Training data generation method and device, unmanned aerial vehicle detection device and storage medium
The drone simulation signal is generated through sampling and signal parameter simulation of the receiving device, which solves the problem of obtaining training data of the drone detection model and improves detection efficiency and generalization capabilities.
Patent Information
- Application Number
- CN202510565375.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-30
- Publication Date
- 2025-08-08
AI Technical Summary
In the prior art, it is difficult to obtain training data for drone detection models, and it is difficult to cover signal changes in different environments and scenarios, affecting detection efficiency and generalization capabilities.
By receiving the sampling parameters of the device and the drone signal parameters, multiple groups of drone reception signals are simulated and interfering signals are superimposed to form the drone simulation signals, and then the training data is determined. Combined with fast Fourier variation and data aggregation processing, a two-dimensional time-frequency power array is generated to obtain the training data.
It realizes the rapid and simple generation of training data for the drone detection model, improves the coverage of training data and the generalization ability of the detection model, and reduces the difficulty of data acquisition.
Smart Images
Figure CN120454903A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of drone technology, and specifically to a training data generation method and device, a drone detection device, and a storage medium. Background Art
[0002] After receiving the radio signal transmitted by the drone through the receiving equipment, the drone is detected, identified, direction-finded and located based on the received signal. This is one of the main technical means of detecting drones in the current anti-drone field.
[0003] Drone signal detection is typically based on a time-frequency transformation of the signal received by the receiving device combined with a pulse train detection algorithm. With the continuous advancement of technology, deep learning-based two-dimensional time-frequency pattern recognition has gradually been introduced to the field of drone signal detection. To improve the efficiency of drone detection, a drone signal detection model (referred to as a drone detection model) can be used to identify two-dimensional time-frequency patterns to detect drones. However, before using the drone detection model to detect drones, it must be trained using training data. The challenge is how to quickly and easily determine the training data for the drone detection model. Summary of the Invention
[0004] In view of the above problems, the embodiments of the present application provide a training data generation method, device, drone detection device and storage medium, which are used to solve the problem in the prior art that it is difficult to obtain training data for drone detection models.
[0005] According to one aspect of an embodiment of the present application, a method for generating training data is provided, wherein the training data is used to train a drone detection model. The method includes: obtaining sampling parameters when a receiving device samples a drone signal and signal parameters of a target drone transmission signal, wherein the drone detection model is used to detect the target drone based on the drone signal sampled by the receiving device, and the signal parameters include fixed item parameters and traversal item parameters; based on the sampling parameters, the fixed item parameters, and the multiple traversal item parameters obtained by traversal, multiple groups of baseband signals are obtained; based on the multiple groups of baseband signals, multiple groups of drones received by the receiving device are simulated. Receive a signal; determine multiple groups of interference signals, wherein at least one parameter of the interference signal style, signal bandwidth, signal frequency deviation and signal strength is different between any two groups of interference signals; select one group of drone receiving signals and one group of interference signals from the multiple groups of drone receiving signals and the multiple groups of interference signals respectively, superimpose the selected drone receiving signals and interference signals to obtain a group of drone simulation signals, and repeat this step to obtain different multiple groups of drone simulation signals; determine multiple groups of training data based on the multiple groups of drone simulation signals, wherein the multiple groups of training data correspond to the multiple groups of drone simulation signals one to one.
[0006] In an optional manner, the sampling parameters include a reference sampling rate and a sampling duration; the fixed item parameters include one or more of a signal period and a signal frame format; and the traversal item parameters include one or more of a signal bandwidth, a signal frequency deviation, and a signal modulation method.
[0007] In an optional manner, the training data corresponding to each group of drone simulation signals is determined by the following steps: performing fast Fourier transform calculation on the drone simulation signal to obtain a two-dimensional time-frequency complex array; performing a modular square operation on each complex number in the two-dimensional time-frequency complex array to obtain a two-dimensional time-frequency power array corresponding to the two-dimensional time-frequency complex array; performing data aggregation processing on the two-dimensional time-frequency power array to obtain training data corresponding to the drone simulation signal.
