A CEEMDAN-sPCA algorithm to correct GNSS-RTK multipath errors
Through signal decomposition and principal component analysis, the CEEMDAN-sPCA algorithm solves the problems of multipath error and noise interference in GNSS-RTK technology, improves monitoring accuracy, and realizes effective extraction and correction of multipath errors.
Patent Information
- Application Number
- CN202311661338.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-05
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2043-12-05
AI Technical Summary
When GNSS-RTK technology monitors dynamic deformation of large-scale civil engineering structures, background noise caused by multipath error and random instrument noise is difficult to eliminate, reducing monitoring accuracy.
Using CEEMDAN-sPCA algorithm, through signal decomposition, screening, reconstruction and principal component analysis, the instrument random noise interference is weakened, the GNSS-RTK multipath error is extracted and corrected, and the IMF component is filtered using the double threshold value of correlation coefficient and effective coefficient, based on STD value Correct the principal component components to obtain the multipath error model.
Effectively separate the white noise inside the instrument, improves the GNSS-RTK monitoring accuracy, and the obtained principal component component has the largest correlation coefficient with the original signal, confirming the reliability and superiority of the CEEMDAN-sPCA algorithm.
Smart Images

Figure CN117665862B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of GNSS positioning technology, and in particular to a CEEMDAN-sPCA algorithm for correcting GNSS-RTK multipath errors. Background Art
[0002] Monitoring the dynamic deformation of large civil engineering structures based on the Global Navigation Satellite System-Real Time Kinetic (GNSS-RTK) technology has many advantages and broad prospects. The measurement accuracy of the GNSS carrier phase differential technology itself can reach 0.5-2.0 mm, but the monitoring process is carried out under the influence of multiple error sources, resulting in the actual monitoring accuracy being lower than this accuracy. GNSS-RTK monitoring errors include multiple error sources such as satellite-related errors, receiver and station-related errors, and signal propagation-related errors. Among them, satellite-related errors, receiver clock errors, ionosphere and tropospheric delays, etc. can be eliminated by differential treatment. Since multipath errors are not correlated between monitoring stations and base stations, they cannot be eliminated by differential treatment. Multipath errors and random noise inside the instrument are the main error sources of GNSS-RTK, collectively referred to as background noise. Even under ideal conditions, the error caused by background noise can reach 2 cm. How to further reduce the background noise of GNSS-RTK and improve its monitoring accuracy is a research hotspot and cutting-edge issue in the field.
[0003] Multipath error has a periodic repetitive characteristic. Therefore, for the GNSS-RTK continuous multi-day fixed-point monitoring application scenario, the multipath error can be corrected by extracting the multipath error model from the multi-day data. The principal component analysis (PCA) algorithm can effectively extract the principal components between sequences, and has a good extraction and correction effect on the multipath error and annual / seasonal characteristic error in the GNSS-RTK sequence. However, due to the interference of instrument random noise in the monitoring signal, the effect of a single PCA algorithm is poor. Therefore, a CEEMDAN-sPCA algorithm is proposed to first weaken the instrument random noise interference, and then extract and correct the GNSS-RTK multipath error. Summary of the invention
[0004] Based on the above objectives, the present invention provides a CEEMDAN-sPCA algorithm to correct GNSS-RTK multipath errors.
[0005] A CEEMDAN-sPCA algorithm corrects GNSS-RTK multipath errors, including the following steps:
[0006] S1: Perform CEEMDAN decomposition on the signal to obtain multiple IMF components and one residual component;
[0007] S2: Screen and reconstruct the IMF components, calculate the correlation coefficient and effective coefficient respectively, and calculate the correlation coefficient threshold, based on which the IMF components are screened and reconstructed;
[0008] S3: Perform S1 and S2 processing on multiple groups of signals respectively to obtain GNSS-RTK multipath residual sequence;
[0009] S4: The signal components of the GNSS-RTK multipath residual sequence are standardized one by one;
[0010] S5: Perform principal component analysis to identify the principal components;
[0011] S6: Calculate the principal component of the STD value correction to obtain the multipath error model.
