A Method and Device for Detecting the Significant Wave Height in Marginal Seas by Spaceborne Altimeter

Abstract

The present invention discloses a method and device for detecting the significant wave height of marginal seas by a spaceborne altimeter, relating to the field of marine remote sensing technology, and aiming to solve the problem of poor accuracy in detecting the significant wave height of marginal seas by a spaceborne altimeter in the prior art. The method includes: obtaining the sub-satellite point echo waveform data collected by the spaceborne altimeter; using the echo waveform classification model constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer to determine the waveform categories of each echo waveform in the sub-satellite point echo waveform data collected by the spaceborne altimeter, and obtaining the first echo waveform classification library; obtaining the first marginal sea echo waveform from the first echo waveform classification library, determining the target characteristic parameters of the first marginal sea echo waveform, and then using the significant wave height inversion model constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer and the wave height data in ERA5 to determine the significant wave height of the first marginal sea echo waveform; thereby improving the accuracy of detecting the significant wave height of marginal seas.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean remote sensing, and in particular to a method and device for detecting the significant wave height of marginal seas by a spaceborne altimeter. Background Art

[0002] The significant wave height (SWH) is a key parameter characterizing ocean waves and is of crucial significance for sea condition monitoring and ocean dynamic research. Satellite radar altimeters are the main tools for measuring sea surface information. The radar altimeter emits microwave pulse signals towards the nadir point and obtains the echo waveform of the sea surface by tracking and processing the energy reflected by the sea surface. In the open ocean, due to the small change in the reflection surface, the echo waveforms are almost the same, which are ocean waveforms and can be represented by the Brown-Hayne model. The leading edge of the echo waveform is the direct source for inverting the significant wave height and can be directly inverted using the waveform retracking method. Most current algorithms invert the SWH by simplifying or improving the original Brown-Hayne model and fitting the original waveform. Such as MLE4, ICE-2, ALES, adaptive algorithms, etc. In marginal seas (non-open sea areas), such as coastal waters, ice sheet edges and other regions, the ocean waveforms are affected by land and sea ice, and the waveforms will generate significant noise but still contain sea surface information. Such waveforms present a non-Brown model morphology and usually cannot be processed by traditional retracking methods.

[0003] Currently, for the processing method of marginal sea waveforms detected by spaceborne radar altimeters, the waveforms are first classified to pick out the marginal sea waveforms, then some processing is performed on the marginal sea waveforms, and then the significant wave height is inverted by retracking; among them, most retracking methods are model-based, and a given echo model is required to fit the actual echo waveform, and an appropriate parameter estimation method needs to be selected to obtain the waveform parameters. Due to iterative calculations, although the accuracy of the fitting result is improved, the calculation speed is relatively slow, and for relatively complex waveform morphologies (such as ice sheet edge regions), the fitting often fails; while the experience-based algorithms have no clear physical meaning, do not require iterative calculations, and have a relatively fast calculation speed, but the accuracy of the results for complex waveform morphologies is relatively low; for example: the SWIM L2 product, which is generally recognized in the market for its high accuracy in detecting the significant wave height of echo waveforms, uses the retracking method to fit the significant wave height of the sea surface echo waveform. Compared with the significant wave height of the echo waveform in the ERA5 reanalysis dataset, the root mean square error is about 1.83 / m, indicating that the detection accuracy of the SWIM L2 product still needs to be improved.

[0004] In view of this, there is an urgent need to design a more advanced method for detecting the significant wave height of marginal seas by a spaceborne altimeter to solve the problem of poor accuracy of the significant wave height of marginal sea waveforms detected by spaceborne radar altimeters in the prior art. Summary of the Invention

[0005] The object of the present invention is to provide a method and device for detecting the significant wave height of marginal seas by a spaceborne altimeter. By using a pre-constructed echo waveform classification model and a significant wave height inversion model, first, based on the echo waveform classification model, the waveform classification of the nadir echo waveform data collected by the spaceborne altimeter is determined to obtain the marginal sea echo waveform. Further, the target characteristic parameters of the marginal sea echo waveform are calculated. Finally, based on the target characteristic parameters, the significant wave height of the marginal sea echo waveform is determined by using the significant wave height inversion model; thereby realizing the efficient detection of the significant wave height of marginal seas by the spaceborne altimeter, improving the detection efficiency and accuracy of the significant wave height of marginal seas; and solving the problem of poor accuracy in detecting the significant wave height of marginal seas by the spaceborne altimeter in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides a method for detecting the significant wave height of marginal seas by a spaceborne altimeter, which may include:

[0008] Obtain the nadir echo waveform data collected by the spaceborne altimeter;

[0009] Using the echo waveform classification model, determine the waveform category of each echo waveform in the nadir echo waveform data collected by the spaceborne altimeter to obtain the first echo waveform classification library; the echo waveform classification model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer.

[0010] Obtain the first marginal sea echo waveform from the first echo waveform classification library;

[0011] Using the preset waveform characteristic parameter determination strategy, determine the target characteristic parameters of the first marginal sea echo waveform; the target characteristic parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters.

[0012] Based on the target characteristic parameters, using the trained significant wave height inversion model, determine the significant wave height of the first marginal sea echo waveform; the significant wave height inversion model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset.

