Underlying surface information fusion snow depth inversion method and system based on SAR (Synthetic Aperture Radar) image

Abstract

The invention discloses an underlying surface information fusion accumulated snow depth inversion method and system based on an SAR image, and relates to the technical field of ice and snow monitoring based on a satellite remote sensing technology, and the method comprises the steps: collecting orbit data, carrying out the preprocessing of the collected orbit data, removing thermal noise and speckle noise, and carrying out the terrain correction. Performing radiation calibration according to a radar equation to obtain a back scattering characteristic parameter and a polarization characteristic parameter, performing characteristic extraction on the corrected data, optimizing a characteristic data set, constructing a neural network model, inputting the optimized characteristic data set into a residual neural network, training to obtain an accumulated snow depth inversion model, and introducing a time variable to obtain an accumulated snow depth inversion model; and predicting the snow depth. According to the method, the advanced SAR data processing technology and the machine learning method are fused, the snow depth inversion and prediction problems are effectively solved, and an important basis is provided for decision support.

Description

Technical Field

[0001] The present invention relates to the technical field of ice and snow monitoring based on satellite remote sensing technology, and in particular to a method and system for snow depth inversion based on underlying surface information fusion of SAR images. Background Art

[0002] Snow is a major component of the surface hydrological system. The snow layer can effectively maintain the surface temperature in winter, and after melting, it provides water resources for organisms in the following year. It is an important input variable for a large number of climate, hydrological, agricultural and ecological models. Accurately obtaining the depth of snow is helpful for the rational use of snow resources, which is very important for human production activities and the study of global climate change. Traditional snow depth measurement methods usually use fixed instruments at ground meteorological base stations, or use portable equipment for field experimental measurements, and these methods usually require a lot of manpower and material resources and a long time. With the development of satellite remote sensing technology, remote sensing observation data can be used to observe the changes in surface snow cover at regional and global scales in real time and quickly.

[0003] At present, the snow depth inversion based on satellite remote sensing technology is mainly divided into two categories: microwave remote sensing and optical remote sensing. In the microwave remote sensing inversion of snow depth, it is divided into passive microwave remote sensing and active microwave remote sensing inversion; the principle of optical remote sensing is mainly to extract relevant data based on the relationship between the spectral characteristics of snow and snow depth, but optical remote sensing is easily affected by solar radiation and fog and it is difficult to explore the internal information of snow, and the accuracy of snow depth inversion is limited; passive microwave can penetrate the surface of snow and obtain information about the inside of snow. It is currently the most common data source for snow parameter research and long-term monitoring such as snow depth, but the existing passive microwave remote sensing images have a low spatial resolution and cannot meet the requirements of high-resolution and high-precision snow parameter research; compared with optical remote sensing images and passive microwave remote sensing, active radar remote sensing usually has the ability to observe the ground all day and all weather, and can penetrate the surface of snow without being affected by foggy weather. In recent years, this method has gradually been used to explore the internal information of snow and invert snow depth. However, at present, since the microwave backscattering coefficient is greatly affected by the underlying surface conditions, it is only suitable for snow depth inversion in a small scale, and the accuracy of snow depth inversion in a large scale needs to be improved. Summary of the invention

[0004] In view of the above-mentioned existing problems, the present invention provides a snow depth inversion method and system based on SAR image-based underlying surface information fusion, so as to solve the problem that the prior art requires a lot of manpower and material resources and a long time, and cannot observe the changes in surface snow on regional and global scales in real time and quickly.

[0005] In order to solve the above technical problems, a snow depth inversion method based on SAR image and underlying surface information fusion is proposed, including:

[0006] Collect orbital data, preprocess the collected orbital data, and remove thermal noise and coherent speckle noise; perform terrain correction, and perform radiation calibration according to the radar equation to obtain backscattering characteristic parameters and polarization characteristic parameters, extract features from the corrected data, and optimize the feature data set; build a neural network model, input the optimized feature data set into the residual neural network, train to obtain the snow depth inversion model, and introduce the time variable to predict the snow depth.

[0007] As a preferred solution of the method for snow depth inversion based on SAR image underlying surface information fusion described in the present invention, the collecting of orbital data includes collecting satellite position data, satellite velocity data, orbital attitude data, time data, orbital dynamic parameters and orbital error correction data, and preprocessing the collected orbital data.

[0008] The satellite position data includes longitude, latitude, altitude, orbit inclination and ascending node longitude; the satellite velocity data includes velocity along the orbit and velocity perpendicular to the orbit; the orbit attitude data includes pitch angle, yaw angle and roll angle; the time data includes ephemeris time and data collection time; the orbit dynamics parameters include orbit semi-major axis, orbit eccentricity, perigee angle and horizontal anomaly angle.

