Unmanned aerial vehicle rotor parameter decoupling estimation method based on OMP algorithm

Abstract

The invention discloses an unmanned aerial vehicle rotor parameter decoupling estimation method based on an OMP algorithm, and the method comprises the steps: obtaining the radar echo of a rotor unmanned aerial vehicle, and completing the preprocessing of signal sampling, pulse compression and the like; an echo matrix is constructed, and motion compensation and filtering of a rigid body part are carried out; estimating the rotating speed of a target rotor according to the jet engine modulation characteristics of the rotor echoes; constructing a phase matching item and estimating the initial phase of the target rotor blade; and designing an over-complete dictionary based on a phase matching item, and reconstructing sparse distribution of scattering points of the target rotor wing by using an OMP algorithm. According to the method, the rotation speed, the blade number and the length of the target rotor can be accurately estimated under the echo azimuth undersampling condition, parameter decoupling is achieved, the calculation complexity can be effectively reduced, the algorithm operation time is shortened, and the defects of an existing method are overcome.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar imaging data processing, and particularly relates to a method for decoupling and estimating unmanned aerial vehicle (UAV) rotor parameters based on the orthogonal matching pursuit (OMP) algorithm. Background Art

[0002] Due to its low price and good practicability, UAVs have been widely used in various fields. However, while bringing convenience, UAVs also bring serious safety problems. In existing low-altitude detection technologies, inverse synthetic aperture radar (ISAR) can reconstruct the radar images of non-cooperative moving targets and shows great application potential in target detection and recognition.

[0003] Traditional ISAR imaging methods usually assume that the target is rigid. However, common small consumer drones are usually driven by rotors, and their radar echoes exhibit special sinusoidal phase modulation, known as the micro-Doppler (m-D) effect. The m-D effect degrades the quality of ISAR images. On the other hand, the m-D characteristics of the target can be used to extract parameters of the drone rotors, such as rotational speed, number of blades, and rotor size. In the past few decades, the analysis of target m-D characteristics has been a major challenge in ISAR research. Since the target m-D frequency is time-varying, early studies used time-frequency analysis methods to extract m-D characteristics. Chen et al. established a signal model for the m-D part and achieved accurate parameter estimation using time-frequency analysis such as the short-time Fourier transform and Wigner-Ville distribution (Reference: Chen V C. Analysis of radar micro-Doppler with time-frequency transform[C]. Proceedings of the Tenth IEEE Workshop on Statistical Signal and Array Processing. IEEE, 2000:463-466.). The time-frequency analysis-based method has a simple structure, but it needs to compromise between time resolution and frequency resolution, and the parameter estimation accuracy is limited. Similarly, the spectral analysis method has also been used to study the m-D effect (Reference: Ren X, Jahangir M, White D, et al. Estimating physical parameters from multi-rotor drone spectrograms[C]. International Conference on Radar Systems (RADAR 2022). IET, 2022:20-25.). It should be noted that when the radar repetition frequency is low and aliasing occurs in the Doppler domain, both time-frequency analysis and spectral analysis are not applicable. Another type of m-D analysis method can be classified as the parametric method. The parametric method fits the echo with a parametric basis function to represent the m-D characteristics. The radar echo of the target rotor exhibits the characteristics of a sinusoidal phase modulation signal, and its instantaneous frequency varies sinusoidally.Therefore, Radon transform-based methods with curve detection capabilities can be used to reconstruct the ISAR images of helicopter rotors to a certain extent. The Complex-valued Inverse Radon Transform (CIRT) can significantly improve the resolution of the reconstructed images because of its coherent integration calculation method (see reference: Bai X, Zhou F, Xing M, et al. High Resolution ISAR Imaging of Targets with Rotating Parts [J]. IEEE Transactions on Aerospace & Electronic Systems, 2011, 47(4): 2530-2543.). For common low-altitude detection radars, their range resolution is limited, and the Pulse Repetition Frequency (PRF) is usually low. The echoes of the helicopter rotor are difficult to distinguish in the range dimension, and severe aliasing usually occurs in the Doppler dimension. Therefore, it is difficult for the method based on real Radon transform to detect the sine curve in the ranging profile, and the CIRT method cannot effectively achieve coherent integration. Similarly, the Hough transform method (see reference: Zhang Q, Yeo T S, Tan H S, et al. Imaging of a Moving Target With Rotating Parts Based on the Hough Transform [J]. IEEE Transactions on Geoscience and Remote Sensing, 2008, 46(1): 291-299.) is also a parametric helicopter image reconstruction method. This method estimates the mapping of helicopter scattering points in the parameter plane by performing a four-dimensional parameter search in the spectrum, inevitably facing a huge computational burden. In summary, the existing parametric methods have high accuracy in m-D feature estimation but are often accompanied by a large amount of computation. In addition, under the condition of azimuth undersampling of low-PRF echoes, the performance of these algorithms will also be significantly reduced.