[0008] In an optional manner, the data aggregation processing is performed on the two-dimensional time-frequency power array to obtain training data corresponding to the UAV simulation signal, including: dividing the two-dimensional time-frequency power array into k n*m data blocks, wherein each of the data blocks includes n consecutive data in the time dimension, and each of the data blocks includes m consecutive data in the frequency dimension; summing the data in each of the data blocks to obtain k sum values corresponding to the k data blocks; and replacing the k data blocks with the k sum values to obtain training data corresponding to the UAV simulation signal.
[0009] In an optional manner, the simulating of multiple groups of drone receive signals received by the receiving device based on the multiple groups of baseband reference signals includes: performing power normalization processing on each group of baseband signals in the multiple groups of baseband signals to obtain multiple groups of baseband reference signals corresponding to the multiple groups of baseband signals; determining the noise power of the receiving device and the power of multiple drone signals; selecting a group of baseband reference signals and a drone signal power from the multiple groups of baseband reference signals and the multiple drone signal powers, multiplying the selected baseband reference signal and the drone signal power to obtain a product, adding the product and the noise power to obtain a group of drone receive signals, and repeating this step to obtain multiple groups of drone receive signals.
[0010] In an optional manner, determining the noise power of the receiving device and the signal powers of multiple drones includes: determining the signal powers of the multiple drones based on the equivalent radiated power of the target drone, the receiving antenna gain of the receiving device, the signal operating frequency wavelength, the transceiver spacing and the link active gain.
[0011] In an optional manner, the signal powers of the multiple drones are determined based on the equivalent radiated power of the target drone, the receiving antenna gain of the receiving device, the signal operating frequency wavelength, the transceiver spacing and the link active gain, including: selecting a parameter from the equivalent radiated power, the signal operating frequency wavelength and the transceiver spacing as a variable, and determining other parameters as fixed values; and determining the signal powers of the multiple drones based on the fixed value and multiple values of the variable.
[0012] According to another aspect of an embodiment of the present application, a training data generating device is provided, which includes a memory, a processor, and a computer program stored in the memory, and the processor executes the computer program to implement the training data generating method as described above.
[0013] According to another aspect of an embodiment of the present application, a drone detection device is provided, wherein the drone detection device is deployed with a drone detection model, wherein the drone detection model is a model obtained by training an original model using training data, and the training data is data generated using the training data generation method as described above.
[0014] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the training data generating method as described above is implemented.
[0015] In actual application scenarios, after the receiving device receives the signal transmitted by the target drone, the drone detection model is required to detect the target drone from the signal received by the receiving device. In an embodiment of the present application, multiple sets of different drone simulation signals are simulated based on the sampling parameters of the receiving device and the signal parameters of the signal transmitted by the target drone, and then the training data of the drone detection model is determined based on the drone simulation signals. Through the above-mentioned digital twin simulation method, combined with the characteristics that some parameters in the drone signal are fixed values and some parameters are variable values, since there is no need to actually control the drone flight, the training data for the drone detection model training can be generated efficiently and quickly, thereby improving the efficiency of generating training data for the drone detection model and reducing the difficulty of obtaining training data.
[0016] Moreover, in the embodiment of the present application, since the drone simulation signal is not generated arbitrarily, but is obtained by imitating the drone signal received by the receiving device as the imitation object, by traversing to obtain multiple traversal item parameters, each group of drone simulation signals corresponds to the drone signal received by the receiving device when the target drone flies in a certain environment and scenario, and the multiple groups of drone simulation signals in the embodiment of the present application correspond to the drone signals received by the receiving device when the target drone flies in different environments and scenarios. That is to say, the training data determined based on the drone simulation signal in the embodiment of the present application covers the drone signals received by the receiving device when the target drone flies in different environments and scenarios. Therefore, after using the training data to train the drone detection model, the generalization ability of the drone detection model can be improved.