[0013] Preferably, the echo waveform classification model may be constructed based on the following steps:

[0014] Obtain the nadir echo waveform data collected by the SWIM spectrometer from the SWIM spectrometer product database;

[0015] Classify the sub-satellite point echo waveform data collected by the SWIM spectrometer based on the waveform geometric features of the sub-satellite point echo waveform data collected by the SWIM spectrometer to obtain a second echo waveform classification library;

[0016] Use a preset waveform feature parameter determination strategy to determine multiple feature parameters corresponding to the echo waveforms of each category in the second echo waveform classification library; at least multiple of the feature parameters include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters;

[0017] Use multiple of the feature parameters as inputs and the echo waveform categories corresponding to the multiple feature parameters as outputs for model training until the model prediction result meets the preset convergence condition to obtain the echo waveform classification model.

[0018] Preferably, the neural network model for model training may include an output layer, a hidden layer, and an output layer; among them, the hidden layer includes three hidden layers, the first hidden layer includes 32 neurons, the second hidden layer includes 16 neurons, and the third hidden layer includes 8 neurons.

[0019] Preferably, the significant wave height inversion model can be constructed based on the following steps:

[0020] Obtain the second marginal sea echo waveform from the second echo waveform classification library;

[0021] According to the second marginal sea echo waveform, obtain the target wave height data with the same geographical space-time as the second marginal sea echo waveform from the ERA5 reanalysis dataset;

[0022] Use the feature parameters of the second marginal sea echo waveform as inputs and the target wave height data as outputs for model training until the model prediction result meets the preset convergence condition to obtain the significant wave height inversion model.

[0023] Preferably, the use of a preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform may include using the formula:

[0024] ;

[0025] Fit the echo waveform to calculate the ICE2 algorithm parameters; where, is the ICE2 algorithm model, is the error function, is the amplitude of the waveform, is the influence of thermal noise, L is the slope of the waveform leading edge, is the midpoint of the leading edge, is the leading edge rise time, is the number of sampling gates corresponding to the waveform data.

[0026] Preferably, the step of determining the target characteristic parameters of the first marginal sea echo waveform by using the preset waveform characteristic parameter determination strategy may include using the formula:

[0027] ;

[0028] calculate the kurtosis of the first marginal sea echo waveform; where is the kurtosis, N is the number of sampling gates of the waveform data, is the normalized waveform data, is the mean value, is the standard deviation.

[0029] Preferably, the step of determining the target characteristic parameters of the first marginal sea echo waveform by using the preset waveform characteristic parameter determination strategy may include using the formula:

[0030] ;

[0031] calculate the height of the first marginal sea echo waveform; where is the height, N is the number of sampling gates of the waveform data, is the normalized waveform data.

[0032] Preferably, the general parameters may include the mean value, standard deviation, skewness, kurtosis and pulse peak value; the OCOG algorithm parameters may include height, width, centroid and half-power point; the ICE2 algorithm parameters may include the waveform leading edge slope and leading edge rise time.

[0033] Preferably, the input of the input layer of the echo waveform classification model is the sub-satellite point echo waveform data, and the output of the output layer is the echo waveform with class labels; the input of the input layer of the significant wave height inversion model is the characteristic parameters of the marginal sea echo waveform, and the output of the output layer is the wave height.

[0034] In a second aspect, the present invention provides a device for detecting the significant wave height of marginal seas by a spaceborne altimeter, which may include:

[0035] a waveform data acquisition module, configured to acquire sub-satellite point echo waveform data collected by a spaceborne altimeter;

[0036] a waveform classification module, configured to use an echo waveform classification model to determine the waveform categories of the respective echo waveforms in the sub-satellite point echo waveform data collected by the spaceborne altimeter, so as to obtain a first echo waveform classification library; the echo waveform classification model is constructed based on the sub-satellite point echo waveform data collected by a SWIM spectrometer;

[0037] Marginal sea waveform acquisition module, which is used to acquire the first marginal sea echo waveform from the first echo waveform classification library;

[0038] Feature parameter determination module, which is used to determine the target feature parameters of the first marginal sea echo waveform by using a preset waveform feature parameter determination strategy; the target feature parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters;

[0039] Significant wave height determination module, which is used to determine the significant wave height of the first marginal sea echo waveform based on the target feature parameters by using a trained significant wave height inversion model; the significant wave height inversion model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset.

[0040] Compared with the prior art, a method for detecting the significant wave height of marginal seas by a spaceborne altimeter provided by the present invention includes acquiring the nadir echo waveform data collected by the spaceborne altimeter; using an echo waveform classification model to determine the waveform categories of each echo waveform in the nadir echo waveform data collected by the spaceborne altimeter, and obtaining a first echo waveform classification library; wherein, the echo waveform classification model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer; acquiring the first marginal sea echo waveform from the first echo waveform classification library; using a preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform; the target feature parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters; finally, based on the target feature parameters, using a trained significant wave height inversion model to determine the significant wave height of the first marginal sea echo waveform; wherein, the significant wave height inversion model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset; based on this, the original echo waveform of the spaceborne radar altimeter can be refinedly classified directly by using the echo waveform classification model, and further the significant wave height of the marginal sea waveform can be efficiently inverted directly by using the significant wave height inversion model, without the need to fit the actual echo waveform with a given echo model and then solve the waveform parameters, avoiding fitting failure; comparing the significant wave height obtained by using the method provided by the present invention with the significant wave height in the ERA5 reanalysis dataset, the root mean square error is 0.46 / m. It can be seen that the method provided by the present invention greatly improves the detection accuracy of the spaceborne radar altimeter for the significant wave height of marginal seas. Description of the Drawings

