconds later than the east node. The algorithm automatically advances the basic signal of the north node by 0.3 microseconds on the time axis and adjusts the signal phase to match the signal of the east node. After calibration, the start time and signal period of the basic signals of all nodes are consistent on the time axis, and the phase shift is completely eliminated, resulting in a time-aligned basic signal.

[0036] In practical applications, to improve the signal processing accuracy of drone detection in key areas, the airspace management department of City A deployed a complete S102 processing flow in its core business district monitoring system: First, an A-brand rubidium atomic clock was deployed in the system control center as a unified time reference. The standard clock signal was distributed to eight radio frequency receiving nodes around the business district via Category 6 shielded network cables. Each node synchronized with the reference clock using built-in time calibration software, ensuring a deviation of less than 0.5 microseconds. When the system captured three raw drone remote control signals (2.4GHz, 5.8GHz, and 915MHz), First, the frequency and modulation characteristics of each signal are separated by signal processing software, and then they are uniformly converted into 70MHz intermediate frequency signals by a hardware frequency conversion module. Subsequently, the intermediate frequency signals are processed by a coherent demodulation module, and the baseband signal containing control commands is obtained after carrier stripping. Finally, based on the node time synchronization data, the time and phase of the baseband signals of each node are adjusted by a timing calibration tool to eliminate the transmission delay deviation caused by the difference in node position. Finally, three time-aligned basic signals are output, providing a unified and accurate signal data source for subsequent RF feature extraction and operator positioning.

[0037] In the overall scheme of step S102 above, time synchronization of each RF receiving node is achieved by deploying a high-precision atomic clock and a precise synchronization protocol, ensuring the time consistency of signal processing; frequency conversion unifies signals from different bands into intermediate frequency signals, reducing the complexity of subsequent signal processing; demodulation extracts the baseband signal containing core information, providing effective data for feature analysis; and timing calibration eliminates phase shifts caused by transmission delays, obtaining a time-aligned base signal. The entire process solves the problems of asynchronous signal time at different nodes, large band differences, and numerous signal clutters, laying a reliable signal foundation for subsequent hardware feature library construction and high-precision positioning. It also improves the efficiency and accuracy of signal processing, making subsequent UAV detection processes smoother and the results more reliable.

[0038] S103. Extract the radio frequency features from the basic signal, associate the radio frequency features with the device identification information in the multi-channel signal data, establish a mapping relationship between the radio frequency features and the drone remote controller model through feature digitization processing and classification analysis, and construct a hardware feature library containing feature templates, threshold parameters and matching algorithms based on the mapping relationship. Optionally, step S103 may specifically include the following steps: S1031. Separate the signal amplitude, frequency offset and modulation period from the basic signal as radio frequency features, and store the radio frequency features in the form of a numerical vector. S1032. Associate the preset device encoding field in the multi-channel signal data with the numerical vector of the radio frequency feature to form a binding dataset of device and feature; S1033. Perform feature digitization processing on the bound dataset, quantize the numerical vector of the radio frequency feature into a fixed-dimensional digital matrix, and perform clustering on the digital matrix to generate feature clusters corresponding to different drone remote controller models. S1034. Establish the mapping relationship between the feature clusters and the UAV remote controller models, and extract the center vector of the feature cluster corresponding to each model based on the mapping relationship to generate feature templates; Specifically, step S1034 may include the following processes: reading the device encoding field from the binding dataset; mapping each feature cluster to a unique UAV remote controller model based on the model information contained in the device encoding field, forming a mapping relationship between model and feature cluster; locating the feature cluster associated with each UAV remote controller model based on the mapping relationship; calculating the arithmetic mean of all numerical vectors within the feature cluster to obtain the center vector representing the feature of each model; standardizing the center vector of each UAV remote controller model, adjusting the numerical precision of all vector elements to a uniform standard, and encapsulating the standardized center vector into a data structure with a fixed format to generate a feature template with a uniform dimension and encoding format.

[0039] S1035. Based on the mapping relationship and the cluster boundary distance, set the threshold parameter, design a matching algorithm based on similarity calculation, integrate the feature template, threshold parameter and matching algorithm, and construct a dynamically updatable hardware feature library.