[0017] The above description is only an overview of the technical solution of the embodiment of the present application. In order to more clearly understand the technical means of the embodiment of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiment of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present application. In addition, the same reference symbols are used to represent the same components throughout the drawings. In the drawings:
[0019] Figure 1 A flow chart of a method for generating training data according to an embodiment of the present application is shown;
[0020] Figure 2 : shows a schematic diagram of a group of baseband reference signals provided by an embodiment of the present application;
[0021] Figure 3 A schematic diagram showing a group of drones receiving signals provided by an embodiment of the present application is shown;
[0022] Figure 4 A schematic diagram of a group of interference signals provided by an embodiment of the present application is shown;
[0023] Figure 5 A schematic diagram showing a set of two-dimensional time-frequency diagram data provided by an embodiment of the present application is shown;
[0024] Figure 6 A schematic diagram of the structure of a training data generating device provided in an embodiment of the present application is shown;
[0025] Figure 7 A schematic structural diagram of a drone detection device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION
[0026] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0027] For the same drone, the drone signals received by the receiving device will be different when flying in different environments and scenarios. For example, electromagnetic interference, terrain obstruction, and weather conditions can all affect the drone signals received by the receiving device. Therefore, in order for the drone detection model to detect drones from different drone signals, it is necessary to use different drone signals as training data to train the drone detection model. This improves the generalization ability of the drone detection model, so that the trained drone detection model can detect drones from different drone signals even when flying in different environments and scenarios.
[0028] However, in order to obtain diverse drone signals as training data for drone detection models, controlling drones to fly in different environments and scenarios to collect actual drone signals presents numerous challenges and limitations. For example, flying in complex environments can result in collisions, signal loss, or inclement weather, posing high safety risks. Actual flight operations require specialized personnel and equipment, resulting in high labor and equipment costs. Therefore, relying on actual flight training data presents challenges such as difficulty in data acquisition and low scenario coverage. Furthermore, the inability to quickly acquire training data impacts the iteration efficiency of the drone detection model.
[0029] Based on this, the present application proposes a training data generation method that generates drone signals through simulation without the need to actually collect drone signals, thereby quickly and easily obtaining training data for training drone detection models. Specifically, in order to obtain drone signals, the receiving device typically collects the signals transmitted by the drone using certain sampling parameters (such as a baseline sampling rate and sampling duration). In addition, the drone signal itself has signal parameters (such as a signal period, a signal pulse width, a signal bandwidth, a signal frame format, and a signal modulation method). Therefore, in an embodiment of the present application, based on the sampling parameters of the receiving device and the signal parameters of the drone signal, multiple sets of different drone received signals received by the receiving device are simulated and generated. Among them, at least one signal parameter differs between any two sets of drone received signals. Since drone signals received by the receiving device often contain interference signals, in an embodiment of the present application, interference signals are superimposed on the multiple sets of simulated drone received signals to obtain multiple sets of drone simulated signals. In an embodiment of the present application, the above method can be used to simulate drone simulated signals that are close to the drone signals actually received by the receiving device. By setting different signal parameters, drone simulated signals corresponding to drone signals when the drone is flying in different environments and scenarios can be obtained. The above method can quickly and easily obtain drone simulation data for training drone detection models.
[0030] Figure 1 A flow chart of the training data generation method provided in an embodiment of the present application is shown, and the method is executed by a training data generation device, which may be a device including one or more processors, such as a server, a computer, or other electronic device with model training capabilities. The processor may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement an embodiment of the present invention, which is not limited here. The one or more processors included in the training data generation device may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs, which is not limited here. Figure 1 As shown, the method includes the following steps:
[0031] Step 110: Acquire sampling parameters when the receiving device samples the drone signal and signal parameters of the target drone transmission signal.
[0032] When using the drone detection model to detect drone signals, the model must detect drone signals received by a specific receiving device. As previously mentioned, the sampling parameters of a receiving device include a baseline sampling rate and sampling duration. Different receiving devices may use different baseline sampling rates and sampling durations to collect signals transmitted by the same drone, resulting in potentially different drone signals. Therefore, if the drone detection model is required to detect drone signals received by a specific receiving device in a practical application, this step involves obtaining the sampling parameters of that receiving device.
[0033] As previously mentioned, drone signal parameters include signal period, pulse width, bandwidth, frame format, and modulation. Signals transmitted by different types and models of drones typically differ in at least one of these parameters. Therefore, when applying a drone detection model to detect drone signals, if the model is required to detect a specific type of drone (i.e., the target drone) from the received drone signal, this step involves obtaining the signal period, pulse width, bandwidth, frame format, and modulation of the signal transmitted by the target drone.
[0034] Step 120: Obtain multiple groups of baseband signals based on the sampling parameters, the fixed item parameters, and the multiple traversal item parameters obtained through traversal.
