Radar rotor target classification and identification method, device, equipment and medium

Abstract

The invention relates to the technical field of classification and recognition, and provides a radar rotor target classification and recognition method, device and equipment and a medium, and the method comprises the steps: carrying out the digital processing of a radar rotor signal, uploading the radar rotor signal after pulse compression to a digital signal processing unit, extracting signal features and track features corresponding to the radar rotor signals in the digital signal processing unit, and identifying a first classification result of the rotor unmanned aerial vehicle according to the signal features and the track features; extracting a micro-motion feature of the radar rotor signal, and extracting a frequency band corresponding to the micro-motion feature through a band-pass filter; extracting a time-frequency graph corresponding to the frequency band uploaded to the radar terminal storage device, and identifying the time-frequency graph to obtain a second classification result of the rotor unmanned aerial vehicle; and fusing the first classification result and the second classification result to obtain a target classification recognition result corresponding to the rotor unmanned aerial vehicle. According to the invention, the accuracy of rotor unmanned aerial vehicle classification and identification can be improved.

Description

Technical Field

[0001] The present invention relates to the field of classification and recognition technology, and in particular to a radar rotor target classification and recognition method, device, equipment and medium. Background Art

[0002] In recent years, micro-motion characteristics have received widespread attention in radar target detection and recognition. Micro-motion refers to the vibration or rotation of the radar target other than the translational motion of the center of mass. The Doppler frequency generated by micro-motion is the micro-Doppler frequency. Rotary-wing small unmanned aerial vehicles have a simple physical structure, low flight altitude, strong maneuverability, wide application, and great threat. How to quickly and accurately identify rotary-wing drones is one of the current research hotspots.

[0003] The existing classification and recognition of rotorcraft UAV targets uses feature combination to build a target classification and recognition library, and realizes target classification and recognition based on deep learning and other means. However, it requires the fusion of radar signal level and track level features, which has a long time problem and cannot better meet the needs of rapid or combat applications. At the same time, the characteristics of rotorcraft UAVs and flying birds are extremely similar, which reduces the probability of rotorcraft UAV classification and recognition, resulting in low accuracy in rotorcraft classification and recognition. Summary of the Invention

[0004] In response to the above problems, embodiments of the present invention provide a radar rotor target classification and identification method, apparatus, device, and medium.

[0005] In a first aspect, an embodiment of the present invention provides a radar rotor target classification and recognition method, comprising:

[0006] Collecting radar rotor signals of the rotary-wing UAV, digitizing the radar rotor signals, and performing pulse compression on the digitized single-frame radar rotor signals;

[0007] Uploading the pulse-compressed radar rotor signal to a digital signal processing unit, extracting signal features corresponding to the radar rotor signal in the digital signal processing unit, extracting track features of the rotary-wing UAV, and identifying a first classification result of the rotary-wing UAV based on the track features and the signal features;

[0008] Extracting micro-motion features of the radar rotor signal, extracting a frequency band corresponding to the micro-motion features through a preset bandpass filter, and uploading the frequency band to a radar terminal storage device using a preset bandwidth transmission;

[0009] Extracting a time-frequency graph corresponding to a frequency band uploaded to a radar terminal storage device through a preset short-time Fourier transform, and identifying the time-frequency graph using a trained deep learning model to obtain a second classification result of the rotary-wing UAV;

[0010] The first classification result and the second classification result are integrated to obtain a target classification recognition result corresponding to the rotary-wing UAV.

[0011] According to an embodiment of the present invention, the identifying a first classification result of the rotary-wing UAV according to the track feature and the signal feature includes:

[0012] Performing feature fusion on the track feature and the signal feature to obtain a fusion feature;

[0013] Using a pre-built classification model to classify and identify the fused features, and obtain a first classification identifier of the rotary-wing UAV;

[0014] A first classification result of the rotary-wing UAV is determined according to the first classification identifier.

[0015] According to an embodiment of the present invention, extracting the micro-motion feature of the radar rotor signal includes:

[0016] Acquiring signal frame data corresponding to the radar rotor signal;

[0017] Determining distance data of the rotary-wing UAV based on the signal frame data;

[0018] Extracting the Doppler characteristics of the radar rotor signal using a pre-built echo baseband model;

[0019] The micro-motion characteristics of the radar rotor signal are determined according to the distance data and the Doppler characteristics.

[0020] According to an embodiment of the present invention, extracting the frequency band corresponding to the micro-motion feature through a preset bandpass filter includes:

[0021] Filtering the radar rotor signal through the bandpass filter to obtain a rotor filtered signal;

[0022] performing a frequency spectrum analysis on the rotor filter signal to obtain a target frequency band;

[0023] determining whether the target frequency band includes the micro-motion feature;

[0024] When the target frequency band includes the micro-motion feature, the target frequency band is used as the frequency band corresponding to the micro-motion feature.

[0025] According to an embodiment of the present invention, the method of using a trained deep learning model to identify the time-frequency graph to obtain a second classification result of the rotary-wing drone includes:

[0026] Extracting the time-frequency graph in the radar terminal storage device according to a preset data period;

[0027] The trained deep learning model is injected into the preset radar equipment terminal backup;

[0028] Using a deep learning model in a radar device terminal backup to classify and identify the time-frequency graph, a second classification identifier of the rotary-wing UAV is obtained;

[0029] A second classification result of the rotary-wing UAV is determined according to the second classification identifier.

[0030] According to an embodiment of the present invention, fusing the first classification result and the second classification result to obtain a target classification recognition result corresponding to the rotary-wing UAV includes:

[0031] extracting a first rotor weight value from the first classification result, and extracting a second rotor weight value from the second classification result;

[0032] When the first classification result is a first identifier and the second classification result is a first identifier, determining a first rotor confidence value according to the first rotor weight value and the second rotor weight value, and determining that a target classification recognition result corresponding to the rotor UAV is a rotor according to the first rotor confidence value;

[0033] When the first classification result is a first identifier and the second classification result is not the first identifier, determining a second rotor confidence value according to the first rotor weight value, and determining, according to the second rotor confidence value, that the target classification recognition result corresponding to the rotor UAV is a target type rotor;

[0034] When the first classification result is not the first identification and the second classification result is the first identification, the third rotor confidence is determined according to the second rotor weight value, and the target classification identification result corresponding to the rotor UAV is determined to be a target type rotor according to the third rotor confidence.