[0040] In the above steps, radio frequency (RF) characteristics are key parameters describing the RF attributes of the basic signal, including signal amplitude, frequency offset, and modulation period. Signal amplitude refers to the intensity of signal vibration, frequency offset refers to the deviation between the actual signal frequency and the reference frequency, and modulation period refers to the time it takes for the signal to complete one modulation change. The numerical vector is a data structure that stores the specific values ​​of the RF characteristics in an ordered array format, used to quantify the RF characteristics. Device identification information is pre-defined information in the multi-channel signal data used to distinguish different drone remote controllers, where the device encoding field is a string or numeric code that uniquely identifies the remote controller. The binding dataset is a data set formed by associating the device encoding field with the corresponding RF characteristic numerical vector, achieving a one-to-one correspondence between devices and characteristics. Feature digitization is the process of converting the RF characteristic numerical vector into a fixed-dimensional digital matrix. The digital matrix is ​​a set of values ​​arranged by rows or columns, with dimensions determined by the RF characteristics. The number of features determines the type of remote controller. Clustering is the operation of grouping similar numerical matrices into the same group using a clustering algorithm. Feature clusters are sets of similar numerical matrices formed after clustering, representing a group of features for a certain type of remote controller. The mapping relationship is the correspondence between feature clusters and drone remote controller models. The center vector is the arithmetic mean of all numerical vectors within a feature cluster, used to represent the core feature of that cluster. The feature template is a fixed-format data structure formed after standardizing the center vector, with unified dimensions and encoding format. The threshold parameter is a judgment standard set based on the boundary distance between feature clusters, used to distinguish the feature differences between different models. The matching algorithm is an algorithm based on similarity calculation, used to determine the similarity between the signal features to be identified and the feature template. The hardware feature library is a database that integrates feature templates, threshold parameters, and matching algorithms, supporting dynamic updates to adapt to new remote controller models.

[0041] In this embodiment, firstly, through step S1031, three core radio frequency features—signal amplitude, frequency offset, and modulation period—are separated from the time-aligned base signal. Then, the specific values ​​of these three features are arranged in the order of "signal amplitude - frequency offset - modulation period" to form a numerical vector and stored. For example, in a drone surveillance scenario in City A, for a previously captured base signal, the signal amplitude is separated to be 1.2V, the frequency offset to be 50kHz, and the modulation period to be 2ms using signal analysis tools. Subsequently, these three values ​​are combined in sequence to generate a numerical vector of radio frequency features [1.2, 50, 2], and the vector is saved in document form.

[0042] Secondly, in step S1032, a preset device code field is extracted from the multi-channel signal data. Each device code field is associated with the corresponding radio frequency feature value vector generated in step S1031 to ensure that each device code is bound to a unique value vector, thus forming a binding dataset of devices and features. For example, in the scenario of city A, the device code field corresponding to the basic signal is read as "D001" from the multi-channel signal data. "D001" is associated with the value vector [1.2,50,2] and recorded as "D001-[1.2,50,2]". At the same time, the same operation is performed on other basic signals, such as associating the device code "D002" with the value vector [1.16,49.6,2.04]. Finally, all association results are integrated to form a binding dataset containing dozens of sets of data.

[0043] Next, in step S1033, feature digitization processing is performed on all radio frequency feature numerical vectors in the bound dataset, converting each numerical vector into a fixed-dimensional numerical matrix with the same dimension as the number of radio frequency features (here, 3 rows and 1 column). Then, the K-means clustering algorithm is used to cluster all numerical matrices, grouping those with similar features into the same feature cluster. For example, in the scenario of city A, the numerical vectors... Convert to of A numerical matrix is ​​a vector of numbers. [2.04] converted to of The numerical matrices were analyzed. The Euclidean distance between the matrices was calculated using the K-means algorithm. It was found that the numerical matrices corresponding to 12 device codes, such as "D001" and "D002", had extremely high similarity. These matrices were classified into the feature cluster C01, while the matrices corresponding to other device codes were classified into different clusters such as C02 and C03.

[0044] Next, in step S1034, the model information contained in the device encoding field of each device is first read from the binding dataset. Based on the model information, each feature cluster is mapped to a unique drone remote controller model, forming a mapping relationship between model and cluster. Then, the cluster associated with each model is located, and the arithmetic mean of all numerical vectors within the cluster is calculated to obtain the center vector. Finally, the center vector is standardized to generate a feature template with a fixed format. For example, in the scenario of city A, the model information corresponding to device codes such as "D001" and "D002" is "M01". Therefore, the feature cluster C01 is mapped to the model "M01". The arithmetic mean of the 12 numerical vectors within the C01 cluster is calculated, where the average signal amplitude is (1.2 + 1.16 + ... + 1.19) / 12 = 1.18V, and the average frequency offset is (50 + 49.6 + ... + 50.2) / 12 = 49.8kHz. The average modulation period is (2+2.04+…+2.01) / 12=2.02ms, resulting in the center vector [1.18,49.8,2.02]. The numerical precision of this vector is adjusted to two decimal places and encapsulated in a fixed format of "Model M01-Feature Template [1.18,49.8,2.02]" to complete the generation of the "M01" model feature template. The same operation is performed on other models such as "M02" and "M03" to generate the corresponding feature templates.