[0035] Among them, after the base sampling rate of the receiving device is determined, its instantaneous receivable bandwidth is also determined accordingly. However, when the instantaneous receivable bandwidth of the receiving device is less than the drone's frequency hopping bandwidth, it means that the receiving device cannot simultaneously cover all of the drone's frequency hopping points, which will cause the drone signals at some frequencies to be unable to be received by the receiving device. Therefore, in actual scenarios, the bandwidth and frequency deviation of the drone signal received by the receiving device may exist in different forms. For example, if the receiving device can only receive signals within the frequency range of -30MHz to 30MHz, and the signal transmitted by the drone is a signal with a frequency of -45MHz to -25MHz, the receiving device can receive signals with a frequency range of -30MHz to -25MHz; if the signal transmitted by the drone is a signal with a frequency of -40MHz to -20MHz, the receiving device can receive signals with a frequency range of -30MHz to -20MHz.
[0036] Therefore, in order to enable the drone detection model to detect drones from drone signals received in different scenarios when it is actually applied, in the embodiment of the present application, the signal parameters of the drone transmission signal are divided into fixed item parameters and ergodic item parameters. Among them, the fixed item parameters may include only the signal period or the signal frame format, or may include both the signal period and the signal frame format. The ergodic item parameters may include one or more of the signal bandwidth, signal frequency deviation, and signal modulation mode. In the embodiment of the present application, preferably, the data types of the sampling parameters and signal parameters are shown in Table 1.
[0037] Table 1: Data types of sampling parameters and signal parameters
[0038]
[0039] In Table 1, the base sampling rate and sampling duration are determined by the receiving device. Here, the base sampling rate and sampling duration of the receiving device are 61.44MSa and 100ms, respectively, as an example. The specific parameters of the signal period, signal pulse width, signal bandwidth, signal frame format, and signal modulation method are determined by the signal transmitted by the target drone. In Table 1, the signal transmitted by the target drone has a signal period of 4ms and 6ms, a signal pulse width of 2.07ms, 3.07ms, 4.07ms, and 5.07ms, a signal bandwidth of 10MHz and 20MHz, a signal frame format of 1 long symbol + 6 short symbols, and signal modulation methods of QPSK and 16QAM as examples.
[0040] After determining Table 1, multiple groups of baseband signals can be determined according to Table 1 in this step. Specifically, when determining a group of baseband signals, its reference sampling rate and sampling duration are determined to be 61.44MSa and 100ms respectively. The signal period includes 4ms and 6ms, the signal pulse width of a single signal pulse can be randomly determined from 2.07ms, 3.07ms, 4.07ms and 5.07ms, the signal bandwidth can be arbitrarily determined from 10MHz and 20MHz, the signal frequency deviation step is 1MHz, the signal time deviation is any value within 10ms, the signal frame format is 1 long symbol + 6 short symbols, and the signal modulation method is selected from QPSK and 16QAM. After determining the above parameters, a group of baseband signals can be determined.
[0041] In this step, by repeating the above steps multiple times and traversing the traversal item parameters, different multiple groups of baseband signals can be obtained. Among them, the multiple groups of baseband signals include baseband signals with bandwidths of 10MHz and 20MHz, respectively, and baseband signals with signal modulation modes of QPSK and 16QAM, respectively. In addition, in view of the fact that the receiving device cannot cover all the frequency hopping points of the drone at the same time, resulting in the drone signals of some frequencies being unable to be received by the receiving device, in an embodiment of the present application, by setting the signal frequency deviation and traversing the frequencies of various drone signals that may be received by the receiving device in 1MHz steps, the baseband signals corresponding to various drone signals can be simulated.
[0042] In the embodiment of the present application, regarding the sampling parameters, the reference sampling rate and sampling duration used for signal generation are determined according to the time width and instantaneous bandwidth of the time-frequency diagram corresponding to the signal transmitted by the target drone received by the receiving device. For a specific drone detection model, the data is an invariant (i.e., a fixed value). Regarding the signal parameters, for drone signals in the LTE-OFDM format, the description of its target characteristics usually includes a signal pulse width set, a signal period, and a signal bandwidth set, wherein the period is a fixed feature and follows a specific law during the entire transmission process. Therefore, the correspondence between the pulse width and the period is designed to be any combination within the tolerable range of the time slot. The signal bandwidth reflects the different working modes of the drone signal and needs to traverse all situations. In a single detection process, the signal has a large uncertainty in time and frequency, and is designed as a traversal item in data generation. The signal modulation needs to be simulated and designed according to the frame format of the target signal. For the simulation of a specific drone signal, its frame format is set as a fixed item, and there is no need to change its frame structure in multiple simulations. The item that needs to be traversed is the modulation scheme adopted by its baseband.
