A method and system for predicting tidal flat sediment components

Abstract

The present invention discloses a method and system for predicting mudflat components, which includes: obtaining sampling samples by field sampling and calculating mudflat component data; performing linear equalization processing on the mudflat component data to obtain first preprocessed data; collecting satellite remote sensing data and preprocessing the satellite remote sensing data to obtain second preprocessed data; segmenting the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set; substituting the data in the training data set into multiple machine learning algorithm models to establish an estimation model for mudflat components according to different light bands; using the data in the test set to verify the fitting degree of the trained multiple machine learning algorithm models, and selecting the optimal machine learning algorithm model according to the fitting degree, and using the optimal machine learning algorithm model to calculate mudflat components. The present invention can improve the simulation accuracy of mudflat components.

Description

Technical Field

[0001] The invention relates to the technical field of tidal flat sediment analysis, and in particular to a tidal flat sediment component prediction method and system. Background Art

[0002] The continuous evolution of tidal flats is one of the typical characteristics of tidal flats. The analysis of sediment components can better deal with the problem of tidal flat evolution, and plays an important role in the study of future changes in tidal flats and the further protection of tidal flats. At present, the existing research methods for sediment component analysis are mostly focused on conventional regression models, while there is little research on the estimation of sediment components through artificial intelligence machine learning.

[0003] In order to better understand the evolution of tidal flats and provide a scientific basis for the protection of tidal flats, it is necessary to accurately predict the changes in the composition of sediment content. At present, satellite remote sensing technology can be used to obtain information on different light bands. Using arcGIS and spatial models, spatial integration analysis of light band data can be performed, but there are major difficulties: the composition of tidal flat sediment varies greatly within a small range, and its relationship with light band data is not completely linear. Traditional prediction models are mostly multivariate linear regression models, which are difficult to handle when faced with nonlinear problems.

[0004] Machine learning algorithm is a data analysis method that uses data to automatically conduct modeling analysis and further predict using the model. It is an effective way to perform regression analysis on nonlinear systems. Research through machine learning methods can effectively overcome the difficulty of establishing complex relationships among multiple factors. Applying machine learning algorithms to the prediction of sediment components can effectively improve accuracy.

[0005] In summary, in order to accurately grasp the characteristics of tidal flat sediment components, it is necessary to propose a method for estimating the proportion of different components of tidal flat sediment, so as to further provide strong technical support for the scientific protection of tidal flats. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a method of using remote sensing to obtain the characteristics of different light bands of the tidal flat, and using a machine learning algorithm model to accurately estimate the different components of tidal flat sediment.

[0007] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0008] A method for predicting tidal flat sediment components of the present invention comprises the following operations:

[0009] Conduct field sampling in the area to be tested to obtain sampling samples, and calculate the sediment component data of the area to be tested based on the sampling samples;

[0010] Performing linear equalization processing on the measured sediment component data to obtain first preprocessed data;

[0011] Collect satellite remote sensing data of tidal flat sediment and pre-process the satellite remote sensing data to obtain second pre-processed data, wherein the second pre-processed data includes information data of different optical bands;

[0012] Segmenting the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set;

[0013] Substitute the data in the training data set into multiple machine learning algorithm models. In multiple machine learning algorithm models, the information data of different bands are used as input variables, and clay Pc', silt Psi', and sand Psa' are used as output variables. An estimation model for sediment components based on different light bands is established. The expression is:

[0014] Pc'=f(B1,B2, ,Bn);

[0015] Psi'=f(B1,B2, ,Bn);

[0016] Psa'=f(B1,B2, ,Bn);

[0017] Among them, f() is the regression function, B1, B2, ,Bn is the information data of different bands, n is the number of different bands;

[0018] The data in the test set is used to verify the degree of fit of multiple trained machine learning algorithm models. According to the degree of fit, the errors of the inversion results of different machine learning algorithm models are compared, and the machine learning algorithm model with the smallest error and the highest accuracy after training is selected. The selected machine learning algorithm model is used to calculate the tidal flat sediment components.

[0019] A further improvement of the present invention is that: field sampling is performed on the area to be tested to obtain a sample, and the sediment component data of the area to be tested is calculated based on the sample, specifically including:

[0020] Sampling multiple points of the area to be tested, each sampling point range includes multiple sampling points, and multiple sampling samples are obtained;

[0021] The sediment particles of the sampled samples were analyzed and recorded to obtain the sediment component data of each sampling point, including clay Pc, silt Psi, and sand Psa.