[0045] Finally, in step S1035, based on the mapping relationship formed in step S1034, the boundary distance between different feature clusters is calculated. For example, the boundary distance between C01 and C02 is 0.5. A threshold parameter is set according to the boundary distance, such as setting the threshold to 0.4, to ensure that similar features can be accurately matched. A matching algorithm based on cosine similarity is designed to calculate the similarity between the features of the signal to be identified and the feature template. All feature templates, threshold parameters, and matching algorithms are integrated to build a dynamically updatable hardware feature library. If a new remote control model appears later, steps S1031-S1034 can be repeated to add new feature templates. For example, in the scenario of city A, after calculating the boundary distance between C01 and C02, the threshold is set to 0.4. When the cosine similarity between the feature vector of the signal to be identified and the template "M01" is ≥0.4, it is determined to be model "M01". The feature templates of all models, the threshold of 0.4, and the cosine similarity algorithm are integrated into the hardware storage module to form a hardware feature library for drone supervision in city A, and a template addition interface is reserved to support dynamic updates.

[0046] In practical applications, to improve the accuracy of drone remote control signal recognition, the airspace management department of City A collected data on 15 mainstream drone remote control models on the market, denoted as M01 to M15. For each model, 20 different devices were selected for signal acquisition. Step S1031 extracted radio frequency features from the basic signal of each device and generated numerical vectors. Step S1032 then associated the device codes with the vectors to form a bound dataset. Subsequently, step S1033 digitized and clustered the datasets, resulting in 15 feature clusters corresponding to M01 to M15. Step S1034 calculated the center vector of each cluster and generated standardized feature templates. Step S1035 set a threshold parameter of 0.38 and designed a cosine similarity matching algorithm. Finally, these were integrated to form a hardware feature library. When new remote control signals are subsequently detected, their radio frequency features are extracted and matched against the templates in the library to quickly identify the remote control model. If new, unmatched features are found, additional data can be collected and the feature library updated to ensure continuous improvement in recognition coverage.

[0047] In the overall scheme of step S103 above, by extracting radio frequency features and associating them with device identifiers, the binding of signal features with specific remote controllers is achieved; through cluster analysis and mapping relationship establishment, radio frequency features can accurately correspond to remote controller models; and by constructing a hardware feature library containing feature templates, threshold parameters, and matching algorithms, a standardized tool is provided for subsequent signal identification. The entire process not only effectively distinguishes remote controller signals from different drone models and eliminates interference signals such as those from civilian communications, but also adapts to new remote controller models through dynamic updates, laying the foundation for accurate signal identification in drone detection and reducing the risk of errors caused by signal misjudgment during subsequent positioning.

[0048] This application provides a flowchart illustrating a specific implementation of a comprehensive optoelectronic platform method for UAV detection, as shown in the embodiments below. Figure 2 As shown, it includes the following: S104. Based on the identification results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, the multipath error is corrected by particle filtering technology and the operator's position at the next moment is predicted. The predicted position is then updated using the corrected multipath error compensation results. This process is iterated until convergence is achieved, thus realizing the integrated detection and positioning of the UAV that combines radio frequency detection and optical observation.

[0049] Optionally, step S104 may specifically include the following steps: S1041. Input the collected basic signal into the hardware feature library, output the identification result of the UAV remote controller model through the matching algorithm, calculate the arrival time difference of the basic signal between different radio frequency receiving nodes, and generate the initial position estimation area by combining the signal propagation speed. S1042. Extract optical observation data corresponding to the recognition result from the spatiotemporal correspondence database. The optical observation data includes the pixel coordinates of the UAV in the video frame. Input the initial position estimation area and the optical observation data into the position prediction unit to generate the operator's initial position prediction point. S1043. Simulate multipath propagation paths based on particle filter model, and calculate the theoretical signal propagation delay between the initial position prediction point and each radio frequency receiving node; Specifically, step S1043 may include the following processes: Based on the initial position prediction point and the position of the radio frequency receiving node, a particle set representing different signal reflection paths is generated, each particle containing a path length parameter and a reflection point position parameter; the matching degree between the path length parameter of each particle in the particle set and the initial position prediction point is calculated, and a preset number of particles with the highest matching degree are retained to form a particle subset; the particles in the particle subset are randomly paired to form multiple particle pairs, and the reflection point position parameters of each particle pair are exchanged, and a new reflection path is generated based on the exchanged reflection point position parameters, and the new particles corresponding to the new reflection path are combined into a new particle set; some particles are randomly selected from the new particle set, and the reflection point position parameters of the selected particles are randomly perturbed to form an optimized particle set; based on the optimized particle set, the path length of each propagation path from the initial position prediction point to each radio frequency receiving node is calculated; according to the electromagnetic wave propagation speed, the path length of each propagation path is converted into a single path propagation time, and the single path propagation time is weighted and merged according to the particle weight to generate the theoretical signal propagation delay.