[0035] According to an embodiment of the present invention, before extracting the frequency corresponding to the micro-motion feature through a preset bandpass filter, the method further includes:

[0036] Determining the center frequency and bandwidth of the bandpass filter according to the frequency range of the micro-motion characteristics;

[0037] Selecting a filter type of a bandpass filter according to the center frequency and the bandwidth;

[0038] The bandpass filter is configured by using the filter type and filter parameters corresponding to the filter type.

[0039] In a second aspect, an embodiment of the present invention provides a radar rotor target classification and identification device, characterized by comprising:

[0040] A radar rotor signal acquisition module is used to collect radar rotor signals of the rotary-wing UAV, digitize the radar rotor signals, and perform pulse compression on the digitized single-frame radar rotor signals;

[0041] a first classification result identification module, configured to upload the pulse-compressed radar rotor signal to a digital signal processing unit, extract signal features corresponding to the radar rotor signal in the digital signal processing unit, extract track features of the rotary-wing UAV, and identify a first classification result of the rotary-wing UAV based on the track features and the signal features;

[0042] a frequency band extraction module, configured to extract the micro-motion characteristics of the radar rotor signal, extract the frequency band corresponding to the micro-motion characteristics through a preset bandpass filter, and upload the frequency band to the radar terminal storage device using a preset bandwidth transmission;

[0043] A second classification result recognition module is used to extract a time-frequency graph corresponding to the frequency band uploaded to the radar terminal storage device through a preset short-time Fourier transform, and use a trained deep learning model to recognize the time-frequency graph to obtain a second classification result of the rotary-wing UAV;

[0044] The classification result fusion module is used to fuse the first classification result and the second classification result to obtain the target classification recognition result corresponding to the rotary-wing UAV.

[0045] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the radar rotor target classification and identification method described in the above aspect.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the radar rotor target classification and identification method described in the above aspects.

[0047] Compared with the prior art, the above technical solution of the present invention has the following beneficial effects:

[0048] The embodiments of the present invention utilize large bandwidth transmission to store and report signals and outputs, creating innovative data resources for radar performance improvement. They focus on the accurate classification of rotary-wing drones, collect time-frequency images of rotary-wing drones in different postures to establish a training library for the image level of rotary-wing drones, which is conducive to the accurate classification of rotary-wing drones. A deep learning method based on time-frequency images is used to establish a stable and reliable learning model. Information communication is established between the radar terminal and the DSP. Based on the target information set by the radar terminal, the DSP uploads the pulse-compressed data by region and time period, reducing the amount of data. Considering the design of a bandpass filter for the rotary-wing drone frequency band for each frame of reported data, frequency analysis and image recognition are more accurate, and data transmission is easier. A fusion classification and discrimination model is designed, and by combining conventional classification and recognition with time-frequency image classification and recognition results, the accuracy and reliability of the rotary-wing target classification and recognition results are further improved. Therefore, the radar rotary-wing target classification and recognition method, device, equipment, and medium proposed in the present invention can solve the problem of low accuracy when performing rotary-wing drone classification and recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 The flowchart of the radar rotor target classification and recognition method according to the first embodiment of the present invention is shown;

[0051] Figure 2 A schematic diagram showing the position of a frame where a target is located according to the first embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a quadrotor drone according to a first embodiment of the present invention is shown;

[0053] Figure 4a A schematic diagram showing a left single-rotor model of a rotary-wing UAV according to the first embodiment of the present invention is shown;

[0054] Figure 4b A schematic diagram showing a right dual-rotor model of a rotary-wing UAV according to the first embodiment of the present invention is shown;

[0055] Figure 5 A three-dimensional schematic diagram showing the MTD detection results of 223,347 frames of a rotary-wing UAV according to the first embodiment of the present invention;

[0056] Figure 6 A schematic diagram showing the display of rotor frequency spectrum characteristics according to the first embodiment of the present invention is shown;

[0057] Figure 7a A schematic diagram showing an amplitude-frequency response curve of the first embodiment of the present invention;

[0058] Figure 7b A schematic diagram showing a phase frequency response curve of the first embodiment of the present invention;

[0059] Figure 7c A schematic diagram showing a time domain signal after bandpass filtering according to the first embodiment of the present invention is shown;

[0060] Figure 7d A schematic diagram showing a frequency signal after bandpass filtering according to the first embodiment of the present invention is shown;

[0061] Figure 8a A schematic diagram showing the STFT results at a sampling rate of 100 MHz according to the first embodiment of the present invention;

[0062] Figure 8b A schematic diagram showing the STFT results at a sampling rate of 200 MHz according to the first embodiment of the present invention;

[0063] Figure 9 A schematic diagram showing the flow of a convolutional neural network model according to the first embodiment of the present invention is shown;

[0064] Figure 10 A schematic diagram showing the process of fusion of classification results according to the first embodiment of the present invention is shown;

[0065] Figure 11 A functional module diagram of a radar rotor target classification and identification device according to a third embodiment of the present invention is shown;

[0066] Figure 12 A schematic diagram of the composition structure of an electronic device for implementing the radar rotor target classification and recognition method according to embodiment 4 of the present invention is shown. DETAILED DESCRIPTION

[0067] The present disclosure will be further described below with reference to the embodiments shown in the accompanying drawings.

[0068] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0069] This paper proposes a radar rotor target classification and recognition method based on classification and recognition technology. Based on deep learning theory and combined with multi-class classification methods, a three-dimensional deep neural network is constructed to achieve classification and recognition of rotorcraft drone targets. Compared with traditional methods, this radar rotor target classification and recognition technology is more efficient and reduces human subjectivity. It has great potential and application prospects in the classification and recognition of rotorcraft drones.

[0070] Example 1

[0071] like Figure 1 As shown, the present invention proposes a radar rotor target classification and recognition method, comprising the following steps:

[0072] S1. Collect radar rotor signals of a rotary-wing UAV, digitize the radar rotor signals, and perform pulse compression on the digitized single-frame radar rotor signals.

[0073] In the embodiment of the present invention, the radar rotor signal refers to signal-level data containing different postures (more than 3) of the rotor UAV. Since the AD signal transmits a large amount of data, a control button for the target of interest is established on the terminal, and the position information of interest (azimuth θ-distance R) and speed are sent to the DSP. After receiving the instruction, the DSP collects signals within the azimuth area [θ-5°, θ+5°] and the distance area [R-100, r+100]. The radar is installed in a suitable position to ensure that the rotor signal can be effectively captured.