[0004] Generally speaking, the research on m-D feature analysis, especially rotor analysis, still faces many difficulties. The existing methods usually require a sufficient azimuth sampling rate and have problems such as lack of utilization of prior knowledge and large computational amount, and further research and improvement are needed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for decoupling and estimating the parameters of a drone rotor based on the OMP algorithm.

[0006] The technical solution for achieving the object of the present invention is as follows: A method for decoupling and estimating the parameters of an unmanned aerial vehicle rotor based on the OMP algorithm, comprising the following steps:

[0007] Step 1: Obtain the radar echo of the rotor unmanned aerial vehicle and complete data preprocessing including signal sampling and pulse compression;

[0008] Step 2: Combine the echo signal model of the rotor component, use the target radar echo in Step 1, construct a target radar echo matrix with one frame of data as a unit, and perform motion compensation and filtering based on the echo of the rigid body part;

[0009] Step 3: Perform a fast Fourier transform on the range cell where the target rotor is located to obtain a spectrum, extract the line spectrum periodicity of the spectrum based on the JEM theory, and estimate the possible values of the target rotor speed;

[0010] Step 4: Based on the rotor speed estimated in Step 3, construct a sine-type phase matching term in combination with the echo signal model of the rotor component where t m represents the slow time, l, ω m , θ 0 respectively represent the blade length, speed, and initial phase angle within the coherent integration time of the rotor, f c is the center frequency of the transmitted signal, and c is the propagation speed of the electromagnetic wave;

[0011] Step 5: Multiply the echo of the range cell where the target rotor is located by the phase matching term and then perform a fast Fourier transform, and calculate the maximum value A max near zero Doppler of the spectrum amplitude; Plot the change trend of A max with the search direction of the initial phase, and complete the estimation of the initial phase of the target rotor blade and determine the correct target rotor speed according to the change trend of A max and in combination with the assumption of the number of blades;

[0012] Step 6: On the basis of obtaining the estimated values of the target rotor speed and the initial phase of the blade, perform the reconstruction of the blade scatter point model based on the OMP algorithm to realize the estimation of the blade length; Input the observed vector h m of the target rotor echo; Input the over-complete dictionary D = [e sin (t m ; Δl), e sin (t m ; 2Δl), …, e sin (t m ; MΔl)], each column atom d m represents the phase matching terms of different scales, M represents the number of atoms in the dictionary, Δl represents the step size, and e sin is the phase matching term; Input the random undersampling matrix S; Initialize the residual r0 = h m , projection matrix F 0 = 0, subspace index set maximum number of iterations K and energy threshold ε 0 , where ε 0 is used to measure the noise level and is calculated from the non-blade area in the parameter plane; let the iteration number t = 1;

[0013] Step 7: Find the atom d t-1 in dictionary D with the largest inner product with the residual r m and its index m, that is, calculate and remove the atom d from dictionary D m , then update the projection matrix F t = F t-1 ∪ d m and the subspace index set Λ t = Λ t-1 ∪ m;

[0014] Step 8: Update the residual and let the iteration number t = t + 1;

[0015] Step 9: Judge the iteration termination condition. If t ≥ K or ||r t || < ε 0 , then jump to Step 10; otherwise, jump to Step 7;

[0016] Step 10: Use and Λ t to reconstruct the distribution function x of the scattering points of the target rotor blade, and complete the estimation of the length of the target rotor blade.