[0043] Step 130: Simulate multiple sets of drone reception signals received by a receiving device based on the multiple sets of baseband signals.
[0044] Specifically, the following steps a1 to a3 can be used to simulate multiple groups of drones receiving signals.
[0045] Step a1: performing power normalization processing on each of the multiple groups of baseband signals to obtain multiple groups of baseband reference signals corresponding to the multiple groups of baseband signals.
[0046] In actual applications, the drone signal received by the receiving device includes the signal transmitted by the drone and the interference signal. To facilitate differentiation, in the embodiments of the present application, the portion of the signal actually received by the receiving device is referred to as the drone received signal. Because the signal transmitted by the drone has a certain power and the receiving device has a certain amount of noise, the actual drone received signal obtained when the receiving device receives the drone transmitted signal has a certain signal-to-noise ratio.
[0047] Therefore, in this step, the power of each group of baseband signals is normalized so that noise can be superimposed on the normalized baseband signals to obtain a signal with a corresponding signal-to-noise ratio to simulate the drone reception signal actually received by the receiving device.
[0048] Figure 2 FIG shows a schematic diagram of a set of baseband reference signals provided by an embodiment of the present application. Figure 2 As shown, the horizontal axis is the sampling time of the baseband reference signal, the sampling time is 100ms, and the vertical axis is the amplitude of the baseband reference signal.
[0049] Step a2: Determine the noise power of the receiving device and the signal power of multiple drones.
[0050] In this step, by determining the noise power of the receiving device and the signal power of multiple drones, different noise ratios can be determined. Specifically, the drone signal power Pr can be determined using the following formula (1).
[0051]
[0052] Among them, PtGt is the equivalent radiated power of the drone, that is, the EIRP value; Gr is the receiving antenna gain; λ is the signal operating frequency wavelength; R is the transmit-receive distance; G c is the link active gain.
[0053] Since there are multiple variables in formula (1), in order to quickly determine the signal powers of multiple different drones, one parameter can be selected as a variable from the drone equivalent radiated power, signal operating frequency wavelength, and transceiver spacing, and the other parameters can be determined as fixed values. By combining the fixed values and multiple values as variables, formula (1) can be used to determine the signal powers of multiple drones.
[0054] For example, PtGt can be used as a variable, and the other parameters in formula (1) can be determined as fixed values. By substituting multiple values of PtGt into formula (1), the signal power of multiple drones can be quickly determined.
[0055] In order to quickly determine the signal power, in some embodiments, a value may be randomly determined when determining the signal power, and the randomly determined value may be determined as the signal power.
[0056] In this step, the noise power can be determined by the following formula (2).
[0057] Pn=kT0BnFnG c (2)
[0058] Where k is the Boltzmann constant, which is 1.3806505×10-23 J / K; T0 is the noise temperature at room temperature, which is 290K; Bn is the bandwidth of the signal transmitted by the target UAV; Fn is the noise coefficient of the receiving link; G c is the link active gain.
[0059] Step a3: Select a group of baseband reference signals and a drone signal power from multiple groups of baseband reference signals and multiple drone signal powers, multiply the selected baseband reference signal and the drone signal power to obtain a product, add the product and the noise power to obtain a group of drone received signals, and repeat this step to obtain multiple groups of drone received signals.
[0060] Since the drone reception signal actually received by the receiving device is a signal with a certain signal-to-noise ratio, in this step, the drone reception signal with a certain signal-to-noise ratio can be obtained by multiplying the power-normalized baseband reference signal by the signal power and adding the noise power. The drone reception signal simulated in the above manner is close to the drone reception signal actually received by the receiving device.
[0061] Figure 3 Schematic diagram of a group of drones receiving signals provided by an embodiment of the present application is shown. Figure 3 As shown, the horizontal axis is the sampling time of the signal and the vertical axis is the signal amplitude.
