An unmanned aerial vehicle individual identification method based on multi-radio frequency fingerprint joint detection

Abstract

The application relates to the technical field of unmanned aerial vehicle identification, and discloses a method for identifying unmanned aerial vehicle individuals based on multi-radio frequency fingerprint joint detection, which comprises a ground monitoring end and a monitoring unmanned aerial vehicle wirelessly connected and used for monitoring in the air, a control signal module used for bidirectional transmission of control signals is arranged on the monitoring unmanned aerial vehicle, and a picture transmission signal module used for unidirectional transmission is arranged on the monitoring unmanned aerial vehicle; the picture transmission signal sent by the monitoring unmanned aerial vehicle in the air is received at the ground monitoring end. Convolutional neural network is used to identify the generated time-frequency data matrix, to determine whether the unmanned aerial vehicle exists; if the unmanned aerial vehicle exists, the model of the unmanned aerial vehicle is identified; if the unmanned aerial vehicle does not exist, a frequency hopping signal is extracted, the extracted frequency hopping signal is intercepted, segmented, and subjected to Hilbert transformation, the RF-DNA of the intercepted signal is extracted, the characteristics of the RF-DNA are classified, and the individual characteristics of the picture transmission module of the unmanned aerial vehicle are identified; the multi-stage classification identification limits the individual identification algorithm within the same model of unmanned aerial vehicle, and reduces the error probability of individual identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle identification, in particular to an unmanned aerial vehicle individual identification method based on multi-radio frequency fingerprint joint detection. BACKGROUND

[0002] With the popularity of commercial small unmanned aerial vehicles, the use of unmanned aerial vehicles is also increasing. In order to protect the airspace safety of a specific area, it is necessary to discover and identify the invading unmanned aerial vehicles in time. At the same time, since the legal use of unmanned aerial vehicles is also very extensive, it is necessary to accurately identify the model and individual identity of the unmanned aerial vehicles, so as to manage the unmanned aerial vehicles differently.

[0003] The control signals used by unmanned aerial vehicles are mostly in conventional civilian frequency bands such as 2.4GHz and 5.8GHz. With the rapid development of open source hardware and the popularity of software radio technology, the technology of cracking unmanned aerial vehicle radio communication protocols has gradually developed. This technology is one of the advanced technologies in the field of unmanned aerial vehicle countermeasures at home and abroad. By cracking the unmanned aerial vehicle signal communication protocol, the control signal is sent to the unmanned aerial vehicle by simulating the remote control, and it will not affect the normal operation of other equipment. However, with the improvement of communication encryption technology, the cracking difficulty is increasing, and it is required to adapt to various unmanned aerial vehicles on the market, which needs to be updated regularly. The unmanned aerial vehicle model is sold, and the cost is high.

[0004] Different manufacturers and different models of unmanned aerial vehicles will use the same model of radio frequency chip. With the progress of chip technology, the difference of the same model of chip will be smaller and smaller, which will increase the error rate of individual identification. At the same time, due to the interference of space radio clutter and the reduction of signal-to-noise ratio, it will also increase the error rate. In order to accurately identify the individual of the unmanned aerial vehicle, the radio frequency fingerprint identification technology can be used. The radio frequency signal emitted by the unmanned aerial vehicle has undergone a series of processes such as encoding, modulation, frequency conversion, switching, etc. Due to the fact that the devices used by each unmanned aerial vehicle cannot be completely consistent, there will be certain differences in performance parameters. These device differences make the radio frequency signal of each unmanned aerial vehicle have its own characteristics. By collecting and processing the radio frequency signal of the unmanned aerial vehicle, the unique fingerprint features of each unmanned aerial vehicle can be extracted. Through machine learning, the individual of the unmanned aerial vehicle can be accurately identified in a large amount of information.

[0005] At present, the cognitive radio protocol cracking technology mainly judges whether the signal is a unmanned aerial vehicle by comparing the passively received environmental radio signal and the unmanned aerial vehicle signal model, and identifies the individual identity of the unmanned aerial vehicle through the link data information. Data cracking has a lag, and the unmanned aerial vehicle manufacturer will change the link protocol or use encryption to transmit, which makes the protocol cracking difficult and involves multiple disciplines, making it difficult to crack the link. Therefore, a unmanned aerial vehicle individual identification method based on multi-radio frequency fingerprint joint detection is needed. SUMMARY

[0006] The purpose of the present application is to provide a multi-radio fingerprint joint detection based unmanned aerial vehicle individual identification method, and the present application distinguishes each unmanned aerial vehicle through individual characteristics, which will increase the error rate. Through multi-level classification identification, the individual identification algorithm is limited within the same model of unmanned aerial vehicle, thereby reducing the error probability of individual identification.