[0041] The accompanying drawings described herein are used to provide a further understanding of the present invention, and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation to the present invention. In the drawings:

[0042] Figure 1 It is a schematic diagram of the comparison result of the significant wave height of the SWIM L2 product marginal sea wave form and the ERA5 significant wave height;

[0043] Figure 2 It is a schematic diagram of the main process of a method for a spaceborne altimeter to detect the significant wave height of marginal seas provided by the present invention;

[0044] Figure 3 It is a schematic diagram of the waveform categories obtained after refined waveform classification using the echo waveform classification model provided by the present invention;

[0045] Figure 4 It is a schematic diagram of the verification effect of the echo waveform classification model in detecting the significant wave height of marginal seas by a spaceborne altimeter provided by the present invention;

[0046] Figure 5 It is a schematic diagram of the comparison result of the significant wave height calculated by using the method for a spaceborne altimeter to detect the significant wave height of marginal seas provided by the present invention and the ERA5 significant wave height;

[0047] Figure 6 It is a schematic diagram of the structure of a device for a spaceborne altimeter to detect the significant wave height of marginal seas provided by the present invention. Detailed implementation manners

[0048] In order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, words such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and roles. For example, the first threshold and the second threshold are only used to distinguish different thresholds, and do not limit their chronological order. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit being different.

[0049] It should be noted that in the present invention, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary" or "for example" aims to present relevant concepts in a specific manner.

[0050] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the former associated object is an "or" relationship. "At least one (item)" or its similar expression refers to any combination of these items, including any combination of single item or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b, and c, where a, b, and c can be single or multiple.

[0051] Currently, the sea surface waveforms detected by spaceborne radar altimeters have not been classified in detail. Their processing methods are often completed through manual screening, which has low work efficiency and is prone to errors. Moreover, when fitting the wave height of the selected waveforms, whether it is the retracking method or the empirical algorithm method, when fitting or iteratively processing the sea surface waveforms detected by spaceborne radar altimeters to obtain the significant wave height, their fitting accuracy needs to be improved. After comparing the significant wave height detected by the SWIM L2 product, which is recognized as having a relatively high accuracy in detecting sea surface echo waveforms in the current market, with the wave height data in the ERA5 reanalysis dataset; please refer to Figure 1 , Figure 1 FIG. is a schematic diagram of the comparison result of the significant wave height detected by the SWIM L2 product and the ERA5 significant wave height. From Figure 1 it can be undoubtedly obtained that: the deviation between the significant wave height of the marginal sea waveform of the SWIM L2 product and the ERA5 significant wave height is 0.36 m, the standard deviation is 1.83 / m, the root mean square error is 1.83 / m, and the correlation coefficient is 0.64; it can be seen that the accuracy of the significant wave height detected by the SWIM L2 product still needs to be improved.

[0052] It should be noted that the ERA5 reanalysis data is a global climate reanalysis dataset released by the European Centre for Medium-Range Weather Forecasts (ECMWF), covering the time from 1979 to the present. This comprehensive ERA5 reanalysis dataset contains a series of meteorological elements, including 10-meter wind, waves, precipitation, temperature, and pressure. It has been widely used in the research of meteorological and ocean environmental factors. The research results show the rationality and reliability of using ERA5 SWH as the ground truth.

[0053] In view of this, the present invention provides a method and device for detecting the significant wave height of marginal seas by spaceborne altimeters. By using an echo waveform classification model to refine the classification of the original echo waveforms of spaceborne radar altimeters, and using a significant wave height inversion model to efficiently invert the significant wave height of marginal sea waveforms, the detection accuracy of the significant wave height of marginal seas by spaceborne radar altimeters is improved.

[0054] Next, the technical solution of the present invention will be described in detail with reference to the accompanying drawings:

[0055] In a first aspect, the present invention provides a method for detecting the significant wave height of marginal seas by a spaceborne altimeter. Please refer to Figure 2 , Figure 2 which is a schematic diagram of the main process of the method for detecting the significant wave height of marginal seas by a spaceborne altimeter provided by the present invention; the execution subject of the method is a server or a terminal device, such as a detection service platform, a radar altimeter, or a SWIM wave spectrometer, etc.

[0056] In Figure 2 , the method may include:

[0057] Step 210: Obtain the nadir echo waveform data collected by the spaceborne altimeter.

[0058] Step 220: Use the echo waveform classification model to determine the waveform category of each echo waveform in the nadir echo waveform data collected by the spaceborne altimeter, and obtain the first echo waveform classification library; the echo waveform classification model is constructed based on the nadir echo waveform data collected by the SWIM wave spectrometer.

