A method and system for identifying a remote control signal of a drone based on a four-dimensional code

Abstract

The application discloses a kind of unmanned plane remote control signal identification method and system based on four-dimensional code, the method includes: the analog signal of unmanned plane is collected and is merged, obtains time domain signal, the combined time-frequency spectrum obtained by the combined time-frequency transform of time domain signal, the combined time-frequency spectrum obtained is based on singular value decomposition denoising, extracts unmanned plane remote control signal using low rank matrix recovery and based on four-dimensional code time-frequency spectrum similarity search, unmanned plane remote control signal identification is carried out.The application solves the technical problems that the difficulty of extracting the shallow features of artificially designed signals in the prior art is greater, and the identification of unmanned plane is not accurate enough.The application uses the method of combined time-frequency transform to analyze the time-frequency of unmanned plane collection signal, uses singular value decomposition for noise reduction processing, and estimates the similarity of time-frequency spectrum into four-dimensional code similarity search by mapping four-dimensional code.

Description

TECHNICAL FIELD

[0001] The present application relates to a UAV identification system and method, in particular to a UAV remote control signal identification method and system based on a four-dimensional code. BACKGROUND

[0002] With the continuous development of UAV technology, the application of UAVs is becoming more and more widespread, from aerial photography, patrol to military reconnaissance, and plays a role in various fields. However, the emergence of UAVs has also brought certain safety hazards. Therefore, how to accurately identify UAVs has become one of the current research hotspots.

[0003] As an important signal processing method, time-frequency spectrum can integrate time domain and frequency domain information of a signal together, and can more comprehensively and intuitively reflect the characteristics of the signal, so it has been widely used in the field of signal processing. In UAV identification, time-frequency spectrum can also be used to describe the characteristics of UAV communication signals. Specifically, UAV communication signals usually have certain modulation characteristics, such as frequency, amplitude, phase, etc., which can be represented by time-frequency spectrum. Document 1 (Li Guangwei. Application of spectrum detection technology in the field of UAV detection and countermeasures [J]) summarizes the application and prospect of electromagnetic spectrum measurement technology in UAV detection and countermeasures, and gives several detection methods, but no specific algorithm verification. Document 2 (Liu Li. Analysis and identification technology of civilian UAV frequency hopping signal [D]) proposes a classification system for frequency hopping signals, which can realize automatic classification of UAV frequency hopping signals, but this method requires a large amount of UAV sample data and the algorithm is complex. The above methods are all realized by extracting artificially designed shallow signal features, and the feature design depends on the experience of researchers, which is difficult to extract, and the recognition dimension is small and the recognition information capacity is also small.

[0004] Four-dimensional code technology has gradually begun to be applied in various aspects of production and life, such as production line part automatic tracking, product traceability, medical emergency, e-commerce, media tourism, etc. However, these four-dimensional code applications are concentrated in recording numbers, letters and words, how to use four-dimensional code for time-frequency spectrum mapping and storage, and use the fast response characteristics of four-dimensional code to realize the rapid and accurate identification of UAV models has important significance. SUMMARY

[0005] The application aims to provide a four-dimensional code-based unmanned aerial vehicle remote control signal recognition method and system, and solve the technical problems of great difficulty in extracting artificial design signal shallow features and inaccurate unmanned aerial vehicle recognition in the prior art.

[0006] To achieve the above-mentioned purpose, the application provides a four-dimensional code-based unmanned aerial vehicle remote control signal recognition method, which comprises the following steps:

[0007] S100, collecting analog signals of an unmanned aerial vehicle and merging to obtain a time domain signal;

[0008] S200, performing combined time-frequency transformation on the obtained time domain signal to obtain a combined time-frequency spectrum TFR 组合 ;

[0009] The combined time-frequency transformation method is as follows:

[0010] S201, performing Chirp-William transformation on the obtained time domain signal to obtain a time-frequency spectrum TFR CWD ;

[0011] S202, performing short-time Fourier transformation on the obtained time domain signal to obtain a time-frequency spectrum TFR STFT ; adopting a threshold segmentation method to perform truncation processing on the time-frequency spectrum TFR STFT , to obtain a truncated time-frequency spectrum TFR 截断 :

[0012]

[0013] In the formula, max() represents the maximum value of elements, || represents the absolute value, ε is a threshold value, here ε = μ·mean(|TFR STFT (t,f)|); μ is a threshold factor, and μ = 0.15; mean() represents the mean value; t is a time variable, and f is a frequency variable.

[0014] S203, performing Hadamard product on the truncated time-frequency spectrum TFR 截断 and the time-frequency spectrum TFR CWD to obtain a combined time-frequency spectrum TFR 组合 .

