Unmanned aerial vehicle detection method and system based on radio intelligent fusion

Abstract

The invention relates to the technical field of data processing, and discloses an unmanned aerial vehicle detection method and system based on radio intelligent fusion. The method comprises the following steps: acquiring a radio signal to obtain I / Q data; performing time-frequency conversion and feature extraction; inputting convolutional neural network identification; obtaining a type by using a clustering joint recognition algorithm; constructing relay network fusion processing to determine a position identity; and calculating a threat index to execute early warning grading to form a defense strategy. According to the invention, cooperative sensing and intelligent fusion of multi-source information can be realized through a multi-stage relay network in a complex electromagnetic environment, the accuracy, coverage and real-time performance of unmanned aerial vehicle detection are improved, and adaptive early warning and defense response are realized based on threat intelligent evaluation.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method and system for detecting unmanned aerial vehicles (UAVs) based on radio intelligent fusion. Background Art

[0002] In the field of UAV detection and defense, existing technologies mainly rely on single detection means, such as radar detection, radio monitoring, or optoelectronic tracking. Radar detection technology uses the principle of electromagnetic wave reflection and can detect UAV targets at relatively long distances, but it faces problems such as difficulty in identifying small targets and susceptibility to interference. Radio monitoring technology can sense the presence of UAVs within a certain range by receiving the communication signals between UAVs and remote controllers, but it has limited effectiveness for encrypted communication or autonomously flying UAVs. Optoelectronic tracking technology relies on infrared or visible light cameras to capture UAV images and has the advantage of intuitive visualization, but it is greatly restricted by environmental conditions such as weather and light. In addition, the existing UAV identification methods usually rely on simple feature matching or threshold judgment, with low identification accuracy and a lack of effective multi-source information fusion mechanism, making it difficult to meet the precise identification requirements in complex electromagnetic environments.

[0003] The main deficiencies of the existing technologies are as follows: First, the perception ability of a single sensor is limited, unable to comprehensively capture the feature information of UAVs, resulting in high false alarm rates and missed detection rates. Second, traditional signal processing methods have insufficient capabilities for processing UAV signals under weak signal and low signal-to-noise ratio conditions, making it difficult to achieve reliable detection in complex environments. Third, there is a lack of an effective multi-sensor collaborative detection mechanism, where each sensor works independently and fails to fully utilize information complementarity. Finally, existing UAV threat assessments are mostly static models, lacking an intelligent dynamic assessment and early warning response mechanism and being unable to respond promptly to rapidly changing UAV threats. These deficiencies severely restrict the actual effectiveness of UAV detection systems in important application scenarios such as no-fly zone monitoring and critical infrastructure protection. Summary of the Invention

[0004] This application provides a method and system for detecting UAVs based on radio intelligent fusion, which can achieve collaborative perception and intelligent fusion of multi-source information through a multi-level relay network in a complex electromagnetic environment, improve the accuracy, coverage, and real-time performance of UAV detection, and realize adaptive early warning and defense responses based on threat intelligent assessment.

[0005] In a first aspect, the present application provides a method for detecting unmanned aerial vehicles (UAVs) based on radio intelligent fusion. The method for detecting UAVs based on radio intelligent fusion includes: collecting radio signals through distributed relay sensor nodes to obtain digital I / Q data; performing time-frequency domain conversion and feature extraction on the digital I / Q data to obtain a set of UAV radio signal feature vectors; inputting the set of UAV radio signal feature vectors into a convolutional neural network for training and recognition to obtain a UAV signal recognition result; obtaining a UAV type recognition result by performing a clustering signal joint recognition algorithm on the UAV signal recognition result; constructing a multi-level relay network to perform collaborative data fusion processing based on the UAV type recognition result to obtain UAV position and identity information; calculating a threat index for the UAV position and identity information and performing early warning classification based on the relay network to obtain a defense response strategy.

[0006] In a second aspect, the present application provides a system for detecting UAVs based on radio intelligent fusion. The system for detecting UAVs based on radio intelligent fusion includes: a collection module, configured to collect radio signals through distributed relay sensor nodes to obtain digital I / Q data; an extraction module, configured to perform time-frequency domain conversion and feature extraction on the digital I / Q data to obtain a set of UAV radio signal feature vectors; an input module, configured to input the set of UAV radio signal feature vectors into a convolutional neural network for training and recognition to obtain a UAV signal recognition result; a recognition module, configured to obtain a UAV type recognition result by performing a clustering signal joint recognition algorithm on the UAV signal recognition result; a fusion module, configured to construct a multi-level relay network to perform collaborative data fusion processing based on the UAV type recognition result to obtain UAV position and identity information; a classification module, configured to calculate a threat index for the UAV position and identity information and perform early warning classification based on the relay network to obtain a defense response strategy.

[0007] In a third aspect, there is provided a computer device, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor invokes the instructions in the memory so that the computer device executes the above-mentioned method for detecting UAVs based on radio intelligent fusion.

[0008] In a fourth aspect, there is provided a computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, and when the instructions are run on a computer, the computer is caused to execute the above-mentioned method for detecting UAVs based on radio intelligent fusion.

[0009] In the technical solution provided by this application, digital I / Q data is acquired by collecting radio signals through distributed relay sensor nodes, achieving seamless coverage of a wide-area space and solving the problem of limited coverage of a single sensor; the digital I / Q data is subjected to time-frequency domain conversion and feature extraction to obtain a set of UAV radio signal feature vectors, fully exploiting the multi-dimensional features of UAV signals and improving the richness and discrimination of feature expression; the set of UAV radio signal feature vectors is input into a convolutional neural network for training and recognition to obtain UAV signal recognition results, enabling the model to automatically learn the deep feature patterns of the signals, and the recognition ability is no longer limited by manually preset feature rules; the preliminary UAV signal recognition results are processed through a clustering signal joint recognition algorithm to obtain UAV type recognition results, skillfully integrating clustering analysis and pattern recognition technologies, and accurately distinguishing UAV models with similar features, significantly reducing the misrecognition rate; a multi-level relay network is constructed based on the UAV type recognition results to perform collaborative data fusion processing to obtain UAV position and identity information, solving the information island problem in traditional single-node detection and achieving effective complementarity and redundant verification of information between different nodes; a threat index is calculated for the UAV position and identity information and an early warning classification based on the relay network is performed to obtain a defense response strategy, constructing a complete intelligent defense link from perception to decision-making, and ensuring that corresponding defense measures are taken for UAVs with different threat levels. The present invention applies artificial intelligence algorithms and models, especially convolutional neural networks and clustering signal joint recognition algorithms, in a specific UAV detection application field, not only achieving transcendence over traditional methods in the feature extraction and recognition stages, but also significantly improving the detection accuracy and anti-interference ability through algorithm features, showing excellent adaptability in complex electromagnetic environments; the collaborative data fusion mechanism of the multi-level relay network fully considers the spatial characteristics of radio signal propagation, utilizes the spatial diversity formed by distributed sensing nodes, and can still maintain a stable and reliable detection effect in a strong interference environment; the adaptive early warning classification strategy based on the threat index further deeply combines artificial intelligence decision-making technology with professional domain knowledge to form an intelligent defense system against UAV threats, which has higher pertinence and resource utilization efficiency compared with traditional static defense strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0011] Figure 1 It is a schematic diagram of an embodiment of the UAV detection method based on radio intelligent fusion in the embodiments of this application; Figure 2 Schematic diagram of an embodiment of the UAV detection system based on radio intelligent fusion in the embodiments of the present application; Figure 3 It is a structural schematic block diagram of a computer device in the embodiments of the present invention. Detailed implementation manners

[0012] The embodiments of the present application provide a UAV detection method and system based on radio intelligent fusion. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the UAV detection method based on radio intelligent fusion in the embodiments of the present application includes: Step S101: Collect radio signals through distributed relay sensor nodes to obtain digital I / Q data; Step S102: Perform time-frequency domain conversion and feature extraction on the digital I / Q data to obtain a UAV radio signal feature vector set; Step S103: Input the UAV radio signal feature vector set into a convolutional neural network for training and recognition to obtain a UAV signal recognition result; Step S104: Use a clustering signal joint recognition algorithm to process the UAV signal recognition result to obtain a UAV type recognition result; Step S105: Construct a multi-level relay network according to the UAV type recognition result to perform collaborative data fusion processing to obtain UAV position and identity information; Step S106: Calculate a threat index for the UAV position and identity information and perform early warning classification based on the relay network to obtain a defense response strategy.

[0014] It can be understood that the execution subject of the present application can be a UAV detection system based on radio intelligent fusion, or a terminal or a server. Specifically, it is not limited here. The embodiments of the present application will be described by taking the server as the execution subject as an example.