[0050] S1044. Compare the theoretical signal propagation delay with the actual arrival time difference to generate a multipath error compensation amount, correct the initial position prediction point based on the multipath error compensation amount, and output the updated prediction position. S1045. Repeat the error compensation and position update operation until the deviation of the predicted position is less than the convergence threshold for two consecutive times, and obtain the UAV integrated detection and positioning result combining radio frequency detection and optical observation.

[0051] In the above steps, the hardware feature library identification result refers to the drone remote controller model information obtained by analyzing the basic signal after inputting it into the hardware feature library through the matching algorithm within the library; signal time difference is the arrival time difference between different radio frequency receiving nodes receiving the same basic signal; optical observation data is drone-related data extracted from the spatiotemporal correspondence database and matched with the identification result time, including the pixel coordinates of the drone in the video frame, which can be mapped to the actual spatial position reference; particle filtering technology is an algorithm that simulates a set of particles to characterize the signal propagation path and achieve error correction and position prediction; multipath error is caused by the signal being reflected by objects such as buildings, forming multiple propagation paths, leading to actual... The deviation between reception time and linear propagation time; the convergence threshold is the standard for judging whether the position prediction result is stable. When the deviation between two consecutive predicted positions is less than the threshold, the positioning result is considered reliable; the initial position estimation area is the approximate spatial range where the operator may exist, calculated based on the signal time difference and signal propagation speed; the initial position prediction point is the operator's position point preliminarily determined by combining the initial position estimation area and optical observation data; the theoretical signal propagation delay is the propagation time of the signal from the predicted position to each radio frequency receiving node, simulated based on the particle filter model; the multipath error compensation amount is a value used to correct the position deviation by comparing the theoretical delay with the actual signal time difference.

[0052] In this embodiment, firstly, through step S1041, the previously processed basic signal is input into the hardware feature library. The matching algorithm in the library compares the radio frequency characteristics of the basic signal with the feature template and outputs the corresponding drone remote controller model identification result. Simultaneously, the arrival time difference of the basic signal received by different radio frequency receiving nodes is calculated. Combined with the electromagnetic wave propagation speed, the relationship "distance = speed × time difference" is used to outline the possible initial position estimation area of ​​the operator. For example, in a drone monitoring scenario in the core business district of city A, the basic signals of three radio frequency receiving nodes on the east, north, and south sides are input into the hardware feature library, and the signal is matched to the B model drone remote controller. It is calculated that the north node receives the signal 0.002 seconds later than the east node, and the south node receives the signal 0.0015 seconds later than the east node. Combined with the electromagnetic wave 3×10 8 With a propagation speed of meters per second, the signal propagation distance difference between the north and east nodes was calculated to be 600 meters, and the distance difference between the south and east nodes was calculated to be 450 meters. Thus, a fan-shaped initial position estimation area with the three nodes as vertices and meeting the distance difference condition was delineated.

[0053] Secondly, through step S1042, optical observation data consistent with the time of the B-type remote controller's recognition result is extracted from the spatiotemporal correspondence database. The pixel coordinates of the UAV in the high-definition video frame within this time period are found. Then, based on the deployment location of the photoelectric sensor and lens parameters, the pixel coordinates are converted into the UAV's position in the actual airspace. Subsequently, the initial position estimation area is spatially correlated with the actual UAV position—due to the signal transmission distance constraint between the UAV and the operator, the estimation range is narrowed by using the UAV position as a reference, ultimately generating the operator's initial position prediction point. For example, in the above scenario, a video frame matching the signal time is extracted from the spatiotemporal correspondence database. The UAV's pixel coordinates in the frame are (X:850, Y:420). Combining the parameters of the photoelectric sensor being deployed on a 100-meter-high building and the lens focal length being 100 millimeters, the actual UAV position is calculated to be 1.2 kilometers east of the business district. Considering the characteristic that the UAV remote controller is usually controllable within 1 kilometer, the portion of the initial position estimation area exceeding the UAV's 1-kilometer range is removed, determining the operator's initial position prediction point as a coordinate point near an office building on the east side of the business district.