[0015] TFR 组合 (t,f) = TFR 截断 (t,f)☉TFR CWD (t,f)

[0016] where is Hadamard product operation, t is time variable, and f is frequency variable;

[0017] S300, performing singular value decomposition on the obtained combined time-frequency spectrum TFR 组合 based on singular value decomposition denoising;

[0018] The singular value decomposition denoising method based on singular value decomposition is as follows:

[0019] S301, performing singular value decomposition on the combined time-frequency spectrum TFR 组合 based on singular value decomposition;

[0020] The combined time-frequency spectrum TFR 组合 is decomposed by singular value decomposition (SVD), that is, TFR 组合 = UΣV H , wherein U and V are both orthogonal matrices, and Σ is a diagonal matrix, the elements on the diagonal are singular values of the combined time-frequency spectrum TFR 组合 , and the upper index H represents conjugate transpose;

[0021] S302, selecting a suitable threshold to perform singular value filtering;

[0022] In the SVD decomposition, the size of the singular value of the combined time-frequency spectrum TFR 组合 represents the energy of the combined time-frequency spectrum TFR 组合 at each frequency, a suitable threshold is selected, and the frequencies corresponding to the singular values less than the threshold are filtered out to obtain filtered U and V, so as to achieve the purpose of denoising;

[0023] The singular values in the matrix Σ that are lower than the threshold are set to 0, and these singular values are filtered out in the SVD decomposition to obtain filtered , so as to achieve the effect of signal denoising;

[0024] S303, reconstructing the signal;

[0025] The filtered U, and V matrices are multiplied back together to obtain the reconstructed , that is,

[0026] S400, using low-rank matrix recovery to extract the unmanned aerial vehicle remote control signal from the reconstructed

[0027] The low-rank matrix recovery is to express the obtained reconstructed as the sum of a low-rank matrix A and a sparse matrix E, wherein the image transmission signal is distributed in the low-rank matrix A, and the remote control signal is distributed in the sparse matrix E; and the reconstructed​ a remote control signal is obtained; the remote control signal is processed by using an alternating direction multiplier method (ADMM) to obtain a time-frequency spectrum after time-frequency filtering;

[0028] The solving of the norm optimization problem is as follows:

[0029]

[0030] In the formula, min A,E () represents the minimum value for the low-rank matrix A and the sparse matrix E, s.t. represents the constraint condition; ||| * represents the core norm, that is, the sum of singular values of the matrix; |||0 represents the 0 norm, that is, the sum of absolute values of all elements in the matrix; λ is a regularization parameter, and is taken as

[0031] The sparse matrix E is obtained by solving (1) by using the alternating direction multiplier method (ADMM), that is, the time-frequency spectrum after time-frequency filtering;

[0032] S500, based on the time-frequency spectrum similarity search of the four-dimensional code, the unmanned aerial vehicle remote control signal is identified;

[0033] The time-frequency spectrum after time-frequency filtering is mapped into a four-dimensional code, and a four-dimensional code word in a four-dimensional code set of the unmanned aerial vehicle remote control signal is compared with the mapped four-dimensional code by using a four-dimensional code-based time-frequency spectrum similarity search method, to determine whether the model of the corresponding unmanned aerial vehicle is found;

[0034] The method for mapping the time-frequency spectrum after time-frequency filtering into a four-dimensional code is as follows:

[0035] a, obtaining to-be-mapped data, the to-be-mapped data including:

[0036] a contour map of a time-frequency plane, a contour map of a time-time-frequency spectrum plane, and a contour map of a frequency-time-frequency spectrum plane;

[0037] The contour map of the time-frequency plane, the contour map of the time-time-frequency spectrum plane, and the contour map of the frequency-time-frequency spectrum plane are obtained by using the following method:

[0038] The time-frequency spectrum after time-frequency filtering is projected onto a time-frequency plane, a time-time-frequency spectrum plane, and a frequency-time-frequency spectrum plane in sequence to obtain

[0039] b, the to-be-mapped data is mapped into a three-dimensional code group in a png image format

[0040] The to-be-mapped data is mapped into a three-dimensional code group, and the format is converted into a png image format, to obtain a three-dimensional code group in a png format;

[0041] c. integrate the obtained png format three-dimensional code group into a cyclic dynamic display three-dimensional code, i.e. a four-dimensional code.

[0042] The time-frequency spectrum similarity search method based on the four-dimensional code is as follows:

[0043] S501, set an initial identification rate level;

[0044] S502, search for a corresponding four-dimensional code in the four-dimensional code set mapped by the unmanned aerial vehicle remote control signal, if yes, go to step S503; if no, go to step S504;

[0045] S503, the four-dimensional code is the sought four-dimensional code, i.e. the model of the corresponding unmanned aerial vehicle is found;

[0046] S504, reduce the identification rate level, i.e. the current identification rate = identification rate - 5%;

[0047] S505, judge whether the current identification rate level is less than the minimum level, if no, go to step S502; if yes, go to step S506;

[0048] S506, discard the four-dimensional code, i.e. the model of the corresponding unmanned aerial vehicle is not found.