[0015] Specifically, radio signals are collected by distributed relay sensor nodes to obtain digitized I / Q data. The distributed relay sensor nodes refer to radio signal receiving devices arranged at different geographical locations, which form a network and can capture radio signals in the environment. The devices include general software radio peripherals and corresponding signal processing modules, and the operating frequencies cover the 2.4 GHz ISM band, the 5.8 GHz band, and the 900 MHz band. These bands are the main bands for the operation of the remote control, video transmission, and telemetry systems of most commercial drones. During the signal acquisition process, the sensor nodes pre-amplify the received analog radio frequency signals through a low-noise amplifier, then filter out out-of-band interference signals through a band-pass filter, mix the radio frequency signals with the carrier signals generated by the local oscillator through a mixer to down-convert the radio frequency signals to intermediate frequency or baseband signals, and digitize the signals into I / Q data streams in complex form at a sampling rate of at least 20 MHz through an analog-to-digital converter.

[0016] Perform time-frequency domain conversion and feature extraction on the digitized I / Q data to obtain the UAV radio signal feature vector set. Convert the I / Q data into a time-frequency domain representation through the short-time Fourier transform to generate a signal spectrogram. Calculate parameters such as power spectral density and frequency occupancy bandwidth for the signal spectrogram to obtain frequency domain feature parameters. At the same time, perform envelope analysis on the I / Q data to extract time domain features, including signal envelope shape features, pulse repetition intervals, etc. Then calculate the high-order cumulants and cyclic spectral features for the time domain features and frequency domain feature parameters to obtain modulation recognition parameters, which are used to identify the modulation types used by UAV signals. Extract MAC layer features from the modulation recognition parameters and the signal message structure to obtain communication protocol features, including frame structure, preamble pattern, etc. Finally, normalize and assign weights to all the extracted features to form the UAV radio signal feature vector set.

[0017] The drone radio signal feature vector set is input into a convolutional neural network for training and recognition to obtain the drone signal recognition result. The convolutional neural network structure consists of 5 convolutional layers, each followed by a max pooling layer and a batch normalization layer. The first convolutional layer uses 32 3×3 convolutional kernels, the second convolutional layer uses 64 3×3 convolutional kernels, the third convolutional layer uses 128 3×3 convolutional kernels, the fourth layer uses 256 3×3 convolutional kernels, and the fifth layer uses 512 3×3 convolutional kernels. Two fully connected layers are designed after the convolutional layers, containing 1024 and 512 neurons respectively. During the training process, the Adam optimizer is adopted, with a learning rate of 0.001, a batch size of 64, and 200 training rounds. The drone radio signal feature vector set is first normalized to conform to the input format of the convolutional neural network, and then sent to the convolutional layer for feature mapping to generate hierarchical feature expressions. After dimensionality reduction and normalization processing by the pooling layer and the batch normalization layer, it is input into the fully connected layer for non-linear transformation. Finally, the probability distribution of each drone type is calculated through the Softmax function, and the category corresponding to the highest probability is selected as the recognition result. The drone type recognition result is obtained by processing the drone signal recognition result through the clustering signal joint recognition algorithm. The clustering signal joint recognition algorithm first configures a two-dimensional self-organizing mapping network with a network size of 10×10 and a total of 100 neurons. Training is carried out in batch mode, with the initial learning rate of 0.5 and exponential decay, and the initial neighborhood radius of 5 and gradual reduction. The known drone samples are input into the trained network, and the distribution of each type of sample on the grid is statistically analyzed to form a grid position probability distribution map. The drone signal recognition result is input into the network, the distance from each neuron is calculated, and the neuron with the smallest distance is found as the matching unit to obtain the position of the signal in the feature space. According to this position, the grid position probability distribution map is queried, and the spatial mapping probability value is calculated. The spatial mapping probability value and the classification probability output by the convolutional neural network are weighted and fused according to the dynamically adjusted weight coefficient to obtain the comprehensive recognition probability. Then, the recognition result is smoothed by the state transition matrix for time series to obtain the drone type recognition result.

[0018] Construct a multi - level relay network based on the UAV type recognition result to perform collaborative data fusion processing, and obtain the UAV position and identity information. The multi - level relay network divides sensor nodes into master nodes, relay nodes, and edge nodes, and establishes communication links between nodes through a hierarchical network topology. Each node is synchronized through the Precision Time Protocol, and the clock deviation is controlled within 100 nanoseconds. According to the geographical locations and signal reception strengths of each node, the trilateration algorithm is used to calculate the position coordinates of the signal source. The feature vectors extracted by each node are transmitted through the network to the upper - level nodes for feature weighted averaging and dimensionality reduction processing to obtain the fused feature vectors. The trust degrees and uncertainties of the independent recognition results of each node are calculated through the evidence theory framework, and the weight coefficients are dynamically adjusted to obtain the comprehensive recognition result. The position coordinates of the signal source are associated and matched with the comprehensive recognition result, and trajectory association is performed through a tracking filter to obtain the UAV position and identity information.

[0019] Calculate the threat index for the UAV position and identity information and perform early - warning classification based on the relay network to obtain the defense response strategy. The threat index calculation comprehensively considers factors such as the UAV type danger coefficient, the proximity to the sensitive area, the UAV size, the approaching speed, and the security level of the sensitive area. The threat index is divided into four levels: normal (0 - 0.25), attention (0.25 - 0.5), warning (0.5 - 0.75), and emergency (0.75 - 1.0). Adjust the monitoring area scanning frequency according to the early - warning level, and send notifications with different priorities to relevant personnel. Activate the corresponding detection equipment through the multi - level relay network for target confirmation to generate target confirmation information. Perform trajectory tracking and position prediction on the UAV position to obtain the trajectory prediction result. Allocate response resources according to the target confirmation information, the trajectory prediction result, and the early - warning level determination result to form the defense response strategy. In a certain actual measurement, when a UAV approached the sensitive area, the threat index calculated by the system was 0.68, which was determined to be the warning level. The optoelectronic detection equipment was immediately activated for confirmation, and it was predicted that it would enter the core area after 10 seconds. Therefore, the electromagnetic interference defense equipment was triggered, successfully preventing the UAV from approaching further.

[0020] In the embodiments of the present application, digital I / Q data is acquired by collecting radio signals through distributed relay sensor nodes, achieving seamless coverage of a wide-area space and solving the problem of limited coverage of a single sensor; the digital I / Q data is subjected to time-frequency domain conversion and feature extraction to obtain a set of UAV radio signal feature vectors, fully exploiting the multi-dimensional features of UAV signals and improving the richness and distinctiveness of feature expression; the set of UAV radio signal feature vectors is input into a convolutional neural network for training and recognition to obtain UAV signal recognition results, enabling the model to automatically learn the deep feature patterns of the signals, and the recognition ability is no longer limited by manually preset feature rules; the preliminary UAV signal recognition results are processed by a clustering signal joint recognition algorithm to obtain UAV type recognition results, cleverly integrating clustering analysis and pattern recognition technologies, and accurately distinguishing UAV models with similar features, significantly reducing the misrecognition rate; a multi-level relay network is constructed based on the UAV type recognition results to perform collaborative data fusion processing to obtain UAV position and identity information, solving the information island problem in traditional single-node detection and achieving effective complementarity and redundant verification of information between different nodes; a threat index is calculated for the UAV position and identity information and an early warning classification based on the relay network is performed to obtain a defense response strategy, constructing a complete intelligent defense link from perception to decision-making, and ensuring corresponding defense measures are taken for UAVs with different threat levels. The present invention applies artificial intelligence algorithms and models, particularly convolutional neural networks and clustering signal joint recognition algorithms, in a specific UAV detection application field, not only achieving transcendence over traditional methods in the feature extraction and recognition stages, but also significantly improving the detection accuracy and anti-interference ability through algorithm features, showing excellent adaptability in complex electromagnetic environments; the collaborative data fusion mechanism of the multi-level relay network fully considers the spatial characteristics of radio signal propagation, utilizes the spatial diversity formed by distributed sensing nodes, and can still maintain a stable and reliable detection effect in a strong interference environment; the adaptive early warning classification strategy based on the threat index deeply combines artificial intelligence decision-making technology with professional domain knowledge, forming an intelligent defense system against UAV threats, which has higher pertinence and resource utilization efficiency compared with traditional static defense strategies.

[0021] In a specific embodiment, the process of executing step S101 may specifically include the following steps: The radio signals in the environment are pre-amplified through a low-noise amplifier to obtain an analog radio frequency signal; The analog radio frequency signal is subjected to out-of-band interference filtering through a band-pass filter to obtain a filtered radio frequency signal; The filtered radio frequency signal is down-converted using a mixer and a local oscillator to obtain an intermediate frequency or baseband signal; Digitize the intermediate frequency or baseband signal through an analog-to-digital converter at a high sampling rate to obtain the original I / Q data stream; Perform digital downsampling and digital filtering on the original I / Q data stream to obtain noise-reduced I / Q data; Mark the noise-reduced I / Q data by adding a timestamp and geographical location information to obtain digitized I / Q data.