[0074] In detail, the collected radar rotor signal of the rotorcraft is converted into a digital-to-analog conversion, that is, the collected analog signal is converted into a digital signal. First, the analog signal of the rotor is obtained through the radar equipment, and the analog signal is sampled using an analog-to-digital converter (ADC). The sampled signal is converted into a discrete digital value, and the digital value represents the amplitude of the analog signal. The quantized signal value is converted into a digital code, and the quantized digital data is stored in a data buffer or memory to form a frame of data, that is, AD single-frame data. AD single-frame data refers to a frame of digital data collected by the analog-to-digital converter (ADC). The AD converter converts the continuous analog signal into a discrete digital signal. Each frame of data usually represents the state of the analog signal at a time point or a time period, thereby converting the analog signal of the radar rotor signal into a single-frame digital signal corresponding to the radar rotor signal. In order to significantly improve the time resolution of the radar acquisition signal, it is necessary to perform pulse compression on the digitized single-frame radar rotor signal to make the detection of weak targets more reliable.

[0075] Specifically, a matched filter is designed based on the transmitted pulse of the radar system. The matched filter is a time-domain inverted version of the transmitted pulse, which is used to improve the resolution in pulse compression. The matched filter is then convolved with a single frame of data in the time domain, or the data and the matched filter are converted to the frequency domain, multiplied, and then an inverse Fourier transform is performed to obtain the signal after pulse compression, thereby obtaining the radar rotor signal after the single-frame radar rotor signal pulse.

[0076] Furthermore, the collected radar rotor signal can be processed in two channels. Normal conventional means are first used to detect and extract the signal characteristics and track characteristics corresponding to the radar rotor signal. Then, the result of whether it is a rotor UAV is determined based on the signal characteristics and track characteristics.

[0077] S2. Upload the pulse-compressed radar rotor signal to a digital signal processing unit, extract the signal features corresponding to the radar rotor signal in the digital signal processing unit, and extract the track features of the rotorcraft, and identify the first classification result of the rotorcraft based on the track features and the signal features.

[0078] In an embodiment of the present invention, when performing dual-channel processing, the pulse-compressed radar rotor signal is first uploaded to the digital signal processing unit (DSP) based on channel one, that is, first ensure that the pulse-compressed signal meets the input format requirements of the DSP, such as data type and sampling rate, and then transmit the compressed signal to the DSP unit through a suitable interface (such as serial bus, Ethernet, etc.), and then process the signal in the DSP.

[0079] Furthermore, the received radar rotor signal is normally processed by MTI+MTD+CFAR+FFT in the DSP. Then, MIT (Moving Target Indication) is used as a signal processing technology to identify and extract moving targets. MTD (Moving Target Detection) detects moving targets and extracts useful information from them. CFAR (Constant False Alarm Rate) is used to maintain a constant false alarm rate in background noise to improve the reliability of target detection. FFT (Fast Fourier Transform) converts the signal from the time domain to the frequency domain, making the analysis of frequency components faster and more efficient.

[0080] In an embodiment of the present invention, the signal characteristics refer to the amplitude change of the radar rotor signal, such as acceleration or deceleration, the spectral characteristics of the signal, the phase characteristics of the signal, and the amplitude intensity (Amplitude Intensity): the intensity of the speed or acceleration change; the trajectory characteristics refer to the speed (Speed) of the rotor UAV at each moment during the flight process, the altitude (Altitude) change in flight altitude, the energy ratio (Energy Ratio), the ratio of energy change, which reflects the flight efficiency, the acceleration (Acceleration) acceleration at each moment during the flight process, the trajectory fitting characteristics (Trajectory Fitting Features): such as the curve fitting error, curvature, etc., the number of scatter points (Number of Data Points) the number or density of data points recorded in the trajectory, wherein MTI+MTD+CFAR+FFT can be applied in the digital signal processing unit (DSP) to extract the signal characteristics and trajectory characteristics corresponding to the radar rotor signal.

[0081] Furthermore, conventional classification and identification methods are used, that is, attribute characteristics including speed, altitude, amplitude intensity, energy ratio, acceleration, track fitting characteristics, number of scattered points, etc. are used to classify and identify the track, and the result of whether it is a rotary-wing UAV can be determined. The rotary-wing target is marked as 1, and the non-rotor-wing target is marked as 0.

[0082] In an embodiment of the present invention, the first classification result refers to the type of classification and identification of the rotary-wing UAV, and the classification types include rotary-wing target (identified as 1), fixed-wing target (identified as 2), flying bird target (identified as 3), and not a rotary-wing target (identified as 0).

[0083] In an embodiment of the present invention, the identifying a first classification result of the rotary-wing UAV according to the track characteristics and the signal characteristics includes:

[0084] Performing feature fusion on the track feature and the signal feature to obtain a fusion feature;

[0085] Using a pre-built classification model to classify and identify the fused features, and obtain a first classification identifier of the rotary-wing UAV;

[0086] A first classification result of the rotary-wing UAV is determined according to the first classification identifier.

[0087] In detail, the UAV's track information, such as motion trajectory, speed, and acceleration, is extracted from the radar, and the frequency domain, time domain, and time-frequency domain features in the rotor signal, such as amplitude and frequency spectrum, are analyzed. The trajectory features are fused with the signal features to generate comprehensive fusion features. Feature fusion can be achieved through splicing, and then through a pre-trained classification model (such as support vector machine, decision tree model), a large amount of labeled rotor UAV data is collected to ensure that the data contains track features and signal features. The pre-processed fusion feature data is then used to train the model, and the model parameters are optimized to improve the classification accuracy.

[0088] Specifically, the new fusion feature data is input into the trained classification model. The model performs classification based on the fusion features and outputs the first classification identification of the rotorcraft UAV. Based on the output of the classification model, the first classification identification of the rotorcraft UAV is obtained, such as rotor target (identification 1), fixed rotor target (identification 2), flying bird target (identification 3), and not rotor target (identification 0).

[0089] Furthermore, the result of determining whether it is a rotary-wing drone solely through conventional classification may be inaccurate. Therefore, it is necessary to use channel 2 to further determine the classification type of the rotary-wing drone based on the radar rotor signal, thereby achieving accurate classification and identification of the rotary-wing drone based on radar equipment.

[0090] S3. Extract the micro-motion feature of the radar rotor signal, extract the frequency band corresponding to the micro-motion feature through a preset bandpass filter, and upload the frequency band to the radar terminal storage device using a preset bandwidth transmission.