[0012] Furthermore, the S1 specifically includes:
[0013] S11: Add adaptive white noise ε0E1ω to the original signal x(t) j (t), and then perform EMD decomposition to obtain the initial IMF component f1 j (t), for f1 j (t) is averaged to obtain the first-order IMF component, which is expressed as:
[0014] S12: The original signal x(t) and the IMF component Subtract and get the new signal to be decomposed
[0015] S13: Add white noise ε1E1ω to r1(t) i (t) Continue to decompose and obtain the second-order IMF component, which is expressed as: Where: E1[r1(t)+ε1E1(ω j (t))] represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε i is the standard deviation of added white noise;
[0016] S14: Repeat the process to obtain the i-th signal to be decomposed:
[0017] S15: Calculate the i+1th order IMF component as:
[0018] S16: Circularly decompose the signal and IMF components until the decomposition is complete and the final residual component is obtained:
[0019] Furthermore, the S2 specifically includes:
[0020] S21: The correlation coefficient is used to screen and identify the real IMF components. The expression of the correlation coefficient is: Among them, x(t) and y(t) represent the time course variable and signal variable respectively, and N represents the signal length;
[0021] S22: The threshold for determining the noise-dominated IMF component based on the correlation coefficient, which is defined as: Among them, ρ i is the correlation coefficient between the ith IMF component and the original signal, λ is the scale factor, which is 8;
[0022] S23: Screening IMF components based on the energy coefficient determination method based on the effective coefficient. The expression of the energy coefficient is: Among them, D i is the energy density, is the average period, and the expression of energy density is: Among them, [f i (t)] is the amplitude of the ith IMF component, and the expression of the average period is: Among them, O i is the number of extreme values of the ith IMF component;
[0023] S24: The effective coefficient is used to determine whether the energy coefficient is a constant, and its expression is: When the effective coefficient Ri>1, the energy coefficient of the i-th item increases by multiples relative to the average value of the first i-1 items, and the first i-1 IMF components are used as white noise characteristic components;
[0024] S25: Based on the correlation coefficient threshold ε and the effective coefficient threshold R i , determine that IMF1~i-1 components are the dominant components of white noise, and subtract the white noise component from the original signal, expressed as: That is, the multipath residual component. Multiple sets of data are processed, the white noise components are removed respectively, and then the signal is reconstructed, and finally the multipath residual sequence [X1 X2X j ……].
[0025] Furthermore, the standardization process in S4 is based on the formula: in, and std(X0) are the mean and standard deviation of X0, respectively.
[0026] Furthermore, the S5 specifically includes:
[0027] The covariance matrix B of the data matrix X is expressed as: B = XT X, B are n×n symmetric matrices, and the orthogonal decomposition of B is expressed as: B = νΛν T , where Λ is a diagonal matrix with eigenvalues λ1>λ2>…>λ k >0, where k≤n; ν T is the eigenvector matrix, The eigenvectors are mutually orthogonal;
[0028] The eigenvalue decomposition of matrix X is: T X=νΛν T , where ν T As an orthogonal basis of X, it is expanded as: X = Aν T ;
[0029] The transpose of an orthogonal matrix is equal to the inverse matrix of the orthogonal matrix, formula X = Aν T Transformed to: X = Aν, where A is the m×k principal component matrix, also expressed as: Among them, a ij is the jth principal component in the i-th row, x ip is the variable corresponding to the i-th row and p-th column of matrix X, ν pj For the j-th row component of the p-th eigenvector (i=1...m, j=1...k, p=1...n), obtain the k-column principal component components.
[0030] Furthermore, S5 also includes a cumulative contribution rate method to ensure that the principal component fully represents the common characteristics of the original variables. The cumulative contribution rate of the characteristic root is used to reflect the ability of the principal component to integrate the original variables. The formula is: When the cumulative contribution rate exceeds the 75% threshold, the corresponding first q-order principal components are the principal component components.
[0031] Furthermore, in S6, in order to avoid a large amplitude difference between the principal component decomposed by PCA and the original signal, the principal component is calculated based on the sPCA algorithm. The sPCA algorithm adds a correction of the principal component based on the STD value on the basis of the classic PCA algorithm, which is expressed as:
[0032] In the formula, x i (t) represents the signal timing sequence, represents the signal mean, and N is the signal length.
