A method for aircraft attitude inversion based on the SBL algorithm

Abstract

This invention proposes an aircraft attitude inversion method based on the SBL algorithm, comprising: constructing an observation angle-RCS model based on sparse reconstruction for the electromagnetic calculation results of a specific aircraft; and inverting the radar observation angle based on the SBL algorithm using the measured RCS sequence data of the radar on the flight target, which helps to extract effective information such as the key actions and attitude changes of the aircraft. Simulation results show that the SBL algorithm can fully utilize the prior information such as sparsity given by the target digital model, achieves good fitting of the RCS curve, and does not require manual setting of sparsity. It accurately characterizes the correspondence between the target angle and RCS, and can effectively invert the target attitude information from the RCS variation law.

Description

Technical Field

[0001] This invention belongs to the field of computer simulation and method optimization technology, specifically relating to an aircraft attitude inversion method based on the SBL algorithm. Background Technology

[0002] As a fundamental target detection device, radar can extract the time-domain and frequency-domain signal characteristics and structural parameters of a target through relevant algorithms, thereby making corresponding judgments about the target's attributes. With the gradual improvement of radar technology at different levels, and based on modern signal processing techniques, many distinguishable radar target characteristic signals have been continuously discovered, leading to the development of related target feature extraction theories and techniques. Among these, radar target feature extraction techniques based on the target's radar cross section (RCS) and its fluctuation characteristics are particularly prominent. The RCS of a target observed by radar is closely related to its observation time and target attitude angle, implicitly containing the target's attitude characteristic information. Therefore, studying and processing the measured target RCS data to extract target attitude angle information becomes crucial.

[0003] Currently, there are two main development trends in RCS feature extraction: one is extracting features with obvious physical significance, such as using RCS to analyze the size, shape, and periodicity of motion of a target; the other is transforming the RCS sequence, extracting the transformed features for classification and recognition, mainly using modern signal processing methods such as wavelet transform and time-frequency analysis, and then performing pattern recognition. Many algorithms have been proposed for these two aspects of RCS feature parameter extraction for space targets. For example, extracting RCS statistical characteristics involves using a large amount of RCS data detected by radar to obtain a statistical model of the echo RCS corresponding to the target size, called a Size Estimation Model (SEM). With SEM, the target size can be estimated by simply inputting the RCS sequence of the detected unknown target into the SEM and matching the features of each target one by one. From the measured RCS echo data, statistical features such as the target's mean, variance, kurtosis, and skewness can be extracted, and further, the target's micro-motion frequency features can be extracted. Traditional methods for extracting micro-motion frequency features mainly include the Autocorrelation Method (AUTOC), the Cyclic Autocorrelation Method (CAUTOC), the Cyclic Average Amplitude Difference Method (CAMDF), and the Analysis of Variance (ANOVA). Among these, CAUTOC and CAMDF algorithms are easy to implement and have simple principles, but they suffer from poor noise resistance and are susceptible to frequency division misjudgments, resulting in poor estimation performance. ANOVA offers some improvement over CAUTOC and CAMDF, exhibiting strong noise resistance and eliminating frequency division misjudgments. However, this algorithm requires a high data rate and relatively long observation periods, posing a challenge to radar performance. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides an aircraft attitude inversion method based on the SBL algorithm. The technical problem to be solved by this invention is achieved through the following technical solution:

[0005] A method for aircraft attitude inversion based on the SBL algorithm includes:

[0006] S100, based on the attitude angle sequence and RCS sequence in the electromagnetic calculation results, constructs a functional relationship expression between angle and RCS through the SBL algorithm, and obtains the parameter vector expression of the angle and RCS model based on the functional relationship expression;

[0007] S200, the observation noise involved in the RCS sequence is introduced into the parameter vector expression of the angle and RCS model to obtain the parameter expression of the angle and RCS model containing the observation noise, and the coefficient vector is solved using the SBL algorithm.

[0008] S300, Based on the coefficient vector, construct an observation angle-RCS model based on sparse reconstruction. Expand and inversely transform the attitude angles in the constructed observation angle-RCS model using the discrete cosine transform matrix to obtain the transformed attitude angle expression, and substitute it into the dictionary matrix to transform the attitude angle inversion problem into the problem of solving a nonlinear representation matrix to obtain the equation to be solved. Solve the equation to be solved using the SBL algorithm to invert the attitude angles, and concatenate all attitude angles to obtain the attitude angle sequence.