[0059] In steps 210 to 220, the working principle of the spaceborne (radar) altimeter is to vertically transmit a signal to the sea surface and then receive the signal; the nadir echo waveform data obtained in the present invention is the echo waveform signal data received after the altimeter vertically transmits a signal to the sea surface, which is the original data; first, use the echo waveform classification model to refine the classification of the original data collected by the spaceborne altimeter, for example, divide it into 9 categories, determine the waveform category of each echo waveform in the echo waveform data, and obtain the first echo waveform classification library; since the echo waveform classification model is a refined classification model obtained by refining the classification and training the neural network model based on the nadir echo waveform data collected by the SWIM wave spectrometer; therefore, compared with the classification method using the prior art, the classification method of the present invention for the echo waveform category in the first echo waveform classification library is more accurate and the classification efficiency is higher.

[0060] Step 230: Obtain the first marginal sea echo waveform from the first echo waveform classification library.

[0061] Step 240: Determine the target characteristic parameters of the first marginal sea echo waveform by using a preset waveform characteristic parameter determination strategy; the target characteristic parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters.

[0062] Step 250: Based on the target characteristic parameters, use the trained significant wave height inversion model to determine the significant wave height of the first marginal sea echo waveform; the significant wave height inversion model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset.

[0063] In steps 230 to 250, since each category of echo waveform in the first echo waveform classification library brings its corresponding identifier, the marginal sea echo waveform can be directly screened and extracted according to the identifier to obtain the first marginal sea echo waveform for which the wave height needs to be inverted in the present invention; further, it is necessary to calculate the target characteristic parameters of the first marginal sea echo according to a preset waveform characteristic parameter determination strategy, such as OCOG algorithm parameters: height A, width Width, center of gravity COG, and half-power point HP; ICE2 algorithm parameters: waveform leading edge slope L, leading edge rise time and so on; finally, based on the target characteristic parameters, use the significant wave height inversion model to invert the significant wave height to obtain the significant wave height of the first marginal sea echo waveform. Since the significant wave height inversion model is an inversion model obtained by training a neural network model based on the nadir echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset; and experimental verification shows that: for the significant wave height of the marginal sea detected by the method provided by the present invention, compared with the significant wave height in the ERA5 reanalysis dataset, the root mean square error is only 0.46 / m, which greatly improves the detection accuracy of the spaceborne radar altimeter for the significant wave height of the marginal sea.

[0064] Based on this, a method for a spaceborne altimeter to detect the significant wave height of the marginal sea provided by the present invention, through refined classification of the original echo waveform of the spaceborne radar altimeter by using an echo waveform classification model, and efficient inversion of the significant wave height of the marginal sea waveform by using a significant wave height inversion model, does not need to fit the actual echo waveform with a given echo model to solve the waveform parameters, avoiding fitting failure; and the root mean square error with the significant wave height in the ERA5 reanalysis dataset is only 0.46 / m, which greatly improves the detection accuracy of the spaceborne radar altimeter for the significant wave height of the marginal sea.

[0065] It should be noted that the BP neural network model has powerful non-linear mapping capabilities and adaptive feature learning, can efficiently process high-dimensional data, support multi-classification problems, and can fit complex functional relationships in regression tasks, flexibly coping with tasks of different scales. Therefore, it performs excellently in classification and regression problems. In theory, using the BP neural network to refine the classification of the sub-satellite point echo waveform of the spaceborne altimeter and then perform effective wave height inversion for the marginal sea waveform can further improve the accuracy. Based on this, the echo waveform classification model and the effective wave height inversion model in the present invention are both obtained by self-learning or semi-supervised learning of the BP neural network model based on relevant data. Compared with the prior art, the present invention can obtain a higher-precision effective wave height of the marginal sea echo waveform by using the waveform classification model and the effective wave height inversion model, solving the problem of poor accuracy of the spaceborne altimeter in detecting the effective wave height of the marginal sea in the prior art.

[0066] In step 220, preferably, the waveform categories of the first echo waveform classification library can be divided into 9 waveform categories; please refer to Figure 3 , Figure 3 which is a schematic diagram of the waveform categories obtained after performing refined waveform classification using the echo waveform classification model provided by the present invention.

[0067] In Figure 3 , category 1 represents the ocean waveform observed over the open ocean; category 2 represents a waveform that protrudes like a spike in shape, mostly appearing in sea ice areas; category 3 represents a waveform with multiple spikes of different heights in shape, mostly appearing in land areas; category 4 represents a waveform with a spike-like protrusion at the front edge and a gentle decline at the rear edge, mostly appearing in sea ice areas; category 5 represents a waveform with a spike-like protrusion at the front edge and a rising-then-falling rear edge, mostly appearing in marginal sea areas where the front edge is affected by land and sea ice; category 6 represents a waveform with a rising front edge and a falling-then-rising rear edge, mostly appearing in marginal sea areas where the rear edge is affected by land and sea ice; category 7 represents a noise waveform with a rising fluctuation, mostly appearing in land areas or areas where the radar is interfered; category 8 represents a noise waveform with a falling fluctuation, mostly appearing in land areas or areas where the radar is interfered; category 9 represents a fluctuating noise waveform, mostly appearing in land areas or areas where the radar is interfered.

