A wind profile radar spectrum peak estimation method based on a three-parameter dynamic cost function

Abstract

The application provides a wind profile radar spectrum peak estimation method based on a three-parameter dynamic cost function, comprising the following steps: step 1, distance and Doppler spectrum data preprocessing, three-point sliding average smoothing processing; step 2, noise level estimation based on statistics; step 3, candidate spectrum peak selection based on spatial Doppler window smoothing; step 4, dynamic weight calculation based on covariance matrix eigenvalue decomposition; step 5, construction of a three-parameter cost function of a distance spectrum peak value RSP power item, a continuity item and a spectrum width consistency item, and optimization; step 6, output of an optimal spectrum peak trajectory, and formation of a complete wind speed profile trajectory. The application can improve the spectrum peak recognition precision and stability in a complex environment, and the u and v wind correlation coefficients of the estimation method of the application reach 0.880 and 0.879 through the data set disclosed by an atmospheric radiation measurement website.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of boundary layer wind field inversion, and particularly relates to a wind profile radar spectrum peak estimation method based on a three-parameter dynamic cost function. BACKGROUND

[0002] Wind profile radar is an important meteorological detection equipment, which is widely used to detect wind speed and direction information in the atmospheric boundary layer. This kind of radar utilizes the scattering effect of clear air turbulence on electromagnetic waves, and obtains the wind speed distribution at different height layers through the Doppler principle, and has the ability of continuous detection with high temporal and spatial resolution. Compared with the traditional wind radar, the wind profile radar realizes real-time and continuous monitoring of the three-dimensional wind field of the atmosphere, and has wide application value in aviation meteorology, numerical prediction, boundary layer research and the like.

[0003] However, in actual application, the atmospheric echo signal received by the wind profile radar is usually very weak, and is easily affected by background noise, ground clutter and radio frequency interference, resulting in the decline of signal quality, the ambiguity of spectrum peak, and further affecting the accurate extraction of wind speed. In order to obtain reliable wind field information, accurate spectrum peak estimation of the power spectrum in the echo signal is required to identify the corresponding Doppler frequency component. However, under the conditions of low signal-to-noise ratio and complex background interference, the traditional spectrum peak estimation method is limited by its fixed parameters and static model, and it is difficult to effectively adapt to the dynamic environment, resulting in the decline of spectrum peak recognition rate and frequency estimation accuracy.

[0004] In recent years, researchers have proposed a series of improved algorithms to improve the performance of spectrum peak estimation. For example, the earliest Vaisala single peak method directly selects the spectrum peak with the maximum amplitude in the power spectrum as the atmospheric signal; the maximum entropy method improves the estimation accuracy by improving the frequency spectrum resolution; the adaptive moment estimation method improves the recognition ability of the spectrum peak by setting a dynamic threshold; the multi-parameter cost function comprehensively considers the signal power and differential wind shear, but the weight factor is usually fixed by experience setting, and the limited parameter selection is difficult to fully characterize the feature distribution of the spectrum peak in complex scenes, which is easy to cause estimation deviation. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a wind profile radar spectrum peak estimation method based on a three-parameter dynamic cost function, which solves the problems of the prior art.

[0006] Step 1, distance and Doppler spectrum data preprocessing: the position of ground clutter is determined by finding the local maximum value point closest to zero frequency, the signal change trend is analyzed by using difference operation, the ground clutter influence area is set to zero and smoothed, Kalman filter is used to denoise the spectrum line of each distance library, the process noise and measurement noise parameters are dynamically set, and finally three-point moving average smoothing processing is performed.

[0007] Step 2, perform statistical analysis. Noise level estimation: Sort the power spectrum of each distance library and calculate the mean of noise samples of different lengths. and variance ,pass The statistic is used to find the value closest to 1, and the corresponding sample length is the optimal number of points for estimating the noise level. Based on the optimal number of points, the noise basis of each distance library is determined.