[0062] In step 130, in order to quickly simulate and obtain multiple sets of drone receiving signals, different signal-to-noise ratios can also be randomly determined, and then the drone receiving signal is determined based on the determined signal-to-noise ratio and baseband signal, thereby obtaining multiple sets of drone receiving signals based on different signal-to-noise ratios.
[0063] Step 140: Determine multiple groups of interference signals.
[0064] Among them, the interference signal belongs to a style including narrowband targeting, broadband interference and communication signal flow. When determining the interference signal, the style of an interference signal can be determined from the above-mentioned multiple styles, and after randomly determining the signal bandwidth of the interference signal, the signal frequency deviation of the interference signal and the signal strength of the interference signal, the corresponding interference signal can be generated. By performing the above operation multiple times, multiple groups of interference signals can be obtained. It is worth noting that in the embodiment of the present application, there is at least one parameter different between any two groups of interference signals among the multiple groups of interference signals, including the style of the interference signal, signal bandwidth, signal frequency deviation and signal strength.
[0065] Figure 4 FIG. 1 shows a schematic diagram of a set of interference signals provided by an embodiment of the present application. Figure 4 As shown, the interference signal is of narrowband targeting type.
[0066] Step 150: Select a group of drone receiving signals and a group of interference signals from the multiple groups of drone receiving signals and the multiple groups of interference signals, respectively, superimpose the selected drone receiving signals and interference signals to obtain a group of drone simulation signals, and repeat this step to obtain different groups of drone simulation signals.
[0067] As mentioned above, the signal actually received by the receiving device includes the drone receiving signal and the interference signal. Therefore, in this step, by superimposing the simulated drone receiving signal and the interference signal, a drone simulation signal close to the drone signal actually received by the receiving device can be simulated.
[0068] In which, when step 150 is repeatedly executed, step 150 may be stopped until each group of drone simulation signals in the multiple groups of drone simulation signals and each group of interference signals in the multiple groups of interference signals are selected out; or step 150 may be stopped when the number of obtained drone simulation signals reaches a preset threshold.
[0069] Step 160: Determine multiple sets of training data based on the multiple sets of drone simulation signals, wherein the multiple sets of training data correspond to the multiple sets of drone simulation signals in a one-to-one manner.
[0070] Specifically, for each group of drone simulation signals, the training data corresponding to the group of drone simulation signals can be determined through the following steps b1 to b3.
[0071] Step b1: Perform fast Fourier transform calculation on the UAV simulation signal to obtain a two-dimensional time-frequency complex array.
[0072] Among them, since the drone simulation signal is based on the digital signal obtained by sampling, a fast Fourier transform calculation is performed on it to convert the digital signal from a one-dimensional time complex sequence into a two-dimensional time-frequency complex array.
[0073] Step b2: performing a modular square operation on each complex number in the two-dimensional time-frequency complex number array to obtain a two-dimensional time-frequency power array corresponding to the two-dimensional time-frequency complex number array.
[0074] The power corresponding to the complex number can be obtained by performing modular square operations on the complex numbers in the two-dimensional time-frequency complex array.
[0075] Step b3: Perform data aggregation processing on the two-dimensional time-frequency power array to obtain training data corresponding to the UAV simulation signal.
[0076] Specifically, this step can be implemented through the following steps b31 to b33.
[0077] Step b31: Divide the two-dimensional time-frequency power array into k n*m data blocks.
[0078] In the time dimension, each data block includes n consecutive data, and in the frequency dimension, each data block includes m consecutive data. k, m, and n are all positive integers, and the values of k, m, and n can be determined as needed.
[0079] Step b32: summing the data in each data block to obtain k sum values corresponding to the k data blocks.
[0080] Step b33: Replace k data blocks with k sum values respectively to obtain training data corresponding to the UAV simulation signal.
[0081] In the embodiments of the present application, by performing data aggregation processing on the two-dimensional time-frequency power array, the data volume of the two-dimensional time-frequency power array can be effectively reduced. Furthermore, by replacing each data block with the sum of the data in that data block, the data volume of the two-dimensional time-frequency power array can be reduced while also ensuring that the accuracy of the resulting training data meets the requirements, thereby avoiding the situation where the resulting training data suffers from significant distortion due to the aggregation processing of the two-dimensional time-frequency power array.
[0082] In order to improve the efficiency of determining the training data, in some embodiments, the maximum value in each data block may be used to replace the data block to obtain the training data.