[0054] Next, through step S1043, a simulation of the signal propagation path from the initial position prediction point to each radio frequency receiving node is constructed based on the particle filter model: First, a set of 1000 particles is generated, each particle representing a possible propagation path, including two parameters: path length (total distance of signal propagation) and reflection point position (coordinates of the signal reflected by the building); the matching degree of the path length of each particle with the straight-line distance from the initial position prediction point to the corresponding node is calculated, and the top 300 particles with the highest matching degree are retained to form a particle subset; the particles in the particle subset are randomly paired in pairs, and the reflection point position parameters of each pair of particles are exchanged to generate 300 new particles to form a new particle set; 100 particles are randomly selected from the new particle set, and their reflection point position parameters are slightly adjusted to form an optimized particle set; finally, based on the optimized particle set, the actual path length from the initial position prediction point to the three radio frequency receiving nodes is calculated for each path, and then the path length is converted into a single path propagation time according to the electromagnetic wave propagation speed. Weights are assigned according to the matching degree of each particle, and the single path propagation times are weighted and merged to obtain the theoretical signal propagation delay. For example, in the above scenario, the straight-line distance from the initial position prediction point to the east node is 800 meters. Among the simulated particles, the path of a certain particle is 850 meters long after being reflected by a nearby tall building. It has a high matching degree and is retained. After pairing and exchanging reflection points and performing minor perturbations, the optimized particle path length is concentrated between 840 and 860 meters. After weighted calculation, the theoretical signal propagation delay from the prediction point to the east node is 0.0028 milliseconds, and the theoretical delays to the north and south nodes are 0.0032 milliseconds and 0.0030 milliseconds, respectively.

[0055] Next, in step S1044, the theoretical signal propagation delay obtained in step S1043 is compared with the actual arrival time difference calculated in step S1041 to find the difference between the two—this difference is the deviation caused by multipath error. Based on the magnitude and direction of the deviation, the multipath error compensation amount is calculated, and this compensation amount is used to adjust the coordinates of the initial position prediction point. For example, if the theoretical delay is larger than the actual time difference, it indicates that the simulated path is too long, and the prediction point needs to be adjusted towards the receiving node. Finally, the updated prediction position is output. For example, in the above scenario, after comparing the theoretical delay with the actual arrival time difference, it is found that the theoretical delay of the east node is 0.0003 milliseconds larger than the actual time difference, and the north node is 0.0002 milliseconds larger. Based on this, the multipath error compensation amount is calculated to move the initial position prediction point 90 meters towards the east node and 60 meters towards the north node. After adjustment, the new prediction position coordinates are obtained.

[0056] Finally, through step S1045, the theoretical delay calculation in step S1043 and the error compensation and position update operations in step S1044 are repeated: after each update, the deviation between the new predicted position and the previous predicted position is calculated. If the deviation is still greater than the convergence threshold, the particle path is simulated again, the delay is calculated, and the position is corrected; until the deviation between two consecutive predicted positions is less than the convergence threshold, the iteration stops. The position obtained at this time is the integrated detection and positioning result of the UAV that integrates radio frequency detection and optical observation results. For example, in the above scenario, the deviation between the predicted position after the first update and the initial predicted point is 120 meters, which is greater than the convergence threshold of 50 meters, so it needs to be iterated again; after the second iteration, the deviation between the new predicted position and the previous position is 35 meters, which is less than the convergence threshold, so the calculation stops, and it is determined that the operator is located on the 15th floor of an office building on the east side of the business district, near the window, thus completing the integrated positioning.

[0057] In practical application, the drone detection system deployed by the airspace management department of City A in the core business district initiated the S104 procedure for a detected illegal drone signal: First, the basic signals of the three radio frequency receiving nodes were input into the hardware feature database to match the signal to a certain model of drone remote controller of Brand C. Simultaneously, the signal time difference between the nodes was calculated, and the initial position estimation area was determined by combining the electromagnetic wave velocity. Next, optical data from the same time period was extracted from the spatiotemporal correspondence database, revealing that the drone was located above the southeast plaza of the business district in the video frame. After converting the actual position, the portion of the drone outside the initial area that exceeded the drone's 1-kilometer control range was identified. After elimination, the initial predicted location was determined to be near a shopping mall on the north side of the square. Then, a particle filter model was activated to generate 1,000 particle simulated signal propagation paths. After screening, pairing, and perturbation optimization, the theoretical signal propagation delay was calculated. After comparing it with the actual time difference, a compensation amount was generated, and the initial predicted point was adjusted 110 meters northeast. After repeating the iteration three times, the deviation between two consecutive predicted positions was reduced to 40 meters, which is less than the convergence threshold of 50 meters. Finally, the operator was located in a shop on the 7th floor of the shopping mall. Based on the positioning result, the staff quickly arrived at the scene and stopped the illegal drone operation in time, verifying the practicality of the process.

[0058] In the overall solution of step S104 above, by integrating the accurate identification results of the hardware feature library, the time difference of multi-node signals, and optical observation data, the multipath error in dense urban environments is effectively corrected by using particle filtering technology, solving the problem that single radio frequency positioning is difficult to cope with non-line-of-sight and multi-reflection scenarios; through iterative iteration until the position prediction converges, the accuracy of operator tracing is greatly improved, meeting the needs of urban management for meter-level positioning; at the same time, the signal analysis of radio frequency detection is combined with the spatial reference of optical observation to form a complete positioning closed loop of "identification-prediction-correction-convergence", avoiding the misjudgment and omission problems that are prone to occur by single technical means, and providing reliable technical support for the illegal management of drones in dense urban environments.