[0008] Step 3: Perform candidate spectral peak selection based on spatial Doppler window smoothing;

[0009] Step 4: Perform dynamic weight calculation based on the eigenvalue decomposition of the covariance matrix;

[0010] Step 5: Construct a three-parameter cost function for the distance spectral peak RSP power term, continuity term, and spectral width consistency term, and optimize it: Construct the RSP power term by quantizing the spectral peak power intensity based on the exponential decay function of the signal-to-noise ratio. By combining wind shear constraints and smoothness constraints, a continuous term is constructed. The current peak spectral width is calculated using the local second moment, and then compared with the reference spectral widths in the historical and neighborhood distance databases using Gaussian similarity to construct a spectral width consistency term. Dynamic weights are used for the RSP power term. Continuous terms And spectral width consistency term An adaptive weighted summation is performed, and the candidate peak with the highest cost value is selected as the optimal spectral peak in the distance library;

[0011] Step 6, Output the optimal spectral peak trajectory: Use a group processing strategy with a group size of N1 (usually 3) distance libraries. Process each distance library one by one starting from the first distance library. Use the speed selected in the previous distance library as a reference for the continuity term. Optimize and select the optimal spectral peak for each distance library through a three-parameter cost function to form a complete wind speed profile trajectory.

[0012] Step 3 includes: calculating the Doppler velocity window. The corresponding speed difference value Where RR is the radar range resolution and vres is the velocity resolution.

[0013] Step 3 also includes: setting a spatial window for each distance library, extracting all spectral data within the spatial window, and calculating the spatial average power spectrum; selecting the top N1 peaks (usually 5) from the average power spectrum as candidate peaks, ensuring that the distance between peaks is greater than the velocity grid difference value. Record the power value and index position of the candidate spectral peaks.

[0014] Step 4 includes:

[0015] Perform an inverse transform on the power spectrum data of each distance gate to construct the covariance matrix. :

[0016] ,

[0017] in, This represents the signal vector of the r-th distance gate in the k-th time window, where K is the number of time windows. This indicates the conjugate transpose.

[0018] Step 4 also includes:

[0019] By analyzing the covariance matrix Eigenvalue decomposition yields a set of eigenvalues ​​arranged in descending order:

[0020] ,

[0021] Where the subscript N represents the dimension of the covariance matrix. The largest eigenvalue corresponds to the energy of the principal component of the signal; It is the Nth eigenvalue;

[0022] By analyzing the intervals between adjacent feature values, a relative gap threshold is used. Automatically divide the signal and noise subspaces and determine the order p of the noise subspace:

[0023] ,

[0024] in Let k be the k-th eigenvalue, ranging from 1 to N, and be the point where k is the maximum value. This means finding a value within the range of variable k that satisfies the subsequent expression. The specific value of k corresponding to when it is true and reaches its maximum value;

[0025] Noise power estimation based on eigenvalue distribution :

[0026] ,

[0027] Where i ranges from 1 to N;

[0028] Further calculation of linear signal-to-noise ratio :

[0029] ,

[0030] And convert it to signal-to-noise ratio in decibels. :

[0031] .

[0032] Step 4 further comprises:

[0033] According to the distribution of all distance gates Statistical quantile as dynamic center value , as input, dynamic weight distribution is realized by mapping with Type function:

[0034] ,

[0035] Wherein, represents the power term weight, and are the upper and lower bounds of the constraint boundary, respectively, and t is the adjustment steepness coefficient; exp is the natural exponential function, represents the signal-to-noise ratio of the rth distance gate;

[0036] The remaining weights are determined by the complementary relationship:

[0037] ,

[0038] Wherein is the sum of the continuity term and the spectral width consistency term; The distribution ratio of the remaining weights between the continuity term and the spectral width consistency term realizes the comprehensive optimization of spectral width consistency and signal stability while ensuring the smoothness of spectral peak continuity, which is determined by verifying a plurality of groups of experimental data as follows:

[0039] ,

[0040] Wherein, , , Respectively represent the weights of the power term , the continuity term and the spectral width consistency term .