[0049] Preferably, in S100, the data collection and combination in S100 includes:

[0050] A data collection system based on the AD9361 zero intermediate frequency receiver architecture is used to collect the signals of a certain model of DJI unmanned aerial vehicle; an analog / digital device in the receiver performs analog-to-digital conversion on the received analog signals, and converts the received signals into two digital baseband signals of in-phase component I and quadrature component Q through quadrature demodulation and digital down-conversion, then the I and Q signals are combined to obtain a time domain signal sig = I + j·Q, j representing an imaginary unit,

[0051] Preferably, in S200, the combined time-frequency transformation is to analyze the obtained time domain signal with time as the horizontal axis and frequency as the vertical axis, and process Num 采 data points in the collected time domain signal, each group of data being Num 组 points, a total of Num 采 / Num 组 times, each time representing a time point; Num 组 points correspond to the signal frequency within the Bw bandwidth, and each point frequency is Bw / Num 组 ; Num 采 / Num 组The combined time-frequency transform processing is performed each time to obtain the combined time-frequency spectrum TFR 组合 The combined time-frequency spectrum TFR is calculated each time 组合 The combined time-frequency spectrum TFR along the time axis 组合 The maximum value of the combined time-frequency spectrum TFR of each time 组合 The maximum value of the combined time-frequency spectrum TFR of each time 组合 .

[0052] Preferably, in S500, the four-dimensional code comprises a time dimension, a frequency dimension, a color dimension and a playing order dimension.

[0053] More preferably, the time dimension and the frequency dimension are corresponding time and frequency in the time-frequency spectrum after the time-frequency filtering.

[0054] More preferably, the color dimension represents the numerical value of the time-frequency spectrum in the contour graph of the time-frequency plane, the high and low of the frequency in the contour graph of the time-time-frequency spectrum plane and the order of time in the contour graph of the frequency-time-frequency spectrum plane.

[0055] More preferably, the playing order is a dynamic graph formed by cyclically and dynamically playing the three-dimensional code mapped from the contour graph of the time-frequency plane, the three-dimensional code mapped from the contour graph of the time-time-frequency spectrum plane and the three-dimensional code mapped from the contour graph of the frequency-time-frequency spectrum plane.

[0056] The application further provides a four-dimensional code-based unmanned aerial vehicle remote control signal identification system of the identification method.

[0057] The application further provides an application of the identification method in monitoring unmanned aerial vehicle remote control signals.

[0058] The four-dimensional code-based unmanned aerial vehicle remote control signal identification method and system solve the technical problems of the prior art that it is difficult to extract the shallow features of artificially designed signals and the identification of unmanned aerial vehicles is not accurate enough, and have the following advantages:

[0059] 1. The application adopts a combined time-frequency transform method to perform time-frequency analysis on the analog signals collected by the unmanned aerial vehicle, overcoming the limitations brought by single linear or nonlinear time-frequency transform. 截断 The combined transform of the linear truncated time-frequency spectrum TFR CWD and the nonlinear time-frequency spectrum TFR 组合 The combined time-frequency spectrum TFR 截断The property of (t,f) also eliminates the TFR CWD The cross term in (t,f) can more accurately depict the characteristics of the signal in the time-frequency domain, and improve the resolution of the combined time-frequency spectrum TFR 组合 .

[0060] 2. For the noise problem existing in the combined time-frequency spectrum TFR 组合 , the application adopts SVD decomposition for noise reduction processing. SVD can not only help remove noise, but also can preserve important information in the data. This is because SVD is based on mathematical principles, which can ensure that no information is lost in the dimension reduction process.

[0061] 3. The application uses a low-rank matrix recovery method to reduce the interference of the image transmission signal on the remote control signal. The collected signal is first converted to the time-frequency domain by using the combined time-frequency transform, and then according to the correlation of the energy distribution of the remote control signal and the image transmission signal in the time-frequency domain, it is respectively divided into sparse matrix and low-rank matrix in the time-frequency domain, realizing the effective separation of the image transmission signal and the remote control signal, thereby efficiently extracting the unmanned aerial vehicle remote control signal.