[0033] In the formula, X PCA_0 Represents the principal component extracted by the PCA algorithm, std(X PCA_0 ) is the standard deviation of the principal components extracted by the PCA algorithm;
[0034] Based on the X calculated by the above sPCA algorithm PCA , reflecting the common components among variables, which is the multipath error correction model.
[0035] Beneficial effects of the present invention:
[0036] The present invention screens and reconstructs IMF components based on dual thresholds of correlation coefficient and effective coefficient, which can weaken high-frequency white noise components to the maximum extent, and corrects the principal component components based on the STD value to avoid large amplitude differences between the principal component components decomposed by PCA and the original signal. Therefore, the principal component components extracted by the sPCA algorithm have better amplitude consistency with the original pure signal than the PCA algorithm.
[0037] In the present invention, the correlation coefficient value between the principal component obtained based on the CEEMDAN-sPCA algorithm and the original simulation signal is the largest, which proves the reliability of the CEEMDAN-sPCA combined algorithm and its superiority over a single algorithm.
[0038] The present invention has a good application effect on the processing of GNSS-RTK continuous multi-day fixed-point monitoring data, and can effectively separate the internal white noise of the instrument and then extract and correct the multipath error. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0040] Figure 1 A schematic diagram of the application process of the CEEMDAN-sPCA joint algorithm according to an embodiment of the present invention;
[0041] Figure 2 A schematic diagram of a simulation signal according to an embodiment of the present invention;
[0042] Figure 3 It is a schematic diagram of a displacement residual sequence after CEEMDAN decomposition and reconstruction of a simulation signal according to an embodiment of the present invention;
[0043] Figure 4 A schematic diagram of a multipath error model extracted by a CEEMDAN-sPCA joint algorithm for a simulation signal according to an embodiment of the present invention;
[0044] Figure 5 Schematic diagram of a timing sequence of a GNSS-RTK on-site measured signal relative displacement according to an embodiment of the present invention;
[0045] Figure 6It is a schematic diagram of the displacement time sequence of the field measured signal after being decomposed and reconstructed by CEEMDAN according to an embodiment of the present invention;
[0046] Figure 7 Schematic diagram of a multipath error model extracted from a field measured signal using a CEEMDAN-sPCA joint algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION
[0047] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.
[0048] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0049] Embodiment 1, based on simulation signal
[0050] like Figure 1-Figure 7 As shown, a CEEMDAN-sPCA algorithm is used to correct GNSS-RTK multipath errors. The simulation signal application specifically includes the following steps:
[0051] S1. Construct digital simulation signals, including:
[0052] S1-1. Construct a digital simulation signal x0 based on a sine function with a frequency of 1 Hz, and its expression is: x0 = 3·sin(2π·t);
[0053] S1-2. Create three groups of variables related to x0, and the expressions are: x3=x0+ω3-ω4. In the formula: ω1 - ω4 are all random variables that obey the normal distribution;
[0054] S1-3, the signal sampling frequency is 0.1s, the duration is 100s, and it is assumed that the amplitude unit of the simulation signal is centimeters (cm), such as Figure 2As shown in the figure (from left to right, from top to bottom: pure sinusoidal signal x, related signal x1, related signal x2, related signal x3), the displacement time history curves of the pure signal x0 and the related signals x1, x2 and x3 are given respectively. The correlation coefficients between the pure signal x0 and the related signals x1, x2 and x3 are: C 0-1 =0.8398, C 0-2 =0.8505 and C 0-3 =0.8441.