[0041] Step 5 comprises:

[0042] The calculation formula of RSP power term Is:

[0043] ,

[0044] Wherein, is the ratio of the power of the current candidate spectral peak to the current distance library noise power bottom , and the calculation formula is:

[0045] ,

[0046] The wind shear term is calculated​ :

[0047] ,

[0048] in, and These are the wind speeds at the current distance and the wind speed at the previous distance, respectively.

[0049] Calculate the smoothness term :

[0050] ,

[0051] in, It is the wind speed at the (i-2)th distance from the reservoir;

[0052] Continuous terms The weighted sum of the wind shear term and the smoothness term:

[0053] ,

[0054] Spectral width consistency term The calculation formula is:

[0055] ,

[0056] in, It is the spectral width of the current peak. It is the reference spectral width. It is the tolerance parameter for spectral width. , where j is the proportionality coefficient.

[0057] Step 5 also includes:

[0058] The total cost function, TotalCost, is obtained by weighted summation of the power term, continuity term, and spectral width consistency term.

[0059] ,

[0060] The optimal spectral peak position is determined by selecting the candidate peak with the highest cost value.

[0061] The present invention also provides an electronic device, including a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method.

[0062] The present invention also provides a storage medium storing a computer program or instructions that, when the computer program or instructions are run on a computer, execute the steps of the method described.

[0063] Compared with the prior art, the present invention has the following beneficial effects: The method of the present invention optimizes the selection of candidate spectral peaks through spatial Doppler window smoothing, accurately divides the signal and noise subspaces, extracts the signal-to-noise ratio features and maps them to dynamic weight factors of the power term, and constructs a three-parameter cost function by combining the distance spectral peak RSP power term, continuity term, and spectral width consistency term for adaptive dynamic summation, thereby achieving optimal estimation of the spectral peak trajectory and improving the accuracy and stability of spectral peak identification in complex environments. Attached Figure Description

[0064] Figure 1 This is a flowchart of a wind profile radar spectral peak estimation method based on a three-parameter dynamic cost function, provided by an embodiment of the present invention.

[0065] Figure 2 This is a comparison chart of the peak estimation results in this invention.

[0066] Figure 3 This is a comparison chart of the peak estimation results in this invention.

[0067] Figure 4 This invention involves estimating the peak values ​​of 16 sets of power spectrum data and comparing them with GPS radiosonde data. Wind scatter plot.

[0068] Figure 5 This invention involves estimating the peak values ​​of 16 sets of power spectrum data and comparing them with GPS radiosonde data. Wind scatter plot. Detailed Implementation

[0069] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0070] like Figure 1 As shown, this embodiment of the invention provides a method for estimating the spectral peak of a wind profiler radar based on a three-parameter dynamic cost function, including the following steps:

[0071] Step 1: Range and Doppler spectral data preprocessing. Ground clutter locations are determined by finding local maxima closest to zero frequency. Differential operations are used to analyze signal variation trends, and the clutter-affected areas are zeroed and smoothed. Kalman filtering is employed to denoise the spectral lines of each range library, dynamically setting process noise and measurement noise parameters, and matrix... Set to:

[0072] ,

[0073] In the formula, Indicates the current distance to the Kudoppler power spectrum The variance of the distance bins, which only scales the diagonal, rather than directly using as a scalar, characterizes the non-stationarity of the turbulence noise in the distance bins. In the case of strong atmospheric turbulence signals large, allowing greater state prediction uncertainty; in the noise-dominated region small, requiring more conservative predicted states. The scaling factor is taken as 0.1, used to balance the need to preserve signal dynamics and suppress excessive prediction jitter.