[0009] As a preferred solution of the method for snow depth inversion based on SAR image underlying surface information fusion described in the present invention, the preprocessing includes orbit correction, thermal noise removal, coherent speckle filtering and decibel processing of the acquired orbit data.

[0010] The orbit correction includes performing error estimation by fitting a polynomial to the orbit error, forming an error data set by calculating the difference between the actual orbit data and the reference data, selecting a polynomial order to fit the error data, and determining the polynomial coefficients using the least squares method, that is, the transpose of the matrix multiplied by the matrix and then multiplied by the coefficient vector is equal to the transpose of the matrix multiplied by the error data vector. After the coefficients are determined, the value of the polynomial is calculated based on the timestamp, and the value is added as a correction amount to the original orbit data for orbit correction.

[0011] The thermal noise removal includes obtaining a thermal noise vector for any given position and time by linear interpolation of two noise vectors closest to each other. Within the detection range, the thermal noise vector derives two noise vectors, which are removed from the detection range by linear interpolation.

[0012] The speckle filtering includes updating the central pixel value through a filter and utilizing the local statistical characteristics of the image to perform speckle filtering on the SAR remote sensing image, and the filtering window is 7×7.

[0013] The decibel processing formula is:

[0014] ε=10log 10 σ o

[0015] Among them, σ o is the dimensionless backscattering coefficient, and ε is the normalized backscattering coefficient.

[0016] As a preferred solution of the method for snow depth inversion based on underlying surface information fusion of SAR images described in the present invention, the terrain correction includes eliminating radiation brightness errors using the SCS+C correction method of the Lambertian reflectivity model based on elevation information DEM, so that objects with the same reflective properties have the same brightness values ​​in the image.

[0017] The SCS+C correction method formula is expressed as:

[0018] ρ H =ρ T (cosθ p cosθ z +C λ ) / (cosγ i +C λ )

[0019] ρ T =a λ +b λ cosγ i

[0020] C λ =a λ / b λ

[0021] Among them, ρ H is the horizontal surface reflection, ρ T is the slope surface reflection, θ p is the slope angle, θ z is the solar zenith angle, γ i is the solar incident angle, C λ is the correction factor, a λ is the slope of the regression line, b λ is the intercept of the regression line.

[0022] The radiation calibration formula is expressed as:

[0023]

[0024] Among them, P d is the backscatter intensity received by the sensor, P t is the transmission power, P n is the additional power, is the perspective antenna gain, is the receiving antenna gain, is the current gain of the radar receiver, G p is the processor constant, R is the distance propagation loss, θ el is the antenna elevation angle, θ oc is the antenna azimuth, L s is the atmospheric loss, L a is the system loss, A is the scattering area, λ is the wavelength, σ o is the dimensionless backscattering coefficient.

[0025] As a preferred solution of the method for snow depth inversion based on underlying surface information fusion of SAR images described in the present invention, obtaining polarization characteristic parameters includes using dual polarization ratio data, and calculating polarization entropy, anisotropy and scattering angle by eigenvalues ​​of three coherence matrix sub-components obtained by eigendecomposing the coherence matrix.

[0026] The formula for obtaining polarization characteristic parameters is:

[0027]

[0028]

[0029] Among them, P i is the ratio of the eigenvalue corresponding to the i-th eigenvector to the total eigenvalues, λ i is the i-th eigenvalue, H is the polarization entropy, A is the anisotropy, α is the scattering angle, i is the variable index, λ1 is the volume scattering of the target, λ2 is the surface scattering of the target, and λ3 is the multiple scattering component of the target.

[0030] As a preferred solution of the method for snow depth inversion based on underlying surface information fusion of SAR images described in the present invention, the optimized feature data set includes preprocessing geographic feature data to obtain slope and aspect data, preprocessing meteorological data, extracting features from the corrected data, and optimizing features using the Pearson correlation coefficient.

[0031] The formula for obtaining slope and aspect data is:

[0032]

[0033] Among them, Slope is the slope, Aspect is the slope direction, x Slope is the horizontal slope in the two-dimensional image. y is the vertical slope in the two-dimensional image.

[0034] The Pearson correlation coefficient formula is:

[0035]

[0036] Among them, Y i and X i is the characteristic parameter, and is the average value of the characteristic parameter, n is the number of parameters, i is the variable index, PCC Y-X is the Pearson correlation coefficient; the Pearson correlation coefficient threshold is selected as 0.9. When the calculated correlation coefficient is greater than or equal to 0.9, the characteristic parameter is judged to be a parameter correlated with the snow depth and is used as the input characteristic parameter of the snow depth inversion model.