[0055] S2, for the related signals x1, x2 and x3, according to the formula formula formula formula formula formula Perform CEEMDAN decomposition. Taking the signal x1 processing as an example, it specifically includes:
[0056] S2-1, add adaptive white noise ε0E1ω to signal x1 j (t), and then perform EMD decomposition to obtain the first-order IMF component f1 j (t). Then for f1 j (t) is averaged to obtain the first-order IMF component
[0057] S2-2, the original signal x(t) and the IMF component Subtract and get the new signal to be decomposed
[0058] S2-3. Add white noise ε1E1ω to r1(t) i (t) Continue to decompose and obtain the second-order IMF component, which is expressed as: Where: E1[r1(t)+ε1E1(ω j (t))] represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε i is the standard deviation of added white noise;
[0059] S2-4, repeat the process to obtain the i-th signal to be decomposed:
[0060] S2-5. Calculate the i+1th order IMF component as:
[0061] S2-6, cyclically decompose the signal and IMF components until the decomposition is complete and the final residual component is obtained:
[0062] S3, IMF component screening and reconstruction, including:
[0063] S3-1, according to the formula and formula Calculate the correlation coefficient and the effective coefficient respectively, and use the formula Calculate the correlation coefficient threshold, then filter and reconstruct the IMF component to obtain the displacement residual sequence X without white noise interference. C =[X1 X2 X3] 1000×3 ;
[0064] S3-2, Figure 3 As shown in the figure (from left to right, from top to bottom: displacement residual sequence X1, displacement residual sequence X2, displacement residual sequence X3), the three groups of displacement residual sequences obtained based on the CEEMDAN algorithm still have some burrs, but the drift error is basically eliminated, showing obvious sinusoidal signal periodic characteristics, and the amplitudes and trends of the three are well consistent;
[0065] S3-3. After calculation, the correlation coefficients between the displacement residual sequences X1, X2 and X3 are: C C1-2 =0.9278, C C1-3 =0.8795 and C C2-3 =0.8996. The correlation coefficients between the displacement residual sequences obtained by the CEEMDAN algorithm are much higher than those between the original signals. The correlation coefficient between sequence 1 and sequence 2 is as high as 0.9278.
[0066] S3-4. After calculation, the correlation coefficients between the original signal x0 and the displacement residual sequences X1, X2 and X3 are: C C1-0 =0.9594, C C2-0 =0.9658 and C C3-0 =0.9318, which fully proves the necessity and effectiveness of CEEMDAN to decompose the white noise component.
[0067] S4. According to the formula The GNSS-RTK displacement residual sequence is standardized. Where: and std(X0) are the mean and standard deviation of X0, respectively.
[0068] S5. For the multipath residual sequence X C =[X1 X2 X3] 1000×3 The constructed data matrix is used for principal component analysis;
[0069] S5 specifically includes:
[0070] S5-1. The covariance matrix B of the data matrix X can be expressed as: B = X TX;
[0071] S5-2, B is a symmetric matrix of n×n, and the orthogonal decomposition of B can be expressed as: B = νΛν T Λ is a diagonal matrix, with eigenvalues λ1>λ2>…>λ k >0(k≤n); ν T is the eigenvector matrix, The eigenvectors are mutually orthogonal;
[0072] S5-3, the eigenvalue decomposition form of matrix X is: T X=νΛν T ;
[0073] S5-4, ν T As an orthogonal basis of X, X can be expanded as: X = Aν T ;
[0074] S5-5. The transpose of an orthogonal matrix is equal to the inverse matrix of the orthogonal matrix, formula X = Aν T It can be converted to: X = Aν. Where: A is the m×k principal component matrix;
[0075] S5-6, formula X = Aν can be further expressed as: Where: a ij is the jth principal component in the i-th row, x ip is the variable corresponding to the i-th row and p-th column of matrix X, ν pj is the j-th row component of the p-th eigenvector (i=1…m, j=1…k, p=1…n);
[0076] S5-7. According to S5-6, a total of k columns of principal component components can be obtained. In order to ensure that the principal component components fully reflect the common characteristics of the original variables, it is necessary to use the principal component components according to the cumulative contribution rate method, and the formula is: When the cumulative contribution rate exceeds the 75% threshold, the corresponding first q-order principal components are the main mode components;
[0077] After calculation, the contribution rate of the first principal component is 93.77%, which is the displacement residual matrix X c The common component X cp_0 .