[0022] A further improvement of the present invention is to collect satellite remote sensing data of tidal flat sand and pre-process the satellite remote sensing data to obtain second pre-processed data, which specifically includes:

[0023] Obtain satellite remote sensing data of tidal flat sediment;

[0024] The collected satellite remote sensing data were segmented using ENVI software, and the light beaches with flooding frequency corresponding to the area to be measured were selected according to the land type to form a remote sensing map.

[0025] Using the resampling function in ArcGIS software, the stacking sampling method is adopted, that is, the nearest interpolation method is used first, and then the bilinear interpolation method is used to resample the remote sensing image, and the mask extraction function is used to extract the features of the remote sensing image, and the extracted features are vectorized to generate a vector map;

[0026] ArcGIS software is used to perform multi-value extraction on the vector map to obtain the second preprocessed data.

[0027] A further improvement of the present invention is that the machine learning algorithm model includes SVR support vector regression, random forest RF, and logistic regression LR.

[0028] A further improvement of the present invention is that the estimation model is:

[0029] Pc'=f(B1,B2,B3,B4,B5,B6,B7);

[0030] Psi'=f(B1,B2,B3,B4,B5,B6,B7);

[0031] Psa'=f(B1,B2,B3,B4,B5,B6,B7);

[0032] Among them, B1 is the coastal band, B2 is the blue band, B3 is the green wave, B4 is the red wave, B5 is the near-infrared band, B6 is shortwave infrared 1, and B7 is shortwave infrared 2.

[0033] The further improvement of the present invention is to use the data in the test set to verify the degree of fit of multiple trained machine learning algorithm models, compare the errors of the inversion results of different machine learning algorithm models according to the degree of fit, and select the machine learning algorithm model with the smallest error and the highest accuracy after training, specifically including:

[0034] Substitute the data of the test data set into multiple trained machine learning algorithm models to obtain the prediction results of the corresponding machine learning algorithm models, including the percentage of clay Pc, silt Psi, and sand Psa content;

[0035] Verify the fit of multiple machine learning algorithm models, including:

[0036] Using the determination coefficient, root mean square error and mean absolute error as evaluation indicators, the prediction results of different machine learning algorithm models are compared to select the optimal machine learning algorithm model.

[0037] The tidal flat sediment component prediction system of the present invention comprises:

[0038] The sampling module is used to obtain sampling samples by field sampling in the area to be tested, and calculate the sediment component data of the area to be tested based on the sampling samples;

[0039] A first preprocessing module is used to perform linear equalization processing on the measured sediment component data to obtain first preprocessing data;

[0040] A second preprocessing module is used to collect satellite remote sensing data of tidal flat sediment and preprocess the satellite remote sensing data to obtain second preprocessed data, wherein the second preprocessed data includes information data of different optical bands;

[0041] A segmentation module, used to segment the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set;

[0042] The model building module is used to establish multiple estimation models for sediment components according to different light bands using multiple machine learning algorithms, including using information data of different bands as input variables in multiple machine learning algorithm models, clay Pc', silt Psi', and sand Psa' as output variables, and establishing estimation models for sediment components according to different light bands. The expression is:

[0043] Pc'=f(B1,B2, ,Bn);

[0044] Psi'=f(B1,B2, ,Bn);

[0045] Psa'=f(B1,B2, ,Bn);

[0046] Among them, f() is the regression function, B1, B2, ,Bn is the information data of different bands, n is the number of different bands;

[0047] The model verification module is used to verify the fit of multiple trained machine learning algorithm models using the data in the test set, and select the optimal algorithm model based on the fit;

[0048] The prediction module is used to select the optimal algorithm model and use the optimal algorithm model to calculate the mudflat sediment components.

[0049] A further improvement of the present invention is that the specific operations performed by the sampling module include:

[0050] Sampling multiple points of the area to be tested, each sampling point range includes multiple sampling points, and multiple sampling samples are obtained;

[0051] The sediment particles of the sampled samples were analyzed and recorded to obtain the sediment component data of each sampling point, including clay Pc, silt Psi, and sand Psa.

[0052] A further improvement of the present invention is that the specific operations performed by the second preprocessing module include:

[0053] Obtain satellite remote sensing data of tidal flat sediment;

[0054] The collected satellite remote sensing data were segmented using ENVI software, and the light beaches with flooding frequencies corresponding to the tested area were selected according to the land types to form remote sensing maps, where the land types included light beaches, salt marsh vegetation belts, etc.

[0055] Using the resampling function in ArcGIS software, the stacking sampling method is adopted, that is, the nearest interpolation method is used first, and then the bilinear interpolation method is used to resample the remote sensing image, and the mask extraction function is used to extract the features of the remote sensing image, and the extracted features are vectorized to generate a vector map;

[0056] ArcGIS software is used to perform multi-value extraction on the vector map to obtain the second preprocessed data.