[0009] Beneficial effects:

[0010] This invention proposes an aircraft attitude inversion method based on the SBL algorithm, comprising: constructing an observation angle-RCS model based on sparse reconstruction for the electromagnetic calculation results of a specific aircraft; and inverting the radar observation angle based on the SBL algorithm using the measured RCS sequence data of the radar on the flight target, which helps to extract effective information such as the key actions and attitude changes of the aircraft. Simulation results show that the SBL algorithm can fully utilize the prior information such as sparsity given by the target digital model, achieves good fitting of the RCS curve, and does not require manual setting of sparsity. It accurately characterizes the correspondence between the target angle and RCS, and can effectively invert the target attitude information from the RCS variation law.

[0011] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of an aircraft attitude inversion method based on the SBL algorithm provided by the present invention;

[0013] Figure 2 This is a schematic diagram of the process of the SBL-based aircraft attitude inversion algorithm provided by the present invention;

[0014] Figure 3 This is a diagram showing the attitude inversion results of paragraph 1 provided by the present invention;

[0015] Figure 4 This is the attitude inversion result diagram of paragraph 2 provided by the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0017] For the same target, when the radar's operating frequency and polarization remain constant, the RCS measured at the same attitude angle will be a fixed value. However, measured data contains noise from various unpredictable factors, causing fluctuations in the dynamic RCS. Therefore, employing appropriate methods to reduce noise in the measured RCS is crucial to ensuring the accuracy of the dynamic RCS model. After obtaining the target's radar line-of-sight angle, a multi-scattering-center parameterized model of the RCS is constructed using the sparsity of the number of scattering centers. A single scattering-center model can be fitted using multiple sinc functions; therefore, we use sinc functions to fit the angle-RCS curve, and the appropriate fitting method is the SBL algorithm.

[0018] like Figure 1 As shown, this invention provides a method for aircraft attitude inversion based on the SBL algorithm, including:

[0019] S100, based on the attitude angle sequence and RCS sequence in the electromagnetic calculation results, constructs a functional relationship expression between angle and RCS through the SBL algorithm, and obtains the parameter vector expression of the angle and RCS model based on the functional relationship expression;

[0020] In theory, for the same target, when the radar's operating frequency, polarization, and other parameters remain unchanged, the RCS measured at the same attitude angle will be a fixed value. However, measured data contains noise from various unpredictable factors, causing fluctuations in the dynamic RCS. Therefore, employing appropriate methods to reduce noise in the measured RCS is essential to ensure the accuracy of the dynamic RCS model.

[0021] In one specific embodiment of the present invention, combined with Figure 1 and Figure 2 S100 includes:

[0022] S110, using the multi-scattering center model theory and the functional form of the single scattering center parameterized model, construct a functional expression for the angle and RCS; wherein, the functional expression for the angle and RCS is the superposition of the time delays of multiple sinc functions;

[0023] According to the theory of multiple scattering centers, the angle-RCS curve of a target can be regarded as a superposition of multiple scattering center models, and the parameterized model of a single scattering center is in the form of an exponential function or a sinc function. For the electromagnetic scattering characteristics of an aircraft target, we only need to focus on the function expression in the range of 0-180°, within which the function is band-limited. As we know from signal processing principles, band-limited signals can be recovered by sinc interpolation. Therefore, multiple sinc functions can be used to fit the angle-RCS curve to obtain its parameterized model. Thus, given the attitude angle sequence and RCS value sequence, the problem of constructing its parameterized model is simply fitting the curve with sinc functions and calculating its coefficients. This problem can be solved using compressed sensing principles. The known RCS sequence corresponds to the observed signal, the time delay of the sinc function can form a dictionary matrix, and the model coefficients meet the sparsity condition, which can be solved using the SBL algorithm in sparse recovery algorithms.