[0017] Furthermore, in Step 6, based on the obtained estimated values of the rotational speed and initial phase of the target rotor blade, perform the reconstruction of the blade scattering point model based on the OMP algorithm; the echo of the target rotor is expressed as:

[0018] h m = SDx + n

[0019] where x is the distribution function of the scattering points on the blade, and n is the noise input to the target rotor echo observation vector; then the reconstruction of the sparse scattering points of the rotor is expressed as:

[0020] min ||x|| 0 s.t. ||h m - SDx|| 2 < ε

[0021] where, ||·|| 0 and ||·|| 2 respectively represent the L 0 norm and L2 The norm, and ε represents the residual.

[0022] An electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the above method are implemented.

[0023] A computer-readable storage medium stores a computer program. When the program is executed by a processor, the steps of the above method are implemented.

[0024] A computer program product includes a computer program. When the computer program is executed by a processor, the steps of the above method are implemented.

[0025] Compared with the prior art, the significant advantages of the present invention are:

[0026] (1) A parameter phase matching term is introduced to complete the estimation of rotor parameters. In addition, a parameter decoupling strategy is designed to make full use of the prior knowledge of the rotor to solve the problem, simplifying the multi-dimensional parameter search problem into a one-dimensional parameter estimation problem. Compared with the existing methods, the method proposed by the present invention has a low computational complexity and a short running time.

[0027] (2) The present invention can accurately estimate the rotor parameters under the condition of azimuth under-sampling. With the help of the compressed sensing theory, an over-complete dictionary based on the phase matching term is designed, and the blade length estimation is transformed into a sparse reconstruction problem, and the OMP algorithm is used to effectively estimate the rotor blade length.

[0028] (3) Experiments prove that the rotor parameter estimation method proposed by the present invention has strong robustness to noise, and the estimation accuracy of the rotor parameters is high enough. Compared with the traditional exhaustive algorithm (refer to the literature: Zhang Q, Yeo T S, Tan H S, et al. Imaging of a Moving Target With Rotating Parts Based on the Hough Transform [J]. IEEE Transactions on Geoscience and Remote Sensing, 2008, 46(1): 291-299; Ji G, Song C, Huo H. Detection and identification of low-slow-small rotor unmanned aerial vehicle using micro-Doppler information [J]. IEEE Access, 2021, 9: 99995-100008.), the running time of this algorithm can be saved by about 70%. Description of the Drawings

[0029] Figure 1 This is the flowchart of the method for decoupling and estimating the parameters of the UAV rotor based on the OMP algorithm in the present invention.

[0030] Figure 2 This is the schematic diagram of the target ISAR observation in Example 1.

[0031] Figure 3 This is the target range-Doppler image of the measured data in Example 1.

[0032] Figure 4 This is the schematic diagram of the JEM characteristics of the measured target rotor echo in Example 1.

[0033] Figure 5 This is the curve of the A max value varying with value in Example 1 under the assumption of two rotor blades.

[0034] Figure 6 This is the curve of the A max value varying with value in Example 1 under the assumption of three rotor blades.

[0035] Figure 7 This is the reconstructed rotor scatter point distribution function of the measured data in Example 1 based on the OMP algorithm. Detailed implementation manner

[0036] The present invention will be further described below in conjunction with the accompanying drawings of the specification.