[0057] A further improvement of the present invention is that the specific operations performed by the model verification module include:

[0058] Substitute the data of the test data set into multiple trained machine learning algorithm models to obtain the prediction results of the corresponding machine learning algorithm models, including the percentage of clay Pc, silt Psi, and sand Psa content;

[0059] Verify the fit of multiple machine learning algorithm models, including:

[0060] Using the determination coefficient, root mean square error and mean absolute error as evaluation indicators, the prediction results of different machine learning algorithm models are compared to select the optimal machine learning algorithm model.

[0061] The beneficial effects of the present invention are as follows: the present invention uses remote sensing means to obtain the characteristics of different light bands of tidal flats, and integrates them on a GIS platform. By utilizing a machine learning algorithm model, accurate estimation of different components of tidal flat sediment can be achieved, eliminating basic data errors and nonlinear data problems in inversion in traditional methods, thereby improving the simulation accuracy of tidal flat sediment components. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0064] like Figure 1 As shown, a method for predicting mudflat components in this embodiment includes the following steps:

[0065] Step S1, conduct field sampling in the area to be tested to obtain sampling samples, and calculate the sediment component data of the area to be tested based on the sampling samples. Specifically including:

[0066] Step S11, sampling multiple points of the area to be tested, each sampling point range includes multiple sampling points, and multiple sampling samples are obtained;

[0067] Step S12, analyze and record the sediment particles of the sampled samples to obtain sediment component data of each sampling point, including clay Pc, silt Psi, and sand Psa.

[0068] Step S2, using the Delta method to perform linear equalization processing on the measured sediment component data to obtain first preprocessed data.

[0069] Step S3, collecting satellite remote sensing data of tidal flat sand and preprocessing the satellite remote sensing data to obtain second preprocessed data. Specifically including:

[0070] Step S31, obtaining satellite remote sensing data of tidal flat sediment, the satellite remote sensing data comes from satellites such as Landset 8 and Sentinel 2;

[0071] Step S32, using ENVI software to segment the collected satellite remote sensing data into regions, selecting the light beach with flooding frequency corresponding to the area to be measured according to the land type, and forming a remote sensing map;

[0072] Step S33, using the resampling function in ArcGIS software, adopting the overlay sampling method, that is, first using the nearest interpolation method and then using the bilinear interpolation method, resampling the remote sensing image, using the mask extraction function to extract features from the remote sensing image, and vectorizing the extracted features to generate a vector map;

[0073] Step S34, using ArcGIS software to perform multi-value extraction on the vector map to points to obtain second preprocessed data, wherein the second preprocessed data includes information data of different optical bands, specifically including: coastal band, blue band, green wave, red wave, near infrared band, short wave infrared 1, and short wave infrared 2.

[0074] This embodiment can retain edge information while improving computational efficiency. Using nearest neighbor interpolation as a preliminary fast interpolation method can quickly generate a larger image and retain edge features in the image. Bilinear interpolation is then used for refinement processing, which further increases the smoothness of the image while maintaining edge clarity, reducing computational complexity and processing time.

[0075] Step S4, segmenting the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set.

[0076] Step S5, substitute the data in the training data set into multiple machine learning algorithm models, the machine learning algorithm models include: SVR support vector regression, random forest RF, logistic regression LR, in SVR support vector regression, random forest RF, logistic regression LR, use the information data of different bands as input variables, clay Pc', silt Psi', sand Psa' as output variables, and establish an estimation model for sediment components according to different light bands, the expression is:

[0077] Pc'=f(B1,B2,B3,B4,B5,B6,B7);

[0078] Psi'=f(B1,B2,B3,B4,B5,B6,B7);

[0079] Psa'=f(B1,B2,B3,B4,B5,B6,B7);

[0080] Among them, f() is the regression function, B1 is the coastal band, B2 is the blue band, B3 is the green band, B4 is the red band, B5 is the near-infrared band, B6 is shortwave infrared 1, and B7 is shortwave infrared 2.

[0081] Step S6, using the data in the test set to verify the degree of fit of the trained multiple machine learning algorithm models, comparing the errors of the inversion results of different machine learning algorithm models according to the degree of fit, screening the machine learning algorithm model with the smallest error and the highest accuracy after training, that is, the optimal machine learning algorithm model, and using the optimal machine learning algorithm model to calculate the tidal flat sediment components.