[0022] Specifically, pre-amplify the radio signals in the environment through a low-noise amplifier to obtain the amplified analog radio frequency signal. A low-noise amplifier is an electronic amplifier, and its main function is to amplify weak radio frequency signals while introducing as little noise as possible. In a drone detection system, the received radio signals are usually weak, and the signal power may be between -90 dBm and -120 dBm. The low-noise amplifier amplifies these weak signals to a level suitable for subsequent processing, usually with a gain between 20 dB and 30 dB, while maintaining a low noise figure, generally controlled below 2 dB. The dynamic range of the amplifier needs to be considered during the amplification process to ensure that the amplifier will not saturate due to an overly strong signal or be overwhelmed by noise due to an overly weak signal.

[0023] Filter out out-of-band interference from the amplified analog radio frequency signal through a band-pass filter to obtain the filtered radio frequency signal. A band-pass filter is an electronic filter that only allows signals within a specific frequency range to pass through and blocks signals of other frequencies. In a drone detection system, the band-pass filter is designed to only pass through frequency bands commonly used by drones, such as the 2.4 GHz ISM band, the 5.8 GHz band, and the 900 MHz band. The bandwidth of the filter needs to be set according to the bandwidth characteristics of the target drone communication signal, usually between 20 MHz and 40 MHz. The filtering process uses a multi-stage band-pass filter structure, which can provide at least 60 dB of out-of-band rejection, effectively reducing the impact of interference signals generated by other electronic devices on the detection system. The filtered signal retains the frequency band characteristics that the drone communication may use, while significantly reducing the interference components of other frequency bands. Then, perform down-conversion processing on the filtered radio frequency signal using a mixer and a local oscillator to obtain an intermediate frequency or baseband signal. A mixer is a circuit that multiplies the input signal by the local oscillator signal to generate sum and difference frequency signals. The local oscillator generates a stable sine wave signal, and its frequency is set according to the target drone communication frequency. During the down-conversion process, the filtered radio frequency signal is mixed with the local oscillator signal to convert the high-frequency radio frequency signal into a lower-frequency intermediate frequency signal (usually in the range of dozens of MHz) or directly into a baseband signal (close to zero frequency). For example, when detecting a 2.4 GHz drone signal, the local oscillator frequency can be set to 2.385 GHz, and an intermediate frequency signal of 15 MHz is obtained through mixing, which is convenient for subsequent digital processing. The phase noise and frequency stability of the local oscillator need to be controlled during the down-conversion process to ensure the quality of the converted signal.

[0024] The intermediate frequency or baseband signal is digitized by an analog-to-digital converter at a high sampling rate to obtain the original I / Q data stream. An analog-to-digital converter is a device that converts a continuous analog signal into a discrete digital signal. In a drone detection system, the analog-to-digital converter needs to have a sufficient sampling rate and bit depth to accurately capture the signal characteristics. The sampling rate is at least 20 MHz, which is sufficient to cover the bandwidth of most commercial drone communication signals according to the Nyquist sampling theorem. The bit depth of analog-to-digital conversion is usually 12 bits or 16 bits to provide sufficient dynamic range and digitization accuracy. The digitization process converts the intermediate frequency or baseband signal into in-phase component (I) and quadrature component (Q) data, that is, I / Q data. This representation preserves the amplitude and phase information of the signal. I / Q data is represented in complex form, where I is the real part and Q is the imaginary part, describing the instantaneous state of the signal.

[0025] The original I / Q data stream is subjected to digital downsampling and digital filtering to obtain noise-reduced I / Q data. Digital downsampling refers to reducing the sampling rate of the signal, aiming to reduce the amount of data and the computational complexity of subsequent processing. In a drone detection system, the original I / Q data stream may have too high a sampling rate. By downsampling, it is reduced to the actual required sampling rate, usually reduced to 1 / 2 or 1 / 4 of the original sampling rate. Anti-aliasing filtering is required before downsampling to prevent high-frequency components from folding into the low-frequency part and causing signal distortion. Digital filtering is to perform band-pass, low-pass or high-pass filtering on the signal in the digital domain to further suppress noise and interference. FIR (finite impulse response) filters are commonly used in drone detection systems because their linear phase characteristics are beneficial to maintaining the signal form characteristics. The noise components in the I / Q data after digital filtering are significantly reduced, and the signal characteristics are more prominent. Timestamp and geographical location information are added to the noise-reduced I / Q data for marking to obtain digitized I / Q data. A timestamp is accurate time information recording the data acquisition moment, with an accuracy usually at the microsecond or nanosecond level, obtained through a GPS receiver or network time protocol (NTP). Geographical location information includes the longitude, latitude and altitude of the sensor node, also obtained through a GPS receiver, with a positioning accuracy at the meter level. The addition of timestamp and geographical location information enables the data collected by multiple sensor nodes to be correlated in time and space, providing a basis for subsequent multi-node collaborative processing and target positioning. During the data marking process, the timestamp and geographical location information are attached to the I / Q data packet header in the form of metadata to form a digitized I / Q data packet, facilitating network transmission and subsequent processing.

[0026] For example, a certain sensor node receives a suspected drone signal with an initial signal strength of -100 dBm. After being amplified by a low-noise amplifier with a 30 dB gain, the signal strength reaches -70 dBm. After being processed by a band-pass filter, the interference signals outside the 2.4 GHz frequency band are attenuated by 60 dB, and the signal-to-noise ratio is increased from 5 dB to 25 dB. Subsequently, the signal is mixed with a local oscillator with a frequency of 2.385 GHz to obtain an intermediate-frequency signal with a center frequency of 15 MHz. The intermediate-frequency signal is digitized by an analog-to-digital converter with 16 bits and a sampling rate of 50 MHz to obtain the original I / Q data stream. The original data is decimated by a factor of 4, and the sampling rate is reduced to 12.5 MHz. At the same time, a 64th-order FIR filter is applied for digital filtering to further increase the signal-to-noise ratio to 30 dB. Finally, a timestamp accurate to the microsecond and geographical location information with a meter-level accuracy are added to complete the preparation of the digitized I / Q data.

[0027] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Convert the digitized I / Q data to a time-frequency domain representation through short-time Fourier transform to obtain a signal spectrogram; Calculate the power spectral density and analyze the frequency occupancy bandwidth of the signal spectrogram to obtain frequency-domain characteristic parameters; Extract time-domain characteristics from the digitized I / Q data through envelope analysis to obtain time-domain characteristics; Calculate the high-order cumulant and cyclic spectrum characteristics of the time-domain characteristics and frequency-domain characteristic parameters to obtain modulation recognition parameters; Extract MAC-layer characteristics from the modulation recognition parameters and the signal message structure to obtain communication protocol characteristics; Normalize and assign weights to the frequency-domain characteristic parameters, time-domain characteristics, modulation recognition parameters, and communication protocol characteristics to obtain a set of UAV radio signal feature vectors.

[0028] Specifically, convert the digitized I / Q data to a time-frequency domain representation through short-time Fourier transform. Short-time Fourier transform is a method for time-frequency analysis of signals, which is realized by performing Fourier transform after windowing the signal. The digitized I / Q data is divided into multiple overlapping time segments, and each segment is multiplied by a window function (such as Hamming window, Blackman window, etc.) and then subjected to discrete Fourier transform. In UAV signal processing, the window length is usually selected as 1024 points, and the overlap rate is 50% to balance the time resolution and frequency resolution. The Fourier transform results of each time segment form a column of the spectrum, and the spectra of multiple time segments are arranged in time order to form a two-dimensional time-frequency spectrogram. The horizontal axis represents time, the vertical axis represents frequency, and the color shade represents the strength of the signal energy. Through the signal spectrogram, the time-varying characteristics and frequency characteristics of the UAV signal are visually presented.

[0029] Perform power spectral density calculation and frequency occupancy bandwidth analysis on the signal spectrogram to obtain frequency-domain characteristic parameters. Power spectral density calculation estimates the energy at each frequency point in the signal spectrogram. Based on Welch's method of the periodogram, the power spectra calculated from multiple time periods are averaged to reduce the influence of noise. The calculation result shows the distribution of signal energy in frequency. Frequency occupancy bandwidth analysis measures the effective bandwidth of the signal power spectrum. The specific method is to determine the continuous frequency range where the signal energy exceeds the noise threshold, or to determine the frequency range that contains 90% of the total signal energy. In addition to bandwidth, parameters such as center frequency, frequency offset, and spectral peak ratio are also calculated. For example, the typical bandwidth of the remote control signal of a certain brand of drone is 4 MHz, while the bandwidth of the video transmission signal reaches 20 MHz, and the center frequency is in the 2.4 GHz band.

[0030] Extract time-domain characteristics from the digitized I / Q data through envelope analysis to obtain time-domain characteristics. Envelope analysis is a method to obtain the instantaneous amplitude change of the signal. The signal envelope is obtained by calculating the modulus of the I / Q data or using the Hilbert transform. The time-domain characteristics extracted from the signal envelope include the start time, duration, rise time, fall time, pulse repetition interval, duty cycle, etc. of the signal. It also includes the statistical characteristics of the signal amplitude, such as mean, variance, kurtosis, skewness, etc. Time-domain characteristics are particularly important for identifying different drone communication signal formats because the remote control signals of drones from different manufacturers usually have different signal transmission timing patterns. For example, the duration of each frame of the control signal of a certain model of drone is 20 milliseconds, the pulse repetition interval is 5 milliseconds, and the rise time is about 0.2 milliseconds. These characteristics are significantly different from those of other models of drones.