[0037] Small rotor unmanned aerial vehicles (UAVs) have characteristics such as low flight altitude, slow flight speed, and small radar cross-section, which pose a challenge to traditional radar detection. The rotor is an important feature of rotor UAVs. Estimating the UAV rotor parameters is an important means for identifying UAVs and classifying threats. The rotation of the UAV rotor causes the micro-Doppler effect in its radar echo. By analyzing the micro-Doppler characteristics of the echo, the UAV rotor parameters can be estimated. As an effective method for analyzing micro-Doppler characteristics, the parametric method can accurately estimate rotor parameters. However, the existing methods have a large amount of calculation and their performance degrades or even fails under the condition of azimuth undersampling of the echo. The present invention can quickly and accurately estimate the rotor parameters of small UAVs under the condition of azimuth undersampling of the echo; the present invention introduces a parametric phase matching term to complete the estimation of rotor parameters and designs a parametric decoupling strategy to simplify the multi-dimensional parameter search problem into a series of one-dimensional parameter estimation problems, effectively reducing the amount of calculation; the present invention designs an over-complete dictionary based on the phase matching term with the help of the compressed sensing theory, transforms the estimation of blade length into a sparse reconstruction problem of the rotor scatter point distribution, and effectively estimates the rotor blade length by using the Orthogonal Matching Pursuit (OMP) algorithm.

[0038] Combined with Figure 1 , a method for decoupling and estimating UAV rotor parameters based on the OMP algorithm includes the following steps:

[0039] Step 1: Use a radar to obtain the echo signal of the rotor UAV target and complete preprocessing work such as signal sampling and pulse compression. Model the rotor echo signal of the target, establish an O-XY plane coordinate system on the target projection plane, where the origin O represents the rotation center of the rigid body. Let the rotation center of the rotor blade be Q(x q , y q ). Then the distance R m from the rotor scatter point P to the radar at time t m,P (t m ) is:

[0040] R m,P (t m ) ≈ R 0 (t m ) + x q θ(t m ) + y q + l p cos(ω m t m - θ 0,P )

[0041] where l p = L p cosβ represents the distance between P and Q on the target projection plane, and Lp is the distance between P and Q on the blade, and β is the elevation angle of the target relative to the radar; R 0 (t m ) is the distance between the radar and the origin O, and θ(t m ) = ω r t m represents the instantaneous rotation angle of the rigid body part of the target at time t m , ω r is the equivalent rotational speed of the target rigid body, ω m is the rotational speed of the rotor, and θ 0,P is the initial rotation phase of the scattering point P.

[0042] Step 2: Construct a radar echo matrix with the target echo signal obtained in Step 1 as a unit. The rotor echo after translational compensation satisfies the following formula:

[0043]

[0044] where t m is the slow time, τ is the fast time, the rotor part consists of K scattering points, σ m,k is the echo amplitude of the k-th scattering point, B is the bandwidth of the transmitted signal, f c is the center frequency of the transmitted signal, and c is the propagation speed of electromagnetic waves. Motion compensation and filtering are performed based on the echo of the rigid body part, and the phase term can be compensated by estimating the Doppler frequency of the target rigid body part.

[0045] Step 3: Perform a Fourier transform in the azimuth direction on the basis of the echo matrix in Step 2 to obtain the range-Doppler image of the target echo. Find the range cell where the rotor is located through the spectral broadening in the azimuth direction of the range-Doppler image; pick out the echo of the range cell where the target rotor is located and perform a Fourier transform. According to the JEM (Jet Engine Modulation) theory, select the spectral lines that show a periodic distribution in the spectrum, and estimate the blade rotational speed of the target rotor based on the distribution period of the spectral lines where is the estimated value of the rotor blade rotational speed, f T is the periodic line spectrum interval, and N is the number of blades. When N is even, α = 1; when N is odd, α = 2. Since the number of blades of the UAV to be measured is often unknown, ω m estimation will have multiple possible results, and it is necessary to further determine the true estimated value of ω m .

