Coherent X-waveband microwave radar average wave period estimation method based on machine learning

Abstract

The invention provides a machine learning-based coherent X-waveband microwave radar average wave period estimation method, relates to the field of radar signal processing and artificial intelligence, and aims to solve the problem of interference caused by nonlinear characteristics in a radar echo wavenumber frequency spectrum in a process of inverting sea wave parameters based on a sea wave spectrum of a coherent microwave radar. The invention provides an average wave period inversion method by analyzing motion characteristics of waves in a speed space-time sequence. The method comprises the following steps: firstly, deriving a speed space-time sequence from a time Doppler spectrum, combining ECMWF data, accurately identifying the positions of a wave crest and a wave trough from the speed space-time sequence through a random forest and a linear regression machine learning integrated model method, and then estimating an average wave period by using a relationship established between a wavelength and a period. The method effectively reduces the error caused by the uncertainty of the energy distribution of different frequency components in the wave number frequency spectrum in the inversion process of the wave period, optimizes the data processing algorithm of the coherent microwave radar, and improves the precision of estimating the average wave period.

Description

Technical Field

[0001] The present invention relates to the technical field of radar, and in particular to a method for estimating the average wave period of a coherent X-band microwave radar based on machine learning. Background Art

[0002] In ocean monitoring, wave parameters, such as the average wave period, are of great significance for ocean engineering, hazard prediction, and oceanographic research. As an effective ocean detection tool, coherent microwave radar has received extensive attention from researchers at home and abroad in recent years. Coherent radar can extract Doppler velocity information from the echo and use the relationship between the wave orbital velocity and the wave height to obtain wave parameters such as the wave spectrum and the average wave period. In addition, the high spatio-temporal resolution enables shore-based coherent microwave radar to detect details of the sea surface, such as breaking phenomena. However, due to the influence of breaking waves, in the wave number frequency spectrum measured by the radar, in addition to the energy along the wave dispersion relationship, non-linear features such as "group lines" and higher harmonic energy also appear, thereby reducing the accuracy of wave parameter inversion, especially the estimation of the average wave period.

[0003] The traditional steps for estimating the average wave period by coherent microwave radar are as follows: First, perform a two-dimensional Fourier transform on the velocity spatio-temporal sequence obtained by the radar to estimate the wave number frequency spectrum. Then, integrate in the wave number domain or the frequency domain to obtain a one-dimensional velocity spectrum. Using the direct transformation relationship between the one-dimensional velocity spectrum and the wave height spectrum, estimate the wave height spectrum from the one-dimensional velocity spectrum, and finally obtain wave parameters such as the average wave period by spectral estimation methods. This method ignores the non-linear wave energy in the radar echo. Due to the uncertainty of the energy distribution of different frequency components in the wave number frequency spectrum, there is often a large error in comparing the wave period calculated in this way with the measurement results of fixed-point devices such as buoys. At present, there is no method that can directly estimate the average wave period from the perspective of the velocity spatio-temporal sequence.

[0004] Machine learning is usually used to predict wave parameters. In the past few decades, various model methods have been studied to predict wave parameters. Currently, there are mainly two methods for estimating wave parameters. The first is a numerical model for wave forecasting based on the physical relationship between the wind field and waves. The numerical model uses terrain and wind field data to solve the physical equations of the interaction between wind and waves for wave forecasting. However, the estimation of wave parameters usually requires a large amount of computing resources and input data, and the efficiency of long-term wave parameter prediction is low. The second is a data-driven model that uses historical wave data for short-term and long-term wave parameter forecasting. Due to the limitations of data variability and the assumptions of linearity and stationarity, time series models have large errors in predicting non-linear and non-stationary wave parameters. Therefore, researchers have adopted machine learning algorithms to improve prediction accuracy. In ocean prediction, using machine learning to predict wave periods can effectively solve the problems of large computational amount, long time consumption, and low accuracy in traditional methods. In addition, various model combination methods of existing machine learning have been used for wave parameter prediction, demonstrating the potential of the integration method in improving prediction accuracy and model robustness. The above methods mainly rely on the quality of input data (such as wind field, terrain, and ocean condition data or historical observation data, etc.). However, when dealing with extreme weather conditions or rare events (such as typhoons, storm surges, etc.), the prediction effect is usually not ideal. To solve this problem, a method for estimating the average wave period of coherent X-band microwave radar based on machine learning is proposed. Summary of the Invention