[0037] As a preferred solution of the method for snow depth inversion based on underlying surface information fusion of SAR images described in the present invention, the predicted snow depth includes: constructing snow depth inversion models based on different neural network model frameworks, selecting a residual neural network ResNet model for training, obtaining a snow depth inversion model, and performing model training, and obtaining the gap between the inversion model result and the measured value, i.e., the inversion accuracy, through testing, adjusting the model parameters according to the results, selecting the model parameters with the highest inversion accuracy, constructing a snow depth inversion model based on the parameters, and inputting the time variable into a long short-term memory network LSTM model for training, obtaining a long time series inversion model of snow depth, and predicting the snow depth.

[0038] The model training includes defining the model structure, inputting the experimental data set after feature optimization, initializing the learning rate, optimizer and number of iterations, capturing the nonlinear relationship between the input feature parameters and the measured snow depth, and obtaining the gap between the inversion model results and the measured values ​​through testing, that is, the accuracy of the inversion results.

[0039] Another object of the present invention is to provide a snow depth inversion system based on underlying surface information fusion of SAR images. The present invention combines SAR data processing technology and machine learning methods to solve the problem of snow depth inversion and prediction; the system of the present invention improves data quality through data acquisition and preprocessing of synthetic aperture radar images, including orbit data collection, thermal noise removal and coherent speckle filtering; performs terrain correction and radiation calibration to ensure data standardization; the system extracts and optimizes advanced polarization feature parameters including dual polarization ratio and Cloude-Pottier polarization decomposition to accurately describe snow characteristics.

[0040] As a preferred solution of the SAR image-based underlying surface information fusion snow depth inversion system described in the present invention, it is characterized by including a data acquisition module, a terrain correction and radiation calibration module, a feature extraction and optimization module, a neural network model construction and training module, and a snow depth prediction module.

[0041] The data acquisition module is used to collect satellite position, speed, orbit attitude, time data and dynamic parameters, and perform orbit correction, thermal noise removal, coherent speckle filtering and decibel processing to reduce errors and noise in the data.

[0042] The terrain correction and radiation calibration module is used to eliminate the radiation brightness error caused by terrain undulations by using the SCS+C correction method to ensure the consistency of reflection properties of different terrains, and convert the backscatter intensity received by the sensor into a dimensionless backscatter coefficient according to the radar equation.

[0043] The feature extraction and optimization module is used to calculate polarization entropy, anisotropy and scattering angle through dual polarization ratio data and coherence matrix feature decomposition, obtain polarization information reflecting snow characteristics, extract slope and slope direction from elevation data, and extract ambient temperature and humidity data, use the Pearson correlation coefficient to screen features, and retain features related to snow depth.

[0044] The neural network model construction and training module is used to define the structure of the residual neural network ResNet model, input the optimized feature data set, initialize the model parameters, capture the nonlinear relationship between the feature parameters and the snow depth through training, and evaluate the accuracy of the model.

[0045] The snow depth prediction module is used to input the SAR remote sensing image based on the trained ResNet model, invert the snow depth, introduce the time variable into the LSTM model, train the prediction model of the snow depth changing with time, and perform long-term prediction of the snow depth.

[0046] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the method described in the method of fusing snow depth inversion based on underlying surface information of SAR images are implemented.

[0047] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the method described in the method of fusing snow depth inversion based on underlying surface information of SAR images are implemented.

[0048] Beneficial effects of the invention: Unlike optical remote sensing technology, which is easily affected by natural environmental conditions such as clouds, fog, and light, the invention can work all day, in all climates, and for a long time. The data obtained is large in scale and the image is rich in effective information. The snow depth inversion based on this is more stable and reliable. Unlike passive microwave remote sensing technology, which has a low spatial resolution, the invention adopts synthetic aperture radar SAR remote sensing technology, which can obtain higher spatial resolution, so that the snow depth inversion is not affected by insufficient spatial resolution, and can perform high-precision and high-resolution snow depth inversion. The snow depth inversion based on this has a higher accuracy rate. guarantee; and it integrates information such as the backscattering characteristics and polarization characteristics of the underlying surface, overcoming the problem of limited snow depth inversion accuracy in a large scale caused by the influence of the underlying surface conditions on SAR images. At the same time, it takes into account the influence of geographical and meteorological factors on snow depth, and constructs a snow depth inversion model based on the fusion of multi-source data, which is conducive to improving the snow depth inversion accuracy; and introduces a machine learning model with memory efficiency, and constructs a long-term series inversion model of snow depth by combining multi-source data and time series, which can realize the prediction of snow depth and has more in-depth applications in disaster warning and other aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0050] Figure 1 An overall flow chart of a method for snow depth inversion based on SAR image-based underlying surface information fusion provided by an embodiment of the present invention.