[0024] The functional expression for angle and RCS can be represented as the superposition of delays of multiple sinc functions. Therefore, the functional expression for angle and RCS is:

[0025] (1)

[0026] in, Represents the RCS value. Indicates the attitude angle. Let represent the coefficient of the i-th sinc function, and M be the total number of sinc functions;

[0027] S120, rewrite the variables in the function expression of angle and RCS in vector form to obtain the vector form function expression of angle and RCS;

[0028] Rewriting the variables in equation (1) in vector form, the function expression of the angle and RCS in vector form is:

[0029] (2)

[0030] in, Represents a static RCS data sequence. This represents a sequence of attitude angles, with values ​​ranging from 1 to 2. , The interpolation precision set for sinc interpolation. ,Pick In engineering applications, it is generally set to 5 times, Let be the sparse coefficient vector to be determined. The dictionary matrix represents the constructed dictionary matrix, since Its precision setting is relatively small, and its column vectors are highly redundant, therefore Most of the sparsity coefficients are 0;

[0031] (3)

[0032] In equation (3), each column of the matrix represents the sinc function for each time delay. For the attitude angle sequence, To set the interpolation precision, a uniform variation is sufficient.

[0033] S130, based on the sparse weighting coefficients in the function expression of angle and RCS, the vector form function expression of angle and RCS is converted into a vector form expression to obtain the parameter vector expression of the angle and RCS model.

[0034] The coefficients in equation (2) are known. It is sparse, and let:

[0035] (4)

[0036] Substituting into equation (2), the expression for the angle and the parameter vector of the RCS model is:

[0037] (5)

[0038] in, , It is a dictionary matrix.

[0039] S200, the observation noise involved in the RCS sequence is introduced into the parameter vector expression of the angle and RCS model to obtain the parameter expression of the angle and RCS model containing the observation noise, and the coefficients are solved using the SBL algorithm;

[0040] In one specific embodiment of the present invention, combined with Figure 1 and Figure 2 S200 includes:

[0041] S210, The observation noise involved in the measured RCS sequence is introduced into the parameter vector expression of the angle and RCS model to obtain the parameter expression of the angle and RCS model that includes the observation noise;

[0042] To account for errors in the modeling, calculation, and measurement processes, an error factor is added to formula (5). Then, the expression for the angle containing observation noise and the parameters of the RCS model is:

[0043] (6)

[0044] In the formula, Indicates observation noise;

[0045] This invention employs the Sparse Bayesian Learning (SBL) algorithm to solve the problem. .

[0046] S220, define the probability density function of the coefficient vector, the probability density function of the observation noise, and the probability density function of the hyperparameters;

[0047] Assume the sparse signal to be reconstructed It follows a complex Gaussian distribution with a mean of 0, and If the inner elements are independent, then the coefficients The probability density function (PDF) is:

[0048] (7)

[0049] in, , , yes The precision (the reciprocal of the variance). It is a by A diagonal array arranged in sequence.

[0050] Observation noise Follows the pattern with a mean of 0 and a precision of The complex Gaussian distribution has the following PDF:

[0051] (8)

[0052] in, .

[0053] and These are called hyperparameters, and they follow a gamma distribution with the following PDFs:

[0054] (9)

[0055] (10)

[0056] in, Represents the gamma function. a and b They are respectively called Shape and dimensional parameters,c and d They are respectively called The shape and dimensional parameters, to ensure and No prior information a , b , c and d The value is usually very small (e.g.) ).

[0057] S230, the posterior probability distribution of the coefficients is solved by using the probability density function of the coefficients, the probability density function of the observation noise, the probability density function of the hyperparameters, and the Gaussian likelihood distribution of the RCS value.

[0058] Based on the above prior information, the Gaussian likelihood distribution of the RCS value y is:

[0059] (11)

[0060] Using (7)-(11), we can obtain that the posterior probability distribution of x is also a complex Gaussian distribution, i.e.

[0061] (12)

[0062] Wherein, the mean vector and the covariance matrix are respectively:

[0063] (13)

[0064] (14)

[0065] Here, H represents the conjugate transpose, using the Woodbury identity.

[0066] (13) and (14) can be rewritten as:

[0067] (15)

[0068] (16)

[0069] Where Q is called the covariance matrix of the observed signal, and its form is:

[0070] (17)

[0071] I represents the identity matrix of appropriate dimensions.

[0072] like Given that, according to the Bayesian method, the maximum a posteriori probability estimate of x is:

[0073] (18)

[0074] S240, based on the posterior probability distribution of the coefficients, updates the hyperparameters cyclically using the EM algorithm until error convergence is achieved, and estimates the coefficient vector based on the hyperparameters at error convergence.