[0062] 4. The application adopts a time-frequency spectrum similarity search method based on four-dimensional code. The similarity of the time-frequency spectrum is estimated by mapping the four-dimensional code. The four-dimensional code of the application expands the data information capacity in the spatial dimension, improves the coding capacity of the data information, and expands the time dimension. The cyclic dynamic png format three-dimensional code group generates several frames of three-dimensional codes which are different in content and capacity according to the obtained data information to be coded, which can be regarded as expanding the dimension of the three-dimensional code in the time dimension, and improving the information capacity of the stored data (the principle is: first, the four-dimensional code used in the application is a dynamic three-dimensional code group, that is, three contour maps of time-frequency spectrum, specifically, contour maps of time-frequency plane, time-time-frequency spectrum plane and frequency-time-frequency spectrum plane, wherein each contour map corresponds to a three-dimensional code, and the three three-dimensional codes are displayed dynamically in a certain order; second, the dynamic three-dimensional code group constituting the four-dimensional code of the application is not limited to three three-dimensional codes, and the information capacity of the four-dimensional code of the application is at least three times that of the conventional three-dimensional code; finally, compared with the existing three-dimensional code, the four-dimensional code of the application has a time dimension, which can carry a larger information capacity, so that more unmanned aerial vehicles can be identified. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 The flow chart of the unmanned aerial vehicle identification method based on four-dimensional code mapping of the application.

[0064] Figure 2 The flow chart of the time-frequency spectrum similarity search method based on four-dimensional code of the application.

[0065] Figure 3 A schematic diagram of the unmanned aerial vehicle recognition system based on four-dimensional code mapping of the present application.

[0066] Figure 4 A combined time-frequency spectrum contour map of the unmanned aerial vehicle collected signal of the present application.

[0067] Figure 5 The time-frequency spectrum after time-frequency filtering and the mapped four-dimensional code of the present application.

[0068] Figure 6 A state diagram of the dynamic four-dimensional code of the present application. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only a part 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.

[0070] Embodiment 1

[0071] An unmanned aerial vehicle remote control signal recognition system based on four-dimensional code, as shown in Figure 3 The schematic diagram of the unmanned aerial vehicle recognition system based on four-dimensional code mapping of the present application, which comprises an unmanned aerial vehicle, a remote control station and a monitoring device, wherein the unmanned aerial vehicle transmits a video transmission signal to the ground remote control station, the remote control station transmits a remote control signal to the unmanned aerial vehicle, and the monitoring device collects the remote control signal and the video transmission signal.

[0072] An unmanned aerial vehicle remote control signal recognition method based on four-dimensional code, as shown in Figure 1 The flowchart of the unmanned aerial vehicle remote control signal recognition method based on four-dimensional code mapping of the present application, which comprises:

[0073] S100, data acquisition and merging

[0074] A data acquisition system based on the architecture of an AD9361 zero intermediate frequency receiver is adopted to collect the analog signals of a certain model of DJI unmanned aerial vehicle; the analog / digital device inside the receiver performs analog-to-digital conversion on the received analog signals (including video transmission signals and remote control signals), and converts the analog signals into two digital baseband signals of in-phase component I and quadrature component Q through quadrature demodulation and digital down-conversion, then the I and Q signals are merged to obtain a time domain signal sig = I + j·Q, j representing an imaginary unit,

[0075] S200, combined time-frequency transformation is performed on the obtained time domain signal to obtain a combined time-frequency spectrum TFR 组合

[0076] A combined time-frequency transform (TFR) was used to analyze the obtained time-domain signal with time on the horizontal axis and frequency on the vertical axis. 8000×2048 data points from the acquired time-domain signal were processed, with each group of data consisting of 2048 points, for a total of 8000 analyses (equivalent to 8000 groups). Each analysis represented one time point; the 2048 points corresponded to signal frequencies within a 40MHz bandwidth, with each point having a frequency of 0.019532625MHz. After 8000 TFR analyses, the combined time-frequency spectrum (TFR) was obtained from each output. 组合 Then calculate the TFR spectrum during each combination. 组合 Combined time spectrum (TFR) along the time axis 组合 The maximum value is the combined time-frequency spectrum (TFR) of each iteration. 组合 The maximum values ​​are concatenated to obtain the combined frequency spectrum (TFR) of all sample data. 组合 .

[0077] like Figure 4 As shown, this invention presents a combined time-frequency contour map of the signals collected by the UAV. (a) is a time-frequency distribution map of the DJI Phantom 4, with the three-dimensional coordinates representing time-frequency, frequency, and time, respectively; (b) is a time-frequency contour map of the DJI Phantom 4, with the horizontal axis representing time and the vertical axis representing frequency; (c) is a time-frequency spectrum contour map of the DJI Phantom 4, with the horizontal axis representing time and the vertical axis representing time-frequency; and (d) is a frequency-time-frequency spectrum contour map of the DJI Phantom 4, with the horizontal axis representing time-frequency and the vertical axis representing frequency. Figure 4 It can be seen that the DJI Phantom 4-Combined time spectrum is distributed in three-dimensional space of time, frequency and time spectrum amplitude, and that the three-dimensional time spectrum is projected onto the time-frequency plane, time-time spectrum plane and frequency-time spectrum plane respectively.