[0051] Figure 2 A schematic diagram of a backbone network ResNet model of a snow depth inversion model of a snow depth inversion method based on underlying surface information fusion of SAR images provided in one embodiment of the present invention.

[0052] Figure 3 A schematic diagram of a backbone network LSTM model of a snow depth long time series inversion model for a snow depth inversion method based on underlying surface information fusion of SAR images provided in one embodiment of the present invention.

[0053] Figure 4 A system solution flow chart of a system for snow depth inversion based on SAR image-based underlying surface information fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an embodiment that is mutually exclusive with other embodiments, either individually or selectively.

[0057] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0058] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0059] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] Example 1, reference Figure 1-Figure 3 , which is the first embodiment of the present invention, provides a snow depth inversion method based on SAR image underlying surface information fusion, including:

[0061] S1: Collect orbit data, pre-process the collected orbit data, and remove thermal noise and coherent speckle noise.

[0062] The collecting of orbital data includes collecting satellite position data, satellite velocity data, orbital attitude data, time data, orbital dynamics parameters and orbital error correction data, and preprocessing the collected orbital data;

[0063] The satellite position data includes longitude, latitude, altitude, orbit inclination and ascending node longitude; the satellite velocity data includes velocity along the orbit and velocity perpendicular to the orbit; the orbit attitude data includes pitch angle, yaw angle and roll angle; the time data includes ephemeris time and data collection time; the orbit dynamics parameters include orbit semi-major axis, orbit eccentricity, perigee angle and horizontal anomaly angle.

[0064] It should be noted that the preprocessing includes performing orbit correction, thermal noise removal, speckle filtering and decibel conversion on the acquired orbit data;

[0065] The orbit correction includes performing error estimation by fitting a polynomial to the orbit error, forming an error data set by calculating the difference between the actual orbit data and the reference data, selecting a polynomial order to fit the error data, determining the polynomial coefficients by the least squares method, that is, the transpose of the matrix multiplied by the matrix and then multiplied by the coefficient vector is equal to the transpose of the matrix multiplied by the error data vector, calculating the value of the polynomial according to the timestamp after the coefficient is determined, and adding the value as a correction amount to the original orbit data for orbit correction;

[0066] The thermal noise removal includes obtaining a thermal noise vector for any given azimuth time by linear interpolation of two noise vectors closest to each other, within the detection range, deriving two noise vectors from the thermal noise vector, and removing them from the detection range by linear interpolation;

[0067] The speckle filtering includes updating the central pixel value through a filter and using the local statistical characteristics of the image to perform speckle filtering on the SAR remote sensing image, and the filter window is 7×7;

[0068] The decibel processing formula is:

[0069] ε=10log 10 σ o

[0070] Among them, σ o is the dimensionless backscattering coefficient, and ε is the normalized backscattering coefficient.

[0071] S2: Perform terrain correction and perform radiation calibration according to the radar equation to obtain backscattering characteristic parameters and polarization characteristic parameters, extract features from the corrected data, and optimize the feature data set.

[0072] Furthermore, the terrain correction includes eliminating the radiation brightness error by using the SCS+C correction method of the Lambertian reflectance model based on the elevation information DEM, so that the objects with the same reflective properties have the same brightness value in the image;

[0073] The SCS+C correction method formula is expressed as:

[0074] ρ H =ρ T (cosθ p cosθ z +C λ ) / (cosγ i +C λ )

[0075] ρ T =a λ +b λ cosγ i

[0076] C λ =a λ / b λ

[0077] Among them, ρ H is the horizontal surface reflection, ρ T is the slope surface reflection, θ p is the slope angle, θ z is the solar zenith angle, γ i is the solar incident angle, C λ is the correction factor, a λ is the slope of the regression line, b λ is the intercept of the regression line;

[0078] The radiation calibration formula is expressed as:

[0079]

[0080] Among them, P d is the backscatter intensity received by the sensor, P t is the transmission power, P n is the additional power, is the perspective antenna gain, is the receiving antenna gain, is the current gain of the radar receiver, G p is the processor constant, R is the distance propagation loss, θ el is the antenna elevation angle, θoc is the antenna azimuth, L s is the atmospheric loss, L a is the system loss, A is the scattering area, λ is the wavelength, σ o is the dimensionless backscattering coefficient.