[0068] Preferably, the echo waveform classification model in step 220 can be constructed based on the following steps:

[0069] (1) Obtain the sub-satellite point echo waveform data collected by the SWIM spectrometer from the SWIM spectrometer product database.

[0070] (2) Based on the waveform geometric features of the sub-satellite point echo waveform data collected by the SWIM spectrometer, classify the sub-satellite point echo waveform data collected by the SWIM spectrometer to obtain a second echo waveform classification library.

[0071] (3) Determine multiple characteristic parameters corresponding to the echo waveforms of each category in the second echo waveform classification library by using a preset waveform characteristic parameter determination strategy; the multiple characteristic parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters.

[0072] (4) Use the multiple characteristic parameters as inputs and the echo waveform categories corresponding to the multiple characteristic parameters as outputs to perform model training until the model prediction result meets the preset convergence condition, thereby obtaining an echo waveform classification model.

[0073] Specifically, the SWIM spectrometer can be a SWIM L1 spectrometer product or a product of other levels; after obtaining the nadir echo waveform data collected by the SWIM spectrometer, first, according to the read SWIM L1 nadir waveform data , normalize it using the following formula.

[0074] (1)

[0075] where is the normalized waveform data, is the number of sampling gates corresponding to the waveform data, is the SWIM L1 nadir waveform data; usually, the number of sampling gates of the SWIM nadir waveform is 512.

[0076] Further, classify the nadir echo waveform data collected by the SWIM spectrometer according to the waveform geometric characteristics of the nadir echo waveform data. The classification method can be realized by refined classification based on the empirical method or the method combining the empirical method with classification tools. In the present invention, after refined classification of the nadir echo waveform data collected by the SWIM spectrometer, 9 categories of echo waveforms as shown in Figure 3 are obtained, thereby establishing a second echo waveform classification library; among them, the echo waveforms of the 5th category and the 6th category can be used as marginal sea echo waveforms.

[0077] Further, determine multiple characteristic parameters corresponding to the echo waveforms of each category in the second echo waveform classification library by using a preset waveform characteristic parameter determination strategy; the above multiple characteristic parameters can be determined by using the methods of the following formula (2) to formula (11); it should be noted that the method for determining the target characteristic parameters of the first marginal sea echo waveform in step 240 is the same as this method, and will not be elaborated here.

[0078] Preferably, the general parameters may include the mean M, standard deviation STD, skewness SKEN, kurtosis KURT, and pulse peak PP; the OCOG algorithm parameters may include height A, width Width, center of gravity COG, and half-power point HP; the ICE2 algorithm parameters may include waveform leading-edge slope L and leading-edge rise time .

[0079] Specifically, using the preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform may include using the formula:

[0080] (2)

[0081] Calculate the kurtosis of the first marginal sea echo waveform; where is the kurtosis, N is the number of waveform data sampling gates, is the normalized waveform data, is the mean, and STD is the standard deviation.

[0082] And using the formula:

[0083] (3)

[0084] (4)

[0085] (5)

[0086] (6)

[0087] Calculate the mean, standard deviation, skewness, and pulse peak respectively; in formulas (3) to (6), is the mean, is the standard deviation, is the skewness, is the pulse peak, N is the number of waveform data sampling gates, is the normalized waveform data.

[0088] Preferably, using the preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform may include using the formula:

[0089] (7)

[0090] Calculate the height of the first marginal sea echo waveform; where is the height, N is the number of waveform data sampling gates, is the normalized waveform data.

[0091] And using the formula:

[0092] (8)

[0093] (9)

[0094] (10)

[0095] The width, center of gravity, and half-power points are calculated respectively; in formulas (8) to (10), Width is the width, COG is the center of gravity, HP is the half-power point, and N is the number of waveform data sampling gates, is the waveform data after normalization.

[0096] Preferably, using a preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform, which may include using the formula:

[0097] (11)

[0098] Fitting the echo waveform to calculate the ICE2 algorithm parameters; where, is the ICE2 algorithm model, is the error function, is the amplitude of the waveform, is the influence of thermal noise, is the slope of the waveform leading edge, is the midpoint of the leading edge, is the rise time of the leading edge, is the number of sampling gates corresponding to the waveform data.

[0099] Furthermore, to construct a refined classification neural network model, a refined echo waveform classification model is constructed using a neural network, and the output of the input layer of the echo waveform classification model is multiple feature parameters corresponding to the echo waveforms of each category in the second echo waveform classification library.

[0100] Preferably, the neural network model for model training may include an output layer, a hidden layer, and an output layer; where the hidden layer includes three hidden layers, the first hidden layer includes 32 neurons, the second hidden layer includes 16 neurons, and the third hidden layer includes 8 neurons.

[0101] Preferably, the formula:

[0102] (12)

[0103] Determine the activation function of the hidden layer of the refined classification neural network model .

[0104] The output of the output layer of the echo waveform classification model is the refined echo category identifier, and the activation function It can be expressed by the following formula:

[0105] (13)

[0106] Furthermore, the Levenberg - Marquardt training algorithm, abbreviated as trainlm, is used to train the neural network algorithm to optimize the weights of the neural network to minimize the loss function. To minimize the loss function, this application needs to calculate the gradient and the Hessian matrix and update the weights until a certain stopping condition is met.