[0031] Calculate high-order cumulants and cyclic spectrum characteristics for the time-domain characteristics and frequency-domain characteristic parameters to obtain modulation recognition parameters. High-order cumulants are the high-order statistical characteristics of the signal, which are not affected by Gaussian noise and are effective for identifying signal modulation methods. The second-order cumulant is equivalent to the autocorrelation function, and the third-order and fourth-order cumulants reflect the non-linear characteristics of the signal. Cyclic spectrum characteristics depict the periodicity of the signal. By calculating the correlation of the signal at different frequency offsets, a two-dimensional cyclic spectrum is obtained, which can distinguish different modulation types. The modulation recognition parameters extracted from high-order cumulants and cyclic spectra include spectral peak value, spectral line density, spectral line distribution, etc., which are used to determine the modulation method adopted by the signal. Because drones from different manufacturers usually use different modulation schemes. For example, the remote control signal of a certain brand of drone uses FHSS (Frequency Hopping Spread Spectrum) technology, and the modulation method is 2FSK, while the video transmission signal uses OFDM modulation.

[0032] Extract MAC layer features from modulation recognition parameters and signal message structures to obtain communication protocol features. MAC layer feature extraction is the analysis of the communication protocol structure, including frame format, preamble pattern, header structure, data payload length distribution, etc. By synchronizing and demodulating the signal, the boundaries of data frames are identified, and then the intra-frame structural features are analyzed. Common UAV communication protocols such as MAVLink, SBUS, DSM, etc. all have specific frame structures and checksum mechanisms. The extracted MAC layer features include frame length statistical characteristics, frame interval statistical characteristics, preamble sequence pattern, frame header field features, etc. These features are particularly effective in distinguishing UAVs from different manufacturers because even if the same physical layer technology is used, different manufacturers usually use proprietary communication protocols. For example, a certain manufacturer's UAV uses a customized communication protocol, with the characteristics that the frame length is fixed at 64 bytes, the preamble is a specific 8-byte sequence, and the frame interval is 20 milliseconds. These features are significantly different from the UAV products of other manufacturers. Normalize and assign weights to the frequency domain feature parameters, time domain features, modulation recognition parameters, and communication protocol features to obtain the UAV radio signal feature vector set. Normalization is to convert various features into the same numerical range, usually the [0,1] interval. Common methods include min-max normalization, Z-score standardization, etc. Only after normalization can the features be effectively compared and fused. Weight assignment assigns different weight coefficients according to the recognition ability of various features, usually determined by feature importance evaluation methods such as information gain, Gini index, etc. The obtained feature vector set is a high-dimensional vector composed of various normalized features combined according to weights. Each vector corresponds to a signal sample, and each element of the vector corresponds to a feature value. The entire feature extraction process forms a conversion from the original I / Q data to structured feature vectors.

[0033] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Normalize the UAV radio signal feature vector set to obtain training data that conforms to the input format of the convolutional neural network; Send the training data into a network structure containing five convolutional layers for feature mapping to obtain hierarchical feature expression data; Perform dimensionality reduction and standardization processing on the hierarchical feature expression data through a max pooling layer and a batch normalization layer to obtain compressed feature expression data; Input the compressed feature expression data into two fully connected layers for non-linear transformation to obtain vector representation data in a high-dimensional feature space; Calculate the probability distribution of each UAV type by passing the vector representation data in the high-dimensional feature space through the Softmax function; Based on the probability distribution of each drone type, select the category corresponding to the highest probability as the recognition result to obtain the drone signal recognition result.

[0034] Specifically, normalize the drone radio signal feature vector set. Normalization means converting feature vectors with different dimensions and ranges into the same numerical interval to make the contributions of each feature to the model relatively balanced. The specific processing methods include min-max normalization and Z-score standardization. Min-max normalization maps the feature values to the [0, 1] interval, and the calculation method is to subtract the minimum value from the feature value and then divide by the difference between the maximum value and the minimum value. Z-score standardization converts the features into a distribution with a mean of 0 and a standard deviation of 1, and the calculation method is to subtract the mean from the feature value and then divide by the standard deviation. In drone signal processing, min-max normalization is usually used for frequency domain features such as center frequency and bandwidth, while Z-score standardization is used for time domain features such as signal strength. After normalization, the feature vectors are reorganized into a format suitable for the input of the convolutional neural network, usually reshaping the one-dimensional feature vectors into two-dimensional matrices or three-dimensional tensors for the network to perform convolution operations.

[0035] Send the training data into a network structure containing five convolutional layers for feature mapping to obtain hierarchical feature expression data. The convolutional layer is the core component of the convolutional neural network, which extracts local features by sliding the convolutional kernel on the input data. In the drone signal recognition network, the first convolutional layer uses 32 3×3 convolutional kernels with a stride of 1 and a padding of 1 to mainly extract low-level features such as edges and textures; the second convolutional layer uses 64 3×3 convolutional kernels to extract more complex local structures; the third convolutional layer uses 128 3×3 convolutional kernels to capture more advanced feature combinations; the fourth convolutional layer uses 256 3×3 convolutional kernels; the fifth convolutional layer uses 512 3×3 convolutional kernels to extract highly abstract feature representations. After each layer of convolution, the ReLU activation function is used to introduce non-linearity and enhance the network's expressive ability. The ReLU function sets negative values to 0 and retains positive values. It has a simple form but a significant effect, which helps to solve the gradient vanishing problem in deep networks. The five convolutional layers extract features step by step to form hierarchical feature expression data, from simple signal features to complex drone feature patterns.

[0036] The hierarchical feature expression data is reduced in dimension and standardized through a max pooling layer and a batch normalization layer to obtain compressed feature expression data. The max pooling layer is a downsampling operation that selects the maximum value within a fixed-size window as the output, thereby reducing the data dimension. In the UAV signal recognition network, a 2×2 max pooling layer with a stride of 2 is connected after each convolutional layer, halving the width and height of the feature map and significantly reducing the subsequent computational amount. The batch normalization layer standardizes each mini-batch of data, stabilizing its distribution within a range with a mean of 0 and a variance of 1, which helps to accelerate network training and improve generalization. Batch normalization is performed after each convolutional layer and before the activation function. The processing includes calculating the mean and variance of the batch data, then performing standardization, and finally introducing learnable scaling and translation parameters to enable the network to adaptively adjust the data distribution. Through the combined processing of max pooling and batch normalization, the hierarchical feature expression data is compressed into a more compact and standardized representation form.

[0037] The compressed feature expression data is input into two fully connected layers for non-linear transformation to obtain vector representation data in a high-dimensional feature space. The fully connected layer connects all input neurons to all output neurons and is used to integrate the local features extracted by the previous convolutional layers to form a global feature representation. In the UAV signal recognition network, the first fully connected layer contains 1024 neurons, receives the compressed feature expression data, performs a linear transformation, introduces non-linearity through the ReLU activation function, and at the same time applies 50% Dropout to randomly deactivate half of the neurons to prevent overfitting; the second fully connected layer contains 512 neurons, further integrates the features, and also uses ReLU activation and 50% Dropout. The two-layer fully connected network maps the compressed feature expression data to a high-dimensional feature space, forming a more abstract vector representation data, which is closer to the semantic-level features of the UAV type.

[0038] The probability distribution of each UAV type is calculated for the vector representation data in the high-dimensional feature space through the Softmax function. The Softmax function is a function that converts an arbitrary real vector into a probability distribution and is commonly used in the output layer of multi-classification problems. It takes the exponent of each element in the input vector and then divides it by the sum of the exponents of all elements to ensure that the sum of the output elements is 1, which can be interpreted as the class probability. In the UAV signal recognition network, the Softmax layer receives the output vector of the second fully connected layer, and the number of neurons is equal to the preset number of UAV types plus one (an additional class represents non-UAV signals). For example, if the system needs to identify 5 UAV types, the Softmax layer has 6 neurons, corresponding to the probabilities of 5 UAV types and 1 non-UAV signal. The Softmax calculation converts the vector representation data into probability values within the range of [0,1], and the sum is 1, which intuitively reflects the likelihood of the signal belonging to each category. Based on the probability distribution of each UAV type, the category corresponding to the highest probability is selected as the recognition result, and the UAV signal recognition result is obtained. This step realizes the conversion from the probability distribution to the decision-making, adopting the maximum a posteriori probability criterion, that is, selecting the category with the highest probability as the recognition result. In practical applications, a probability threshold is often set, such as 0.75. Only when the highest probability exceeds the threshold is the recognition result confirmed, otherwise it is marked as "unknown type" to improve the reliability of the recognition. For the recognition of consecutive frames, a time smoothing strategy can also be adopted, such as the majority voting method or the exponential weighted average method, to comprehensively obtain a more stable recognition result by combining the results of multiple frames. The output UAV signal recognition result includes the identified UAV type identifier, recognition confidence, and recognition timestamp, providing the basic input for the subsequent clustering signal joint recognition.