[0081] Obtaining polarization characteristic parameters includes using dual polarization ratio data and calculating polarization entropy, anisotropy and scattering angle by eigenvalues ​​of three coherence matrix subcomponents obtained by eigendecomposing the coherence matrix;

[0082] The formula for obtaining polarization characteristic parameters is:

[0083]

[0084] Among them, P i is the ratio of the eigenvalue corresponding to the i-th eigenvector to the total eigenvalues, λ i is the i-th eigenvalue, H is the polarization entropy, A is the anisotropy, α is the scattering angle, i is the variable index, λ1 is the volume scattering of the target, λ2 is the surface scattering of the target, and λ3 is the multiple scattering component of the target.

[0085] Furthermore, the optimized feature data set includes preprocessing the geographic feature data to obtain slope and aspect data, preprocessing the meteorological data, extracting features from the corrected data, and optimizing features using the Pearson correlation coefficient;

[0086] In the auxiliary data preprocessing process, taking into account the influence of meteorological factors and topography on snow depth, the original elevation data and reanalyzed meteorological data were introduced when constructing the experimental data set. After preprocessing of the original elevation data, geographical feature data such as slope and aspect can be obtained. After preprocessing of the reanalyzed meteorological data, meteorological characteristics such as ambient temperature, ambient humidity, snow layer temperature, and snow density can be obtained.

[0087] The formula for obtaining slope and aspect data is:

[0088]

[0089] Among them, Slope is the slope, Aspect is the slope direction, x Slope is the horizontal slope in the two-dimensional image. y is the vertical slope in the two-dimensional image;

[0090] The Pearson correlation coefficient formula is:

[0091]

[0092] Among them, Y i and X iis the characteristic parameter, and is the average value of the characteristic parameter, n is the number of parameters, i is the variable index, PCC Y-X is the Pearson correlation coefficient; the Pearson correlation coefficient threshold is selected as 0.9. When the calculated correlation coefficient is greater than or equal to 0.9, the characteristic parameter is judged to be a parameter with a strong correlation with the snow depth and is used as the input characteristic parameter of the snow depth inversion model.

[0093] S3: Build a neural network model, input the optimized feature data set into the residual neural network, train to obtain the snow depth inversion model, and introduce the time variable to predict the snow depth.

[0094] Furthermore, the prediction of snow depth includes constructing snow depth inversion models based on different neural network model frameworks, selecting a residual neural network ResNet model for training, obtaining a snow depth inversion model, and performing model training, and obtaining the gap between the inversion model results and the measured values, i.e., the inversion accuracy, through testing, adjusting the model parameters according to the results, selecting the model parameters with the highest inversion accuracy, constructing a snow depth inversion model based on the parameters, and inputting the time variables into a long short-term memory network LSTM model for training, obtaining a long time series inversion model of snow depth, and predicting the snow depth.

[0095] The backbone network of the snow depth inversion model is the Resnet model, such as Figure 2 , the model consists of a standard residual module, a downsampled residual module and a fully connected layer; in the standard residual module, the input data is transformed through two 3×3 convolutional layers to obtain the output after feature transformation; at the same time, the input data is directly passed through the direct connection edge and then added to the initial output to obtain the final output. In the downsampled residual block, a 1x1 convolutional layer is added to the direct connection edge to adjust the input dimension to match the output dimension for addition operation. The formula is:

[0096] y=F(x,W i )+x

[0097] y=F(x,W i )+W s (x)

[0098] Among them, W i is the weight parameter of the convolutional layer in the residual block, x is the input data, y is the final output, F is the output after feature transformation, and W s is the weight parameter of the convolutional layer.

[0099] The model training includes defining the model structure, inputting the experimental data set after feature optimization, initializing the learning rate, optimizer and number of iterations, capturing the nonlinear relationship between the input feature parameters and the measured snow depth, and obtaining the gap between the inversion model results and the measured values ​​through testing, that is, the accuracy of the inversion results.

[0100] The backbone network of the snow depth prediction model is the LSTM model, such as Figure 3 The model consists of a forget gate, an input gate, and an output gate. The data is read in the forget gate, and a value between 0 and 1 is output to each number in the cell state. 1 means it is completely retained, and 0 means it is completely discarded, which is used to decide whether to discard information. In the input gate, the old cell state is multiplied by the forget gate, the information to be discarded is discarded, and the new cell state generated in the input gate is added to obtain the updated cell state. In the output gate, it is determined which part of the cell state is output, and the cell state is processed to obtain the final output content.