[0078] S6. In order to avoid the large amplitude difference between the main component decomposed by PCA and the original signal, the sPCA algorithm is further proposed. According to the formula Calculate the STD value to correct the principal component and obtain the multipath error model;
[0079] S6 specifically includes:
[0080] S6-1. The standard deviation can reflect the discreteness of the signal itself, and its expression is: Where: x i (t) represents the signal timing sequence, represents the signal mean, N is the signal length;
[0081] S6-2, the principal component modified based on STD value can be expressed as: In the formula, X cp_0 Represents the principal component extracted by the PCA algorithm, std(X PCA_0 ) is the standard deviation of the principal components extracted by the PCA algorithm;
[0082] S6-3. Multipath error model X extracted by CEEMDAN-sPCA joint algorithm for simulation signals cp ,like Figure 4 As shown in the figure, compared with the three groups of displacement residual sequences obtained by the CEEMDAN algorithm, the burrs of the multipath error components obtained by the CEEMDAN-sPCA algorithm are significantly reduced, the periodic characteristics of the sinusoidal signal are obvious, and its value and trend show good consistency with the original pure signal;
[0083] S6-4, after calculation, the correlation coefficient C between the principal component and the original pure signal CP1-0 Up to 0.9830;
[0084] S6-5. After calculation, the correlation coefficient between the single sPCA algorithm result and the pure signal is 0.9539, and the correlation coefficient (mean) between the single CEEMDAN algorithm result and the pure signal is 0.9523;
[0085] S6-6. The above results confirm that the sPCA algorithm, CEEMDAN algorithm and CEEMDAN-sPCA combined algorithm can all effectively extract common components between variables, and the principal component obtained by the CEEMDAN-sPCA algorithm has the largest correlation coefficient with the pure signal, which confirms the reliability of the CEEMDAN-sPCA combined algorithm and its superiority over a single algorithm.
[0086] Example 2, based on field measured GNSS-RTK signals
[0087] like Figure 1-Figure 7 As shown in Figure 1, a CEEMDAN-sPCA algorithm is used to correct GNSS-RTK multipath errors. The application of measured signals specifically includes the following steps:
[0088] S1. Analyze the relative displacement sequence of GNSS-RTK field measured signals of the same period, elevation direction and duration of 10800s for 3 days, such as Figure 5 As shown ( Figure 5From top to bottom, they are the relative displacement sequence on the first day, the relative displacement sequence on the second day, and the relative displacement sequence on the third day);
[0089] S2. For the 3-day relative displacement sequence, according to the formula formula formula formula formula formula Perform CEEMDAN decomposition. Taking the CEEMDAN processing of the relative displacement sequence on the first day as an example, it specifically includes:
[0090] S2-1, add adaptive white noise ε0E1ω to signal x1 j (t), and then perform EMD decomposition to obtain the first-order IMF component f1 j (t). Then for f1 j (t) is averaged to obtain the first-order IMF component
[0091] S2-2. Original signal x(t) and IMF components Subtract and get the new signal to be decomposed
[0092] S2-3. Add white noise ε1E1ω to r1(t) i (t) Continue to decompose and obtain the second-order IMF component Where: E1[r1(t)+ε1E1(ω j (t))] represents an EMD decomposition, ω j (t) is the white noise added for the jth time, ε i is the standard deviation of added white noise;
[0093] S2-4, and so on, we get the i-th signal to be decomposed:
[0094] S2-5. Then, the i+1th order IMF component is obtained as:
[0095] S2-6. Repeat the steps of S2-4 and S2-5 until the decomposition is complete, and obtain the final residual component:
[0096] S3, IMF component screening and reconstruction, according to the formula and formula Calculate the correlation coefficient and the effective coefficient respectively, and use the formula Calculate the correlation coefficient threshold, then filter and reconstruct the IMF components to obtain the GNSS-RTK displacement residual sequence, including:
[0097] S3-1, according to the formula Japanese style Calculate the correlation coefficient and the effective coefficient respectively, and use the formula Calculate the correlation coefficient threshold;
[0098] S3-2. After calculation, the correlation coefficient thresholds of the 3-day data are: ε C1 =0.4285 / 8=0.0536, ε C2 =0.4579 / 8=0.0572 and ε C3 =0.4663 / 8=0.0583; According to the effective coefficient threshold, in the 3-day sequence, IMF1~6 components are dominated by white noise, and IMF7 and subsequent components are dominated by multipath errors;
[0099] S3-3, removing white noise components from the three sets of data and reconstructing them as multipath residual sequences;
[0100] S3-4, the displacement time program after the field measured signal is decomposed and reconstructed by CEEMDAN is as follows Figure 6 As shown, the three groups of multipath residual sequences are also distributed within the range of ±0.02m and have good consistency;
[0101] S3-5. After calculation, the three groups of multipath residual sequences are also distributed within the range of ±0.02m and have good consistency. The correlation coefficients are: C C1-2 =0.8989, C C1-3 =0.8557 and C C2-3 =0.9489;
[0102] S3-6. The correlation coefficients between multipath residual sequences are greatly improved compared with the original signals, which proves that the CEEMDAN algorithm can effectively reduce white noise interference and help obtain multipath residual sequences with higher correlation, which is beneficial to the next step of principal component analysis.