[0078] The method for the combined time-frequency transformation processing is as follows:

[0079] S201. Perform Joe-Williams CWD transform on the obtained time-domain signal to obtain the time-frequency spectrum (TFR). CWD

[0080] S202. Perform a Short-Time Fourier Transform (STFT) on the obtained time-domain signal to obtain the time-frequency spectrum (TFR). STFT

[0081] To remove irrelevant information and improve the accuracy and reliability of time-frequency analysis, a threshold segmentation method is used to segment the time-frequency spectrum (TFR). STFT The truncation process is performed to obtain the truncated frequency spectrum (TFR). 截断 .

[0082] Truncation Time Spectrum TFR 截断 It can be represented as:

[0083]

[0084] In the formula, max() represents the maximum value of elements, || represents the absolute value; ε is a threshold value, here ε = μ·mean(|TFR STFT (t,f)|); μ is a threshold factor, μ = 0.15; mean() represents the mean value; t is the time variable, and f is the frequency variable.

[0085] S203, truncate the time-frequency spectrum TFR 截断 and the time-frequency spectrum TFR CWD to obtain the combined time-frequency spectrum TFR 组合

[0086] The combined time-frequency spectrum TFR 组合 is expressed as:

[0087] TFR 组合 (t,f) = TFR 截断 (t,f)☉TFR CWD (t,f)

[0088] In the formula, ☉ is the Hadamard product operation, t is the time variable, and f is the frequency variable.

[0089] The combination of linear time-frequency transform and nonlinear time-frequency transform is adopted to balance the time-frequency resolution and overcome interference. Here, the combination transform of the linear truncated time-frequency spectrum TFR 截断 and the nonlinear time-frequency spectrum TFR CWD is adopted, that is, the Hadamard product of TFR 截断 (t,f) and TFR CWD (t,f) is performed to obtain the combined time-frequency spectrum TFR 组合 , which not only retains the properties of TFR 截断 (t,f), but also eliminates the cross terms in TFR CWD (t,f), and can more accurately depict the characteristics of the signal in the time-frequency domain.

[0090] S300, based on singular value decomposition denoising 组合

[0091] For random noise existing in the combined time-frequency spectrum TFR 组合 , a singular value decomposition denoising method can be used to improve the accuracy and reliability of the combined time-frequency transform. The specific steps are as follows:

[0092] S301, calculate the singular value decomposition of the combined time-frequency spectrum TFR 组合

[0093] The combined time-frequency spectrum TFR 组合 ​​SVD decomposition is performed, i.e. TFR 组合 = U∑V H , where U and V are orthogonal matrices, ∑ is a diagonal matrix, the elements on the diagonal are singular values of the combined time-frequency spectrum TFR 组合 , and the upper index H represents the conjugate transpose.

[0094] S302, selecting an appropriate threshold for singular value filtering

[0095] In SVD decomposition, the size of the singular value represents the energy of TFR 组合 at each frequency.

[0096] Therefore, by selecting an appropriate threshold, the frequencies corresponding to the singular values less than the threshold can be filtered out, and the filtered U and V are obtained, achieving the purpose of noise reduction. Assuming that the noise level of the combined time-frequency spectrum TFR 组合 is sigma, the threshold is set to k x sigma, where k is a constant, and here k = 2.

[0097] The singular values in the ∑ matrix that are lower than the threshold are set to 0, and these singular values are filtered out in SVD decomposition, obtaining the filtered to achieve the effect of signal noise reduction.

[0098] S303, reconstructing the signal to obtain the reconstructed

[0099] The filtered U, and V matrices are multiplied back together to obtain the reconstructed , i.e.

[0100] S400, using low-rank matrix recovery for unmanned aerial vehicle remote control signal extraction

[0101] According to the difference in the correlation between the remote control signal and the video transmission signal in the time-frequency domain, the low-rank matrix recovery method is used to distribute the parts with strong correlation to the low-rank matrix, and the parts with weak correlation are stored in the sparse matrix. The video transmission signal is continuously present within the bandwidth range, while the remote control signal is a step-like diagonal (or folded) line in the time-frequency domain, with the frequency changing over time. Therefore, in separating the remote control signal and the video transmission signal, most of the video transmission signal is distributed in the low-rank matrix A, and by solving the norm optimization problem, the video transmission signal is suppressed; most of the remote control signal is distributed in the sparse matrix E, and by using the alternating direction multiplier method ADMM, the remote control signal is further processed to obtain the time-frequency filtered time-frequency spectrum, as shown in the time-frequency filtered time-frequency spectrum shown in Figure 5 (a).