[0083] Figure 5 Schematic diagram of a set of two-dimensional time-frequency diagram data provided by an embodiment of the present application is shown. Figure 5 As shown in Figure 2, the two-dimensional time-frequency graph data is a set of training data. The signal parameters corresponding to the two-dimensional time-frequency graph data can be found in Table 2. It is worth noting that the array data obtained after data aggregation processing of the two-dimensional time-frequency power array and the two-dimensional time-frequency graph data are different only in the data representation format, and the two contain the same content.
[0084] Table 2: Signal parameters corresponding to the two-dimensional time-frequency graph data
[0085]
[0086]
[0087] In actual application scenarios, after the receiving device receives the signal transmitted by the target drone, the drone detection model is required to detect the target drone from the signal received by the receiving device. In an embodiment of the present application, multiple sets of different drone simulation signals are simulated based on the sampling parameters of the receiving device and the signal parameters of the signal transmitted by the target drone, and then the training data of the drone detection model is determined based on the drone simulation signals. Through the above-mentioned digital twin simulation method, combined with the characteristics that some parameters in the drone signal are fixed values and some parameters are variable values, since there is no need to actually control the drone flight, the training data for the drone detection model training can be generated efficiently and quickly, thereby improving the efficiency of generating training data for the drone detection model and reducing the difficulty of obtaining training data.
[0088] Moreover, in the embodiment of the present application, since the drone simulation signal is not generated arbitrarily, but is obtained by imitating the drone signal received by the receiving device as the imitation object, by traversing to obtain multiple traversal item parameters, each group of drone simulation signals corresponds to the drone signal received by the receiving device when the target drone flies in a certain environment and scenario, and the multiple groups of drone simulation signals in the embodiment of the present application correspond to the drone signals received by the receiving device when the target drone flies in different environments and scenarios. That is to say, the training data determined based on the drone simulation signal in the embodiment of the present application covers the drone signals received by the receiving device when the target drone flies in different environments and scenarios. Therefore, after using the training data to train the drone detection model, the generalization ability of the drone detection model can be improved.
[0089] Figure 6 A structural diagram of a training data generating device provided in an embodiment of the present application is shown. The specific embodiment of the present application does not limit the specific implementation of the training data generating device.
[0090] like Figure 6 As shown, the training data generating device 200 may include: a processor 202 and a memory 204 .
[0091] The memory 204 is used to store a computer program 206. The memory 204 may include a high-speed RAM memory, or may also include a non-volatile memory, such as at least one disk memory. The computer program 206 may include computer-executable instructions.
[0092] The processor 202 is configured to execute the computer program 206 to implement the above-mentioned training data generation method embodiment.
[0093] Processor 202 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in training data generation device 200 may be processors of the same type, such as one or more CPUs, or may be processors of different types, such as one or more CPUs and one or more ASICs.
[0094] Figure 7 A schematic structural diagram of a drone detection device provided in an embodiment of the present application is shown. The specific embodiments of the present application do not limit the specific implementation of the drone detection device.
[0095] like Figure 7 As shown, the drone detection device 300 is deployed with a drone detection model 301. The drone detection model 301 is a model obtained by training an original model using training data, and the training data is data generated using the above-mentioned training data generation method.
[0096] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned training data generation method embodiment is implemented.
[0097] An embodiment of the present application provides a computer program, which can be executed by a processor to implement the above-mentioned training data generation method embodiment.
[0098] An embodiment of the present application provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned training data generation method embodiment is implemented.
[0099] In the several embodiments provided in this application, if any function is implemented in the form of a software function module / unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the technical solution of this application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or other electronic device) to execute all or part of the steps of the method described in each embodiment of this application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store computer program code.
[0100] The algorithm or demonstration provided here are not inherently relevant to any particular computer, virtual system or other equipment. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present application embodiment is not directed to any specific programming language yet. It should be understood that various programming languages can be utilized to realize the content of the present application described here, and the above description of specific languages is for the purpose of disclosing the best mode of implementation of the present application.
[0101] It should be noted that the above embodiments illustrate rather than limit the present application, and that a person skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between brackets should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present application may be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In claims that list several means, several units or modules of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments should not be understood as limiting the order of execution unless otherwise specified.
[0102] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.