[0082] in:

[0083] The formula for the coefficient of determination is: ;

[0084] The formula for the root mean square error is: ;

[0085] The formula for mean absolute error is: ;

[0086] Among them, R 2 is the coefficient of determination, RMSE is the root mean square error, MAE is the mean absolute error, P is the percentage of sediment components in the existing data, specifically representing clay Pc, silt Psi, and sand Psa, The percentage of sediment components inverted from the test data, specifically clay Pc', silt Psi', sand Psa', is the average percentage of sediment components in the existing data, and n is the total number of sampling points.

[0087] The above method is implemented by the tidal flat sediment component prediction system of the present invention in this embodiment, and the system includes:

[0088] The sampling module is used to obtain sampling samples by field sampling in the area to be tested, and calculate the sediment component data of the area to be tested based on the sampling samples;

[0089] A first preprocessing module is used to perform linear equalization processing on the measured sediment component data to obtain first preprocessing data;

[0090] A second preprocessing module is used to collect satellite remote sensing data of tidal flat sediment and preprocess the satellite remote sensing data to obtain second preprocessed data;

[0091] A segmentation module, used to segment the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set;

[0092] The model building module is used to establish multiple estimation models for sediment components according to different light bands using multiple machine learning algorithms, including using information values ​​of different bands as input variables in multiple machine learning algorithm models, clay Pc', silt Psi', and sand Psa' as output variables, and establishing estimation models for sediment components according to different light bands. The expression is:

[0093] Pc'=f(B1,B2,B3,B4,B5,B6,B7);

[0094] Psi'=f(B1,B2,B3,B4,B5,B6,B7);

[0095] Psa'=f(B1,B2,B3,B4,B5,B6,B7);

[0096] Among them, f() is the regression function, B1 is the coastal band, B2 is the blue band, B3 is the green band, B4 is the red band, B5 is the near-infrared band, B6 is short-wave infrared 1, and B7 is short-wave infrared 2;

[0097] The model verification module is used to verify the fit of multiple trained machine learning algorithm models using the data in the test set, and select the optimal algorithm model based on the fit;

[0098] The prediction module is used to select the optimal algorithm model and use the optimal algorithm model to calculate the mudflat sediment components.

[0099] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless defined similarly here.

[0100] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting tidal flat sediment components, characterized by: The following operations are included: Conduct field sampling in the area to be tested to obtain sampling samples, and calculate the sediment component data of the area to be tested based on the sampling samples; Performing linear equalization processing on the measured sediment component data to obtain first preprocessed data; Collect satellite remote sensing data of tidal flat sediment and pre-process the satellite remote sensing data to obtain second pre-processed data, wherein the second pre-processed data includes information data of different optical bands; Segmenting the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set; Substitute the data in the training data set into multiple machine learning algorithm models. In multiple machine learning algorithm models, the information data of different bands are used as input variables, and clay Pc', silt Psi', and sand Psa' are used as output variables. An estimation model for sediment components based on different light bands is established. The expression is: Pc’=f(B1,B2, ,Bn); <h2 style=";text-align:left;direction:ltr">Psi'=f(B1,B2,<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (Bn) <h2 style=";text-align:left;direction:ltr">Psa'=f(B1,B2,<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (Bn) Among them, f() is the regression function, B1, B2, ,Bn is the information data of different bands, n is the number of different bands; Use the data in the test set to verify the degree of fit of multiple trained machine learning algorithm models. According to the degree of fit, compare the errors of the inversion results of different machine learning algorithm models, select the machine learning algorithm model with the smallest error and the highest accuracy after training, and use the selected machine learning algorithm model to calculate the mudflat sediment components; The collecting of satellite remote sensing data of tidal flat sediment and preprocessing the satellite remote sensing data to obtain second preprocessed data specifically includes: Obtain satellite remote sensing data of tidal flat sediment; The collected satellite remote sensing data were segmented using ENVI software, and the light beaches with flooding frequency corresponding to the area to be measured were selected according to the land type to form a remote sensing map. Using the resampling function in ArcGIS software, the stacking sampling method is adopted, that is, the nearest interpolation method is used first, and then the bilinear interpolation method is used to resample the remote sensing image, and the mask extraction function is used to extract the features of the remote sensing image, and the extracted features are vectorized to generate a vector map; ArcGIS software is used to perform multi-value extraction on the vector map to obtain the second preprocessed data.

2. A method for predicting tidal flat sediment composition according to claim 1, characterized in that: The method of obtaining sampling samples by field sampling in the area to be tested and calculating the sediment component data of the area to be tested according to the sampling samples specifically includes: Sampling multiple points of the area to be tested, each sampling point range includes multiple sampling points, and multiple sampling samples are obtained; The sediment particles of the sampled samples were analyzed and recorded to obtain the sediment component data of each sampling point, including clay Pc, silt Psi, and sand Psa.