[0074] Update phase is the covariance of the observation noise, expressed as:

[0075] ,

[0076] where is the standard deviation of the current distance bin Doppler power spectrum .

[0077] Finally, a three-point sliding average smoothing process is performed.

[0078] Step 2, noise level estimation based on statistical quantities . For each distance bin power spectrum, sort the values and calculate the mean and variance of different length noise samples. Find the value closest to 1 through the statistical quantities , and the corresponding sample length is the optimal estimation point of the noise level. Thus, the noise floor of each distance bin is determined.

[0079] Step 3, candidate spectrum peak selection based on spatial Doppler window smoothing. Calculate the Doppler velocity window , is the radar range resolution, unit , the corresponding velocity grid value , is the velocity resolution, unit .

[0080] For the current distance bin to be processed , according to the preset window size , a spatial window containing the current distance bin and its adjacent distance bins is determined. The starting distance bin index of the spatial window is denoted as , the ending distance bin index is denoted as , and the number of distance bins contained in the window is:

[0081] ,

[0082] Extract the index range from the preprocessed Doppler velocity spectrum matrix : The spectral line data corresponding to all distance bins form a spatial window data set:

[0083] ,

[0084] wherein, represents the power spectrum value on the rth distance bin and the mth Doppler unit, and M is the total number of velocity units.

[0085] On this basis, the set X is averaged in the distance bin dimension to obtain the spatially smoothed average power spectrum :

[0086] ,

[0087] wherein, represents the value of the spatially smoothed average power spectrum at the mth velocity unit.

[0088] After obtaining the above spatially smoothed average power spectrum, the power spectrum can be further sorted and filtered, and the top five peaks in the average power spectrum are selected as candidate spectral peaks to ensure that the distance between the peaks is greater than the Doppler window constraint , and the power value and index position of the candidate spectral peak are recorded.

[0089] Step 4, dynamic weight calculation based on eigenvalue decomposition of covariance matrix. The Doppler power spectrum data of each distance gate is inversely transformed into a time domain signal, a covariance matrix is constructed, and eigenvalue decomposition is performed:

[0090] ,

[0091] wherein, represents the signal vector of the rth distance gate under the kth time window, and K is the number of time windows, represents the conjugate transpose.

[0092] By performing eigenvalue decomposition on the covariance matrix , a descendingly arranged eigenvalue set is obtained:

[0093] ,

[0094] wherein, the subscript N is the dimension of the covariance matrix, is the maximum eigenvalue, and corresponds to the signal principal component energy.

[0095] By analyzing the interval between adjacent eigenvalues, a relative gap threshold is used to automatically divide the signal and noise subspaces and determine the noise subspace order p:

[0096] ,

[0097] where, is the kth eigenvalue, and its value range is from 1 to N, Based on eigenvalue variation analysis and multiple sets of data test to find the optimal solution 0.3, which can achieve better results in signal and noise separation.

[0098] Estimate noise power based on eigenvalue distribution :

[0099] ,

[0100] where the threshold position p is determined by the relative gap between adjacent eigenvalues.

[0101] Further calculate the linear signal-to-noise ratio :

[0102] ,

[0103] And convert it to decibel form signal-to-noise ratio :

[0104] ,

[0105] According to the distribution of all distance gates , the quantile point (experimentally verified as the 70th percentile) is taken as the dynamic center value . Take as input, realize dynamic weight allocation through type function mapping:

[0106] ,

[0107] where, represents the power term weight, and are constraint boundaries, taking values of 0.8 and 0.2 respectively, is the adjustment steepness coefficient, represents the signal-to-noise ratio of the rth distance gate.

[0108] The remaining weights are determined by complementary relationship:

[0109] ,

[0110] where is the sum of continuity term and spectral width consistency term. The distribution ratio of the remaining weights between the continuity term and the spectral width consistency term, while ensuring the smoothness of spectral peak continuity, realizes the comprehensive optimization of spectral width consistency and signal stability, which is determined by verifying multiple sets of experimental data as:

[0111] ,

[0112] where, , , denote the weights of power constraint, continuity constraint and spectral width constraint, respectively.