[0075] Hyperparameters can be obtained using the EM algorithm:

[0076] (19)

[0077] (20)

[0078] in It is by A vector consisting of the diagonal elements.

[0079] Specifically, S240 includes:

[0080] S241, Input RCS value and dictionary matrix;

[0081] S242, Initialize hyperparameters , ;

[0082] S243, when ,implement:

[0083] (1) Based on and Update separately and ;

[0084] (2) Based on the updated and renew and ;

[0085] (3) Based on the updated , and convergence threshold Calculate the signal iteration error ,judge Whether it is true; if Proceed to the next iteration; if Exit the iteration and output. ;

[0086] S244, Output Coefficient .

[0087] S300: The attitude angles in the RCS measured sequence are expanded and inversely transformed using the discrete cosine transform matrix to obtain the transformed attitude angles. These transformed attitude angles are then substituted into the dictionary matrix to transform the problem of inverting attitude angles into the problem of solving a nonlinear representation matrix, thus obtaining the equation to be solved. The SBL algorithm is used to solve the equation to be solved, thereby inverting the attitude angles. All attitude angles are then concatenated to obtain the attitude angle sequence.

[0088] When radar detects external targets, the observed RCS time series is related to the changes in its attitude angle, which is the angle between the target axis vector and the radar line-of-sight vector in the projectile coordinate system. Based on the mapping relationship between angle and RCS, theoretically, the target attitude change can be directly inferred from the RCS sequence.

[0089] In one specific embodiment of the present invention, S300 includes:

[0090] S310, Based on the coefficient vector, construct an observation angle-RCS model based on sparse reconstruction. Expand the attitude angles in the constructed observation angle-RCS model using the discrete cosine transform matrix to obtain the expanded sequence. ;

[0091] The observation angle-RCS model based on sparse reconstruction is expressed as:

[0092]

[0093] In the formula, This is a measured RCS sequence with noise. Observation noise;

[0094] Therefore, the target attitude angle inversion problem is: In ,parameter , Given the given information, solve the equation to find the answer. .

[0095] This invention employs Discrete Cosine Transform (DCT) to analyze the signal using a set of orthogonal cosine basis functions. The signal is decomposed. For slowly varying signals dominated by low frequencies, the energy is highly concentrated in the low-frequency coefficients, while the amplitude of the high-frequency coefficients decays exponentially. Therefore, the transformed signal is generally sparsity. Furthermore, the smooth oscillatory characteristics of the cosine basis function are highly compatible with the time-domain correlation of the slowly varying signal. The DCT basis uses an even-symmetric extension (mirror boundary) by default, avoiding high-frequency components introduced by discontinuities at the signal boundaries, further enhancing the sparse distribution of the coefficients. The attitude angle change sequence is smooth and continuous, therefore, it can be transformed into a sparse signal using DCT.

[0096] In equation (21) Expanding using the discrete cosine transform matrix, the expansion sequence is as follows:

[0097]

[0098] here , is the normalization factor.

[0099] S320, for the unfolded sequence Perform an inverse transformation to obtain the attitude angle expression, and then rewrite it to obtain the rewritten expression of the attitude angle;

[0100] Transformed sequence It is sparse, and the expression for the attitude angle after inverse transformation is:

[0101]

[0102] make The above formula can be written as:

[0103] (twenty two)

[0104] make , .

[0105] S330, Substitute the rewritten expression of the attitude angle into the dictionary matrix, thereby transforming the attitude angle inversion problem into solving the expansion matrix. The problem is to obtain the measured RCS sequence and expansion matrix. Relational expressions;

[0106] Substituting equation (22) into The formula yields:

[0107] (twenty three)

[0108] Inversion angle The problem is transformed into a solution Based on equations (21) and (23), the relational expression is as follows:

[0109] (twenty four)

[0110] S340, the nonlinear term of the relational expression described in S330 is applied using the Taylor formula. The expansion is obtained by expanding the expression, which gives the RCS sequence expression containing the initial values ​​of the coefficient vector to be solved, and then it is transformed into the equation to be solved.

[0111] because about It is nonlinear and difficult to solve. Now, Using Taylor's formula Expanding on this point, the expansion formula is:

[0112]

[0113]

[0114]

[0115] The expression for the RCS sequence containing the initial values ​​of the coefficient vector to be determined is:

[0116]

[0117] In the formula: Let be the initial value of the sparse vector to be determined.