3. A method for predicting tidal flat sediment composition according to claim 1, characterized in that: The machine learning algorithm models include SVR support vector regression, random forest RF, and logistic regression LR.

4. A method for predicting tidal flat sediment composition according to claim 3, characterized in that: The estimation model is: Pc'=f(B1,B2,B3,B4,B5,B6,B7); Psi'=f(B1,B2,B3,B4,B5,B6,B7); Psa'=f(B1,B2,B3,B4,B5,B6,B7); Among them, B1 is the coastal band, B2 is the blue band, B3 is the green wave, B4 is the red wave, B5 is the near-infrared band, B6 is shortwave infrared 1, and B7 is shortwave infrared 2.

5. A method for predicting tidal flat sediment composition according to claim 1, characterized in that: The method of using the data in the test set to verify the degree of fit of the trained multiple machine learning algorithm models, comparing the errors of the inversion results of different machine learning algorithm models according to the degree of fit, and screening the machine learning algorithm model with the smallest error and the highest accuracy after training, specifically includes: Substitute the data of the test data set into multiple trained machine learning algorithm models to obtain the prediction results of the corresponding machine learning algorithm models, including the percentage of clay Pc, silt Psi, and sand Psa content; Verify the fit of multiple machine learning algorithm models, including: Using the determination coefficient, root mean square error and mean absolute error as evaluation indicators, the prediction results of different machine learning algorithm models are compared to select the optimal machine learning algorithm model.

6. The tidal flat sediment composition prediction system is characterized by: The system comprises: The sampling module is used to obtain sampling samples by field sampling in the area to be tested, and calculate the sediment component data of the area to be tested based on the sampling samples; A first preprocessing module is used to perform linear equalization processing on the measured sediment component data to obtain first preprocessing data; A second preprocessing module is used to collect satellite remote sensing data of tidal flat sediment and preprocess the satellite remote sensing data to obtain second preprocessed data, wherein the second preprocessed data includes information data of different optical bands; A segmentation module, used to segment the first preprocessed data and the second preprocessed data to obtain a training data set and a test data set; The model building module is used to establish multiple estimation models for sediment components according to different light bands using multiple machine learning algorithms, including using information data of different bands as input variables in multiple machine learning algorithm models, clay Pc', silt Psi', and sand Psa' as output variables, and establishing estimation models for sediment components according to different light bands. The expression is: Pc’=f(B1,B2, ,Bn); <h2 style=";text-align:left;direction:ltr">Psi'=f(B1,B2,<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (Bn) <h2 style=";text-align:left;direction:ltr">Psa'=f(B1,B2,<h2 style=";text-align:left;direction:ltr"> <h2 style=";text-align:left;direction:ltr"> (Bn) Among them, f() is the regression function, B1, B2, ,Bn is the information data of different bands, n is the number of different bands; The model verification module is used to verify the fit of multiple trained machine learning algorithm models using the data in the test set, and select the optimal algorithm model based on the fit; The prediction module is used to select the optimal algorithm model and use it to calculate the mudflat sediment components; The specific operations performed by the second preprocessing module include: Obtain satellite remote sensing data of tidal flat sediment; The collected satellite remote sensing data were segmented using ENVI software, and the light beaches with flooding frequency corresponding to the area to be measured were selected according to the land type to form a remote sensing map. Using the resampling function in ArcGIS software, the stacking sampling method is adopted, that is, the nearest interpolation method is used first, and then the bilinear interpolation method is used to resample the remote sensing image, and the mask extraction function is used to extract the features of the remote sensing image, and the extracted features are vectorized to generate a vector map; ArcGIS software is used to perform multi-value extraction on the vector map to obtain the second preprocessed data.

7. The tidal flat sediment composition prediction system according to claim 6, characterized in that: The specific operations performed by the sampling module include: Sampling multiple points of the area to be tested, each sampling point range includes multiple sampling points, and multiple sampling samples are obtained; The sediment particles of the sampled samples were analyzed and recorded to obtain the sediment component data of each sampling point, including clay Pc, silt Psi, and sand Psa.

8. The system for predicting tidal flat sediment composition according to claim 6, characterized in that: The specific operations performed by the model verification module include: Substitute the data of the test data set into multiple trained machine learning algorithm models to obtain the prediction results of the corresponding machine learning algorithm models, including the percentage of clay Pc, silt Psi, and sand Psa content; Verify the fit of multiple machine learning algorithm models, including: Using the determination coefficient, root mean square error and mean absolute error as evaluation indicators, the prediction results of different machine learning algorithm models are compared to select the optimal machine learning algorithm model.

Images