[0039] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Configure a two-dimensional self-organizing mapping network for the UAV signal recognition result, and train the network by setting the batch processing mode, initial learning rate, learning rate decay function, initial neighborhood radius, and neighborhood shrinkage function to obtain a topology-preserving mapping network; Input the known UAV sample features into the topology-preserving mapping network for mapping, and count the distribution of various samples on the grid to obtain a grid position probability distribution map; Input the UAV signal recognition result into the topology-preserving mapping network, calculate the distance values from each neuron, and select the neuron with the smallest distance as the matching unit to obtain the position of the signal in the feature space; Query the grid position probability distribution map according to the position of the signal in the feature space, and obtain the spatial mapping probability value through weighted summation processing of the neighboring regions; Fuse the spatial mapping probability value and the classification probability value in the UAV signal recognition result according to the dynamically adjusted weight coefficient to obtain the comprehensive recognition probability; Apply the state transition matrix to the comprehensive recognition probability for time - series smoothing processing, and conduct voting statistics in combination with the recognition results of historical frames to obtain the UAV type recognition result.

[0040] Specifically, configure a two - dimensional self - organizing map network for the UAV signal recognition result. The key first step is to train the network by setting the batch processing mode, initial learning rate, learning rate decay function, initial neighborhood radius, and neighborhood shrinkage function. The two - dimensional self - organizing map network is an unsupervised learning artificial neural network that can map high - dimensional input data to a low - dimensional space (usually two - dimensional) while maintaining the topological structure of the data. In the UAV detection method, the structure of the self - organizing map network is a 10×10 two - dimensional grid with a total of 100 neurons. Each neuron has a weight vector with the same dimension as the input data. The network training adopts the batch processing mode, that is, processing multiple samples at a time and cumulatively updating, which is more stable than single - sample updating. The initial learning rate is set to 0.5 and gradually decreased through an exponential decay function. The specific formula is learning rate = initial learning rate×exp(-current iteration number / total iteration number), enabling large - step adjustments in the early stage and fine - tuning in the later stage. The initial neighborhood radius is set to 5, indicating that the influence range covers 5 cells around the center point and linearly decreases to 1 during training to ensure global sorting in the early stage and local fine - tuning in the later stage. During the training process, for each input sample, find the neuron that best matches it (referred to as the winning neuron), and then update the weight vectors of this neuron and the neurons within its neighborhood to make them closer to the input sample. After 1000 rounds of training, the network converges to form a topological - preserving map of the input data space.

[0041] Input the known UAV sample features into the topological - preserving map network for mapping, and count the distribution of various samples on the grid to obtain the grid position probability distribution map. This step is a process of constructing the association relationship between categories and grid positions. The known UAV samples refer to the UAV signal feature data sets with confirmed types, usually from laboratory acquisitions or on - site records. Input these samples into the trained topological - preserving map network. For each sample, find the best - matching unit in the network, that is, calculate the Euclidean distance between the sample and the weight vector of each neuron, and select the neuron with the smallest distance as the activation point. Then, count the activation distribution of samples of each type of UAV on the grid, calculate the frequency of each grid position being activated by various samples, and convert the frequency into a probability, that is, the number of activation times of a certain type of sample at this position divided by the total number of samples of this type. A three - dimensional data structure is formed: grid coordinate x, grid coordinate y, and category, and the value is the probability that the position is activated by the corresponding category. This grid position probability distribution map visually shows the distribution areas of different types of UAVs in the feature space and provides a spatial reference for subsequent sample classification.

[0042] Input the UAV signal recognition result into the topologically preserving mapping network, calculate the distance values from each neuron, select the neuron with the minimum distance as the matching unit, and obtain the position of the signal in the feature space. The UAV signal recognition result contains the feature representation extracted by the convolutional neural network. Input it into the trained topologically preserving mapping network and calculate the distance from each neuron weight vector. The distance calculation usually uses the Euclidean distance, which is the square root of the sum of the squares of the differences between the corresponding elements of the two vectors. After the calculation, select the neuron with the minimum distance as the best matching unit, that is, the mapping position of the signal in the two-dimensional feature space. This position is a two-dimensional coordinate (x, y), and the value ranges of both x and y are from 0 to 9 (corresponding to a 10×10 grid). In this way, the high-dimensional UAV signal features are mapped to a specific position in the two-dimensional space, facilitating visual analysis and comparison with the distribution of known samples.

[0043] Query the grid position probability distribution map according to the position of the signal in the feature space, and obtain the spatial mapping probability value through weighted summation processing in the neighboring area. This step infers the possible category that the signal belongs to based on its spatial position. First, according to the position (x, y) of the signal in the feature space obtained in the previous step, query the probability values of each category at this position in the grid position probability distribution map. Considering the continuity and smoothness of the feature mapping, not only query the probability of the exact matching position, but also consider the probability distribution of neighboring positions, and comprehensively consider it through the weighted summation method. The specific method is to take all the grid points within a certain radius (usually 2) centered on the signal mapping position, and perform weighted summation on the probability values of each category at each grid point. The weight decreases with the increase of the distance, and the Gaussian weight function is usually used. The calculated result is a probability vector, and each element corresponds to the probability of a UAV type, reflecting the possibility of category attribution based on the spatial position. This neighborhood weighting process helps to smooth local fluctuations and improve the stability of classification.

[0044] The spatial mapping probability value and the classification probability value in the UAV signal recognition result are weighted and fused according to a dynamically adjusted weight coefficient to obtain a comprehensive recognition probability. This step fuses the classification result of the convolutional neural network and the spatial clustering result of the self-organizing mapping, and the two complement each other. The classification probability value in the UAV signal recognition result is the probability distribution output by the Softmax layer of the convolutional neural network, while the spatial mapping probability value is the class distribution inferred based on the position of the signal in the feature space. The two probabilities are fused by weighted averaging, and the weight coefficient is dynamically adjusted according to the signal-to-noise ratio. When the signal-to-noise ratio is high, the weight of the classification result of the convolutional neural network is larger, for example, the weight is 0.7; when the signal-to-noise ratio is low, the weight of the spatial mapping result increases, for example, the weight of the convolutional neural network drops to 0.3. The fusion formula is: comprehensive probability = α × convolutional neural network probability + (1 - α) × spatial mapping probability, where α is the dynamically adjusted weight coefficient. The fused comprehensive recognition probability combines the advantages of the two methods, making use of both the sensitivity of the convolutional neural network to features and the ability of the self-organizing mapping to capture spatial distributions.

[0045] Apply the state transition matrix to the comprehensive recognition probability for time series smoothing, and combine the recognition results of historical frames for voting statistics to obtain the UAV type recognition result. This step takes into account the temporal continuity of the UAV signal and improves the stability of recognition through time series smoothing. The state transition matrix describes the transition probability from the UAV type in the current frame to the type in the next frame, usually obtained by statistical analysis of historical data. The method of applying the state transition matrix is to adjust the comprehensive recognition probability of the current frame according to the recognition result and state transition probability of the previous frame, strengthening the category judgment that is continuous in time. Time series smoothing also includes performing sliding window voting statistics on the recognition results of multiple consecutive frames. The window size is usually 5 to 10 frames, and the category with the highest frequency within the window is taken as the result. This processing in the time dimension effectively reduces instantaneous interference and misjudgment, improves the overall stability and reliability of recognition, and the output UAV type recognition result more accurately reflects the actual type of the detected target.

[0046] In a specific embodiment, the process of performing step S105 may specifically include the following steps: Assign roles of master node, relay node, and edge node to the sensor nodes, and establish communication links between nodes through a hierarchical network topology structure to obtain a multi-level relay network architecture; Perform precise time protocol synchronization processing on each node in the multi-level relay network architecture to keep the clock deviation of each node within a specified range, obtaining a time-synchronized node network; Perform trilateration calculations on the UAV type recognition result based on the geographical locations and signal reception strengths of each node to obtain the signal source position coordinates; The feature vectors extracted by each node are transmitted to the upper-level node through a multi-level relay network architecture for feature weighted averaging and dimensionality reduction processing to obtain a fused feature vector; The trust and uncertainty of the independent recognition results of each node are calculated through the evidence theory framework, and the weights are dynamically adjusted according to the signal-to-noise ratio and distance to obtain a comprehensive recognition result; The position coordinates of the signal source are associated and matched with the comprehensive recognition result, and trajectory association is performed through a tracking filter to obtain the position and identity information of the UAV.