[0118] make ,

[0119]

[0120] Define vector , .

[0121] ,

[0122] make , symbol Let the product of Hadamard and the equation to be solved be:

[0123] (25)

[0124] S350, Substitute the initial value of the coefficient vector to be solved into the equation to be solved, and use the SBL algorithm to iteratively solve the equation to be solved, thereby retrieving the attitude angle;

[0125] S360, concatenate all attitude angles in sequence to obtain the attitude angle sequence.

[0126] This invention employs the EM algorithm, which updates parameters iteratively and alternately to correct inaccurate prior initial values. Substituting into equation (25), and estimating using the SBL algorithm, The solution is then used as a priori and substituted back into the equation. This iterative process is repeated to obtain a better estimate.

[0127] To verify the effectiveness of the angle-RCS model construction and inversion method based on the SBL algorithm proposed in this invention, the following simulation comparison experiments were set up.

[0128] Two segments were randomly selected from the dynamic RCS sequence of the flight target measurement, and the algorithm of this invention was used for inversion to verify its effectiveness. The total sampling time for each segment was set to 50 seconds, meaning one RCS sequence contained 100 data points. (Refer to...) Figure 3 , Figure 4 According to Table 1, the present invention can effectively retrieve target attitude information.

[0129] Table 1: Errors in using the inversion results for the two sets of data

[0130]

[0131] This invention proposes an aircraft attitude inversion method based on the SBL algorithm, comprising: constructing an observation angle-RCS model based on sparse reconstruction for the electromagnetic calculation results of a specific aircraft; and inverting the radar observation angle based on the SBL algorithm using the measured RCS data of the radar on the flying target, which helps to extract effective information such as the key actions and attitude changes of the aircraft. Simulation results show that the SBL algorithm can fully utilize the prior information such as sparsity given by the target digital model, achieves good fitting of the RCS curve, and does not require manual setting of sparsity. It accurately characterizes the correspondence between the target angle and RCS, and can effectively invert the target attitude information from the RCS variation law.

[0132] It is worth noting that the terms "first" and "second" in this invention are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0133] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for aircraft attitude inversion based on the SBL algorithm, characterized in that, include: S100, based on the attitude angle sequence and RCS sequence in the electromagnetic calculation results, constructs a functional relationship expression between angle and RCS through the SBL algorithm, and obtains the parameter vector expression of the angle and RCS model based on the functional relationship expression; S200, the observation noise involved in the RCS sequence is introduced into the parameter vector expression of the angle and RCS model to obtain the parameter expression of the angle and RCS model containing the observation noise, and the coefficient vector is solved using the SBL algorithm. S300, Based on the coefficient vector, construct an observation angle-RCS model based on sparse reconstruction. Expand and inversely transform the attitude angles in the constructed observation angle-RCS model using the discrete cosine transform matrix to obtain the transformed attitude angle expression, and substitute it into the dictionary matrix to transform the attitude angle inversion problem into the problem of solving a nonlinear representation matrix to obtain the equation to be solved. Solve the equation to be solved using the SBL algorithm to invert the attitude angles, and concatenate all attitude angles to obtain the attitude angle sequence.

2. The aircraft attitude inversion method based on the SBL algorithm according to claim 1, characterized in that, S100 includes: S110, using the multi-scattering center model theory and the functional form of the single scattering center parameterized model, construct a functional expression for the angle and RCS; wherein, the functional expression for the angle and RCS is the superposition of the time delays of multiple sinc functions; S120, rewrite the variables in the function expression of angle and RCS in vector form to obtain the vector form function expression of angle and RCS; S130, based on the sparse weighting coefficients in the function expression of angle and RCS, the vector form function expression of angle and RCS is converted into a vector form expression to obtain the parameter vector expression of the angle and RCS model.

3. The aircraft attitude inversion method based on the SBL algorithm according to claim 2, characterized in that, The functional expression for the angle and RCS in S110 is as follows: in, Represents the RCS value. Indicates the attitude angle. Let represent the coefficient of the i-th sinc function, and M be the total number of sinc functions; The function expression for the angle in vector form and RCS in S120 is: in, Represents a static RCS data sequence. This represents a sequence of attitude angles, with values ​​ranging from 1 to 2. , The interpolation precision set for sinc interpolation. ,Pick In engineering applications, it is set as follows 5 times, Let be the sparse coefficient vector to be determined. Represents the constructed dictionary matrix; The parameter vector expressions for the angle and RCS model in S130 are as follows: in, , It is a dictionary matrix.