[0102] where the recovery of the low-rank matrix is to obtain the reconstructed This is represented as the sum of a low-rank matrix A and a sparse (noise) matrix E. The low-rank matrix A is then recovered by solving a norm optimization problem, further suppressing the image transmission signal. The norm optimization problem is solved as follows:

[0103]

[0104] In the formula, min A,E () indicates finding the minimum value of a low-rank matrix A and a sparse matrix E, where st represents the constraint condition; |||| * The expression `||||0` represents the kernel norm, which is the sum of the singular values ​​of the matrix; `||||0` represents the zero norm, which is the sum of the absolute values ​​of all elements in the matrix; `λ` is a regularization parameter used to adjust the weight relationship between low-rank terms and sparse terms.

[0105] Among them, the sparse matrix E is obtained by solving equation (1) using the alternating direction multiplier method (ADMM), which is the time spectrum after time-frequency filtering.

[0106] S500, based on time-spectrum similarity search using four-dimensional codes, for UAV remote control signal identification.

[0107] The obtained time-frequency filtered time spectrum is mapped into a four-dimensional code. Then, a time-frequency similarity search method based on the four-dimensional code is used to compare the mapped four-dimensional code with the four-dimensional code words in the four-dimensional code set of UAV remote control signals (the four-dimensional code set refers to the four-dimensional codes of many corresponding UAVs collected and organized into a set before the test) to determine whether the corresponding UAV model is found.

[0108] The specific steps for mapping the obtained time-frequency filtered time spectrum into a four-dimensional code are as follows:

[0109] a. Obtain the data to be mapped, including:

[0110] The time spectrum after time-frequency filtering is projected onto the time-frequency plane, the time-time spectrum plane, and the frequency-time spectrum plane, respectively, to obtain the contour plots of the time-frequency plane (set the flag sequence number to CM1), the time-time spectrum plane (set the flag sequence number to CM2), and the frequency-time spectrum plane (set the flag sequence number to CM3) in sequence (or, in other words, the contour plot of the time spectrum is the contour plot of the three-dimensional time spectrum in the three-dimensional space of time, frequency, and time spectrum).

[0111] b. Map the data to be mapped separately to obtain 3D code groups in PNG format.

[0112] The contour map of the time-frequency plane, the contour map of the time-time frequency spectrum plane and the contour map of the frequency-time frequency spectrum plane are respectively mapped into three-dimensional codes to generate a three-dimensional code group, and the format is converted into a png image format to obtain a png format three-dimensional code group;

[0113] c. The png format three-dimensional code group is integrated into a cyclic dynamic display three-dimensional code, i.e. a four-dimensional code, and the dynamic playing order of the four-dimensional code is CM1>CM2>CM3 in sequence.

[0114] As shown in Figure 5 , the time-frequency spectrum after time-frequency filtering and the mapped four-dimensional code, wherein (a) is a time-frequency spectrum map; (b) is a four-dimensional code map mapped from the time-frequency spectrum, which is a four-dimensional code based on dynamic display of the three-dimensional code group, and the four-dimensional code includes a time dimension, a frequency dimension, a color dimension and a playing order. The time dimension and the frequency dimension are corresponding time and frequency in the time-frequency spectrum after time-frequency filtering; the color dimension represents the numerical value of the time-frequency spectrum in the contour map of the time-frequency plane, the high and low of the frequency in the contour map of the time-time frequency spectrum plane, and the sequence of the time in the contour map of the frequency-time frequency spectrum plane; the playing order is a dynamic map formed by cyclic dynamic playing of the three-dimensional code mapped from the contour map of the time-frequency plane, the three-dimensional code mapped from the contour map of the time-time frequency spectrum plane and the three-dimensional code mapped from the contour map of the frequency-time frequency spectrum plane.

[0115] As shown in Figure 6 , the state diagram of the dynamic four-dimensional code. Wherein (a) is a contour map of the time-frequency plane; (b) is a contour map of the time-time frequency spectrum plane; (c) is a contour map of the frequency-time frequency spectrum plane.

[0116] The time-frequency spectrum similarity search method based on the four-dimensional code is adopted, as shown in Figure 2 , the flow chart of the time-frequency spectrum similarity search method based on the four-dimensional code, the features in the extracted time-frequency contour map (i.e. the contour map of the time-frequency plane, the contour map of the time-time frequency spectrum plane and the contour map of the frequency-time frequency spectrum plane) are identified to obtain the model of the unmanned aerial vehicle as DJI sprite 4. The search method is as follows:

[0117] S501, set the initial identification rate level, such as 90%, i.e. the highest identification rate level;

[0118] S502, whether the corresponding four-dimensional code is searched in the four-dimensional code set mapped from the unmanned aerial vehicle remote control signal; if yes (the corresponding four-dimensional code is searched), go to step S503; if no (the corresponding four-dimensional code is not searched), go to step S504;

[0119] S503, the four-dimensional code is the four-dimensional code sought, here is the four-dimensional code mapped by DJI Spark 4, and the corresponding model of the unmanned aerial vehicle is DJI Spark 4;

[0120] S504, the recognition rate level is reduced, i.e. the current recognition rate = recognition rate - 5%;

[0121] S505, it is judged whether the current recognition rate level is less than the minimum level (e.g. the minimum level of the recognition rate is 60%), if not (the recognition rate level >= the minimum level), step S502 is turned; if yes (the recognition rate level < the minimum level), step S506 is turned.