[0007] The present application is implemented in the following steps:

[0008] S1: including a ground monitoring end and a monitoring unmanned aerial vehicle in the air for monitoring, a control signal module for bidirectional transmission of control signals is arranged on the monitoring unmanned aerial vehicle, and a picture transmission signal module for unidirectional transmission, the ground monitoring end receives the picture transmission signal sent by the monitoring unmanned aerial vehicle in the air, and samples the signal after band pass filtering;

[0009] S2: the sampled signal is segmented, and short-time Fourier transform is performed on each segment to obtain a time-frequency signal matrix; specifically as formula (1);

[0010] Formula (1)

[0011] Wherein, is the original sampling signal, is the discrete time, is the angular frequency, is the window function;

[0012] S3: using a convolutional neural network to identify the time-frequency data matrix generated in step S2, to determine whether there is an unmanned aerial vehicle, if there is, to identify the model of the unmanned aerial vehicle;

[0013] S4: if not, extract the frequency hopping signal, intercept, segment and perform Hilbert transform on the extracted frequency hopping signal, extract the RF-DNA of the intercepted signal, and then classify its characteristics through k-nearest neighbor algorithm to identify the individual characteristics of the radio frequency module of the unmanned aerial vehicle control signal;

[0014] S5: extract the OFDM signal cyclic prefix, intercept, segment and perform Hilbert transform on the separated signal sequence, extract the RF-DNA fingerprint of the intercepted signal, and then classify its characteristics through k-nearest neighbor algorithm to identify the individual characteristics of the unmanned aerial vehicle picture transmission module;

[0015] S6: integrate the classification and identification results in steps S3-S5, and combine the unmanned aerial vehicle model and individual database of the unmanned aerial vehicle manufacturer to determine the identity of the unmanned aerial vehicle individual.

[0016] According to the identification results of the three aspects of the model of the unmanned aerial vehicle, the individual characteristics of the unmanned aerial vehicle control signal radio frequency module, and the individual characteristics of the unmanned aerial vehicle signal radio frequency module, the results of the unmanned aerial vehicle individual identification are determined, and the determination basis for a specific unmanned aerial vehicle individual A includes the spectral waterfall diagram classification identification result, the frequency hopping signal classification identification result, and the OFDM signal classification identification result.

[0017] The spectral waterfall diagram classification identification result includes two cases of conforming to the common characteristics of the unmanned aerial vehicle signal but not conforming to the model characteristics of the unmanned aerial vehicle A, and conforming to the model characteristics of the unmanned aerial vehicle A.

[0018] The frequency hopping signal classification identification result includes two cases of conforming to the individual characteristics of the unmanned aerial vehicle A and not conforming to the individual characteristics of the unmanned aerial vehicle A.

[0019] The OFDM signal classification identification result includes two cases of conforming to the individual characteristics of the unmanned aerial vehicle A and not conforming to the individual characteristics of the unmanned aerial vehicle A.

[0020] Further, in step S4, the RF-DNA method of the frequency hopping signal is specifically executed according to the following steps:

[0021] S 4。1 : After removing the direct current component from the original sampling signal, taking the absolute value, using the triangular threshold segmentation method to segment the signal, obtaining the time period in which the effective signal appears (there may be multiple time periods);

[0022] S 4。2 : Extract the time period between 0.1ms and 10ms in length, check the signal bandwidth of the time period in the time-frequency matrix, and if it is between 0.5MHz and 3MHz, it is considered to be a frequency hopping signal of the unmanned aerial vehicle;

[0023] S 4。3 : The signal sequence corresponding to the time period is intercepted from the original sampling;

[0024] S 4。4 : The original sampling signal in the time period is intercepted, segmented, and RF-DNA is extracted.

[0025] Specifically executed according to the following steps:

[0026] First, after removing the direct current component from the original sampling signal, taking the absolute value, using the triangular threshold segmentation method to segment the signal, obtaining the time period in which the effective signal appears, wherein the time period is at least one time period;

[0027] Extract the time period between 0.1ms and 10ms in length, check the signal bandwidth of the time period in the time-frequency matrix, and if it is between 0.5MHz and 3MHz, it is determined to be a frequency hopping signal of the unmanned aerial vehicle;

[0028] Extracting the signal sequence corresponding to the time period from the original sampling; the original sampling signal in the time period is intercepted, segmented, and RF-DNA is extracted.