[0107] (14)

[0108] (15)

[0109] (16)

[0110] (17)

[0111] In formulas (14) to (17), is the loss function,[[]] is the target output,[[]] is the actual output,[[]] is the number of samples,[[]] is the gradient vector,[[]] is the grid weight vector,[[]] is the Hessian matrix which is a second - derivative matrix,[[]] is the updated weight vector,[[]] is the current weight vector,[[]] is a small positive number,[[]] is the identity matrix,[[]] is the first - order partial derivative of with respect to is the second - order partial derivative of with respect to is used to balance the influence of Newton's method and gradient descent, and its initial value is usually set as a constant less than 1; , is the actual input, and f is the activation function .

[0112] It should be noted that the Levenberg-Marquardt Training Algorithm is usually referred to as the "Levenberg-Marquardt algorithm" or simply the "LM algorithm" in Chinese. It is an optimization algorithm used to solve nonlinear least squares problems and is commonly used in training neural networks, especially suitable for small datasets and networks with fewer parameters, as it generally has a faster convergence speed and better convergence accuracy than other optimization algorithms (such as gradient descent).

[0113] More specifically, to facilitate optimizing the performance of the model and improving the model prediction quality, the following settings can also be made:

[0114] 1. The training objective is that the mean squared error is lower than 0.001; when the mean squared error of the network drops below this value, the training will automatically stop, thus avoiding unnecessary calculations.

[0115] 2. For the training progress display, when every 1000 iterations are completed, the network will show the current training results, including the mean squared error and the training progress; this helps to monitor the model learning process and ensure that the training is moving towards the expected goal.

[0116] 3. The maximum number of training iterations is set to 30,000 times, and the training will proceed until this upper limit is reached, until the target mean squared error is satisfied or other stop conditions are met.

[0117] 4. The learning rate is 0.1, which controls the amplitude of each parameter update. An appropriate learning rate helps to ensure that the model learns stably and converges effectively.

[0118] 5. The maximum number of allowed validation failures during training is 100 times; if the performance on the validation set fails to improve during consecutive training iterations, the count will increase; when the count exceeds 100 times, the training will stop prematurely to avoid overfitting.

[0119] Through the above parameter settings, the network model provided by the present invention aims to be trained efficiently and accurately to achieve the predetermined performance goals.

[0120] Furthermore, according to the established neural network model, 70% of the data in the refined waveform category library is used for training and 30% is used for testing to obtain a refined echo waveform classification model.

[0121] Specifically, the training set accounts for 70% of the data volume of the sample dataset, and the test set accounts for 30% of the data volume of the sample dataset; the proportions of the data from the 1st category to the 9th category in the training set and in the test set are shown in Table 1.

[0122] Table 1 Proportion of Data of Category 1 to Category 9 in the Training Set and in the Test Set

[0123]

[0124] Based on the data volume proportion set described in Table 1, the neural network model is trained to obtain an echo waveform classification model; and the waveform classification refinement degree of this echo waveform classification model is verified, and its classification effect is as Figure 4 shown Figure 4 This is a schematic diagram of the verification effect of the echo waveform classification model for detecting the significant wave height of the marginal sea by the spaceborne altimeter provided by the present invention.

[0125] From Figure 4 The results shown can undoubtedly be obtained: among them, except for Category 4, the verification effects of the other categories are all 100%, and the accuracy rate of Category 4 is 84.6%; since the present invention requires the classification data of Category 5 and Category 6, the above classification results do not affect the marginal sea waveform classification effect required by this application. It can be seen that the echo waveform classification model is adopted in the method for detecting the significant wave height of the marginal sea by the spaceborne altimeter provided by the present invention, and the waveform classification is carried out on the sub-satellite point echo waveform data collected by the spaceborne altimeter, and its classification accuracy can be improved.

[0126] Furthermore, the significant wave height inversion model in step 250 can be constructed based on the following steps:

[0127] First, obtain the second marginal sea echo waveform from the second echo waveform classification library, and then obtain the target wave height data with the same geographical space-time as the second marginal sea echo waveform from the ERA5 reanalysis dataset according to the second marginal sea echo waveform; then, taking the characteristic parameters of the second marginal sea echo waveform as the input and the target wave height data as the output, carry out model training until the model prediction result meets the preset convergence condition, and obtain the significant wave height inversion model. Based on this, the obtained significant wave height inversion model can be applied to the radar altimeter to carry out the inversion of the significant wave height, and its accuracy is higher; at the same time, there is no need to use formula fitting calculation anymore, and the inversion model can directly output the significant wave height data, and its inversion efficiency is also greatly improved.

[0128] Specifically, after the BP neural network model is set in the same setting manner as how to train the model to obtain the echo waveform classification model above, model training is carried out; the difference is that in the training process of obtaining the significant wave height inversion model, the input of the training model is the characteristic parameters of the second marginal sea echo waveform, and the output of the output layer is the target wave height data, that is, the significant wave height.