[0059] The following is a complete embodiment for steps S101 to S104: like Figure 3As shown, City A deploys a comprehensive optoelectronic platform for drone detection in its core business district. Initially, five radio frequency (RF) receiving nodes and two high-definition optoelectronic sensors are deployed around the business district. When a drone flies illegally, each RF node captures its original remote control signal, generating multi-channel signal data with microsecond-level timestamps. Simultaneously, the optoelectronic sensors are triggered to capture high-definition video streams of the suspected airspace. The system matches the signal timestamps with the video frame timestamps to establish a spatiotemporal correspondence database. Subsequently, the platform uses a high-precision atomic clock as a reference and achieves time synchronization among the five RF nodes through a precise protocol. Based on the frequency and modulation characteristics of the signal, frequency conversion and demodulation are performed to obtain a time-aligned base signal. Next, RF features such as amplitude and frequency offset are extracted from the base signal, and device identification information in the signal is associated. After digital processing and classification analysis, a mapping relationship between these features and remote controllers of drone models B and C is established, constructing a hardware feature library containing feature templates and matching algorithms. Finally, based on the feature library recognition results, it was determined that the drone was a B-type remote controller, and the drone's position data in the video was based on the multi-node signal time difference and the drone's position data. The error caused by the multipath reflection of the building was corrected by particle filtering technology. The operator's position was iteratively updated repeatedly until the prediction deviation met the convergence condition twice in a row, and finally the accurate positioning of the drone and the operator was achieved.

[0060] The integrated optoelectronic platform method for UAV detection provided in this application achieves comprehensive optimization of UAV detection through multi-source data fusion and precise processing. The spatiotemporal correlation of signals and video provides a foundation for data verification, high-precision time synchronization ensures the consistency of signal analysis, and the hardware feature library improves the accuracy of remote control signal recognition, effectively distinguishing interference signals. Combining particle filtering error correction and iterative optimization solves the positioning deviation problem caused by non-line-of-sight and multipath reflections in complex urban environments. Integrating the advantages of radio frequency detection and optical observation forms a complete detection closed loop, significantly improving the reliability and accuracy of UAV and operator positioning, and providing strong technical support for urban airspace supervision.

[0061] Figure 4 This is a schematic diagram illustrating a specific embodiment of an integrated optoelectronic platform system for UAV detection provided in this application. (Refer to...) Figure 4 The system may include: The acquisition module 41 is used to acquire the raw signal of the UAV remote control signal, generate multi-channel signal data with precise timestamps based on the raw signal, synchronously trigger the photoelectric platform to capture high-definition video streams in suspicious airspace, and establish a spatiotemporal correspondence database of the multi-channel signal data and the high-definition video streams. The conversion module 42 is used to configure a high-precision atomic clock as a unified time reference, and to achieve time synchronization of each radio frequency receiving node through a precision time synchronization protocol. Based on the frequency characteristics and modulation characteristics in the multi-channel signal data, the signal is frequency converted and demodulated to obtain a time-aligned basic signal. The construction module 43 is used to extract radio frequency features from the basic signal, associate the radio frequency features with the device identification information in the multi-channel signal data, establish a mapping relationship between radio frequency features and UAV remote controller model through feature digitization processing and classification analysis, and construct a hardware feature library containing feature templates, threshold parameters and matching algorithms based on the mapping relationship. The update module 44 is used to correct multipath errors and predict the operator's position at the next moment by using particle filtering technology, based on the recognition results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, and then using the corrected multipath error compensation results to update the predicted position. This process is iterated until convergence, thereby realizing the integrated detection and positioning of UAVs that combines radio frequency detection and optical observation.

[0062] The integrated optoelectronic platform system for UAV detection in this application embodiment is used to implement the aforementioned integrated optoelectronic platform method for UAV detection. Therefore, the specific implementation of the integrated optoelectronic platform system for UAV detection can be found in the embodiment section of the integrated optoelectronic platform method for UAV detection above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0063] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the integrated optoelectronic platform method for UAV detection described above.

[0064] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described integrated optoelectronic platform methods for UAV detection.

[0065] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0066] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the integrated optoelectronic platform method for UAV detection.