4. The aircraft attitude inversion method based on the SBL algorithm according to claim 1, characterized in that, S200 includes: S210, The observation noise involved in the measured RCS sequence is introduced into the parameter vector expression of the angle and RCS model to obtain the parameter expression of the angle and RCS model that includes the observation noise; S220, define the probability density function of the coefficient vector, the probability density function of the observation noise, and the probability density function of the hyperparameters; S230, using the probability density function of the coefficients, the probability density function of the observation noise, the probability density function of the hyperparameters, and the Gaussian likelihood distribution of the RCS value, the posterior probability distribution of the coefficient vector is solved. S240, based on the posterior probability distribution of the coefficient vector, updates the hyperparameters cyclically using the EM algorithm until error convergence is achieved, and estimates the coefficient vector based on the hyperparameters at error convergence.

5. The aircraft attitude inversion method based on the SBL algorithm according to claim 4, characterized in that, The expressions for the angle and RCS model parameters in S210, which include observation noise, are as follows: In the formula, Indicates observation noise; The probability density function of the coefficient vector in S220 is expressed as: in , , yes accuracy, It is a by A diagonal array arranged in sequence; The probability density function of the observation noise is expressed as: in, ; The probability density function of the hyperparameters in S230 is: in, Represents the gamma function. a and b They are respectively called Shape and dimensional parameters, c and d They are respectively called Shape and dimensional parameters; The Gaussian likelihood distribution of the RCS value in S230 is as follows: The posterior probability distribution of the coefficients is: in, Represents the mean vector. This represents the covariance matrix, with the superscript H indicating the conjugate transpose. , where I represents the identity matrix of appropriate dimensions.

6. The aircraft attitude inversion method based on the SBL algorithm according to claim 5, characterized in that, S240 includes: S241, Input RCS value and dictionary matrix; S242, Initialize hyperparameters , ; S243, when ,implement: (1) Based on and Update separately and ; (2) Based on the updated and renew and ; (3) Based on the updated , and convergence threshold Calculate the signal iteration error ,judge Whether it is true; if Proceed to the next iteration; if Exit the iteration and output. ; S244, Output coefficient vector .

7. The aircraft attitude inversion method based on the SBL algorithm according to claim 1, characterized in that, The S300 includes: S310, Based on the coefficient vector, construct an observation angle-RCS model based on sparse reconstruction. Expand the attitude angles in the constructed observation angle-RCS model using the discrete cosine transform matrix to obtain the expanded sequence. ; S320, for the unfolded sequence Perform an inverse transformation to obtain the attitude angle expression, and then rewrite it to obtain the rewritten expression of the attitude angle; S330, Substitute the rewritten expression of the attitude angle into the dictionary matrix, thereby transforming the attitude angle inversion problem into solving the expansion matrix. The problem is to obtain the measured RCS sequence and expansion matrix. Relational expressions; S340, the nonlinear term of the relational expression described in S330 is applied using the Taylor formula. The expansion is obtained by expanding the expression, which gives the RCS sequence expression containing the initial values ​​of the coefficient vector to be solved, and then it is transformed into the equation to be solved. S350, Substitute the initial value of the coefficient vector to be solved into the equation to be solved, and use the SBL algorithm to iteratively solve the equation to be solved, thereby retrieving the attitude angle; S360, concatenate all attitude angles in sequence to obtain the attitude angle sequence.

8. The aircraft attitude inversion method based on the SBL algorithm according to claim 7, characterized in that, The observation angle-RCS model based on sparse reconstruction in S310 is expressed as follows: In the formula, This is a measured RCS sequence with noise. Observation noise; Expanding the sequence in S310 Represented as: In the formula, , is the normalization factor; The attitude angle expression in S320 is: The rewritten expression for the attitude angle in S320 is: In the formula, ; Measured RCS sequence and expansion matrix in S330 The relational expression is: The expansion in S340 is: The RCS sequence expression containing the initial values ​​of the coefficient vector to be determined in S340 is: ； In the formula, The initial values ​​for the sparse vector to be determined; The equation to be solved in S340 is: In the formula, , , , , This represents the Hadamard product.

Images