[0046] Step 4: Based on the target rotor speed parameter estimated in Step 3, construct a sinusoidal phase matching term which satisfies:

[0047]

[0048] where represents the estimated value of the initial phase of the target rotor blade, represents the estimated value of the length of the target rotor blade, and λ is the wavelength of the radar transmitted signal.

[0049] Step 5: Based on the phase matching term in Step 4, let be a relatively small value, such as where Δr is the length of a range cell; let take uniformly spaced values within [0, 2π) with a step size of , such as At this time, the phase matching term can be expressed as Perform phase matching on the rotor echo signal and perform a fast Fourier transform, that is, H m,n (f) = FFT[h m (t m ) ⊙ e sin,n (t m )]. Finally, near the zero Doppler frequency, within a range such as [-0.2PRF, 0.2PRF], calculate and record the maximum amplitude of the spectrum H m,n (f), that is, Amax = max(|H m,n (f)|), f ∈ [-0.2PRF, 0.2PRF]. Traverse within [0, 2π) and repeat the above calculations to plot the variation curve of A max . The corresponding at the peak of the curve is the estimated value of the initial phase of the rotor blade. In addition, due to the fact that there are multiple possible results for the estimation of the rotational speed ω m in Step 3, all possible results of ω m in Step 5 are calculated using the above phase matching method. When the estimated value of ω m is close to the true value, the variation trend of A max will show peaks equal to the number of blades; conversely, when the estimated value of ω m deviates from the true value, the variation trend of A max will show multiple disordered peaks, which is contrary to the assumption of the number of blades. Thus, the correct estimated value of ω m and the number of blades can be determined to solve the rotational speed ambiguity problem.

[0050] Step 6: On the basis of obtaining the estimated values of the rotational speed and initial phase of the target rotor blade, perform the reconstruction of the blade scatter point model based on the OMP algorithm. The target rotor echo can be expressed as:

[0051] h m = SDx + n

[0052] Among them, h m is the observed signal vector, S is the random undersampling matrix, x is the distribution function of the scattering points on the blade, n is the noise input target rotor echo observation vector, D represents the overcomplete dictionary, and its columns of the original d m sub - atoms are phase - matching terms of different scales, that is, D = [e sin (t m ; Δl), e sin (t m ; 2Δl), …, e sin (t m ; MΔl)]. Among them, e sin is the phase - matching term, M represents the number of atoms in the dictionary, Δl represents the step size, and take Δl = 0.005m. Then the reconstruction of the sparse scattering points of the rotor can be expressed as:

[0053] min||x|| 0 s.t.||h m - SDx|| 2 <ε

[0054] Among them, ||·|| 0 and ||·|| 2 respectively represent the L 0 norm and the L 2 norm, and ε represents the residual.

[0055] Input the target rotor echo observation vector h m , the overcomplete dictionary D and the random undersampling matrix S; initialize the residual r 0 = h m , the projection matrix F 0 = 0, the subspace index the maximum number of iterations K and the energy threshold ε 0 . Among them, K is related to the sparsity of the scene and can be obtained through the radar wavelength and the target prior knowledge; the energy threshold ε 0 can be calculated from the area in the parameter plane that does not contain the blade; let the iteration number t = 1.

[0056] Step 7: Find the atom d t-1 in the dictionary D with the largest inner product with the residual r m and its index m, that is, calculate and remove the atom d m from the dictionary D, then update the projection matrix F t = F t-1 ∪d m and the subspace index set Λ t = Λ t-1 ∪m;

[0057] Step 8: Update the residual And let the iteration number \(t = t + 1\);

[0058] Step 9: Judge the iteration termination condition. If \(t\geq K\) or \(\|r\) t \| \lt \varepsilon\) 0 , then jump to Step 10; otherwise, jump to Step 7;

[0059] Step 10: Use and \(\Lambda\) t to reconstruct the target rotor scatter point distribution function \(x\) and complete the estimation of the target rotor blade length.