[0103] S4. According to the formula The GNSS-RTK displacement residual sequence is standardized. Where: and std(X0) are the mean and standard deviation of X0 respectively;
[0104] S5, performing principal component analysis on the data matrix consisting of the 3-day multipath residual sequence;
[0105] S5 specifically includes:
[0106] S5-1. The covariance matrix B of the data matrix X can be expressed as: B = X T X;
[0107] S5-2, B is a symmetric matrix of n×n, and the orthogonal decomposition of B can be expressed as: B = νΛν T Λ is a diagonal matrix, with eigenvalues λ1>λ2>…>λ k >0(k≤n); ν T is the eigenvector matrix, The eigenvectors are mutually orthogonal;
[0108] S5-3, the eigenvalue decomposition form of matrix X is: T X=νΛν T ;
[0109] S5-4, ν T As an orthogonal basis of X, X can be expanded as: X = Aν T ;
[0110] S5-5. The transpose of an orthogonal matrix is equal to the inverse matrix of the orthogonal matrix, formula X = Aν T It can be converted to: X = Aν. Where: A is the m×k principal component matrix;
[0111] S5-6, formula X = Aν can be further expressed as: Where: a ij is the jth principal component in the i-th row, x ip is the variable corresponding to the i-th row and p-th column of matrix X, ν pj is the j-th row component of the p-th eigenvector (i=1…m, j=1…k, p=1…n);
[0112] S5-7. According to S5-6, a total of k columns of principal component components can be obtained. In order to ensure that the principal component components fully reflect the common characteristics of the original variables, the principal component components are determined according to the cumulative contribution rate method, and the formula is: Generally speaking, when the cumulative contribution rate exceeds the 75% threshold, the corresponding first q-order principal components are the main mode components. After calculation, the contribution rate of the first principal component is 93.47%, and this principal component can be used as the principal component of the three groups of multipath residual sequences.
[0113] S6. According to the formula Calculate the principal component of the STD value correction to obtain the multipath error model;
[0114] S6 specifically includes:
[0115] S6-1. Calculate the signal standard deviation, the expression is: Where: x i (t) represents the signal timing sequence, represents the signal mean, N is the signal length;
[0116] S6-2, the principal component modified based on STD value can be expressed as: Where, X cp_0 Represents the principal component extracted by the PCA algorithm, std(X PCA_0 ) is the standard deviation of the principal components extracted by the PCA algorithm;
[0117] S6-3. Multipath error model X extracted by field measured signals using CEEMDAN-sPCA joint algorithm PCA ;like Figure 7 As shown in Figure 2, the multipath residuals extracted by the sPCA algorithm have similar amplitude and trend.
[0118] Good consistency;
[0119] S6-4. After calculation, the correlation coefficients between the multipath error extracted by the sPCA algorithm and the 3-day multipath residual are: C CP-1 =0.9483, C CP-2 =0.9819 and C CP-3 =0.9668;
[0120] S6-5. The multipath errors separated by the CEEMDAN-sPCA algorithm have a high correlation coefficient with the 3-day multipath residual sequence, which confirms the effectiveness of the CEEMDAN-sPCA algorithm.
[0121] Conclusion of the practical application:
[0122] Through the analysis results of digital simulation signals and on-site GNSS-RTK measured signals, it is confirmed that the use of CEEMDAN-sPCA algorithm to extract GNSS-RTK multipath errors can minimize the interference of instrument white noise, effectively extract the GNSS-RTK multipath error correction model, and improve the GNSS-RTK monitoring accuracy.
[0123] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.