Claims
1. A method for generating training data for training a drone detection model, characterized in that: The method comprises: Obtaining sampling parameters when a receiving device samples a drone signal and signal parameters of a target drone transmission signal, wherein the drone detection model is used to detect the target drone based on the drone signal sampled by the receiving device, and the signal parameters include fixed item parameters and ergodic item parameters; Obtaining multiple groups of baseband signals based on the sampling parameters, the fixed item parameters, and the multiple traversal item parameters obtained through traversal; Simulating a plurality of sets of drone reception signals received by the receiving device according to the plurality of sets of baseband signals; Determining multiple groups of interference signals, wherein any two groups of interference signals among the multiple groups of interference signals differ in at least one parameter of the interference signal pattern, signal bandwidth, signal frequency deviation, and signal strength; Selecting a group of drone received signals and a group of interference signals from the multiple groups of drone received signals and the multiple groups of interference signals, respectively, superimposing the selected drone received signals and interference signals to obtain a group of drone simulation signals, and repeating this step to obtain multiple different groups of drone simulation signals; A plurality of sets of training data are determined based on the plurality of sets of drone simulation signals, wherein the plurality of sets of training data correspond one-to-one to the plurality of sets of drone simulation signals.
2. The method according to claim 1, characterized in that The sampling parameters include a base sampling rate and a sampling duration; The fixed item parameters include one or more of a signal period and a signal frame format, and the traversal item parameters include one or more of a signal bandwidth, a signal frequency deviation and a signal modulation mode.
3. The method according to claim 1, characterized in that The training data corresponding to each set of drone simulation signals is determined by the following steps: Perform fast Fourier transform calculation on the UAV simulation signal to obtain a two-dimensional time-frequency complex array; Performing a modular square operation on each complex number in the two-dimensional time-frequency complex number array to obtain a two-dimensional time-frequency power array corresponding to the two-dimensional time-frequency complex number array; Data aggregation processing is performed on the two-dimensional time-frequency power array to obtain training data corresponding to the UAV simulation signal.
4. The method according to claim 3, characterized in that The performing data aggregation processing on the two-dimensional time-frequency power array to obtain training data corresponding to the UAV simulation signal includes: Divide the two-dimensional time-frequency power array into k n*m data blocks, wherein each data block includes n consecutive data in the time dimension and m consecutive data in the frequency dimension; performing summation processing on the data in each of the data blocks respectively to obtain k sum values corresponding to the k data blocks; The k data blocks are replaced respectively with the k sum values to obtain training data corresponding to the UAV simulation signal.
5. The method according to claim 1, wherein The simulating, based on the multiple groups of baseband reference signals, multiple groups of drone reception signals received by the receiving device includes: performing power normalization processing on each of the multiple groups of baseband signals to obtain multiple groups of baseband reference signals corresponding to the multiple groups of baseband signals; determining a noise power of the receiving device and a plurality of drone signal powers; A group of baseband reference signals and a drone signal power are selected from the multiple groups of baseband reference signals and the multiple drone signal powers, the selected baseband reference signal and the drone signal power are multiplied to obtain a product, the product and the noise power are added to obtain a group of drone received signals, and this step is repeated to obtain multiple groups of drone received signals.
6. The method according to claim 5, characterized in that Determining the noise power of the receiving device and the power of the multiple drone signals includes: The signal powers of the multiple drones are determined according to the equivalent radiated power of the target drone, the receiving antenna gain of the receiving device, the signal operating frequency wavelength, the transceiver spacing and the link active gain.
7. The method according to claim 6, characterized in that The determining of the signal powers of the multiple drones according to the equivalent radiated power of the target drone, the receiving antenna gain of the receiving device, the signal operating frequency wavelength, the transceiver spacing, and the link active gain includes: Selecting one parameter from the equivalent radiated power, the signal operating frequency wavelength, and the transmit / receive distance as a variable, and determining the other parameters as fixed values; Determine multiple drone signal powers based on the fixed value and multiple values of the variable.
8. A training data generating device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the training data generating method according to any one of claims 1 to 7.
9. A drone detection device, characterized in that: The drone detection device is deployed with a drone detection model, which is a model obtained by training the original model using training data, and the training data is data generated using the training data generation method described in 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 training data generating method according to any one of claims 1 to 7 is implemented.