[0060] The present invention will be described in detail below in conjunction with embodiments and drawings.

[0061] Embodiment

[0062] Combined with Figure 1 , a method for decoupling and estimating UAV rotor parameters based on the OMP algorithm includes the following steps:

[0063] Step 1: Use radar to obtain the echo signal of the rotor UAV target and complete the preprocessing work including signal sampling and pulse compression. The ISAR observation of the target is as Figure 2 shown. Model the rotor echo signal of the target, establish an O-XY plane coordinate system on the target projection plane, where the origin O represents the rotation center of the rigid body. Let the rotation center of the rotor blade be Q(x q , y q ). Then the distance \(R\) m from the rotor scatter point P to the radar at time \(t\) m,P (t m ) is:

[0064] \(R\) m,P (t m ) \(\approx R\) 0 (t m ) + x q \(\theta(t\) m ) + y q + l p \(\cos(\omega\) m t m - \(\theta\) 0,P )

[0065] where \(l\) p = \(L\) p \(\cos\beta\) represents the distance between P and Q on the target projection plane, \(L\) p is the distance between P and Q on the blade, \(\beta\) is the elevation angle of the target relative to the radar; \(R\) 0 (t m ) is the distance between the radar and the origin O, and \(\theta(t\) m ) = \(\omega\) r tm represents the instantaneous rotation angle of the target rigid body part at time t m , ω r is the equivalent rotational speed of the target rigid body, ω m is the rotational speed of the rotor, θ 0,P is the initial rotation phase of the scatter point P

[0066] In this embodiment, the used radar is of the Frequency Modulated Continuous Wave (FMCW) system. The signal form of the transmitted waveform is a frequency-modulated continuous sawtooth wave, with the starting frequency of 77 GHz, that is, f c is 77 GHz; the effective bandwidth is 2.8 GHz, the range cell length is 0.0625 m, the PRF is 10 KHz, and one frame of data contains 2048 chirp signals in total; the used unmanned aerial vehicle is the DJI Mini2 small unmanned aerial vehicle, and its rotor diameter is about 0.12 m. Therefore, the echo of the rotor blade accounts for about 2 range cells

[0067] Step 2: Construct a radar echo matrix with the target echo signal obtained in Step 1 as a unit. The rotor echo after translational compensation satisfies the following formula

[0068]

[0069] where t m is the slow time, τ is the fast time, the rotor part consists of K scatter points, and σ m,k is the echo amplitude of the k-th scatter point, B is the bandwidth of the transmitted signal, f c is the center frequency of the transmitted signal, and c is the propagation speed of the electromagnetic wave. Based on the echo of the rigid body part, motion compensation and filtering are performed, and the phase term can be compensated by estimating the Doppler frequency of the target rigid body part

[0070] Step 3: The range-Doppler image of the echo matrix in Step 2 is as Figure 3 shown. It can be judged from the spectrum broadening situation that the range cells where the target rotor part is located are the 208-209 and 211-212 range cells. Taking the 208th range cell as an example, the echo of this range cell is subjected to a fast Fourier transform, and the result is as Figure 4 shown; the spectral lines showing periodic distribution in the spectrum are selected, and combined with the JEM theory, the rotational speed of the rotor blade of the target can be estimated to be about where is the estimated value of the rotational speed of the rotor blade, f T is the periodic line spectrum interval, N is the number of blades, when N is even, α = 1, and when N is odd, α = 2. From Figure 4 it can be known that f T= 112.13 Hz. Since the rotors of common small rotor UAVs are mostly two-blade or three-blade, combining this prior information, the possible values of the target rotor speed can be calculated as (corresponding to two blades) and (corresponding to three blades).

[0071] Step 4: Based on the target rotor speed parameter estimated in Step 3, construct a sine-type phase matching term which satisfies:

[0072]

[0073] where represents the estimated value of the initial phase of the target rotor blade, represents the estimated value of the target rotor blade length, λ = c / f c is the radar transmitted signal wavelength.