[0113] Step 5, three-parameter cost function of distance spectrum peak RSP power item, continuity item, spectral width consistency item is constructed and optimized. The spectral peak power intensity is quantified based on the exponential decay function of signal-to-noise ratio to construct RSP power item , the power item quantifies the power intensity of the spectral peak. The formula is as follows:

[0114] ,

[0115] where, is the ratio of the power of the current candidate spectral peak to the current distance library noise power bottom, which is used to measure the prominence of the spectral peak relative to the noise, and the formula is:

[0116] ,

[0117] Through the exponential decay function, the spectral peak with higher SNR value will be given higher weight, which ensures the priority of the peak value of strong signal.

[0118] The continuity item is constructed by combining the wind shear constraint and the smoothness constraint to ensure the physical reasonableness of the wind speed between adjacent distance libraries , the continuity item consists of two parts: wind shear item and smoothness item. Large differences in wind speed will lead to large wind shear, and its expression is:

[0119] ,

[0120] where, and are the wind speeds of the current distance library and the previous distance library, respectively. is the Doppler velocity window, which controls the punishment strength of the wind shear item when the wind speed difference is large.

[0121] In order to further smooth the wind speed change, curvature is used to calculate the smoothness constraint:

[0122] ,

[0123] where, , , are the wind speeds of the current and its previous two distance libraries, respectively, is the Doppler velocity window.

[0124] The final continuity item The weighted sum of the wind shear term and the smoothness term is:

[0125] ,

[0126] where 0.7 and 0.3 are the weighting coefficients of the wind shear term and the smoothness term, respectively, indicating that the wind shear has a greater contribution to the continuity constraint. Through experimental verification, the wind shear term has a more significant impact on the continuity of the wind speed, and therefore is given a higher weight, while the smoothness term is used to further smooth the speed variation between adjacent distance bins, and therefore is given a lower weight.

[0127] The spectral width consistency term Based on the local second moment calculation, this term is used to ensure that the spectral width of the current peak is consistent with the spectral widths of the historical peaks and the adjacent distance bins. The calculation method is:

[0128] ,

[0129] where is the spectral width of the current peak, is the reference spectral width, which is obtained by integrating the spectral widths of the previously selected spectral peaks of the current distance bin and the strongest spectral peak of the adjacent distance bin, and then filtered by the median filter, so that it can adaptively reflect the spectral width characteristics of the local meteorological target, is the tolerance parameter of the spectral width, which is set as a fixed proportion coefficient of the reference spectral width , that is, where the proportion coefficient , preferably .

[0130] The power term, the continuity term and the spectral width consistency term are weighted and summed to obtain the total cost function:

[0131] ,

[0132] The optimal spectral peak position is determined by selecting the candidate peak with the maximum cost value.

[0133] Step 6, output the optimal spectral peak trajectory. Use the grouping processing strategy, the group size is 3 distance bins, start processing from the first distance bin, use the selected speed of the previous distance bin as the reference of the continuity term, and select the optimal spectral peak of each distance bin through the three-parameter cost function optimization to form a complete wind speed profile trajectory.

[0134] The experiment adopts the RWP915 wind profile radar data provided by the official website of ARM and the sounding instrument data, the true value of the sounding instrument only contains the data of three-dimensional wind components u, v wind, therefore, the vertical beam of the RWP915 radar data set at 18:39:40 on August 31, the north beam of 42 seconds and the east beam of 48 seconds are selected, and after the data processing and spectrum peak estimation by the method of the application, the three beams are synthesized, and are decomposed into three-dimensional wind components u, v. Figure 2 is the spectrum peak estimation u wind result comparison chart in the application (u, v wind speed components are three-beam synthesis, the sounding instrument data can be used as the true value of the atmospheric wind speed at the collection time point, and the higher the coincidence degree, the better the algorithm effect of the application), Figure 3 is the spectrum peak estimation v wind result comparison chart in the application, it can be seen that the application has stronger robustness and constraint conditions, and has better effect in spectrum peak detection and wind speed estimation, and the obtained wind profile has higher coincidence degree with the measured data.