[0029] Further, in step S 3.4 , the method is specifically implemented as follows:

[0030] S 4.1 : Hilbert transform is performed on the original sampling signal sequence in the time period, wherein the specific transformation of the signal includes signal amplitude, phase and frequency;

[0031] S 4.2 : The variance, skewness and kurtosis values of the signal are obtained through the transformation of the signal amplitude, phase and frequency.

[0032] Further, in step S5, the method for extracting the OFDM signal is specifically implemented as follows:

[0033] S 5。1 : Find the time period in which the signal with a bandwidth not less than 10MHz, including 2.4GHz and 5.8GHz frequency bands, or not less than 5MHz, including 800MHz and 900MHz frequency bands, appears, then intercept the original sampling sequence in the time period, and perform autocorrelation operation to check whether there is a periodic autocorrelation peak value, if there is, record the distance between adjacent peak values as k, otherwise, eliminate the signal;

[0034] S 5。2 : A rectangular window with a length of k is used to intercept the signal, and the sampling signal is x(n) , and the rectangular window function is w(n), as shown in formula (2)-formula (3);

[0035] Formula (2)

[0036] Formula (3)

[0037] Wherein, n is the independent variable.

[0038] Further, the application provides a computer readable storage medium, which stores a computer program, and the program is executed by a host controller to realize the method according to any one of the above methods.

[0039] Compared with the prior art, the application has the following advantages:

[0040] 1. The model and individual of the unmanned aerial vehicle can be identified, and the probability of misjudgment can be significantly reduced through three radio frequency fingerprint characteristics, so that the individual of the unmanned aerial vehicle can be accurately identified, and the differentiated control of the unmanned aerial vehicle is provided.

[0041] 2. Distinguishing each drone by individual characteristics will lead to an increase in error rate. By multi-level classification recognition, the individual recognition algorithm is limited within the same model of drones, reducing the error probability of individual recognition. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0043] Figure 1 is the CNN structure and parameter diagram of the present application;

[0044] Figure 2 is the RF-DNA feature extraction diagram of the present application;

[0045] Figure 3 is the waveform diagram of the function y(n) of the present application

[0046] Figure 4 is the system method flow diagram of the present application. DETAILED DESCRIPTION

[0047] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0048] Please refer to Figures 1-4 , a multi-radio fingerprint joint detection based drone individual recognition method,

[0049] In the present embodiment,

[0050] The specific execution is as follows:

[0051] S1: including ground monitoring end and wireless connection in air monitoring monitoring unmanned plane, on the monitoring unmanned plane is equipped with control signal module for bidirectional transmission control signal, and unidirectional transmission of image transmission signal module, in the ground monitoring end receives the image transmission signal sent by the air monitoring unmanned plane, and after band pass filtering, the signal is sampled;

[0052] S2: the sampled signal is segmented, and short-time Fourier transform is performed on each segment to obtain a time-frequency signal matrix; specifically as formula (1);

[0053] Formula (1)

[0054] Among them, is the original sampling signal, is the discrete time, is the angular frequency, is the window function;

[0055] S3: as Figure 1 , using convolution neural network to identify the time-frequency data matrix generated in step S2, to determine whether there is unmanned plane, if there is, the model of unmanned plane is identified;

[0056] S4: if not, extract the frequency hopping signal, intercept, segment and perform Hilbert transform on the extracted frequency hopping signal, extract the RF-DNA of the intercepted signal, and then classify its characteristics through k nearest neighbor algorithm to identify the individual characteristics of the unmanned plane control signal radio frequency module;

[0057] S5: extract the OFDM signal cyclic prefix, intercept, segment and perform Hilbert transform on the separated signal sequence, extract the RF-DNA fingerprint of the intercepted signal, and then classify its characteristics through k nearest neighbor algorithm to identify the individual characteristics of the unmanned plane image transmission module;

[0058] S6: comprehensive classification and identification results in steps S3-S5, combined with the unmanned plane model, individual and other databases of unmanned plane manufacturers to determine the identity of the unmanned plane individual.