[0047] Specifically, the role assignment of sensor nodes as master nodes, relay nodes, and edge nodes, and the establishment of communication links between nodes through a hierarchical network topology structure to obtain a multi-level relay network architecture is the first step in constructing a collaborative detection network. Sensor nodes refer to radio signal receiving devices distributed in the monitoring area, which are divided into three roles according to their functions and locations: master nodes are responsible for global data fusion and decision-making, usually configured with higher computing capabilities and storage resources, and are located at the top layer of the network; relay nodes are responsible for data forwarding and regional fusion, are located in the middle layer of the network, and play the role of a communication bridge between upper and lower layer nodes; edge nodes focus on local signal acquisition and preliminary processing, directly perceive the environment, and are located at the bottom layer of the network. The role assignment is based on the hardware performance, geographical location, and coverage of the nodes, and is determined by evaluating factors such as the computing power, storage capacity, communication bandwidth, and power persistence of the nodes. The hierarchical network topology structure is a tree-like structure, where edge nodes are connected to relay nodes, and relay nodes are connected to master nodes, forming a multi-level data aggregation path. The communication links between nodes use secure encrypted wireless communication technologies, such as dedicated encrypted WiFi or low-power wide-area network technologies, to ensure the security and reliability of data transmission. Precise time protocol synchronization processing is performed on each node in the multi-level relay network architecture to keep the clock deviation of each node within the specified range, obtaining a node network with time synchronization. Time synchronization is the basis of collaborative detection because accurate timestamps are crucial for the association of multi-source signals and target positioning. The precise time protocol is a high-precision time synchronization method, and the IEEE 1588 Precision Time Protocol (PTP) is used in the UAV detection network. The synchronization process first determines the master clock node, usually selecting the master node as the time reference, and then synchronizes the time information to each level of nodes through a hierarchical transmission method. The synchronization message contains the send timestamp, receive timestamp, and propagation delay estimate. The receiving node calculates the deviation between its own clock and the master clock through this information and makes corresponding adjustments. Considering the uncertainty of wireless communication, the method of taking the average of multiple measurements is used to reduce random errors, and an outlier detection mechanism is implemented to eliminate significantly incorrect synchronization data. Through precise time protocol synchronization processing, the clock deviation of each node is controlled within 100 nanoseconds, meeting the accuracy requirements for UAV signal association and positioning.

[0048] Trilateration calculation is performed on the UAV type recognition results based on the geographical locations and signal reception strengths of each node to obtain the position coordinates of the signal source. Trilateration is a distance-based positioning method that determines the spatial position of a target by measuring the distances from the target to three or more nodes at known locations. In radio detection, it is difficult to directly measure the distance, and it is usually estimated indirectly through the signal reception strength. First, according to the radio propagation model, the signal strength decays as the distance increases, and the decay follows the inverse square law or a more complex model that takes into account path loss, shadowing effect, and multipath effect. The received signal strength is related to the transmit power, antenna gain, propagation distance, and environmental factors. By reverse calculation, the distance from the transmitting source to the receiving point is deduced based on the received signal strength. When three or more nodes receive the same signal simultaneously and calculate their respective estimated distances, the position of the signal source can be determined by solving a geometric equation system. The specific calculation uses the weighted least squares method, where the weights are proportional to the signal strength and the historical positioning accuracy of the nodes to reduce the impact of measurement errors. The finally obtained UAV position coordinates include three dimensions: longitude, latitude, and altitude, and the accuracy is closely related to the node deployment density and signal conditions.

[0049] The feature vectors extracted from each node are transmitted to the upper-level node through a multi-level relay network architecture for feature weighted averaging and dimensionality reduction processing to obtain the fused feature vector. The feature vector is the UAV feature extracted by each node based on the locally received signal, including frequency domain features, time domain features, modulation features, and protocol features, etc. The feature vector is uploaded through the multi-level relay network. The edge node sends the feature vector to the relay node it is connected to. The relay node collects the feature vectors of the lower-level nodes, performs preliminary fusion, and then uploads them to the master node. Feature weighted averaging is a feature-level fusion method that calculates the weighted average of the same type of features provided by different nodes according to the weights. The weight coefficients are determined according to the signal-to-noise ratio, signal integrity, and historical recognition accuracy of the nodes. Nodes with high signal-to-noise ratio, high signal integrity, and high historical accuracy obtain higher weights. Feature dimensionality reduction is the process of extracting key information from a high-dimensional feature space. Commonly used methods include principal component analysis (PCA) or linear discriminant analysis (LDA), etc. It reduces the data dimension while retaining the main feature variations to improve the efficiency of subsequent processing. The fused feature vector synthesizes the observation information of multiple nodes and can more comprehensively reflect the characteristics of the UAV than the single-node feature.

[0050] For the independent recognition results of each node, the belief degree and uncertainty are calculated through the evidence theory framework, and the weights are dynamically adjusted according to the signal-to-noise ratio and distance, resulting in a comprehensive recognition result. The evidence theory framework is a method for dealing with uncertainty and fusing multi-source information, which is particularly suitable for processing the judgment results of different nodes on the same target. In UAV recognition, each node obtains the judgment result of the UAV type based on local observations, including category judgment and confidence. In evidence theory, the basic probability assignment function (BPA) is used to represent the support degree of the node for each category. The belief degree represents the direct support for a certain category, and the uncertainty represents the reserved judgment when the evidence is insufficient. The BPA function is defined as follows: ; where is the basic probability assignment of node i to category , is the signal-to-noise ratio of node i, is the distance from node to the target, is the distance influence factor, is the probability that node i judges the target belongs to category . The denominator is the normalization term of the weighted sum of all nodes, N is the total number of nodes participating in the fusion, is the distance from node to the target.

[0051] When multiple nodes have different judgments on the same target, the BPAs of each node are fused through the Dempster combination rule to obtain the comprehensive BPA: ; Here and are the BPAs of different nodes respectively, and the summation range is all set combinations with an intersection of . The summation range in the denominator is the combination with an empty intersection, representing the degree of conflict. In this way, by comprehensively considering the judgments of each node, a more reliable UAV type recognition result is obtained.

[0052] Associate the signal source position coordinates with the comprehensive recognition result, perform trajectory association through a tracking filter, and obtain the UAV position and identity information. Association matching is the process of binding the positioning result and the recognition result to ensure that the position and identity information correspond to the same target. When there are multiple targets in the environment, association matching becomes particularly important. The basic principle of association is spatio-temporal consistency, that is, within the same time window, signals at similar positions are likely to come from the same target. A tracking filter is a recursive estimation method used to extract the true state of a target from noisy observation data. In UAV detection, the Kalman filter or particle filter is commonly used. The tracking process first establishes a UAV motion model to describe the dynamic change rules of state variables such as position, velocity, and acceleration, and then compares the new observation data with the model prediction value to update the state estimate. Tracking filtering not only smooths the observation noise but also provides a prediction of the target's motion trend, facilitating the early judgment of potential threats. Through continuous tracking, the motion trajectory of the UAV is formed. Combining with the identity information, it comprehensively describes the activity characteristics of the target UAV, providing a basis for subsequent threat assessment and defense decision-making.

[0053] In a certain UAV detection experiment, 1 master node, 3 relay nodes, and 8 edge nodes were deployed to form a three-level relay network architecture. Time synchronization of all nodes was achieved through the IEEE 1588 Precision Time Protocol, and the clock deviation was controlled within 75 nanoseconds. When a UAV entered the monitoring area, 5 edge nodes simultaneously received the signal, and the signal strengths were -65dBm, -68dBm, -72dBm, -75dBm, and -80dBm respectively. The distance from the node to the UAV was estimated according to the radio propagation model, and the UAV position coordinates were calculated through trilateration. At the same time, the feature vectors extracted by the 5 nodes were uploaded to the master node through the relay nodes. After weighted average and PCA dimensionality reduction processing, a fused feature vector was obtained. Each node independently gave a judgment on the UAV type. Among them, 3 nodes judged it as model A, with confidence levels of 0.85, 0.78, and 0.72 respectively; 1 node judged it as model B, with a confidence level of 0.65; 1 node judged it as model C, with a confidence level of 0.60. Applying the evidence theory framework, considering the signal-to-noise ratios of each node (22dB, 20dB, 18dB, 15dB, and 12dB respectively) and the distance, the BPA was calculated and fused. The final comprehensive recognition result was model A, with a confidence level of 0.89. The position coordinates were associated with the recognition result, and trajectory tracking was performed through the Kalman filter to finally obtain the UAV position and identity information.

[0054] In a specific embodiment, the process of executing step S106 may specifically include the following steps: Perform weighted calculation based on the UAV position and identity information in combination with the UAV type danger coefficient, proximity, size, approach speed, and regional security level to obtain a threat index; Threshold division is performed on the threat index, and the threat levels are divided into four levels: normal, attention, warning, and emergency, to obtain the judgment result of the early warning level; According to the judgment result of the early warning level, the scanning frequency of the monitoring area is adjusted, and notifications with different priorities are sent to relevant personnel to obtain preliminary response measures; Based on the judgment result of the early warning level, corresponding detection devices are activated through a multi-level relay network for target confirmation to obtain target confirmation information; Track tracing is performed on the position of the drone, and position and intention prediction are carried out through Kalman filtering to obtain the track prediction result; The target confirmation information, the track prediction result, and the judgment result of the early warning level are combined for response resource allocation to obtain a defense response strategy.