[0091] In an embodiment of the present invention, the micro-motion feature refers to the vibration or rotation of the radar target other than the translational motion of the center of mass. The Doppler frequency generated by the micro-motion is the micro-Doppler frequency. Different micro-motions will produce different micro-Dopplers. The micro-Doppler effect can reflect the geometric composition and motion characteristics of the target structural components and is an essential characteristic of the target. The micro-Doppler effect is mainly caused by the rotation and vibration of the rotor, and is characterized by subtle changes in the signal frequency. The rotation frequency is analyzed, which is usually manifested as an obvious frequency component in the spectrum. The vibration mode is used to identify the vibration characteristics of the rotor or drone structure, which may appear as higher-frequency details in the spectrum.

[0092] In an embodiment of the present invention, extracting the micro-motion feature of the radar rotor signal includes:

[0093] Acquiring signal frame data corresponding to the radar rotor signal;

[0094] Determining distance data of the rotary-wing UAV based on the signal frame data;

[0095] Extracting the Doppler characteristics of the radar rotor signal using a pre-built echo baseband model;

[0096] The micro-motion characteristics of the radar rotor signal are determined according to the distance data and the Doppler characteristics.

[0097] In detail, the frame and two-dimensional information (distance and Doppler) of the detection target are confirmed by multi-frame data (greater than 10 frames). The signal frame data includes echo signals at multiple time points, which records the signal strength and frequency information reflected by the drone. Then, by calculating the distance of the signal frame data, the echo signal corresponding to each signal frame data is Fourier transformed to calculate the distance of the detection target, such as Figure 2 As shown in the figure, it is a schematic diagram of the position of the frame where the target is located. The horizontal axis represents the frame number where the detected target is located, and the vertical axis represents the distance value corresponding to each frame number. In this way, the relationship mapping diagram between different frame data and the distance of the detected target can be observed, and the pre-built echo baseband model is used to process the radar signal to extract the Doppler frequency characteristics. The echo baseband model describes the micro-motion characteristics of the collected signal.

[0098] Specifically, each single rotor of a quadcopter drone has two blades (two blades are equally divided into one circle, and three blades are divided into three equal parts into one circle). Figure 3 As shown in the figure, it is a schematic diagram of a quadrotor drone. The initial rotation angles of the blades always differ by 2π / N, where N represents the number of blades. The initial rotation angle of the kth blade can be expressed as Then the corresponding N blade total echo signal baseband model is The phase function is Among them S N (t) is the echo signal of the Nth blade at time t, S L (t) is the echo signal of the Lth blade at time t, λ is the wavelength, R0 is the initial distance, v is the speed, β is the angle, w is the angular velocity, j is the imaginary unit, then based on the above derivation process, the echo baseband model based on multiple M rotors and multiple N blades can be expressed as The phase function is Based on the echo baseband model theory, the rotor vibration can be seen, but the actual collected signal may not necessarily show micro-multiple changes. Other processing is required to see micro-multiple changes. Therefore, the micro-multiple features corresponding to the collected signal should be determined based on the echo baseband model. The rotor itself will change, so its own w m t will also have phase changes, so it is necessary to extract the micro-features in the collected signal.

[0099] Furthermore, if Figure 4aAs shown in FIG, it is a schematic diagram of the left single rotor model of the rotary wing UAV, which shows the corresponding relationship between time and time domain signal, that is, as time increases, it reflects the change of the time domain signal of the rotary wing UAV; Figure 4b The figure shows the schematic diagram of the right double-rotor model of the rotor UAV, which shows the corresponding relationship between time and time signal, that is, as time increases, the change of the time domain signal of the rotor UAV is reflected. Figure 4a and Figure 4b For example, at 0.01s, the time domain signals corresponding to the left single rotor and the right double rotor are inconsistent, that is, Figure 4b The corresponding time domain signal is compared with Figure 4a The micro-motion characteristics of the radar rotor signal are extracted based on the echo signal baseband model, that is, the micro-Doppler characteristics and time-frequency characteristics of the rotor UAV are theoretically analyzed, such as Figure 5 As shown in Figure 1, it is a three-dimensional schematic diagram of the MTD detection results of the rotorcraft UAV with 223347 frames. The horizontal axis represents the channel, the vertical axis represents the distance, and the vertical axis represents the amplitude, that is, the distance and amplitude of the rotorcraft UAV are reflected on different channels. By representing the receiving channel of the radar system, the data involving multipath reflection or multiple receivers can be used to observe the micro-Doppler characteristics and time-frequency characteristics of the rotorcraft UAV on different channels.

[0100] Furthermore, for the analysis and confirmation of the micro-Doppler characteristics and time-frequency characteristics of the rotary-wing UAV, since the overall frequency of the fuselage and rotor of the rotary-wing UAV is relatively low, it is necessary to use a bandpass filter to extract the frequency band where the rotary-wing UAV is located to facilitate spectrum analysis and identification.

[0101] In the embodiment of the present invention, the frequency band refers to a frequency having rotor spectrum characteristics, that is, a frequency band in which the rotor UAV is located.

[0102] In an embodiment of the present invention, extracting the frequency band corresponding to the micro-motion feature by using a preset bandpass filter includes:

[0103] Filtering the radar rotor signal through the bandpass filter to obtain a rotor filtered signal;

[0104] performing a frequency spectrum analysis on the rotor filter signal to obtain a target frequency band;

[0105] determining whether the target frequency band includes the micro-motion feature;

[0106] When the target frequency band includes the micro-motion feature, the target frequency band is used as the frequency band corresponding to the micro-motion feature.

[0107] Specifically, the radar rotor signal is filtered by a pre-designed bandpass filter to obtain a rotor filter signal, and the rotor filter signal is Fourier transformed to obtain spectrum data. According to the spectrum analysis results, the frequency range containing the micro-motion feature is determined, and it is judged whether the target frequency band contains the expected micro-motion feature. When the target frequency band contains the micro-motion feature, the frequency band is output as the frequency band corresponding to the micro-motion feature, such as Figure 6 As shown in the figure, it is a schematic diagram of the rotor spectrum characteristics. The simulation is carried out using the intermediate frequency sampling rate fs = 80e6, bandwidth B = 10e6, carrier frequency 16e9, radial speed of the quadcopter 50m / s, and blade length 0.3m. The frequency band where the rotorcraft is located can be obtained. Figure 6 The box in the middle is the frequency band where the rotary-wing drone is located.