[0129] To verify the accuracy of the method provided by the present invention for detecting the significant wave height of marginal seas from the sea surface, the present application compared the significant wave height of marginal seas obtained by inverting using the method provided by the present invention with the wave height data in the ERA5 reanalysis dataset, and obtained a significant wave height with higher accuracy than the SWIM L2 product; please refer to Figure 1 and Figure 5 , Figure 5 which is a schematic diagram of the comparison result between the significant wave height calculated by a method for detecting the significant wave height of marginal seas using a spaceborne altimeter provided by the present invention and the ERA5 significant wave height.

[0130] From Figure 5 it can be undoubtedly obtained that: the deviation between the significant wave height calculated by a method for detecting the significant wave height of marginal seas using a spaceborne altimeter provided by the present invention and the ERA5 significant wave height is 0 m, the standard deviation is 0.46 / m, the root mean square error is 0.46 / m, and the correlation coefficient is 0.96. Further comparing Figure 5 with Figure 1 it can be obtained that: both the deviation and the root mean square error are better than the detection accuracy of the SWIM L2 product; thus, the method provided by the present invention can effectively improve the detection accuracy of the significant wave height of marginal seas and can solve the problem of poor accuracy in detecting the significant wave height of marginal seas by spaceborne altimeters in the prior art.

[0131] In the second aspect, a device for detecting the significant wave height of marginal seas using a spaceborne altimeter according to the present invention, please refer to Figure 6 , Figure 6 which is a schematic structural diagram of a device for detecting the significant wave height of marginal seas using a spaceborne altimeter provided by the present invention.

[0132] In Figure 6 the device may include:

[0133] A waveform data acquisition module 610, configured to acquire the nadir echo waveform data collected by the spaceborne altimeter.

[0134] A waveform classification module 620, configured to use an echo waveform classification model to determine the waveform category of each echo waveform in the nadir echo waveform data collected by the spaceborne altimeter, and obtain a first echo waveform classification library; the echo waveform classification model is constructed based on the nadir echo waveform data collected by the SWIM spectrometer.

[0135] A marginal sea waveform acquisition module 630, configured to acquire the first marginal sea echo waveform from the first echo waveform classification library.

[0136] A feature parameter determination module 640 is configured to determine target feature parameters of the first marginal sea echo waveform by using a preset waveform feature parameter determination strategy; the target feature parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters.

[0137] An effective wave height determination module 650 is configured to determine the effective wave height of the first marginal sea echo waveform by using the trained effective wave height inversion model based on the target feature parameters; the effective wave height inversion model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset.

[0138] Based on this, a device for detecting the effective wave height of a marginal sea by a spaceborne altimeter provided by the present invention first uses a waveform data acquisition module 610 to acquire sub-satellite point echo waveform data collected by the spaceborne altimeter, and then uses a waveform classification module 620 to determine the waveform categories of each echo waveform in the sub-satellite point echo waveform data collected by the spaceborne altimeter by using an echo waveform classification model, so as to obtain a first echo waveform classification library; the echo waveform classification model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer; then uses a marginal sea waveform acquisition module 630 to obtain a first marginal sea echo waveform from the first echo waveform classification library; and then based on a feature parameter determination module 640, uses a preset waveform feature parameter determination strategy to determine the target feature parameters of the first marginal sea echo waveform; the target feature parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters; finally, an effective wave height determination module 650 is used to determine the effective wave height of the first marginal sea echo waveform by using the trained effective wave height inversion model based on the target feature parameters; the effective wave height inversion model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis dataset; thereby greatly improving the detection accuracy of the spaceborne radar altimeter for the effective wave height of the marginal sea and solving the problem of poor detection accuracy of the spaceborne altimeter for the effective wave height of the marginal sea in the prior art.

[0139] Although the present invention has been described in connection with various embodiments herein, however, in the process of implementing the claimed invention, those skilled in the art can understand and realize other variations of the disclosed embodiments by viewing the drawings, the disclosure content, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude a plurality of situations. A single processor or other unit can implement several functions recited in the claims. Certain measures are recited in mutually different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0140] Although the present invention has been described in connection with specific features and their embodiments, it will be apparent that various modifications and combinations can be made without departing from the spirit and scope of the invention. Accordingly, the specification and drawings are merely exemplary illustrations of the invention defined by the appended claims and are considered to cover any and all modifications, variations, combinations or equivalents within the scope of the invention. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for detecting significant wave height in marginal seas using a satellite-borne altimeter, characterized in that: include: Obtain sub-satellite point echo waveform data collected by the satellite-borne altimeter; Determine the waveform category of each echo waveform in the sub-satellite point echo waveform data collected by the satellite altimeter by using the echo waveform classification model, and obtain a first echo waveform classification library; the echo waveform classification model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer; Acquire a first marginal sea echo waveform from the first echo waveform classification library; Using a preset waveform characteristic parameter determination strategy to determine target characteristic parameters of the first marginal sea echo waveform; the target characteristic parameters at least include general parameters, OCOG algorithm parameters, and ICE2 algorithm parameters; using a preset waveform characteristic parameter determination strategy to determine the target characteristic parameters of the first marginal sea echo waveform includes using the formula: ； Fit the echo waveform to calculate the ICE2 algorithm parameters; And using the formula: ； Calculating and obtaining the kurtosis of the first marginal sea echo waveform; And using the formula: ； Calculating and obtaining the height of the first edge sea echo waveform; in, is the ICE2 algorithm parameter, is the error function, is the amplitude of the waveform, Thermal noise influence L is the slope of the leading edge of the waveform, The midpoint of the leading edge, is the leading edge rise time, is the number of sampling gates corresponding to the waveform data, is the kurtosis, N is the number of waveform data sampling gates, is the normalized waveform data, is the mean, is the standard deviation, is the height; Based on the target characteristic parameters, the effective wave height of the first marginal sea echo waveform is determined using the trained effective wave height inversion model; the effective wave height inversion model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis data set.