[0059] According to the identification results of the model of unmanned plane, the individual characteristics of unmanned plane control signal radio frequency module and the individual characteristics of unmanned plane image signal radio frequency transmission module, the result of unmanned plane individual identification is determined, and the determination basis for a specific unmanned plane individual A includes the classification and identification results of spectrum waterfall diagram, frequency hopping signal and OFDM signal;

[0060] Among them, the classification and identification results of spectrum waterfall diagram include two cases that meet the common characteristics of unmanned plane signal, but do not meet the model characteristics of unmanned plane A, and meet the model characteristics of unmanned plane A;

[0061] The frequency hopping signal classification identification result includes two cases: the individual characteristics conforming to the UAV A and the individual characteristics not conforming to the UAV A.

[0062] The OFDM signal classification identification result includes two cases: the individual characteristics conforming to the UAV A and the individual characteristics not conforming to the UAV A; as shown in Table 1.

[0063] Table 1 Decision rule for whether the UAV individual A is found

[0064]

[0065] Further, as Figure 2 In step S4, the RF-DNA method for extracting the frequency hopping signal is specifically performed according to the following steps:

[0066] S 4。1 : After removing the direct current component from the original sampling signal, taking the absolute value, using the triangular threshold segmentation method to segment the signal, obtaining the time period in which the effective signal appears (there may be multiple time periods);

[0067] S 4。2 : Extract the time period with a duration of 0.1ms-10ms, check the signal bandwidth of the time period in the time-frequency matrix, if it is between 0.5MHz-3MHz, it is considered to be a UAV frequency hopping signal;

[0068] S 4。3 : Extract the signal sequence corresponding to the time period from the original sampling;

[0069] S 4。4 : The original sampling signal in the time period is intercepted, segmented, and RF-DNA is extracted.

[0070] Specifically, the following steps are performed:

[0071] First, remove the direct current component from the original sampling signal, take the absolute value, use the triangular threshold segmentation method to segment the signal, obtain the time period in which the effective signal appears, and the time period is at least one time period;

[0072] Extract the time period with a duration of 0.1ms-10ms, check the signal bandwidth of the time period in the time-frequency matrix, if it is between 0.5MHz-3MHz, it is determined to be a UAV frequency hopping signal;

[0073] Extract the signal sequence corresponding to the time period from the original sampling; the original sampling signal in the time period is intercepted, segmented, and RF-DNA is extracted.

[0074] Further, in step S 3.4 , the following steps are specifically performed:

[0075] S 4.1 : Hilbert transform is performed on the original sampling signal sequence in the time period, wherein the transform of the signal specifically includes signal amplitude, phase and frequency;

[0076] S 4.2 : the variance, skewness and kurtosis values of the signal are obtained through the transform of the signal amplitude, phase and frequency.

[0077] Further, in step S5, the method for extracting the OFDM signal is specifically executed according to the following steps:

[0078] S 5。1 : a time period in which a signal with a bandwidth not less than 10MHz, including the 2.4GHz and 5.8GHz frequency bands, or not less than 5MHz, including the 800MHz and 900MHz frequency bands, appears is found, then the original sampling sequence in the time period is intercepted, and autocorrelation operation is performed to check whether there is a periodic autocorrelation peak value, if there is, the distance between adjacent peak values is recorded as k, otherwise the signal is rejected;

[0079] S 5。2 : a rectangular window with a length of k is used to intercept the signal, and the sampling signal is x(n) , and the rectangular window function is w(n), as shown in formula (2) to formula (3);

[0080] Formula (2)

[0081] Formula (3)

[0082] Wherein, n is the independent variable. The waveform is approximately trapezoidal, as shown in the figure. The upper side length of the trapezoid is approximately an integer multiple of k. Within the range of the independent variable n corresponding to the upper side of the trapezoid, an integer multiple of the original signal Figure 3 is intercepted, and the sequence is segmented in units of k to obtain a plurality of segmented sequences, and the RF-DNA feature k is extracted. x n In this embodiment, the application provides a computer readable storage medium, which stores a computer program, and the program is executed by a host controller to realize the method including but not limited to any one of the above methods.