[0108] Specifically, if Figure 7a As shown in the figure, it is a schematic diagram of the amplitude frequency response curve, which shows the response degree of different frequency signals. The horizontal axis represents the normalized frequency and the vertical axis represents the amplitude value. As the frequency increases, the amplitude also changes. Figure 7b As shown, it is a schematic diagram of the phase frequency response curve, which shows the phase offset of different frequency signals and how the phase difference between the input signal and the output signal changes with frequency. The horizontal axis represents the normalized frequency and the vertical axis represents the phase value. As the frequency increases, the phase also changes. Then, the frequency band where the rotorcraft is located is identified through the bandpass filter, as shown in the figure. Figure 7c As shown in the figure, it is a schematic diagram of the time domain signal after bandpass filtering. The bandpass filter allows the signal within a specific frequency range to pass through, while blocking the signal of other frequencies. After the bandpass filter is processed, only the signal components within the specified frequency range are retained, and other frequency components are suppressed or filtered out. Then, only the signal components within the specific frequency range are retained in the time domain waveform, as shown in the figure. Figure 7d As shown in the figure, it is a schematic diagram of the frequency signal after bandpass filtering. After the bandpass filter is processed, only the frequency components within the specified frequency range (i.e., the passband range) in the spectrum will be retained, and other frequency components (i.e., those below the lower limit frequency and above the upper limit frequency) will be significantly suppressed or completely filtered out. Figure 7d The frequency response of an ideal bandpass filter is shown in Figure 1, where the frequency components within the passband are retained and other frequencies are suppressed, showing only the frequency band where the rotorcraft is located.

[0109] Furthermore, the frequency band is uploaded to the radar terminal storage device using a preset bandwidth transmission, and a communication channel or system with a larger transmission bandwidth is used to transmit data. The advantages of large-bandwidth transmission include increasing the data transmission rate and supporting higher data volume or higher-quality signal transmission. The larger the bandwidth of the radar signal, the higher the resolution that can be provided, which helps to detect and identify targets. Using large-bandwidth transmission, the signal and output are stored and reported, which creates innovative data resources for improving radar performance.

[0110] Furthermore, after filtering the radar rotor signal through a bandpass filter, the frequency band where the rotor UAV is located is identified. The micro-motion component in the echo signal returned by the target detected by the radar is non-stationary and nonlinear, so it is necessary to use short-time Fourier transform (STFT) to analyze and extract features from the frequency band collected by the target.

[0111] S4. Extract the time-frequency graph corresponding to the frequency band uploaded to the radar terminal storage device through a preset short-time Fourier transform, and use the trained deep learning model to identify the time-frequency graph to obtain a second classification result of the rotorcraft.

[0112] In the embodiment of the present invention, since the micro-motion components in the echo signal returned by the target detected by the radar are non-stationary and nonlinear, the traditional Fourier transform cannot be applied to process such non-stationary signals, while the short-time Fourier transform (STFT) can perform joint time-frequency analysis on the non-stationary signal, and provide the frequency information of the signal related to time while retaining the time domain information. STFT realizes the time-frequency analysis of the non-stationary signal by adding a time-limited window function h(t) to the signal to be processed y(t). It mainly performs Fourier transform on y(t) within the time sliding window. It can be found that as the time window becomes smaller, the non-stationary signal can be converted into a stationary signal within the time window for analysis and processing. In order to make the Fourier transform have time characteristics, the signal x(u) is pre-windowed at each time instant t. The Fourier transform at this time is called STFT, which is expressed as Where h(t) is the short time analysis window, window h * (ut) can effectively suppress the side lobes of the signal time spectrum near the analysis time point u=t, thereby obtaining the frequency diagram after short-time Fourier transform under different sampling rates;

[0113] In detail, such as Figure 8a The figure below is a schematic diagram of the STFT result when the sampling rate is 100 MHz. It shows the changes of the signal in two dimensions: time and frequency. By showing the spectrum changes of the signal on the time axis, it helps to analyze how the frequency components of the signal change over time, that is, showing the spectrum changes in different time windows. Figure 8aIn the display, the spectrum changes in the case of 0-70us. For example, in the case of 0-10us, the spectrum changes in the range of -20-10; Figure 8b As shown in the figure, it is a schematic diagram of the STFT result under the sampling rate of 200Mhz, showing the display change of the spectrum under the condition of 0-70us. For example, under the condition of 0-10us, the change range of the spectrum is -20-20. It can be seen that Figure 8b The results shown are the signal output results of the rotary-wing UAV.

[0114] Furthermore, based on the time domain image output by the short-time Fourier transform, a pre-trained deep learning model can be used to analyze the time domain image in order to classify and identify the target classification results.

[0115] In an embodiment of the present invention, the second classification result refers to the type of classification and identification of the rotary-wing UAV based on the deep learning model, and the classification types include rotary-wing target (identified as 1), fixed-rotor target (identified as 2), flying bird target (identified as 3), and not a rotary-wing target (identified as 0).

[0116] In an embodiment of the present invention, the use of a trained deep learning model to identify the time-frequency graph to obtain a second classification result of the rotary-wing UAV includes:

[0117] Extracting the time-frequency graph in the radar terminal storage device according to a preset data period;

[0118] The trained deep learning model is injected into the preset radar equipment terminal backup;

[0119] Using a deep learning model in a radar device terminal backup to classify and identify the time-frequency graph, a second classification identifier of the rotary-wing UAV is obtained;

[0120] A second classification result of the rotary-wing UAV is determined according to the second classification identifier.

[0121] In detail, offline learning is carried out based on the deep learning model to form a training model that can be used for testing. All the extracted frame data are divided into 70% training set, 20% validation set, and 10% test set. The CNN model is used to perform image layer convolution on the extracted rotorcraft time-frequency images to establish a convolutional neural network model. The convolutional neural network can be selected to have at least 1 convolution layer, 1 pooling layer, and 1 fully connected layer. The activation function of the intermediate layer is ReLU. The BatchNormalization algorithm is used after the convolution layer and the pooling layer to accelerate training. The loss function uses cross entropy. In order to avoid overfitting, the Dropout regularization method is used. Among them, the commonly used size of the convolution kernel in the convolution layer is 3×3, the stride is 1, and zero padding is used. The maximum pooling layer is selected as the pooling layer with a size of 2×2 and a stride of 2. Figure 9 The figure shows a flow chart of the convolutional neural network model. The image data in the training set is input, and the convolutional neural network outputs features. Then, the classification results corresponding to the image data are determined based on the output features. Then, the trained deep learning model is used to perform time-frequency spectrum classification and recognition of the type of the target of interest.