[0074] Step 5: Based on the phase matching term in Step 4, let be a relatively small value, such as where Δr is the length of a range cell; let uniformly take values within [0, 2π) with a step size of such as At this time, the phase matching term can be expressed as Perform phase matching on the rotor echo signal and perform a fast Fourier transform, that is, H m,n (f) = FFT[h m (t m ) ⊙ e sin,n (t m )]. Finally, near the zero Doppler frequency, within the range of [-0.2PRF, 0.2PRF], calculate and record the maximum amplitude of the spectrum H m,n (f), that is, Amax = max(H m,n (f)), f ∈ [-0.2PRF, 0.2PRF]. Traverse within [0, 2π) and repeat the above calculation, draw the variation curve of A max . The corresponding to the peak of the curve is the estimated value of the initial phase of the rotor blade. Under the two-blade assumption, the obtained variation curve of A max is as shown in Figure 5 , Figure 5 the curve has two obvious peaks, which is consistent with the assumption; under the three-blade assumption, the obtained variation curve of A max is as shown in Figure 6 , Figure 6There are multiple irregular peaks in the curve. Therefore, it can be judged that the target rotor is a two-blade rotor, and the initial phases of the blades are approximately 0.74 rad and 3.88 rad respectively.

[0075] Step 6: On the basis of obtaining the estimated values of the rotational speed and initial phase of the target rotor blades, perform the reconstruction of the blade scatter point model based on the OMP algorithm. The echo of the target rotor can be expressed as:

[0076] h m = SDx + n

[0077] where h m is the observed signal vector, S is the random undersampling matrix, x is the distribution function of the scatter points on the blade, n is the noise input to the target rotor echo observation vector, D represents the overcomplete dictionary, and its columns d m subs are phase matching terms of different scales, that is, D = [e sin (t m ; Δl), e sin (t m ; 2Δl), …, e sin (t m ; MΔl)]. Where e sin is the phase matching term, M represents the number of atoms in the dictionary, and Δl represents the step size. Take Δl = 0.005 m. Then the reconstruction of the sparse scatter points of the rotor can be expressed as:

[0078] min ||x|| 0 s.t. ||h m - SDx|| 2 < ε

[0079] where ||·|| 0 and ||·|| 2 represent the L 0 norm and the L 2 norm respectively, and ε represents the residual.

[0080] Input the target rotor echo observation vector h m , that is, the range cell where the target rotor is located, with a size of 1×2048; input the overcomplete dictionary D and the random undersampling matrix S; initialize the residual r 0 = h m , the projection matrix F 0 = 0, the subspace index the maximum number of iterations K and the energy threshold ε 0 . Among them, K is related to the sparsity of the scene and can be obtained through the radar wavelength and the prior knowledge of the target. The radar wavelength is λ = c / f c= 4 mm. The length of the blades of small rotor unmanned aerial vehicles is usually between 5 - 20 cm. From this, the magnitude of K can be inferred. In this example, K = 18 is taken; the energy threshold ε 0 can be calculated from the area in the parameter plane that does not contain the blades; let the number of iterations t = 1.

[0081] Step 7: Find the atom d t-1 in the dictionary D that has the largest inner product with the residual r m and its index m, that is, calculate and remove the atom d from the dictionary D m , and then update the projection matrix F t = F t-1 ∪d m and the subspace index set Λ t = Λ t-1 ∪m;

[0082] Step 8: Update the residual and let the number of iterations t = t + 1;

[0083] Step 9: Judge the iteration termination condition. If t ≥ K or ||r t || < ε 0 , then jump to Step 10; otherwise, jump to Step 7;

[0084] Step 10: Use and Λ t to reconstruct the target rotor scatter point distribution function x. The result is as Figure 7 shown. From the trend characterized by the scatter points sparsity, it can be judged that the length of the blade is about 0.05 m.