[0067] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0068] The above provides a detailed description of the integrated optoelectronic platform method, system, electronic device, and storage medium for UAV detection provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A comprehensive optoelectronic platform method for UAV detection, characterized in that, include: The system collects raw signals from the drone's remote control system, generates multi-channel signal data with precise timestamps based on the raw signals, synchronously triggers the photoelectric platform to capture high-definition video streams in suspicious airspace, and establishes a spatiotemporal correspondence database between the multi-channel signal data and the high-definition video streams. A high-precision atomic clock is configured as a unified time reference, and each radio frequency receiving node is synchronized through a precision time synchronization protocol. Based on the frequency and modulation characteristics in the multi-channel signal data, the signal is frequency converted and demodulated to obtain a time-aligned base signal. Radio frequency (RF) features are extracted from the basic signal, and the RF features are associated with the device identification information in the multi-channel signal data. Through feature digitization processing and classification analysis, a mapping relationship between RF features and UAV remote controller model is established. Based on the mapping relationship, a hardware feature library containing feature templates, threshold parameters, and matching algorithms is constructed. Based on the identification results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, the multipath error is corrected by particle filtering technology and the operator's position at the next moment is predicted. The predicted position is then updated using the corrected multipath error compensation results. This process is iterated until convergence, thereby realizing the integrated detection and positioning of UAVs that combines radio frequency detection and optical observation.

2. The method according to claim 1, characterized in that, Based on the identification results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, multipath errors are corrected using particle filtering technology to predict the operator's position at the next moment. The predicted position is then updated using the corrected multipath error compensation results. This process is iterated until convergence, achieving integrated UAV detection and positioning combining radio frequency detection and optical observation, including: The collected basic signals are input into the hardware feature library, and the identification result of the UAV remote controller model is output through the matching algorithm. The arrival time difference of the basic signals between different radio frequency receiving nodes is calculated, and the initial position estimation area is generated by combining the signal propagation speed. Optical observation data corresponding to the recognition result is extracted from the spatiotemporal correspondence database. The optical observation data includes the pixel coordinates of the UAV in the video frame. The initial position estimation area and the optical observation data are input into the position prediction unit to generate the operator's initial position prediction point. Based on the particle filter model, multipath propagation paths are simulated, and the theoretical signal propagation delay between the initial position prediction point and each radio frequency receiving node is calculated. The theoretical signal propagation delay is compared with the actual arrival time difference to generate a multipath error compensation amount. Based on the multipath error compensation amount, the initial position prediction point is corrected, and the updated prediction position is output. Repeat the error compensation and position update operations until the deviation of the predicted position is less than the convergence threshold for two consecutive times, and obtain the integrated detection and positioning result of the UAV combining radio frequency detection and optical observation.

3. The method according to claim 2, characterized in that, Based on the particle filter model to simulate multipath propagation paths, the theoretical signal propagation delay between the initial position prediction point and each radio frequency receiving node is calculated, including: Based on the initial position prediction point and the position of the radio frequency receiving node, a set of particles representing different signal reflection paths is generated, and each particle contains a path length parameter and a reflection point position parameter. Calculate the matching degree between the path length parameter of each particle in the particle set and the initial position prediction point, and retain a preset number of particles with the highest matching degree to form a particle subset; The particles in the particle subset are randomly paired to form multiple particle pairs. The reflection point position parameters of each particle pair are exchanged. A new reflection path is generated based on the exchanged reflection point position parameters. The new particles corresponding to the new reflection path are combined into a new particle set. Randomly select some particles from the new particle set, randomly perturb the reflection point position parameters of the selected particles to form an optimized particle set, and calculate the path length of each propagation path from the initial position prediction point to each radio frequency receiving node based on the optimized particle set. Based on the propagation speed of electromagnetic waves, the path length of each propagation path is converted into a single-path propagation time. The single-path propagation times are then weighted and combined according to particle weights to generate the theoretical signal propagation delay.

4. The method according to claim 1, characterized in that, Radio frequency (RF) features are extracted from the basic signal, and these RF features are associated with device identification information in the multi-channel signal data. Through feature digitization processing and classification analysis, a mapping relationship between RF features and UAV remote controller models is established. Based on this mapping relationship, a hardware feature library containing feature templates, threshold parameters, and matching algorithms is constructed, including: The signal amplitude, frequency offset, and modulation period are separated from the basic signal as radio frequency features, which are stored in the form of numerical vectors. The preset device encoding field in the multi-channel signal data is associated with the numerical vector of the radio frequency feature to form a device-feature binding dataset; The bound dataset is subjected to feature digitization processing, the numerical vector of the radio frequency feature is quantized into a fixed-dimensional digital matrix, and the digital matrix is ​​clustered to generate feature clusters corresponding to different drone remote controller models. Establish a mapping relationship between the feature clusters and the drone remote controller models, and extract the center vector of each model corresponding to the feature cluster based on the mapping relationship to generate feature templates; Based on the mapping relationship and the cluster boundary distance, a threshold parameter is set, and a matching algorithm based on similarity calculation is designed. The feature template, threshold parameter and matching algorithm are integrated to build a dynamically updatable hardware feature library.