[0135] In order to improve the scientificity and physical adaptability of the spectrum peak estimation algorithm evaluation of the application, and test the comprehensive performance of the application in different weather scenarios, 16 groups of RWP915 wind profile radar echo data (in August 2008) are selected, Figure 4 is the u wind scatter diagram for comparing the spectrum peak estimation of 16 groups of power spectrum data with the GPS sounding instrument data in the application (the closer the scatter to the diagonal line, the more accurate the spectrum peak recognition of the algorithm, and the smaller the wind speed error), Figure 5 is the v wind scatter diagram for comparing the spectrum peak estimation of 16 groups of power spectrum data with the GPS sounding instrument data in the application, which shows the wind speed component inversion result of the application on the whole verification sample. The closer the scatter to the diagonal line, the more accurate the spectrum peak recognition of the algorithm, the smaller the wind speed error, and the adaptability and robustness of the application under different weather conditions are embodied. The results show that the application can more accurately recognize the atmospheric spectrum peak in multiple experiments, the wind speed estimation result is highly consistent with the sounding instrument data, the error is small and the stability is good, the scatter and the reference line have high matching degree, and the correlation coefficients of u and v wind are 0.88 and 0.879 respectively.

[0136] The application provides a spectrum peak estimation method for wind profile radar based on a three-parameter dynamic cost function, and there are many methods and approaches for realizing the technical scheme, and the above description is only the preferred embodiment of the application, and it should be pointed out that, for ordinary technical personnel in the technical field, some improvements and refinements can be made without departing from the principle of the application, and these improvements and refinements should also be regarded as the protection range of the application. The components not explicitly described in the embodiment can be realized by the existing technology.

Claims

1. A method for wind profile radar spectrum peak estimation based on a three- parameter dynamic cost function, characterized in that, The method comprises the following steps: Step 1, distance and Doppler spectrum data preprocessing: the ground clutter position is determined by finding the local maximum value point closest to zero frequency, the signal change trend is analyzed by using differential operation, the ground clutter influence area is set to zero and smoothed, Kalman filter is used to denoise the spectrum line of each distance bin, the process noise and measurement noise parameters are dynamically set, and finally three-point sliding average smoothing processing is performed; Step 2, perform statistical analysis. Noise level estimation: Sort the power spectrum of each distance library and calculate the mean of noise samples of different lengths. and variance ,pass The statistic is used to find the value closest to 1, and the corresponding sample length is the optimal number of points for estimating the noise level. Based on the optimal number of points, the noise basis of each distance library is determined. Step 3, candidate spectrum peak selection based on spatial Doppler window smoothing is performed; Step 4, dynamic weight calculation based on covariance matrix eigenvalue decomposition is performed; Step 5, construct a three-parameter cost function of the RSP power term, continuity term, and spectral width consistency term, and optimize: the RSP power term is constructed based on the exponential decay function of the signal-to-noise ratio to quantify the spectral peak power intensity , the continuity term is constructed in combination with the wind shear constraint and the smoothness constraint , the spectral width consistency term is constructed by comparing the Gaussian similarity of the current peak spectral width calculated by the local second moment with the reference spectral width of the historical and adjacent distance library ; the RSP power term , the continuity term , and the spectral width consistency term are adaptively weighted and summed by using a dynamic weight, and the candidate peak with the maximum cost value is selected as the optimal spectral peak of the distance library; Step 6, outputting the optimal spectrum peak trajectory: using a grouping processing strategy, the group size is N1 distance bins, starting from the first distance bin, the selected velocity of the previous distance bin is used as the reference of the continuity term, the optimal spectrum peak of each distance bin is selected by three-parameter cost function optimization, and a complete wind speed profile trajectory is formed; Step 3 comprises: calculating Doppler velocity windows corresponding velocity bin values where RR is the radar range resolution and vres is the velocity resolution. The step 3 further comprises: setting a spatial window for each distance bin, extracting all spectral line data within the spatial window, and calculating a spatial average power spectrum; selecting the first N1 peaks in the average power spectrum as candidate spectral peaks, with the distance between the peaks being greater than the velocity grid difference value , and recording the power values and index positions of the candidate spectral peaks.