[0005] The present invention provides a method for estimating the average wave period of coherent X-band microwave radar based on machine learning, aiming to directly estimate the average wave period from the perspective of velocity spatio-temporal sequence and machine learning, retaining most of the echo information and significantly improving the accuracy of coherent microwave radar in predicting the average wave period. This method provides an effective means for estimating wave periods for existing coherent microwave radar ocean observations, thus laying a more accurate theoretical foundation for the subsequent inversion of sea wave parameters.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: A method for estimating the average wave period of coherent X-band microwave radar based on machine learning, comprising the following steps: Step1: Obtain a single Doppler spectrum and a time Doppler spectrum using a coherent X-band microwave radar; Step2: Estimate the left boundary and the right boundary of the signal in the echo Doppler spectrum, and extract the Doppler frequency shift of the echo Doppler spectrum by the method of moment estimation ; Step3: Estimate the spatio-temporal sequence of the radial velocity from the Doppler frequency shift , and extract the feature parameters related to the wave characteristics from it; Step4. Calculate the minimum peak distance based on the eigenvalue extracted from the multi-group velocity spatio-temporal sequence and the average wave period data provided by ECMWF, and establish a random forest and linear regression integrated model; subsequently, input the eigenvalue extracted from the velocity spatio-temporal sequence of the multi-group data into this model to predict the minimum peak distance; and effectively identify the positions of wave peaks and wave troughs according to the minimum peak distance; Step5. Estimate the average wave period through the relationship between wavelength and period, and use a machine learning integrated model to estimate the average wave period.

[0007] The feature parameters related to the wave characteristics in Step3 above include mean value, standard deviation, maximum value, and minimum value fluctuation characteristics.

[0008] The specific steps of Step2 above are: Based on the single Doppler spectrum obtained in Step1, the center frequency The calculation formula is: ;(1) In the formula, is the Doppler frequency, is the Doppler spectrum.

[0009] The specific steps of Step3 above are: The velocity spatio-temporal sequence is estimated from the average Doppler shift The specific calculation method is as follows: ;(2) In the formula, is the electromagnetic wavelength, is the radial distance, is the sampling time.

[0010] The specific steps of Step4 above are: Conduct a correlation analysis on the fluctuation characteristics extracted from the velocity spatio-temporal sequence to screen out the feature subset that is significantly correlated with the target variable; set a correlation threshold to eliminate low-correlation features; the fluctuation characteristics after extraction and screening provide input for the machine learning integrated model, and the model predicts the minimum peak distance by learning the relationship between features and the target variable, and effectively identifies the positions of wave peaks and wave troughs according to the minimum peak distance.

[0011] The specific steps of Step5 above are: Use the relationship between the wavelength and the period to estimate the average wave period. According to the linear wave theory, the relationship between wavelength and period can be expressed by the following formula: ; (3) Among them, is the phase velocity of the corresponding wave.

[0012] In the above Step1, solve the relationship between the radial velocity of the radar echo and the center frequency of the Doppler spectrum therebetween.

[0013] The solving process of the relationship between the radial velocity of the above radar echo and the center frequency of the Doppler spectrum is as follows: The radar orbital velocity can be expressed as: ; In the formula is the radial distance; is the time; obtain the horizontal and vertical components of the orbital velocity; and then the relationship between the radial velocity of the radar echo and the center frequency of the Doppler spectrum is.

[0014] The horizontal and vertical components of the above orbital velocity are and : ; ; Among them, is the incident angle of the radar irradiating the sea surface.

[0015] In the above Step4, the eigenvalues extracted according to multiple groups of velocity spatio-temporal sequences include the mean value, standard deviation, maximum value, and minimum value.