[0122] Specifically, the training set, validation set, and test set are used to obtain the final learning model, which is then fed into the existing radar device terminal for backup. The information sent by the terminal is used to generate a bandpass filter. After bandpass filtering of a single frame of data, the data is quickly stored in the radar terminal storage device using large-bandwidth transmission technology. The backed-up learning model is used to test the reported multi-frame data within a cycle (set to be no less than 3s). The pulse-compressed data can be uploaded in time periods to reduce the amount of data, and then the judgment result of whether it is a rotorcraft is output. The rotor target is identified as 1, and the non-rotor target is identified as 0.

[0123] Furthermore, conventional classification and recognition methods are used, that is, attribute characteristics including speed, altitude, amplitude intensity, energy ratio, acceleration, track fitting characteristics, number of scatter points, etc. are used to classify and recognize the track to obtain the first classification result, and the second classification result is obtained through the deep learning model. Then, the conventional classification and recognition and time-frequency graph classification and recognition results are combined to further improve the accuracy and reliability of the rotor target classification and recognition results.

[0124] S5. Fusing the first classification result and the second classification result to obtain a target classification recognition result corresponding to the rotary-wing UAV.

[0125] In the embodiment of the present invention, the target classification and recognition result refers to the judgment result of whether the detection target is a rotary-wing drone obtained by combining the first classification result and the second classification result, which can more accurately identify the detection target.

[0126] In the embodiment of the present invention, the fusing of the first classification result and the second classification result to obtain the target classification recognition result corresponding to the rotary-wing UAV includes:

[0127] extracting a first rotor weight value from the first classification result, and extracting a second rotor weight value from the second classification result;

[0128] When the first classification result is a first identifier and the second classification result is a first identifier, determining a first rotor confidence value according to the first rotor weight value and the second rotor weight value, and determining that a target classification recognition result corresponding to the rotor UAV is a rotor according to the first rotor confidence value;

[0129] When the first classification result is a first identifier and the second classification result is not the first identifier, determining a second rotor confidence value according to the first rotor weight value, and determining, according to the second rotor confidence value, that the target classification recognition result corresponding to the rotor UAV is a target type rotor;

[0130] When the first classification result is not the first identification and the second classification result is the first identification, the third rotor confidence is determined according to the second rotor weight value, and the target classification identification result corresponding to the rotor UAV is determined to be a target type rotor according to the third rotor confidence.

[0131] In detail, the general classification and identification types include rotor targets (marked as 1), fixed-wing targets (marked as 2), flying bird targets (marked as 3), etc. Combined with the discrimination and confirmation results of the time-frequency graph classification and identification scores, the weight value of the current track point classification as a rotor is 0.4, that is, the first rotor weight value is 0.4; the weight of the time-frequency graph classification as a rotor is 0.6, that is, the second rotor weight value is 0.6.

[0132] Specifically, if the discrimination results are consistent, the discrimination type is displayed as rotor (confidence 100%), that is, the first classification result is identification 1 and the second classification result is also identification 1, indicating that the discrimination results are consistent, then the first rotor confidence is 100%, and the discrimination type is displayed as rotor; if only the track point classification is judged as rotor, the target type rotor (confidence 40%) is displayed, that is, the first classification result is identification 1, but the second classification result is not identification 1, indicating that there is only track point, then the second rotor confidence is 40%, and the discrimination type is displayed as target type rotor; if only the time-frequency graph classification is judged as rotor, the target type rotor (confidence 60%) is displayed, that is, the first classification result is not identification 1, but the second classification result is identification 1, indicating that there is only time-frequency graph, then the third rotor confidence is 60%), and the discrimination type is displayed as target type rotor.

[0133] Furthermore, the accuracy and reliability of rotor target classification and recognition results can be improved by combining conventional classification and recognition with time-frequency graph classification and recognition results. Figure 10 The figure shows a flow chart of classification result fusion. It uses dual-channel processing. On the one hand, the normal signal feature extraction and track feature extraction are performed through the AD single-frame data after pulse compression, and then normal routine detection is performed. On the other hand, the information sent by the terminal is used to generate a bandpass filter. After the single-frame data is bandpass filtered, the data is quickly stored in the radar terminal storage device using large-bandwidth transmission technology. The backed-up learning model is used to test the reported multi-frame data within a cycle (set to be no less than 3s) and output the judgment result of whether it is a rotorcraft. The rotor target is identified as 1, and the non-rotor target is identified as 0. The terminal judgment result is combined for fusion recognition output.

[0134] Example 2

[0135] In order to better understand the present invention, a second embodiment is used below to further explain how the present invention performs noise reduction on audio data.

[0136] In an embodiment of the present invention, in order to address the problems of a large frequency bandwidth and a narrow frequency band of a rotary-wing UAV under a Doppler radar system, due to the analysis and confirmation of the characteristics of the rotary-wing UAV, the frequency of the fuselage and rotor of the rotary-wing UAV is relatively low as a whole, and the approximate speed of its rotor can be calculated by knowing the true speed of the rotary-wing UAV, which often revolves around the main frequency. Therefore, in order to process the target signal more finely and accurately, feedback is provided by paying attention to the true speed of the target. Therefore, it is considered to design a bandpass filter for the frequency band of the rotary-wing UAV for the data reported in each frame, so that frequency analysis and image recognition are more accurate and data transmission is easier.

[0137] In an embodiment of the present invention, before extracting the frequency corresponding to the micro-motion feature through a preset bandpass filter, the method further includes:

[0138] Determining the center frequency and bandwidth of the bandpass filter according to the frequency range of the micro-motion characteristics;

[0139] Selecting a filter type of a bandpass filter according to the center frequency and the bandwidth;

[0140] The bandpass filter is configured by using the filter type and filter parameters corresponding to the filter type.

[0141] Specifically, a center frequency is selected based on the frequency range of the micro-motion signature. This frequency is the midpoint of the bandpass filter, and the bandpass filter will filter around this frequency. The bandwidth, or the width of the frequency range the filter can pass, is determined. This is the frequency interval centered around the center frequency. The wider the bandwidth, the larger the frequency range the filter allows. The appropriate filter type is then selected based on the desired filtering characteristics and design requirements. Common filter types include Butterworth filters, Chebyshev filters, and Elliptic filters. Each filter has unique frequency response characteristics, such as smoothness and steepness. The filter type then determines the filter's cutoff frequency and order.