[0083]

[0084] ​​The above merely describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for identifying a UAV individual based on multi-radio frequency fingerprint joint detection, characterized in that: Specifically, the following steps are performed: S1: a ground monitoring end and a monitoring unmanned aerial vehicle (UAV) for monitoring in the air are connected wirelessly, the monitoring UAV is provided with a control signal module for bidirectional transmission of control signals and a unidirectional transmission of image transmission signal module, the image transmission signal module is configured to receive the image transmission signal transmitted by the monitoring UAV in the air, and the received signal is sampled after band pass filtering; S2: the sampled signal is segmented, and a short-time Fourier transform is performed on each segment to obtain a time-frequency signal matrix; Specifically, as shown in formula (1); Formula (1) wherein is the original sampled signal, is the discrete time, is the angular frequency, is the window function; S3: a convolutional neural network is used to identify the time-frequency data matrix generated in step S2 to determine whether the UAV exists, and if the UAV exists, the model of the UAV is identified; S4: if the UAV does not exist, a frequency hopping signal is extracted, the extracted frequency hopping signal is intercepted, segmented, and subjected to Hilbert transform, the RF-DNA of the intercepted signal is extracted, and the characteristics of the RF-DNA are classified by using a k-nearest neighbor algorithm to identify the individual characteristics of the UAV control signal radio frequency module; the RF-DNA extraction method of the frequency hopping signal is performed according to the following steps: S 4.1 : The absolute value of the original sampling signal after removing the DC component is taken, and the signal is segmented using the triangular threshold segmentation method to obtain the time period when the effective signal appears; S 4.2 : Extract the time period between 0.1ms~10ms, check the signal bandwidth of the time period in the time-frequency matrix, if between 0.5MHz~3MHz, consider it as a UAV frequency hopping signal; S 4.3 : truncate the signal sequence corresponding to the time period from the original sample; S 4.4 : The original sampled signal in this time period is intercepted, segmented, and RF-DNA is extracted, which is specifically performed according to the following steps: First, the original sampling signal is removed from the direct current component and the absolute value is taken, the signal is segmented using a triangular threshold segmentation method, and the time period in which the effective signal appears is obtained, wherein the time period is at least one time period; Then, a time period with a length of 0.1ms-10ms is extracted, and the signal bandwidth of the time period is checked in the time-frequency matrix, if the signal bandwidth is between 0.5MHz-3MHz, it is determined that it is a UAV frequency hopping signal; the signal sequence corresponding to the time period is intercepted from the original sampling; The original sampling signal in the time period is intercepted, segmented, and the RF-DNA is extracted according to the following steps: the original sampling signal sequence in the time period is subjected to Hilbert transform, wherein the specific transformation of the signal includes signal amplitude, phase and frequency; the variance, skewness and kurtosis values of the signal are obtained through the transformation of the signal amplitude, phase and frequency; S5: the OFDM signal cyclic prefix is extracted, the separated signal sequence is intercepted, segmented, and subjected to Hilbert transform, the RF-DNA fingerprint of the intercepted signal is extracted, and the characteristics of the RF-DNA fingerprint are classified by using a k-nearest neighbor algorithm to identify the individual characteristics of the UAV image transmission module; the following steps are performed: S 5.1 : find the time period in which the signal with bandwidth no less than 10MHz, including 2.4GHz and 5.8GHz frequency band, or no less than 5MHz, including 800MHz and 900MHz frequency band, then intercept the original sampling sequence in the time period, and perform autocorrelation operation, check whether there is a periodic autocorrelation peak, if there is, record the distance between adjacent peaks as k, otherwise, eliminate the signal; S 5.2 : using a rectangular window of length k to truncate the signal, let the sampled signal be x(n) , the rectangular window function is w(n), as in equation (2) - equation (3); Equation (2) Equation (3) Wherein, n is the independent variable; S6: the classification and identification results in steps S3-S5 are integrated, and the identity of the UAV individual is determined in combination with the UAV model, individual database of the UAV manufacturer. 2.The method of claim 1, wherein, In step S6, the recognition results of the UAV model, the individual characteristics of the UAV control signal radio frequency module, and the individual characteristics of the UAV image signal radio frequency transmission module are combined to determine the result of the UAV individual recognition. For a specific UAV individual A, the determination basis includes the spectral waterfall graph classification identification result, the frequency hopping signal classification identification result, and the OFDM signal classification identification result. The spectral waterfall graph classification identification result includes two cases: the common characteristics of the UAV signal are consistent, but the model characteristics of the UAV A are not consistent, and the model characteristics of the UAV A are consistent. The frequency hopping signal classification recognition result includes two cases: individual characteristics conforming to the UAV A and individual characteristics not conforming to the UAV A. The OFDM signal classification recognition result includes two cases: individual characteristics conforming to the UAV A and individual characteristics not conforming to the UAV A.

3. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by the host controller, implements a method including any of claims 1-2.

Images