[0124] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A CEEMDAN-sPCA algorithm to correct GNSS-RTK multipath errors. It is characterized in that The following steps are involved: S1: Perform CEEMDAN decomposition on the signal to obtain multiple IMF components and one residual component; S2: Screen and reconstruct the IMF components, calculate the correlation coefficient and effective coefficient respectively, and calculate the correlation coefficient threshold, based on which the IMF components are screened and reconstructed; S3: Perform S1 and S2 processing on multiple groups of signals respectively to obtain GNSS-RTK multipath residual sequence; S4: Perform standardization processing on the GNSS-RTK multipath residual sequence signal, and the standardization processing is based on the formula: ,in, and std( X 0 ) are respectively X 0 The mean and standard deviation of S5: Perform principal component analysis to identify the principal component components, including: The data matrix X The covariance matrix of B It is expressed as: , B for n × n The symmetric matrix of B The orthogonal decomposition of is expressed as: ,in, is a diagonal matrix, with eigenvalues Composition, among which, k ≤ n ; is the eigenvector matrix, , the eigenvectors are mutually orthogonal; The eigenvalue decomposition of matrix X is: ,in, As X The orthogonal basis expansion of is: ; The transpose of an orthogonal matrix is equal to the inverse matrix of the orthogonal matrix. Converts to: ,in, A for m × k The principal component matrix of is also expressed as: ,in, For the i Line j principal components, For the matrix X No. i Row, No. p The variables corresponding to the columns, For the p The first feature vector j Row Component( i =1… m , j =1… k , p =1… n ), get k Column principal component weights; The S5 also includes a cumulative contribution rate method to ensure that the principal component fully represents the common characteristics of the original variables. The cumulative contribution rate of the characteristic root is used to reflect the ability of the principal component to integrate the original variables. The formula is: , when the cumulative contribution rate exceeds the 75% threshold, the corresponding first q-order principal components are the principal component components; S6: Calculate the principal component corrected by the STD value to obtain a multipath error model. The principal component is calculated based on the sPCA algorithm. The sPCA algorithm adds the principal component corrected based on the STD value on the basis of the classic PCA algorithm, which is expressed as: , where Indicates the sequence of signal timing, represents the signal mean, N is the signal length; , where Represents the principal component extracted by the PCA algorithm, The standard deviation of the principal components extracted by the PCA algorithm; Based on the above sPCA algorithm, , reflecting the common components among variables, which is the multipath error correction model.
2. According to claim 1, a CEEMDAN-sPCA algorithm for correcting GNSS-RTK multipath errors, It is characterized in that The S1 specifically includes: S11: For the original signal Add adaptive white noise , and then perform EMD decomposition to obtain the initial IMF component ,right After averaging, the first-order IMF component is obtained, which is expressed as: ; S12: The original signal With IMF component Subtract and get the new signal to be decomposed ; S13: Yes Add white noise Continue to decompose and get the second-order IMF component, expressed as: , where: Represents an EMD decomposition. For the j The white noise added, is the standard deviation of added white noise; S14: Repeat the process to obtain i Signals to be decomposed: ; S15: Calculate the i The +1-order IMF component is: ; S16: Circularly decompose the signal and IMF components until the decomposition is complete and the final residual component is obtained: .
3. According to claim 2, a CEEMDAN-sPCA algorithm for correcting GNSS-RTK multipath errors, It is characterized in that The S2 specifically includes: S21: The correlation coefficient is used to screen and identify the real IMF components. The expression of the correlation coefficient is: ,in, and Represent the time course variable and signal variable respectively, and N represents the signal length; S22: The threshold for determining the noise-dominated IMF component based on the correlation coefficient, which is defined as: ,in, For the i The correlation coefficient between the IMF component and the original signal is is the scale factor, the value is 8; S23: Screening IMF components based on the energy coefficient determination method based on the effective coefficient. The expression of the energy coefficient is: ,in, is the energy density, is the average period, and the expression of energy density is: ,in, It is i The amplitude of the IMF component and the expression of the average period are: ,in, It is i The number of extreme values of an IMF component; S24: The effective coefficient is used to determine whether the energy coefficient is a constant, and its expression is: , when the effective coefficient >1, No. i The energy coefficient of the term is relative to the previous i -1 The average value of the item increases by multiples, i -1 IMF component as white noise characteristic component; S25: Based on correlation coefficient threshold and effective coefficient threshold , determine IMF1~ i The -1 component is the dominant component of white noise. The original signal minus the white noise component is expressed as: , which is the multipath residual component, processes multiple sets of data, removes the white noise component respectively, and then reconstructs the signal, and finally obtains the multipath residual sequence .