[0142] For example, a certain micro-motion feature may occur between 1kHz and 5kHz, so 3kHz is selected as the center frequency of the bandpass filter. According to the frequency range of the micro-motion feature, a bandwidth of 2kHz can be selected. From 1kHz to 5kHz, an appropriate filter type is selected, such as the Butterworth filter, because it has a smooth frequency response. According to the center frequency of 3kHz and the bandwidth of 2kHz, the parameters of the Butterworth filter, such as the order of the filter, are configured to ensure that it can effectively pass the frequency range of 1kHz to 5kHz, so that unnecessary frequency components can be accurately filtered out, and only the frequency range corresponding to the micro-motion feature can be retained, thereby extracting the target micro-motion feature.

[0143] The embodiments of the present invention utilize large bandwidth transmission to store and report signals and outputs, creating innovative data resources for radar performance improvement. They focus on the accurate classification of rotary-wing drones, collect time-frequency images of rotary-wing drones in different postures to establish a training library for the image level of rotary-wing drones, which is conducive to the accurate classification of rotary-wing drones. A deep learning method based on time-frequency images is used to establish a stable and reliable learning model. Information communication is established between the radar terminal and the DSP. Based on the target information set by the radar terminal, the DSP uploads the pulse-compressed data by region and time period, reducing the amount of data. Considering the design of a bandpass filter for the rotary-wing drone frequency band for each frame of reported data, frequency analysis and image recognition are more accurate, and data transmission is easier. A fusion classification and discrimination model is designed, and by combining conventional classification and recognition with time-frequency image classification and recognition results, the accuracy and reliability of the rotary-wing target classification and recognition results are further improved. Therefore, the radar rotary-wing target classification and recognition method, device, equipment, and medium proposed in the present invention can solve the problem of low accuracy when performing rotary-wing drone classification and recognition.

[0144] Example 3

[0145] like Figure 11 As shown, this embodiment also provides a functional module diagram of a radar rotor target classification and identification device.

[0146] The radar rotor target classification and identification device 100 described in this embodiment can be installed in an electronic device. Depending on the functionality implemented, the radar rotor target classification and identification device 100 may include a radar rotor signal acquisition module 101, a first classification result identification module 102, a frequency band extraction module 103, a second classification result identification module 104, and a classification result fusion module 105. The modules described in the present invention, also referred to as units, refer to a series of computer program segments that can be executed by an electronic device processor and perform a fixed function, and are stored in the electronic device's memory.

[0147] In this embodiment, the functions of each module / unit are as follows:

[0148] The radar rotor signal acquisition module 101 is used to collect the radar rotor signal of the rotary-wing UAV, digitize the radar rotor signal, and perform pulse compression on the digitized single-frame radar rotor signal;

[0149] The first classification result identification module 102 is configured to upload the pulse-compressed radar rotor signal to a digital signal processing unit, extract signal features corresponding to the radar rotor signal in the digital signal processing unit, and extract track features of the rotary-wing UAV, and identify a first classification result of the rotary-wing UAV based on the track features and the signal features;

[0150] The frequency band extraction module 103 is used to extract the micro-motion characteristics of the radar rotor signal, extract the frequency band corresponding to the micro-motion characteristics through a preset bandpass filter, and upload the frequency band to the radar terminal storage device using a preset bandwidth transmission;

[0151] The second classification result recognition module 104 is configured to extract a time-frequency graph corresponding to a frequency band uploaded to the radar terminal storage device through a preset short-time Fourier transform, and recognize the time-frequency graph using a trained deep learning model to obtain a second classification result of the rotary-wing UAV;

[0152] The classification result fusion module 105 is used to fuse the first classification result and the second classification result to obtain a target classification recognition result corresponding to the rotary-wing UAV.

[0153] In detail, the modules described in the radar rotor target classification and identification device 100 described in the embodiment of the present invention adopt the same technical means as the radar rotor target classification and identification method described in Example 1 and Example 2 when in use, and can produce the same technical effects, which will not be repeated here.

[0154] Example 4

[0155] like Figure 12 As shown, this embodiment also provides a computer electronic device, which may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as a radar rotor target classification and recognition program.

[0156] In some embodiments, the processor 10 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting the various components of the entire electronic device using various interfaces and lines, and executing the programs or modules stored in the memory 11 (for example, executing a radar rotor target classification and recognition program, etc.), as well as calling the data stored in the memory 11, to perform various functions of the electronic device and process data.

[0157] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 can also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the memory 11 can also include both an internal storage unit of the electronic device and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device, such as the code of the radar rotor target classification and recognition program, but can also be used to temporarily store data that has been output or is to be output.

[0158] The communication bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0159] The communication interface 13 is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0160] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0161] For example, although not shown, the electronic device may further include a power source (such as a battery) for powering various components. Preferably, the power source may be logically connected to the at least one processor 10 via a power management device, thereby implementing functions such as charge management, discharge management, and power consumption management through the power management device. The power source may further include any components such as one or more DC or AC power sources, a recharging device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device may further include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0162] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0163] The radar rotor target classification and recognition program stored in the memory 11 of the electronic device is a combination of multiple instructions. When running in the processor 10, it can achieve the following:

[0164] Collecting radar rotor signals of the rotary-wing UAV, digitizing the radar rotor signals, and performing pulse compression on the digitized single-frame radar rotor signals;

[0165] Uploading the pulse-compressed radar rotor signal to a digital signal processing unit, extracting signal features corresponding to the radar rotor signal in the digital signal processing unit, extracting track features of the rotary-wing UAV, and identifying a first classification result of the rotary-wing UAV based on the track features and the signal features;

[0166] Extracting micro-motion features of the radar rotor signal, extracting a frequency band corresponding to the micro-motion features through a preset bandpass filter, and uploading the frequency band to a radar terminal storage device using a preset bandwidth transmission;

[0167] Extracting a time-frequency graph corresponding to a frequency band uploaded to a radar terminal storage device through a preset short-time Fourier transform, and identifying the time-frequency graph using a trained deep learning model to obtain a second classification result of the rotary-wing UAV;

[0168] The first classification result and the second classification result are integrated to obtain a target classification recognition result corresponding to the rotary-wing UAV.