[0122] S506, the four-dimensional code is discarded, i.e. the corresponding model of the unmanned aerial vehicle is not found.

[0123] The four-dimensional code words in the four-dimensional code set of the unmanned aerial vehicle remote control signal can be obtained according to the time-frequency spectrum mapping of various unmanned aerial vehicle remote control signals tested in advance, and form the four-dimensional code set of the unmanned aerial vehicle remote control signal. Each four-dimensional code word in the four-dimensional code set corresponds to the four-dimensional code of the time-frequency spectrum mapping of the unmanned aerial vehicle remote control signal one by one.

[0124] In practical application, the similar four-dimensional code words in the four-dimensional code set can be retrieved in time by using the fast search characteristics of the four-dimensional code, and the corresponding model of the unmanned aerial vehicle, i.e. DJI Spark 4, is identified.

[0125] Although the content of the present application has been described in detail by the above preferred embodiments, it should be recognized that the above description should not be considered as a limitation of the present application. After reading the above content, various modifications and alternatives of the present application will be obvious to those skilled in the art. Therefore, the protection scope of the present application should be defined by the appended claims.

Claims

1. A method for identifying UAV remote control signals based on four-dimensional codes, characterized in that, The method comprises: S100, collecting and merging analog signals of the unmanned aerial vehicle to obtain a time domain signal; S200, performing a combined time-frequency transform on the obtained time-domain signal to obtain a combined time-frequency spectrum TFR 组合 ; The combined time-frequency transform method is as follows: S201. performing a Choi-Williams transform on the obtained time-domain signal to obtain a time-frequency spectrum TFR CWD ; S202, performing short-time Fourier transform (STFT) on the obtained time-domain signal to obtain a time-frequency spectrum (TFR) STFT ; performing truncation processing on the time-frequency spectrum (TFR) by using a threshold segmentation method to obtain a truncated time-frequency spectrum (TFR) STFT ; performing truncation processing on the time-frequency spectrum (TFR) by using a threshold segmentation method to obtain a truncated time-frequency spectrum (TFR) 截断 : where max() denotes the maximum of the elements, | | denotes the absolute value; ε is a threshold value, here ε = μ · mean(|TFR STFT (t, f) |); μ is a threshold factor, taken μ = 0.15; mean() denotes the mean value; t is the time variable, f is the frequency variable; S203、cutting the time-frequency spectrum TFR 截断 with the time-frequency spectrum TFR CWD performing Hadamard product to obtain the combined time-frequency spectrum TFR 组合 ; TFR 组合 (t, f) = TFR 截断 (t, f) Θ TFR CWD (t, f) In the formula, is Hadamard product operation, t is a time variable, and f is a frequency variable; S300, obtaining the combined time-frequency spectrum TFR 组合 singular value decomposition based denoising; The singular value decomposition-based denoising method is as follows: S301、performing the combination time-frequency spectrum TFR 组合 performing singular value decomposition; combining the time-frequency spectrum TFR 组合 singular value decomposition, i.e. combining the time-frequency spectrum TFR 组合 = U∑V H wherein U and V are both orthogonal matrices, ∑ is a diagonal matrix, the elements on the diagonal are singular values of the combined time-frequency spectrum TFR 组合 and the upper index H denotes the conjugate transpose; S302, selecting a suitable threshold for singular value filtering; In SVD decomposition, the combined time-frequency spectrum (TFR) 组合 The magnitude of the singular values ​​represents the combined time-frequency spectrum (TFR). 组合 For each frequency of energy, a suitable threshold is selected, and the frequencies corresponding to singular values ​​below the threshold are filtered out to obtain the filtered U and V, thus achieving the purpose of noise reduction. The singular values lower than the threshold in the matrix are set to 0, and the singular values are filtered in the SVD decomposition to obtain filtered To achieve the effect of signal noise reduction; S303, reconstructing the signal to obtain a reconstructed The filtered U, and V matrices are multiplied back together to obtain the reconstructed i.e. S400, reconstructing the low-rank matrix from the reconstruction extracting a drone remote control signal from the reconstructed low-rank matrix The low-rank matrix recovery is to reconstruct the obtained is expressed as a sum of a low-rank matrix A and a sparse matrix E, wherein the image transmission signal is distributed in the low-rank matrix A and the remote control signal is distributed in the sparse matrix E; suppressing the graph signal in the reconstruction by solving a norm optimization problem a remote control signal is obtained by solving a norm optimization problem; and the remote control signal is processed by an alternating direction multiplier method (ADMM) to obtain a time-frequency spectrum after time-frequency filtering. The norm optimization problem is solved as follows: In the formula, min A，E () indicates finding the minimum value of a low-rank matrix A and a sparse matrix E, where st represents the constraint condition; || || * The expression `|| ||0` represents the nuclear norm, which is the sum of the singular values ​​of a matrix; `|| ||0` represents the zero norm, which is the sum of the absolute values ​​of all elements in the matrix; `λ` is the regularization parameter, taking a value of... The alternating direction multiplier method ADMM is used to solve formula (1) to obtain a sparse matrix E, that is, a time-frequency spectrum after time-frequency filtering; S500, four-dimensional code-based time-frequency spectrum similarity search is performed to identify the unmanned aerial vehicle remote control signal; The time-frequency spectrum after time-frequency filtering is mapped into a four-dimensional code, and a four-dimensional code-based time-frequency spectrum similarity search method is used to compare the mapped four-dimensional code with four-dimensional code words in a four-dimensional code set of the unmanned aerial vehicle remote control signal to determine whether the model of the corresponding unmanned aerial vehicle is found. The method for mapping the time-frequency spectrum after time-frequency filtering into a four-dimensional code is as follows: a. Obtain to-be-mapped data, which includes a contour map of a time-frequency plane, a contour map of a time-time-frequency spectrum plane, and a contour map of a frequency-time-frequency spectrum plane; The contour map of the time-frequency plane, the contour map of the time-time-frequency spectrum plane, and the contour map of the frequency-time-frequency spectrum plane are obtained by the following method: The time-frequency spectrum after time-frequency filtering is projected onto the time-frequency plane, the time-time-frequency spectrum plane, and the frequency-time-frequency spectrum plane in sequence to obtain; b. Map the to-be-mapped data respectively to obtain a three-dimensional code set in png format The to-be-mapped data is mapped into a three-dimensional code set, and the three-dimensional code set is converted into a png image format to obtain a three-dimensional code set in png format; c. Integrate the three-dimensional code set in png format into a three-dimensional code for cyclic dynamic display, that is, a four-dimensional code; The four-dimensional code-based time-frequency spectrum similarity search method is as follows: S501, set an initial identification rate level; S502, search for a corresponding four-dimensional code in a four-dimensional code set mapped from the unmanned aerial vehicle remote control signal, if yes, go to step S503; if no, go to step S504; S503, the four-dimensional code is the four-dimensional code to be found, that is, the model of the corresponding unmanned aerial vehicle is found; S504, reduce the identification rate level, that is, the current identification rate = identification rate - 5%; S505, determine whether the current identification rate level is less than the minimum level, if no, go to step S502; if yes, go to step S506; S506, discard the four-dimensional code, that is, the model of the corresponding unmanned aerial vehicle is not found.