2. The method for detecting significant wave height of marginal seas by a satellite-borne altimeter as claimed in claim 1, characterized in that: The echo waveform classification model is constructed based on the following steps: Obtain sub-satellite point echo waveform data collected by the SWIM spectrometer from the SWIM spectrometer product database; Based on the waveform geometric characteristics of the sub-satellite point echo waveform data collected by the SWIM spectrometer, the sub-satellite point echo waveform data collected by the SWIM spectrometer is classified and processed to obtain a second echo waveform classification library; Using a preset waveform characteristic parameter determination strategy, a plurality of characteristic parameters corresponding to the echo waveforms of each category in the second echo waveform classification library are determined; the plurality of characteristic parameters at least include general parameters, OCOG algorithm parameters and ICE2 algorithm parameters; The plurality of characteristic parameters are used as input, and the echo waveform categories corresponding to the plurality of characteristic parameters are used as output, and model training is performed until the model prediction result meets the preset convergence condition, thereby obtaining the echo waveform classification model.

3. The method for detecting significant wave height of marginal seas by a satellite-borne altimeter as claimed in claim 2, characterized in that: The neural network model for model training includes an output layer, a hidden layer and an output layer; wherein the hidden layer includes three hidden layers, the first hidden layer includes 32 neurons, the second hidden layer includes 16 neurons, and the third hidden layer includes 8 neurons.

4. The method for detecting significant wave height of marginal seas by a satellite-borne altimeter as claimed in claim 2, characterized in that: The significant wave height inversion model is constructed based on the following steps: Acquire a second marginal sea echo waveform from the second echo waveform classification library; According to the second marginal sea echo waveform, target wave height data having the same geographical time and space as the second marginal sea echo waveform is obtained from the ERA5 reanalysis data set; The characteristic parameters of the second marginal sea echo waveform are used as input and the target wave height data is used as output to perform model training until the model prediction result meets the preset convergence condition, thereby obtaining the effective wave height inversion model.

5. The method for detecting significant wave height of marginal seas by a satellite-borne altimeter as claimed in claim 1, characterized in that: The general parameters include mean, standard deviation, skewness, kurtosis and pulse peak value; the OCOG algorithm parameters include height, width, center of gravity and half power point; the ICE2 algorithm parameters include waveform leading edge slope and leading edge rise time.

6. The method for detecting significant wave height of marginal seas by using a satellite-borne altimeter as claimed in claim 1, characterized in that: The input layer of the echo waveform classification model is the sub-satellite point echo waveform data, and the output layer is the echo waveform with a category identifier; the input layer of the effective wave height inversion model is the characteristic parameters of the marginal sea echo waveform, and the output layer is the wave height.

7. A device for detecting significant wave height of marginal seas using a satellite-borne altimeter, characterized in that: include: A waveform data acquisition module, wherein the waveform data acquisition module is used to acquire sub-satellite point echo waveform data collected by a satellite-borne altimeter; A waveform classification module, wherein the waveform classification module is used to determine the waveform category of each echo waveform in the sub-satellite point echo waveform data collected by the satellite altimeter using an echo waveform classification model, and obtain a first echo waveform classification library; the echo waveform classification model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer; An edge sea waveform acquisition module, the edge sea waveform acquisition module is used to acquire a first edge sea echo waveform from the first echo waveform classification library; A characteristic parameter determination module, the characteristic parameter determination module is used to determine the target characteristic parameters of the first marginal sea echo waveform by using a preset waveform characteristic parameter determination strategy; the target characteristic parameters at least include general parameters, OCOG algorithm parameters and ICE2 algorithm parameters; the use of the preset waveform characteristic parameter determination strategy to determine the target characteristic parameters of the first marginal sea echo waveform includes using the formula: ； Fit the echo waveform to calculate the ICE2 algorithm parameters; And using the formula: ； Calculating and obtaining the kurtosis of the first marginal sea echo waveform; And using the formula: ； Calculating and obtaining the height of the first edge sea echo waveform; in, is the ICE2 algorithm parameter, is the error function, is the amplitude of the waveform, Thermal noise influence L is the slope of the leading edge of the waveform, The midpoint of the leading edge, is the leading edge rise time, is the number of sampling gates corresponding to the waveform data, is the kurtosis, N is the number of waveform data sampling gates, is the normalized waveform data, is the mean, is the standard deviation, is the height; An effective wave height determination module, the effective wave height determination module is used to determine the effective wave height of the first marginal sea echo waveform based on the target characteristic parameters and using the trained effective wave height inversion model; the effective wave height inversion model is constructed based on the sub-satellite point echo waveform data collected by the SWIM spectrometer and the wave height data in the ERA5 reanalysis data set.

Images