[0016] The present invention provides a method for estimating the average wave period of a coherent X-band microwave radar based on machine learning, having the following technical effects: 1. By analyzing the velocity spatio-temporal sequence and combining machine learning technology, key information can be effectively extracted from the radar echo, establishing a direct relationship between the radar echo velocity spatio-temporal sequence and the wave period parameter, reducing the error caused by the spectral energy distribution in the traditional method.

[0017] 2. Combining ECMWF data, the model can effectively identify the positions of wave crests and wave troughs, and accurately infer the average wave period using the relationship between wavelength and period. This process helps to improve the estimation accuracy of the average wave period of the coherent microwave radar in ocean remote sensing, making the remote sensing data more reliable.

[0018] 3. While removing interference, this method can retain most of the echo information, enhancing the integrity of the data after signal processing. By optimizing the signal processing flow, this method effectively balances noise suppression and information retention, ensuring the efficiency and accuracy of wave parameter estimation, thus providing more valuable data support for ocean engineering and monitoring.

[0019] The present invention proposes a method for estimating the mean wave period by combining coherent X-band microwave radar and machine learning, aiming to improve the accuracy of wave period estimation in ocean remote sensing. By analyzing the spatio-temporal velocity sequence and applying advanced machine learning techniques, this method effectively extracts the key information in the radar echo, reducing the errors brought by traditional methods. At the same time, combined with ECMWF data, the model can accurately identify the positions of wave crests and wave troughs, and infer a more accurate mean wave period from the relationship between wavelength and period, increasing the credibility of remote sensing data. In addition, this method retains most of the echo information during the effective removal of interference, achieving a good balance between noise suppression and information retention. This innovative method provides more accurate and reliable data support for ocean engineering and monitoring, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the drawings and embodiments: Figure 1 is the algorithm flowchart of the present invention; Figure 2 is the scattering geometry diagram of the radar irradiating the sea surface; Figure 3 is the spatio-temporal displacement sequence of the simulated sea surface; Figure 4 is a single Doppler spectrum of the echo signal of the sea surface detected by the coherent microwave radar, marking the estimated left and right boundaries and the center frequency; Figure 5 is the time-Doppler spectrum of the echo signal of the sea surface detected by the coherent microwave radar, marking the estimated left and right boundaries and the center frequency; Figure 6 is the spatio-temporal velocity sequence of the sea surface, marking the estimated positions of wave crests and wave troughs; Figure 7 is the comparison of the estimated mean wave period of the simulated sea surface. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To make the objectives, technical solutions and advantages of the present invention clearer, the following content will systematically and completely describe the specific technical solutions of the present invention in conjunction with the drawings provided according to the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0022] Example 1: As Figure 1 shown, the method for estimating the average wave period of a coherent X-band microwave radar based on machine learning includes the following steps: Step1. Obtain a single Doppler spectrum and a time Doppler spectrum using a coherent X-band microwave radar; Step2. Estimate the left boundary and the right boundary of the signal in the echo Doppler spectrum, and extract the Doppler frequency shift of the echo Doppler spectrum by the method of moment estimation ; Step3. Estimate the spatio-temporal sequence of the radial velocity from the Doppler frequency shift , and extract the characteristic parameters related to the wave characteristics therefrom; Step4. Calculate the minimum peak distance according to the eigenvalues extracted from multiple groups of velocity spatio-temporal sequences and the average wave period data provided by ECMWF, and establish a random forest and linear regression integrated model; subsequently, input the eigenvalues extracted from the velocity spatio-temporal sequences of multiple groups of data into the model to predict the minimum peak distance; and effectively identify the positions of wave peaks and wave valleys according to the minimum peak distance; Step5. Estimate the average wave period through the relationship between the wavelength and the period, and use a machine learning integrated model to estimate the average wave period. This method effectively reduces the estimation uncertainty of the energy distribution in the wave spectrum, and while accurately retaining the non-linear characteristics, improves the estimation accuracy of the wave period.