5. The method according to claim 4, characterized in that, Establish a mapping relationship between the feature clusters and the UAV remote controller models, and extract the center vector of each model corresponding to the feature cluster based on the mapping relationship to generate feature templates, including: Read the device encoding field from the bound dataset, and map each feature cluster to a unique drone remote controller model based on the model information contained in the device encoding field, thus forming a mapping relationship between model and feature cluster. Based on the mapping relationship, the feature clusters associated with each drone remote controller model are located, and the arithmetic mean of all numerical vectors within the feature clusters is calculated to obtain the center vector representing the features of each model. The center vector of each drone remote controller model is standardized, the numerical precision of all vector elements is adjusted to a uniform standard, and the standardized center vector is encapsulated into a data structure with a fixed format to generate a feature template with uniform dimensions and encoding format.

6. The method according to claim 1, characterized in that, A high-precision atomic clock is configured as a unified time reference. A precise time synchronization protocol is used to synchronize the time of each radio frequency receiving node. Based on the frequency and modulation characteristics of the multi-channel signal data, frequency conversion and demodulation processing are performed to obtain a time-aligned base signal, including: A highly stable atomic clock is deployed as a unified time reference. The atomic clock outputs a standard clock signal to all radio frequency receiving nodes and distributes the standard clock signal through a wired transmission link, so that the deviation between the local clock of each radio frequency receiving node and the unified time reference is less than a microsecond threshold, thereby completing the time synchronization between nodes. Frequency characteristic parameters and modulation waveform characteristic parameters are separated from the multi-channel signal data, and the original signals of different bands are converted into a unified intermediate frequency signal using a frequency conversion unit; Demodulation is performed on the unified intermediate frequency signal to remove the carrier component and extract the baseband signal, resulting in a basic signal containing timing information; The basic signal is time-calibrated based on the time synchronization results between nodes to eliminate phase offset caused by transmission delay between nodes and generate a time-aligned basic signal.

7. The method according to claim 1, characterized in that, The process involves acquiring raw signals from UAV remote control signals, generating multi-channel signal data with precise timestamps based on these raw signals, synchronously triggering an optoelectronic platform to capture high-definition video streams in suspicious airspace, and establishing a spatiotemporal correspondence database between the multi-channel signal data and the high-definition video streams, including: The system captures the raw signal emitted by the drone remote controller at multiple radio frequency receiving nodes. The raw signal contains frequency bands and modulation waveforms, and the initial time information of the raw signal is recorded by the signal acquisition unit. The initial time information is input into the time stamping unit, and a unique time identifier is added to each of the original signals in combination with the global clock reference to generate multi-channel signal data with precise timestamps. Based on the time identifier sequence in the multi-channel signal data, the photoelectric sensor is controlled to perform directional scanning of the suspicious airspace and capture a high-definition video stream containing UAV images; The time identifiers of the multi-channel signal data are matched and mapped with the frame timecodes of the high-definition video stream to form a spatiotemporal correspondence database containing signal feature fields and video frame coordinates.

8. A comprehensive optoelectronic platform system for unmanned aerial vehicle (UAV) detection, characterized in that, include: The acquisition module is used to acquire the raw signals of the UAV remote control signal, generate multi-channel signal data with precise timestamps based on the raw signals, synchronously trigger the photoelectric platform to capture high-definition video streams in suspicious airspace, and establish a spatiotemporal correspondence database of the multi-channel signal data and the high-definition video streams. The conversion module is used to configure a high-precision atomic clock as a unified time reference, and to achieve time synchronization of each radio frequency receiving node through a precision time synchronization protocol. Based on the frequency characteristics and modulation characteristics in the multi-channel signal data, the module performs frequency conversion and signal demodulation processing to obtain a time-aligned base signal. The module is used to extract radio frequency features from the basic signal, associate the radio frequency features with the device identification information in the multi-channel signal data, establish a mapping relationship between radio frequency features and UAV remote controller model through feature digitization processing and classification analysis, and construct a hardware feature library containing feature templates, threshold parameters and matching algorithms based on the mapping relationship. The update module is used to correct multipath errors and predict the operator's position at the next moment by using particle filtering technology, based on the recognition results of the hardware feature library, combined with the signal time difference in the basic signals of multiple nodes and the optical observation data in the spatiotemporal correspondence database, and then using the corrected multipath error compensation results to update the predicted position. This process is iterated until convergence, thereby realizing the integrated detection and positioning of UAVs that combines radio frequency detection and optical observation.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the integrated optoelectronic platform method for unmanned aerial vehicle (UAV) detection as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the integrated optoelectronic platform method for UAV detection as described in any one of claims 1 to 7.

Images