2. The identification method according to claim 1, characterized in that, In S100, the collection and merging of the analog signals of the unmanned aerial vehicle comprises: An AD9361 zero intermediate frequency receiver architecture-based data acquisition system is adopted to collect analog signals of a certain model of DJI unmanned aerial vehicle; an analog / digital device inside the receiver performs analog-to-digital conversion on the received analog signals, and converts the received analog signals into two-channel digital baseband signals of in-phase component I and quadrature component Q through quadrature demodulation and digital down-conversion, then the I and Q two-channel digital baseband signals are merged to obtain the time domain signal, sig = I + j·Q, j representing an imaginary unit, 3. The identification method according to claim 1, characterized in that, In S500, the four-dimensional code comprises a time dimension, a frequency dimension, a color dimension, and a playing order dimension.

4. The identification method according to claim 3, characterized in that, The time dimension and the frequency dimension are corresponding time and frequency in the time-frequency spectrum after time-frequency filtering.

5. The identification method according to claim 3, characterized in that, The color dimension represents the numerical size of the time-frequency spectrum in the contour map of the time-frequency plane, the height of the frequency in the contour map of the time-time-frequency spectrum plane, and the chronological order of the time in the contour map of the frequency-time-frequency spectrum plane.

6. The identification method according to claim 3, characterized in that, The playing order is a dynamic picture formed by cyclically and dynamically playing the three-dimensional code mapped from the contour map of the time-frequency plane, the three-dimensional code mapped from the contour map of the time-time-frequency spectrum plane, and the three-dimensional code mapped from the contour map of the frequency-time-frequency spectrum plane.

7. A four-dimensional code-based unmanned aerial vehicle remote control signal recognition system, characterized in that, The system comprises: a UAV, a remote control station, and a monitoring device; The monitoring device implements the four-dimensional code-based UAV remote control signal identification method according to any one of claims 1-6.

8. Application of the identification method according to any one of claims 1-6 to the monitoring device for identifying a UAV remote control signal.

Images