[0101] Among them, the forget gate formula is expressed as:

[0102] f(t)=σ(W f ·[h(t-1),x(t)]+b f )

[0103] Among them, the input gate formula is expressed as:

[0104] i(t)=σ(W i ·[h(t-1),x(t)]+b i )

[0105] C(t)′=tanh(W C ·[h(t-1),x(t)]+b C )

[0106] C(t)=f(t)·C(t-1)+i(t)·C(t)'

[0107] Among them, the output gate formula is expressed as:

[0108] o(t)=σ(W o ·[h(t-1),x(t)]·+b o )

[0109] h(t)=o(t)·tanh(C(t))

[0110] Among them, f(t) is the forget gate function, σ is the activation function, W f is the weight parameter of the forget gate, x(t) is the current input data, b fis the forget gate bias parameter, h(t-1) is the output of the hidden layer at the previous moment, i(t) is the input gate function, C(t)′ is the new cell state generated by the input gate, C(t) is the updated cell state, W i and W C is the input gate weight parameter, b i and b C is the input gate bias parameter, C(t-1) is the old cell state, tanh is the hyperbolic tangent activation function, o(t) is the control signal of the output gate, W o is the output gate weight parameter, b o is the output gate bias parameter, and h(t) is the final output.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0112] Example 2 is an embodiment of the present invention, which provides a snow depth inversion method based on SAR image-based underlying surface information fusion. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0113] First, the Sentinel-1 synthetic aperture radar data is selected. The C-band SAR sensor on the Sentinel-1 satellite provides dual polarization capabilities of VV and VH polarization. The primary data products of the satellite after preliminary processing and multi-view processing are used as SAR remote sensing image data with a resolution of 20m. The remote sensing image processing platform ENVI is used to preprocess the image data, including orbit correction, thermal noise removal, radiation calibration, coherent speckle filtering, terrain correction, decibel processing and polarization decomposition, and backscattering features and polarization features are extracted from the radar image data. The original elevation data used in this example is SRTMDEM 90M geographic feature data with a resolution of 90m. After preprocessing, geographic features such as elevation, slope and slope aspect are extracted from the original elevation data. Rp5 is used in this example for meteorological data. This data provides historical monitoring data of most meteorological stations in the world with a time resolution of 2 hours. After preprocessing, meteorological features such as ambient temperature, ambient humidity, snow layer temperature and snow density are extracted from the data. The original experimental data set is composed of the characteristics of SAR remote sensing images when snow is accumulated, the characteristics of SAR remote sensing images when there is no snow, as well as geographical characteristics and meteorological characteristics:

[0114] X=[X1,X2,X3,X4]

[0115] Among them, X1 is the backscattering characteristics and polarization characteristics extracted from the SAR remote sensing image when snow is accumulated, X2 is the backscattering characteristics and polarization characteristics extracted from the SAR remote sensing image when there is no snow, that is, the underlying surface characteristics, X3 is the geographical characteristics extracted from the original elevation data, and X4 is the meteorological characteristics extracted from the meteorological data;

[0116] Secondly, the Pearson correlation coefficient between each feature and the snow depth was calculated. The original experimental data set after feature optimization was used as the experimental data set, and the training set and the test set were randomly divided according to the ratio of 8:2. The data were input into the snow depth inversion model with the ResNet model as the backbone network. After the training was completed, the snow depth inversion model was obtained. 2 Evaluate the model:

[0117]

[0118] Among them, d i is the actual observed value of snow depth, is the snow depth inversion value, is the actual average observed value of snow depth, RMSE is the root mean square error, R 2 is the determination coefficient, n is the number of parameters, and i is the variable index.

[0119] The RMSE of the snow depth inversion model is 2.03 cm and R 2 The inversion accuracy is 0.87 and 85.49%, which shows that the snow depth inversion model constructed in this example can effectively capture the relationship between snow depth and input features.

[0120] On the basis of the above experimental data set, the time variable is added to further analyze the long-term trend of snow depth series. The newly constructed experimental data set is input into the snow depth prediction model with LSTM model as the backbone network. After training, the snow depth prediction model can be obtained. Similarly, RMSE and R 2 The evaluation model has an RMSE of 2.57 cm and R 2 The prediction accuracy is 82.63%, which shows that the snow depth prediction model constructed in this example can effectively predict the snow depth and has a wide range of application scenarios in disaster warning.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0122] Example 3, reference Figure 4 , which is the third embodiment of the present invention, provides a snow depth inversion system based on SAR image underlying surface information fusion, including a data acquisition module 100, a terrain correction and radiation calibration module 200, a feature extraction and optimization module 300, a neural network model construction and training module 400 and a snow depth prediction module 500.