[0023] The characteristic parameters related to the wave characteristics in Step3 above include the mean value, standard deviation, maximum value, and minimum value fluctuation characteristics.

[0024] The specific steps of Step2 above are: Based on the single Doppler spectrum obtained in Step1, the center frequency is calculated by the formula: ; (1) In the formula, is the Doppler frequency, is the Doppler spectrum.

[0025] The specific steps of Step3 above are: The spatio-temporal sequence of velocity is estimated from the average Doppler frequency shift , and the specific calculation method is as follows: ; (2) In the formula, is the electromagnetic wavelength, is the radial distance, is the sampling time.

[0026] The specific steps of the above Step 4 are as follows: Perform a correlation analysis on the fluctuation features extracted from the velocity spatio-temporal sequence to screen out the feature subset significantly correlated with the target variable; set a correlation threshold to eliminate low-correlation features, thereby improving the effectiveness of the subsequent model; the fluctuation features after extraction and screening provide the input for the machine learning ensemble model, and the model predicts the minimum peak distance by learning the relationship between the features and the target variable through training, and effectively identifies the positions of peaks and troughs based on the minimum peak distance.

[0027] The specific steps of the above Step 5 are as follows: Use the wavelength and the period to estimate the average wave period. According to the linear wave theory, the relationship between the wavelength and the period can be expressed by the following formula: ;(3) where is the phase velocity of the corresponding wave.

[0028] In the above Step 1, solve the relationship between the radial velocity of the radar echo and the center frequency of the Doppler spectrum.

[0029] The relationship solving process between the radial velocity of the above radar echo and the center frequency of the Doppler spectrum is as follows: The radar orbital velocity can be expressed as: ; In the formula is the radial distance; is the time; obtain the horizontal and vertical components of the orbital velocity; and then the relationship between the radial velocity of the radar echo and the center frequency of the Doppler spectrum is.

[0030] The horizontal and vertical components of the above orbital velocity are and : ; ; where is the incident angle of the radar illuminating the sea surface.

[0031] Embodiment 2: As Figure 1 shown, the method for estimating the average wave period of a coherent X-band microwave radar based on machine learning includes the following steps: Step 1: The coherent microwave radar can obtain the dynamic information of the ocean surface. When irradiating the main wave direction of the sea surface with an X-band radar, as Figure 2 shown, the range resolution of the radar is set to 2.5 m, and the Doppler spectrum sampling rate is 0.0005 s. The water particles on the sea surface propagate in the form of continuous waves, and their orbital motion causes the frequency shift of the Doppler spectrum. As Figure 1 shown, the radar orbital velocity can be expressed as: ; where is the radial distance; is the time; and are the horizontal and vertical components of the orbital velocity respectively: ; ; where is the incident angle of the radar irradiating the sea surface. The relationship between the radial velocity of the radar echo and the center frequency of the Doppler spectrum is: ; Step 2: The displacement sequence of the sea surface reflects the wave characteristics of the sea surface. The simulated space-time displacement sequence of the sea surface is as Figure 3 shown; Step 3: Obtain the time Doppler spectrum from the acquired echo, estimate the left and right boundaries of the signal, and then use the moment estimation method to extract the Doppler frequency shift of the echo Doppler spectrum, and further obtain the single Doppler spectrum of the sea surface echo and its time Doppler spectrum. As Figure 4 and Figure 5 shown, the left and right boundaries are marked with red lines, and the center frequency is marked with green lines; Step 4: The velocity space-time sequence is derived from the time Doppler spectrum, and the characteristic parameters related to the wave characteristics are extracted from it. By extracting the characteristic values (such as mean, standard deviation, maximum value, minimum value, etc.) of multiple groups of velocity space-time sequences and the average wave period data of ECMWF, the minimum peak distance can be calculated and a model can be established. First, obtain the data and extract the characteristics, and then calculate the minimum peak distance of each group of data. Then, use machine learning (random forest and linear regression integrated model) to train the model, using the extracted characteristics as input and the minimum peak distance as output. After training, predict the new data to obtain the minimum peak distance, and thereby effectively identify the positions of wave crests and wave troughs. As Figure 6 shown, the wave crests are marked with red asterisks, and the wave troughs are marked with blue asterisks; Step 5: Estimate the average wave period using the relationship between wavelength and period, and calculate the average wave period based on the machine learning integration model. The results for wind speeds from 4 m / s to 7 m / s are as follows Figure 7 shown. The blue line represents the result obtained from the target spectrum calculation, and the red line represents the result inverted by the proposed method. The comparison shows that this method not only accurately captures the nonlinear characteristics of the waves but also maintains high prediction accuracy in complex wave phenomena. Compared with traditional estimation methods, this integration model combines the advantages of multiple algorithms and significantly improves the estimation accuracy of the wave period.