[0055] Specifically, weighted calculation is performed by combining the drone's position and identity information with the drone type risk coefficient, proximity, size, approach speed, and regional security level to obtain the threat index, which is a key step in realizing intelligent early warning. The threat index is a quantitative indicator that measures the degree of threat posed by a drone to a specific area and is calculated by comprehensively considering multiple factors. The drone type risk coefficient is a pre-defined weight value based on the drone's model, use, and potential harmfulness. For example, the risk coefficient of a military drone is higher than that of a civilian drone, and the risk coefficient of a drone equipped with a camera device is higher than that of an ordinary toy drone. The proximity refers to the shortest distance between the drone and the boundary of the sensitive area. The closer the distance, the greater the threat. The size of the drone refers to its physical size or weight. Large drones usually have stronger payload capabilities and potential destructive power. The approach speed refers to the speed component of the drone towards the sensitive area. The faster the speed, the less time left for the response system, and the greater the threat. The regional security level is the pre-defined importance of different regions. For example, key infrastructure such as nuclear facilities and government buildings has a higher security level. These factors are weighted and calculated to obtain the comprehensive threat index, and the calculation formula is: ; Among them, represents the threat index (Threat Index), and its value range is [0, 1]; represents the drone type risk coefficient (Drone Category), usually preset within the range of [0, 1]; represents the proximity (ApproachProximity), which is obtained by converting the actual distance into a normalized value in the range of [0, 1]. The closer the distance, the larger the value; represents the size of the drone (Size), which is also normalized to the range of [0, 1]; represents the approach speed (Speed Vector), which is normalized to the range of [0, 1]; Represents the security area level, preset within the range of [0, 1]. The weight coefficient , , , and correspond to the importance weights of each factor respectively, and satisfy + + + + = 1, ensuring that the threat index is between 0 and 1.

[0056] Perform threshold division on the threat index, divide the threat level into four levels: normal, attention, warning, and emergency, and obtain the judgment result of the early warning level. Threshold division is the process of mapping continuous threat index values to discrete early warning levels. Based on practical application experience and security requirements, the threat index range from 0 to 1 is divided into four sub-ranges: 0 to 0.25 is the normal level, indicating no obvious threat; 0.25 to 0.5 is the attention level, indicating the existence of potential threats that need attention; 0.5 to 0.75 is the warning level, indicating the existence of medium threats that require preparation of defense measures; 0.75 to 1.0 is the emergency level, indicating the existence of serious threats that require immediate action. Threshold division is not a simple uniform division, but is adjusted according to actual security requirements and response capabilities. For example, in scenarios with high security requirements, the threshold of the emergency level can be reduced to trigger high-level early warnings in advance. The judgment result of the early warning level is a discrete value, which is directly related to the subsequent response strategy.

[0057] Adjust the scanning frequency of the monitored area according to the judgment result of the early warning level, and send notifications with different priorities to relevant personnel to obtain preliminary response measures. Scanning frequency adjustment is a strategy for dynamically allocating detection resources according to the threat level. Higher threat areas receive higher scanning frequencies to improve detection accuracy and timeliness. The specific adjustment method is: maintain the basic scanning frequency in the normal level area, such as once every 10 seconds; double the scanning frequency in the attention level area, such as once every 5 seconds; quadruple the scanning frequency in the warning level area, such as once every 2.5 seconds; increase the scanning frequency to the highest in the emergency level area, such as once every 1 second. Notification sending is the process of conveying early warning information to relevant personnel, and different levels correspond to different notification methods and recipient groups. The normal level is usually only recorded in the system log without sending active notifications; the attention level sends low-priority notifications to monitoring operators, such as system interface prompts; the warning level sends high-priority notifications to security supervisors and response personnel, such as text messages, emails, etc.; the emergency level sends emergency notifications to all security personnel and management, such as phone calls, alarm sounds, etc., to ensure immediate attention. Preliminary response measures include resource preparation and preliminary defense actions, laying the foundation for subsequent in-depth responses.

[0058] Based on the warning level determination result, corresponding detection devices are activated through a multi-level relay network for target confirmation to obtain target confirmation information. Target confirmation is a cross-verification of the radio detection result, providing supplementary evidence by activating other types of detection devices. Different warning levels trigger different levels of confirmation mechanisms: the attention level may only activate the radio sensing function of adjacent nodes; the warning level, in addition to activating more radio nodes, also activates optoelectronic devices such as visible light cameras and infrared cameras for visual confirmation; the emergency level activates all available resources, including high-precision radars, acoustic detectors, etc. The multi-level relay network plays the role of command transmission and data recovery in this process, sending activation commands from the master node to the detection devices at the corresponding levels and summarizing the detection results upward along the network. The target confirmation information includes the fusion results of multiple sensors, such as the target contour features in the optoelectronic image, radar echo characteristics, etc., combined with the radio features, greatly improving the reliability of the judgment.

[0059] Track the position information of the UAV, predict the position and intention through Kalman filtering, and obtain the trajectory prediction result. Trajectory tracking is a process of continuously recording the position change of the UAV, forming time-series position data. Kalman filtering is a recursive estimation algorithm that optimally estimates the observation data containing random errors and is widely used in the field of target tracking. Kalman filtering includes two main steps: the prediction step predicts the current state based on the state and motion model at the previous moment; the update step corrects the predicted value according to the new observation data. In UAV trajectory tracking, the state variables usually include position, velocity, and acceleration, and the motion model is determined according to the flight characteristics of the UAV, such as a uniform motion model or a uniformly accelerated model. Position prediction is to calculate the possible position of the UAV at a future moment, deduced based on the current state and motion trend. Intention prediction is the inference of the UAV's behavior purpose, such as judging whether the UAV is passing by, hovering for observation, or preparing to invade, achieved by analyzing the trajectory pattern and regional relationship. The trajectory prediction result includes the predicted position coordinates, velocity vector, and possible behavior intention, providing forward-looking information for response decision-making.

[0060] The target confirmation information is combined with the trajectory prediction result and the early warning level determination result to perform response resource allocation, obtaining a defense response strategy. Response resource allocation is a process of reasonably scheduling limited defense resources according to the threat situation, including passive monitoring resources and active defense devices. The target confirmation information provides evidence of the certainty of the current threat, the trajectory prediction result provides the threat development trend, and the early warning level determination result provides the threat severity. The combination of the three forms a complete situation awareness to guide the resource allocation decision. The specific allocation strategy varies according to the early warning level: only basic monitoring resources are allocated at the normal level; the monitoring density is increased at the attention level, and some defense devices are prepared; most defense devices are activated at the warning level, and it enters the preparation state; all defense resources are fully deployed at the emergency level, and it enters the highest alert state. The defense response strategy is a detailed action plan, including the defense measures to be taken, the execution order, the responsible personnel, and the expected effects, ensuring coordinated actions in the face of drone threats.

[0061] The above describes the drone detection method based on radio intelligent fusion in the embodiments of the present application. Next, the drone detection system based on radio intelligent fusion in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the drone detection system based on radio intelligent fusion in the embodiments of the present application includes: An acquisition module, configured to acquire radio signals through distributed relay sensor nodes to obtain digitized I / Q data; An extraction module, configured to perform time-frequency domain conversion and feature extraction on the digitized I / Q data to obtain a set of drone radio signal feature vectors; An input module, configured to input the set of drone radio signal feature vectors into a convolutional neural network for training and recognition to obtain a drone signal recognition result; A recognition module, configured to obtain a drone type recognition result through a clustering signal joint recognition algorithm for the drone signal recognition result; A fusion module, configured to construct a multi-level relay network according to the drone type recognition result to perform collaborative data fusion processing to obtain drone position and identity information; A grading module, configured to calculate a threat index for the drone position and identity information and perform early warning grading based on the relay network to obtain a defense response strategy.

[0062] Through the collaborative cooperation of the above-mentioned various components, digital I / Q data is acquired by collecting radio signals through distributed relay sensor nodes, achieving seamless coverage of the wide-area space and solving the problem of limited coverage of a single sensor; the digital I / Q data is subjected to time-frequency domain conversion and feature extraction to obtain a set of UAV radio signal feature vectors, fully exploiting the multi-dimensional features of UAV signals and improving the richness and discrimination of feature expression; the set of UAV radio signal feature vectors is input into a convolutional neural network for training and recognition to obtain UAV signal recognition results, enabling the model to automatically learn the deep feature patterns of signals, and the recognition ability is no longer limited by manually preset feature rules; the preliminary UAV signal recognition results are processed through a clustering signal joint recognition algorithm to obtain UAV type recognition results, skillfully integrating clustering analysis and pattern recognition technologies, and accurately distinguishing UAV models with similar features, significantly reducing the false recognition rate; a multi-level relay network is constructed according to the UAV type recognition results to perform collaborative data fusion processing to obtain UAV position and identity information, solving the information island problem in traditional single-node detection and achieving effective complementarity and redundant verification of information between different nodes; a threat index is calculated for the UAV position and identity information and an early warning classification based on the relay network is performed to obtain a defense response strategy, constructing a complete intelligent defense link from perception to decision-making, ensuring corresponding defense measures are taken for UAVs with different threat levels. The present invention applies artificial intelligence algorithms and models, especially convolutional neural networks and clustering signal joint recognition algorithms, in a specific UAV detection application field, not only achieving superiority over traditional methods in the feature extraction and recognition stages, but also significantly improving the detection accuracy and anti-interference ability through algorithm features, showing excellent adaptability in complex electromagnetic environments; the collaborative data fusion mechanism of the multi-level relay network fully considers the spatial characteristics of radio signal propagation, utilizes the spatial diversity formed by distributed sensing nodes, and can still maintain stable and reliable detection effects in a strong interference environment; the adaptive early warning classification strategy based on the threat index deeply combines artificial intelligence decision-making technology with professional domain knowledge, forming an intelligent defense system against UAV threats, which has higher pertinence and resource utilization efficiency compared with traditional static defense strategies.