[0169] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0170] Furthermore, if the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0171] Example 5

[0172] This embodiment provides a storage medium storing a computer program. When the computer program is executed by a processor, the steps of the radar rotor target classification and recognition method described above are implemented.

[0173] These program codes can also be loaded onto a computer or other programmable data processing device so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 The steps of a specified function in a process or multiple processes.

[0174] Storage media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of storage media can include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0175] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the module division is merely a logical function division, and other division methods may be used in actual implementation.

[0176] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0177] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional modules.

[0178] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0179] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is not limited only according to the above description, and it is intended that all changes within the meaning and scope of equivalent elements within the scope of protection are included in the present invention.

[0180] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0181] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. Terms such as "first" and "second" are used to indicate names and do not imply any particular order.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A radar rotor target classification and recognition method, characterized in that: The method comprises: Collecting radar rotor signals of the rotary-wing UAV, digitizing the radar rotor signals, and performing pulse compression on the digitized single-frame radar rotor signals; Uploading the pulse-compressed radar rotor signal to a digital signal processing unit, extracting signal features corresponding to the radar rotor signal in the digital signal processing unit, extracting track features of the rotary-wing UAV, and identifying a first classification result of the rotary-wing UAV based on the track features and the signal features; Extracting micro-motion features of the radar rotor signal, extracting a frequency band corresponding to the micro-motion features through a preset bandpass filter, and uploading the frequency band to a radar terminal storage device using a preset bandwidth transmission; Extracting a time-frequency graph corresponding to a frequency band uploaded to a radar terminal storage device through a preset short-time Fourier transform, and identifying the time-frequency graph using a trained deep learning model to obtain a second classification result of the rotary-wing UAV; The first classification result and the second classification result are integrated to obtain a target classification recognition result corresponding to the rotary-wing UAV.

2. The radar rotor target classification and identification method according to claim 1, characterized in that: The identifying a first classification result of the rotary-wing UAV according to the track feature and the signal feature includes: Performing feature fusion on the track feature and the signal feature to obtain a fusion feature; Using a pre-built classification model to classify and identify the fused features, and obtain a first classification identifier of the rotary-wing UAV; A first classification result of the rotary-wing UAV is determined according to the first classification identifier.

3. The radar rotor target classification and recognition method according to claim 1, wherein: The extracting of the micro-motion feature of the radar rotor signal includes: Acquiring signal frame data corresponding to the radar rotor signal; Determining distance data of the rotary-wing UAV based on the signal frame data; Extracting the Doppler characteristics of the radar rotor signal using a pre-built echo baseband model; The micro-motion characteristics of the radar rotor signal are determined according to the distance data and the Doppler characteristics.

4. The radar rotor target classification and recognition method according to claim 3, characterized in that: The extracting the frequency band corresponding to the micro-motion feature by using a preset bandpass filter includes: Filtering the radar rotor signal through the bandpass filter to obtain a rotor filtered signal; performing a frequency spectrum analysis on the rotor filter signal to obtain a target frequency band; determining whether the target frequency band includes the micro-motion feature; When the target frequency band includes the micro-motion feature, the target frequency band is used as the frequency band corresponding to the micro-motion feature.

5. The radar rotor target classification and recognition method according to claim 1, wherein: The method of using the trained deep learning model to identify the time-frequency graph to obtain a second classification result of the rotary-wing UAV includes: Extracting the time-frequency graph in the radar terminal storage device according to a preset data period; The trained deep learning model is injected into the preset radar equipment terminal backup; Using a deep learning model in a radar device terminal backup to classify and identify the time-frequency graph, a second classification identifier of the rotary-wing UAV is obtained; A second classification result of the rotary-wing UAV is determined according to the second classification identifier.

6. The radar rotor target classification and recognition method according to claim 1, wherein: The first classification result and the second classification result are fused to obtain a target classification recognition result corresponding to the rotary-wing UAV, including: extracting a first rotor weight value from the first classification result, and extracting a second rotor weight value from the second classification result; When the first classification result is a first identifier and the second classification result is a first identifier, determining a first rotor confidence value according to the first rotor weight value and the second rotor weight value, and determining that a target classification recognition result corresponding to the rotor UAV is a rotor according to the first rotor confidence value; When the first classification result is a first identifier and the second classification result is not the first identifier, determining a second rotor confidence value according to the first rotor weight value, and determining, according to the second rotor confidence value, that the target classification recognition result corresponding to the rotor UAV is a target type rotor; When the first classification result is not the first identification and the second classification result is the first identification, the third rotor confidence is determined according to the second rotor weight value, and the target classification identification result corresponding to the rotor UAV is determined to be a target type rotor according to the third rotor confidence.

7. The radar rotor target classification and recognition method according to claim 4, characterized in that: Before extracting the frequency corresponding to the micro-motion feature through a preset bandpass filter, the method further includes: Determining the center frequency and bandwidth of the bandpass filter according to the frequency range of the micro-motion characteristics; Selecting a filter type of a bandpass filter according to the center frequency and the bandwidth; The bandpass filter is configured by using the filter type and filter parameters corresponding to the filter type.

8. A radar rotor target classification and identification device, characterized in that: The device comprises: A radar rotor signal acquisition module is used to collect radar rotor signals of the rotary-wing UAV, digitize the radar rotor signals, and perform pulse compression on the digitized single-frame radar rotor signals; a first classification result identification module, configured to upload the pulse-compressed radar rotor signal to a digital signal processing unit, extract signal features corresponding to the radar rotor signal in the digital signal processing unit, extract track features of the rotary-wing UAV, and identify a first classification result of the rotary-wing UAV based on the track features and the signal features; a frequency band extraction module, configured to extract the micro-motion characteristics of the radar rotor signal, extract the frequency band corresponding to the micro-motion characteristics through a preset bandpass filter, and upload the frequency band to the radar terminal storage device using a preset bandwidth transmission; A second classification result recognition module is used to extract a time-frequency graph corresponding to the frequency band uploaded to the radar terminal storage device through a preset short-time Fourier transform, and use a trained deep learning model to recognize the time-frequency graph to obtain a second classification result of the rotary-wing UAV; The classification result fusion module is used to fuse the first classification result and the second classification result to obtain the target classification recognition result corresponding to the rotary-wing UAV.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the radar rotor target classification and recognition method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the radar rotor target classification and recognition method according to any one of claims 1 to 7 are implemented.

Images