[0123] The data acquisition module 100 is used to collect satellite position, velocity, orbit attitude, time data and dynamic parameters, and perform orbit correction, thermal noise removal, coherent speckle filtering and decibel processing to reduce errors and noise in the data.

[0124] The terrain correction and radiation calibration module 200 is used to eliminate the radiation brightness error caused by terrain undulations by using the SCS+C correction method to ensure the consistency of reflection properties of different terrains, and convert the backscatter intensity received by the sensor into a dimensionless backscatter coefficient according to the radar equation.

[0125] The feature extraction and optimization module 300 is used to calculate polarization entropy, anisotropy and scattering angle through dual polarization ratio data and coherence matrix feature decomposition, obtain polarization information reflecting snow characteristics, extract slope and slope direction from elevation data, and extract ambient temperature and humidity data, and use the Pearson correlation coefficient to screen features to retain features that are strongly correlated with snow depth.

[0126] The neural network model construction and training module 400 is used to define the structure of the residual neural network ResNet model, input the optimized feature data set, initialize the model parameters, capture the nonlinear relationship between the feature parameters and the snow depth through training, and evaluate the accuracy of the model.

[0127] The snow depth prediction module 500 is used to input the SAR remote sensing image based on the trained ResNet model, invert the snow depth, introduce the time variable into the LSTM model, train the prediction model of the snow depth changing with time, and perform long-term prediction of the snow depth.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0129] Embodiment 4, the fourth embodiment of the present invention, is different from the first three embodiments in that:

[0130] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0131] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0132] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0133] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A snow depth inversion method based on SAR imagery and fusion of underlying surface information is characterized by: include, Collect orbit data, pre-process the collected orbit data, and remove thermal noise and coherent speckle noise; Perform terrain correction and perform radiation calibration according to the radar equation to obtain backscattering characteristic parameters and polarization characteristic parameters, extract features from the corrected data, and optimize the feature data set; A neural network model was constructed, and the optimized feature data set was input into the residual neural network. The snow depth inversion model was trained and the time variable was introduced to predict the snow depth.

2. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 1 is characterized by: The collecting of orbital data includes collecting satellite position data, satellite velocity data, orbital attitude data, time data, orbital dynamics parameters and orbital error correction data, and preprocessing the collected orbital data; The satellite position data includes longitude, latitude, altitude, orbit inclination and ascending node longitude; the satellite velocity data includes velocity along the orbit and velocity perpendicular to the orbit; the orbit attitude data includes pitch angle, yaw angle and roll angle; the time data includes ephemeris time and data collection time; the orbit dynamics parameters include orbit semi-major axis, orbit eccentricity, perigee angle and horizontal anomaly angle.

3. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 2 is characterized by: The preprocessing includes performing orbit correction, thermal noise removal, speckle filtering and decibel conversion on the acquired orbit data; The orbit correction includes performing error estimation by fitting a polynomial to the orbit error, forming an error data set by calculating the difference between the actual orbit data and the reference data, selecting a polynomial order to fit the error data, determining the polynomial coefficients by the least squares method, that is, the transpose of the matrix multiplied by the matrix and then multiplied by the coefficient vector is equal to the transpose of the matrix multiplied by the error data vector, calculating the value of the polynomial according to the timestamp after the coefficient is determined, and adding the value as a correction amount to the original orbit data for orbit correction; The thermal noise removal includes obtaining a thermal noise vector for any given azimuth time by linear interpolation of two noise vectors closest to each other, within the detection range, deriving two noise vectors from the thermal noise vector, and removing them from the detection range by linear interpolation; The speckle filtering includes updating the central pixel value through a filter and using the local statistical characteristics of the image to perform speckle filtering on the SAR remote sensing image, and the filter window is 7×7; The decibel processing formula is: ε=10log 10 s o Among them, σ o is the dimensionless backscattering coefficient, and ε is the normalized backscattering coefficient.

4. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 3 is characterized by: The terrain correction includes eliminating the radiation brightness error by using the SCS+C correction method of the Lambertian reflectivity model based on the elevation information DEM, so that the objects with the same reflective properties have the same brightness value in the image; The SCS+C correction method formula is expressed as: r H =ρ T (cosθ p cosθ z +C λ ) / (cosγ i +C λ ) r T =a λ +b λ cosγ i C λ =a λ / b λ Among them, ρ H is the horizontal surface reflection, ρ T is the slope surface reflection, θ p is the slope angle, θ z is the solar zenith angle, γ i is the solar incident angle, C λ is the correction factor, a λ is the slope of the regression line, b λ is the intercept of the regression line; The radiation calibration formula is expressed as: Among them, P d is the backscatter intensity received by the sensor, P t is the transmission power, P n is the additional power, is the perspective antenna gain, is the receiving antenna gain, is the current gain of the radar receiver, G p is the processor constant, R is the distance propagation loss, θ el is the antenna elevation angle, θ oc is the antenna azimuth, L s is the atmospheric loss, L a is the system loss, A is the scattering area, λ is the wavelength, σ o is the dimensionless backscattering coefficient.

5. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 4 is characterized by: Obtaining polarization characteristic parameters includes using dual polarization ratio data and calculating polarization entropy, anisotropy and scattering angle by eigenvalues ​​of three coherence matrix subcomponents obtained by eigendecomposing the coherence matrix; The formula for obtaining polarization characteristic parameters is: Among them, P i is the ratio of the eigenvalue corresponding to the i-th eigenvector to the total eigenvalues, λ i is the i-th eigenvalue, H is the polarization entropy, A is the anisotropy, α is the scattering angle, i is the variable index, λ1 is the volume scattering of the target, λ2 is the surface scattering of the target, and λ3 is the multiple scattering component of the target.

6. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 5 is characterized by: The optimization of the feature data set includes preprocessing the geographic feature data to obtain slope and aspect data, preprocessing the meteorological data, extracting features from the corrected data, and optimizing the features using the Pearson correlation coefficient; The formula for obtaining slope and aspect data is: Among them, Slope is the slope, Aspect is the slope direction, x Slope is the horizontal slope in the two-dimensional image. y is the vertical slope in the two-dimensional image; The Pearson correlation coefficient formula is: Among them, Y i and X i is the characteristic parameter, and is the average value of the characteristic parameter, n is the number of parameters, i is the variable index, PCC Y-X is the Pearson correlation coefficient; the Pearson correlation coefficient threshold is selected as 0.

9. When the calculated correlation coefficient is greater than or equal to 0.9, the characteristic parameter is judged to be a parameter correlated with the snow depth and is used as the input characteristic parameter of the snow depth inversion model.

7. The snow depth inversion method based on SAR image underlying surface information fusion as claimed in claim 6 is characterized by: The snow depth prediction includes: constructing snow depth inversion models based on different neural network model frameworks, selecting a residual neural network ResNet model for training, obtaining a snow depth inversion model, and performing model training, and obtaining the gap between the inversion model result and the measured value, i.e., the inversion accuracy, through testing, adjusting the model parameters according to the results, selecting the model parameters with the highest inversion accuracy, constructing a snow depth inversion model based on the parameters, and inputting the time variable into a long short-term memory network LSTM model for training, obtaining a snow depth long time series inversion model, and predicting the snow depth; The model training includes defining the model structure, inputting the experimental data set after feature optimization, initializing the learning rate, optimizer and number of iterations, capturing the nonlinear relationship between the input feature parameters and the measured snow depth, and obtaining the gap between the inversion model results and the measured values ​​through testing, that is, the accuracy of the inversion results.

8. A system using the method for snow depth inversion based on SAR image and underlying surface information fusion as claimed in any one of claims 1 to 7, characterized in that: It includes data acquisition module, terrain correction and radiation calibration module, feature extraction and optimization module, neural network model construction and training module and snow depth prediction module; The data acquisition module is used to collect satellite position, velocity, orbit attitude, time data and dynamic parameters, and perform orbit correction, thermal noise removal, coherent speckle filtering and decibel processing to reduce errors and noise in the data; The terrain correction and radiation calibration module is used to eliminate the radiation brightness error caused by terrain undulations by using the SCS+C correction method to ensure the consistency of reflection properties of different terrains, and convert the backscatter intensity received by the sensor into a dimensionless backscatter coefficient according to the radar equation; The feature extraction and optimization module is used to calculate polarization entropy, anisotropy and scattering angle through dual polarization ratio data and coherence matrix feature decomposition, obtain polarization information reflecting snow characteristics, extract slope and slope direction from elevation data, and extract ambient temperature and humidity data, and use Pearson correlation coefficient to screen features and retain features related to snow depth; The neural network model construction and training module is used to define the structure of the residual neural network ResNet model, input the optimized feature data set, initialize the model parameters, capture the nonlinear relationship between the feature parameters and the snow depth through training, and evaluate the accuracy of the model; The snow depth prediction module is used to input the SAR remote sensing image based on the trained ResNet model, invert the snow depth, introduce the time variable into the LSTM model, train the prediction model of the snow depth changing with time, and perform long-term prediction of the snow depth.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for snow depth inversion based on SAR image-based underlying surface information fusion according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for snow depth inversion based on SAR image-based underlying surface information fusion according to any one of claims 1 to 7 are implemented.

Images