[0085] It should be understood that, in order to streamline the present invention and help those skilled in the art understand various aspects of the present invention, in the above description of the exemplary embodiments of the present invention, various features of the present invention are sometimes described in a single embodiment or with reference to a single figure. However, the present invention should not be construed as meaning that the features included in the exemplary embodiments are all essential technical features of the patent claims.

Claims

1. A method for decoupling and estimating UAV rotor parameters based on OMP algorithm, characterized in that: The following steps are involved: Step 1: Obtain the radar echo of the rotorcraft drone and complete data preprocessing including signal sampling and pulse compression; Step 2: Combined with the rotor component echo signal model, the target radar echo in step 1 is used to construct a target radar echo matrix in units of one frame of data, and motion compensation and filtering are performed based on the echo of the rigid body part; Step 3: Perform fast Fourier transform on the distance unit where the target rotor is located to obtain the spectrum, extract the periodic line spectrum in the spectrum based on the JEM theory, and estimate the possible values ​​of the target rotor speed; Step 4: Based on the estimated rotor speed in step 3, a sinusoidal phase matching term is constructed in combination with the rotor component echo signal model. where t m Indicates slow time, l, ω m , θ0 represent the blade length, rotation speed and initial phase angle of the rotor during the coherent integration time, respectively, and f c is the center frequency of the transmitted signal, c is the propagation speed of the electromagnetic wave; Step 5: Multiply the range unit echo of the target rotor by the phase matching term and perform a fast Fourier transform, and calculate the maximum value A of the spectrum amplitude near zero Doppler max ; Draw A max With the change trend of the initial phase search direction, according to A max The changing trend of the target rotor blades is combined with the assumption of the number of blades to estimate the initial phase of the target rotor blades and determine the correct target rotor speed. Step 6: Based on the target rotor speed and the estimated value of the initial phase of the blade, the blade scattering point model reconstruction based on the OMP algorithm is performed to realize the blade length estimation; the target rotor echo observation vector h is input m ; Input overcomplete dictionary D = [e sin (t m ; Δl),e sin (t m ; 2Δl),…,e sin (t m ; MΔl)], each column of atoms d m represents phase matching items of different scales, M represents the number of atoms in the dictionary, Δl represents the step size, and e sin is the phase matching term; input the random undersampling matrix S; initialize the residual r0 = h m , projection matrix F0 = 0, subspace index set The maximum number of iterations K and the energy threshold ε0, where ε0 is used to measure the noise level and is calculated from the non-blade area in the parameter plane; let the number of iterations t = 1; Step 7: Find the residual r in the dictionary D t-1 The atom with the largest inner product d m and its index m, that is, calculate And remove the atom d from the dictionary D m , then update the projection matrix F t =F t-1 ∪d m and the subspace index set Λ t =Λ t-1 ∪m; Step 8: Update the residual r t =h m -F t (F t H F t ) -1 F t H h m , and let the number of iterations t = t + 1; Step 9: Determine the iteration termination condition, if t ≥ K or || r t ||<ε0, jump to step 10; otherwise, jump to step 7; Step 10: Use x=(F t H F t ) -1 F t H h m and Λ t Reconstruct the target rotor scattering point distribution function x and complete the estimation of the target rotor blade length.

2. The method for decoupling and estimating the rotor parameters of a UAV based on the OMP algorithm according to claim 1 is characterized in that: In step 6, based on the target rotor blade speed and initial phase estimation, the blade scattering point model reconstruction based on the OMP algorithm is performed; the target rotor echo is expressed as: h m =SDx+n Where x is the distribution function of the scattering points on the blade, and n is the noise input target rotor echo observation vector; the reconstruction of the rotor sparse scattering points is expressed as: min||x||0s.t.||h m -SDx||2ϵ Among them, ||·||0 and ||·||2 represent the L0 norm and L2 norm respectively, and ε represents the residual.

3. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 2 are implemented.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

5. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

Images