2. The method of claim 1, wherein, Step 4 comprises: Inverse transform the power spectrum data for each range gate to construct a covariance matrix : ， wherein, denotes the signal vector of the rth distance gate in the kth time window, K is the number of time windows, denotes the conjugate transpose.

3. The method of claim 2, wherein, Step 4 further comprises: By performing eigenvalue decomposition on the covariance matrix a set of eigenvalues in descending order is obtained: ， wherein subscript N is the dimension of the covariance matrix, is the maximum eigenvalue, corresponding to the signal principal component energy; is the Nth eigenvalue; By analyzing the interval between adjacent eigenvalues, using the relative gap threshold Automatic division of signal and noise subspace, determine the noise subspace order p: ， wherein is the kth eigenvalue, with a value range from 1 to N, the maximum point with respect to k represents the specific value of k found in the value range of variable k that can make the subsequent expression hold true and reach the maximum value; Estimating noise power based on eigenvalue distribution : ， Wherein, the value range of i is 1 to N; Further computing linear signal-to-noise ratio : ， and converted to decibel form signal-to-noise ratio : 。 4. The method of claim 3, wherein, Step 4 further comprises: Based on all distances to the door Distribution, statistical quantiles as dynamic central values ,by For input, through Type function mapping enables dynamic weight allocation: ， wherein, represents a power term weight, and are the upper and lower bounds of the constraint boundary, respectively, t is the adjustment steepness coefficient; exp is the natural exponential function, represents the signal-to-noise ratio of the rth distance gate; The remaining weights are determined by the complementary relationship: ， wherein is the sum of continuity term and spectral width consistency term; the distribution ratio of the remaining weight between the continuity term and the spectral width consistency term realizes the comprehensive optimization of spectral width consistency and signal stability while ensuring the smoothness of spectral peak continuity, which is determined as: ， wherein , , represent the weights of the power terms , the continuity terms and the spectral width consistency terms , respectively.

5. The method of claim 4, wherein, Step 5 comprises: RSP power term The calculation formula is: ， wherein, the power of the current candidate spectral peak the ratio of the current distance bin noise power floor is calculated as: ， Computing the wind shear term : ， wherein, and are the current distance bin wind speed and the previous distance bin wind speed, respectively; computing a smoothness term : ， wherein, is the wind speed of the i-2nddistance bin; continuity term is a weighted sum of the wind shear term and the smoothness term: ， Spectrum width consistency term The calculation formula is: ， wherein, is the spectral width of the current peak, is the spectral width of the reference, is a tolerance parameter for the spectral width, where j is a proportionality factor.

6. The method of claim 5, wherein, Step 5 further comprises: The power term, the continuity term and the spectrum width consistency term are weighted and summed to obtain the total cost function TotalCost: ， The optimal spectrum peak position is determined by selecting the candidate peak with the maximum cost value.

7. An electronic device, comprising: A processor and a memory are included, and the memory stores program code, when the program code is executed by the processor, the processor executes the steps of the method as claimed in any one of claims 1 to 6.

8. A storage medium, characterized by The computer program or instructions are stored, and when the computer program or instructions are run on the computer, the steps of the method as claimed in any one of claims 1 to 6 are executed.

Images