Claims

1. A coherent X-band microwave radar average wave period estimation method based on machine learning, characterized in that: The following steps are involved: Step 1, using coherent X-band microwave radar to obtain a single Doppler spectrum and a time Doppler spectrum; Step 2: Estimate the left edge of the signal in the echo Doppler spectrum and right border , and extract the Doppler frequency shift of the echo Doppler spectrum by moment estimation method ; Step 3, by Doppler frequency shift Estimation of the space-time series of radial velocity , and extract characteristic parameters related to wave characteristics from them; Step 4: Calculate the minimum peak distance based on the characteristic values ​​extracted from multiple sets of velocity time-space sequences and the average wave period data provided by ECMWF, and establish a random forest and linear regression integrated model; then, input the characteristic values ​​extracted from the velocity time-space sequences of multiple sets of data into the model to predict the minimum peak distance; and effectively identify the positions of the peaks and troughs based on the minimum peak distance; Step 5: Estimate the average wave period through the relationship between wavelength and period, and use a machine learning integrated model to estimate the average wave period.

2. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 1, characterized in that: The characteristic parameters related to the wave characteristics in Step 3 include mean, standard deviation, maximum and minimum fluctuation characteristics.

3. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 2, characterized in that: The specific steps of Step 2 are: Based on the single Doppler spectrum obtained in Step 1, the center frequency The calculation formula is: ； (1) In the formula, is the Doppler frequency, is the Doppler spectrum.

4. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 3, characterized in that: The specific steps of Step 3 are: Velocity space-time series The average Doppler shift It is estimated that the specific calculation method is as follows: ； (2) In the formula, is the electromagnetic wavelength, is the radial distance, is the sampling time.

5. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 4, characterized in that: The specific steps of Step 4 are: Perform correlation analysis on the fluctuation features extracted from the velocity space-time series to screen out the feature subsets that are significantly correlated with the target variable; set the correlation threshold to eliminate low-correlation features; The extracted and screened fluctuation features provide input for the machine learning integration model. The model predicts the minimum peak distance by training the relationship between the learning features and the target variable, and effectively identifies the locations of peaks and troughs based on the minimum peak distance.

6. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 5, characterized in that: The specific steps of Step 5 are: Using wavelength and cycle According to linear wave theory, the relationship between wavelength and period can be expressed by the following formula: ；(3) in, is the phase velocity of the corresponding wave.

7. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 6, characterized in that: In Step 1, the radial velocity of the radar echo and the center frequency of the Doppler spectrum are solved. The relationship between.

8. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 7, characterized in that: The radial velocity of the radar echo and the center frequency of the Doppler spectrum The relationship between is solved as follows: The radar track velocity can be expressed as: ； In the formula is the radial distance; is time; the horizontal and vertical components of the orbital velocity are obtained; and then the radial velocity of the radar echo and the center frequency of the Doppler spectrum are obtained The relationship between.

9. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 8, characterized in that: The horizontal and vertical components of the orbital velocity are and : ; ； in, is the angle of incidence of the radar illuminating the sea surface.

10. The method for estimating the average wave period of coherent X-band microwave radar based on machine learning according to claim 9, characterized in that: In the Step 4, the characteristic values ​​extracted from the multiple groups of speed spatiotemporal sequences include mean, standard deviation, maximum value and minimum value.

Images