[0063] Referring to Figure 3 , an embodiment of the present invention further provides a computer device, which may be a server, and its internal structure may be as Figure 3As shown in the figure. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0064] Those skilled in the art can understand that Figure 3 the structure shown in the figure is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0065] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0066] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0067] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, systems, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0068] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0069] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

[0070] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0071] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A UAV detection method based on radio intelligent fusion, characterized in that: The UAV detection method based on radio intelligent fusion includes: Collect radio signals through distributed relay sensor nodes to obtain digital I / Q data; Performing time-frequency domain conversion and feature extraction on the digitized I / Q data to obtain a feature vector set of the UAV radio signal; Inputting the drone radio signal feature vector set into a convolutional neural network for training and recognition to obtain a drone signal recognition result; The UAV signal recognition result is obtained by using a clustering signal joint recognition algorithm to obtain a UAV type recognition result; A multi-level relay network is constructed based on the drone type identification result to perform collaborative data fusion processing to obtain the drone location and identity information; The threat index is calculated for the position and identity information of the UAV and a warning classification based on the relay network is performed to obtain a defense response strategy.

2. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The method of collecting radio signals through distributed relay sensor nodes to obtain digital I / Q data includes: The radio signal in the environment is pre-amplified by a low-noise amplifier to obtain an analog radio frequency signal; The analog radio frequency signal is filtered out of out-band interference by a bandpass filter to obtain a filtered radio frequency signal; Down-converting the filtered radio frequency signal using a mixer and a local oscillator to obtain an intermediate frequency or baseband signal; The intermediate frequency or baseband signal is digitized at a high sampling rate through an analog-to-digital converter to obtain an original I / Q data stream; Performing digital downsampling and digital filtering on the original I / Q data stream to obtain noise-reduced I / Q data; The noise reduction I / Q data is marked by adding a timestamp and geographic location information to obtain the digitized I / Q data.

3. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The step of performing time-frequency domain conversion and feature extraction on the digitized I / Q data to obtain a set of feature vectors of the UAV radio signal includes: Performing time-frequency domain conversion on the digitized I / Q data by short-time Fourier transform to obtain a signal spectrogram; Performing power spectrum density calculation and frequency occupied bandwidth analysis on the signal spectrum to obtain frequency domain characteristic parameters; Extracting time domain features from the digitized I / Q data by envelope analysis to obtain time domain features; Performing high-order cumulants and cyclic spectrum feature calculations on the time domain features and the frequency domain feature parameters to obtain modulation recognition parameters; Extracting MAC layer features from the modulation identification parameters and signal message structure to obtain communication protocol features; The frequency domain feature parameters, the time domain features, the modulation identification parameters and the communication protocol features are normalized and weighted to obtain the UAV radio signal feature vector set.

4. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The step of inputting the drone radio signal feature vector set into a convolutional neural network for training and recognition to obtain a drone signal recognition result includes: Normalizing the UAV radio signal feature vector set to obtain training data that conforms to the convolutional neural network input format; The training data is sent into a network structure including five convolutional layers for feature mapping to obtain hierarchical feature expression data; Performing dimension reduction and standardization processing on the hierarchical feature expression data through a maximum pooling layer and a batch normalization layer to obtain compressed feature expression data; Inputting the compressed feature expression data into two fully connected layers for nonlinear transformation to obtain vector representation data in a high-dimensional feature space; The vector representation data in the high-dimensional feature space is subjected to probability distribution calculation by a Softmax function to obtain the probability distribution of each drone type; Based on the probability distribution of each drone type, the category corresponding to the highest probability is selected as the recognition result to obtain the drone signal recognition result.

5. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The method of obtaining the drone type identification result by using the clustering signal joint identification algorithm to identify the drone signal includes: A two-dimensional self-organizing mapping network is configured for the drone signal recognition result, and a network training is performed by setting a batch mode, an initial learning rate, a learning rate decay function, an initial neighborhood radius, and a neighborhood shrinkage function to obtain a topology-preserving mapping network; Input known drone sample features into the topology-preserving mapping network for mapping, count the distribution of various samples on the grid, and obtain a grid position probability distribution map; Input the drone signal recognition result into the topology preserving mapping network, calculate the distance value with each neuron, select the neuron with the smallest distance as the matching unit, and obtain the position of the signal in the feature space; Querying the grid position probability distribution map according to the position of the signal in the feature space, and obtaining a spatial mapping probability value through weighted summation of neighboring regions; The spatial mapping probability value and the classification probability value in the drone signal recognition result are weighted and fused according to a dynamically adjusted weight coefficient to obtain a comprehensive recognition probability; The comprehensive recognition probability is subjected to time series smoothing processing by applying a state transfer matrix, and voting statistics are performed in combination with the historical frame recognition results to obtain the drone type recognition result.

6. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The method of constructing a multi-level relay network according to the drone type identification result to perform collaborative data fusion processing to obtain the drone location and identity information includes: Assign the roles of master node, relay node and edge node to sensor nodes, establish inter-node communication links through hierarchical network topology, and obtain a multi-level relay network architecture; Performing precise time protocol synchronization processing on each node in the multi-stage relay network architecture so that the clock deviation of each node is kept within a specified range, thereby obtaining a time-synchronized node network; Perform trilateral measurement calculation on the drone type identification result according to the geographical location of each node and the signal reception strength to obtain the signal source location coordinates; The feature vectors extracted from each node are transmitted to the upper-level node through the multi-level relay network architecture, and feature weighted averaging and dimensionality reduction processing are performed to obtain a fused feature vector; The trust and uncertainty of each node's independent recognition results are calculated through the evidence theory framework, and the weight is dynamically adjusted according to the signal-to-noise ratio and distance to obtain a comprehensive recognition result. The signal source position coordinates are associated and matched with the comprehensive recognition result, and the trajectory is associated through a tracking filter to obtain the position and identity information of the drone.

7. The method for detecting unmanned aerial vehicles based on radio intelligent fusion according to claim 1 is characterized in that: The step of calculating the threat index of the drone position and identity information and performing early warning classification based on the relay network to obtain a defense response strategy includes: A threat index is obtained by performing a weighted calculation based on the drone location and identity information combined with the drone type danger factor, proximity, size, approach speed and regional safety level; Thresholds are set for the threat index, and the threat level is divided into four levels: normal, attention, warning, and emergency, to obtain a warning level determination result; Adjust the scanning frequency of the monitoring area according to the warning level determination result, and send different priority notifications to relevant personnel to obtain preliminary response measures; Based on the warning level determination result, corresponding detection equipment is activated through a multi-level relay network to perform target confirmation, thereby obtaining target confirmation information; Tracking the position of the UAV, predicting the position and intention through Kalman filtering, and obtaining a trajectory prediction result; The target confirmation information and the trajectory prediction result are combined with the warning level determination result to perform response resource allocation to obtain the defense response strategy.

8. A UAV detection system based on radio intelligent fusion, used to implement the UAV detection method based on radio intelligent fusion as described in any one of claims 1 to 7, characterized in that: The UAV detection system based on radio intelligent fusion includes: A collection module is used to collect radio signals through distributed relay sensor nodes to obtain digital I / Q data; An extraction module, used for performing time-frequency domain conversion and feature extraction on the digitized I / Q data to obtain a feature vector set of the UAV radio signal; An input module, used to input the drone radio signal feature vector set into a convolutional neural network for training and recognition, to obtain a drone signal recognition result; An identification module, used to identify the drone signal through a clustering signal joint identification algorithm to obtain a drone type identification result; A fusion module is used to construct a multi-level relay network to perform collaborative data fusion processing according to the drone type identification result to obtain the drone location and identity information; The classification module is used to calculate the threat index of the drone position and identity information and perform early warning classification based on the relay network to obtain a defense response strategy.

9. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and the processor implements the drone detection method based on radio intelligent fusion as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the processor executes the UAV detection method based on radio intelligent fusion as described in any one of claims 1 to 7.

Images