Remote sensing image inversion method based on spatial interpolation and machine learning and related equipment

Abstract

The invention relates to the technical field of data processing, in particular to a remote sensing image inversion method and related equipment based on spatial interpolation and machine learning, and the method comprises the steps: selecting a first preset number of sampling points in a to-be-detected region, and obtaining the target information of the sampling points; based on the target information of the sampling points, target information of a second preset number of random points is generated through a spatial interpolation technology; extracting a spectral value of each wave band in the remote sensing image; performing correlation analysis based on the spectral value and the target parameter, and determining an optimal waveband combination; constructing a data set based on the spectral value of the optimal waveband combination and the target parameter, training and verifying a target machine learning algorithm, and obtaining an inversion model for determining the target parameter based on the spectral value; and performing target parameter inversion based on the inversion model and the remote sensing image to be inverted. Therefore, on the basis of ensuring the inversion precision of the remote sensing image, the data demand can be greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a remote sensing image inversion method and related equipment based on spatial interpolation and machine learning. Background Art

[0002] Remote sensing images, especially hyperspectral remote sensing images, contain rich spectral information. After being inverted, they have important application values in many fields such as environmental monitoring, ecological research, and resource exploration. However, in the prior art, the inversion generally requires a large number of field sampling points as training data to ensure the accuracy of relevant models. This not only increases the data acquisition cost and workload, but also it is difficult to obtain sufficient samples in complex terrain areas, resulting in great deficiencies in the application of remote sensing images. Therefore, there is an urgent need for a method to reduce the data requirements for remote sensing image inversion while ensuring accuracy. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a remote sensing image inversion method and related equipment based on spatial interpolation and machine learning to overcome the problem of high data requirements in the current application of remote sensing image inversion.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] In the first aspect, the present application provides a remote sensing image inversion method based on spatial interpolation and machine learning, including:

[0006] Select a first preset number of sampling points in the area to be measured, and obtain the target information of the sampling points, where the target information includes target parameters and geographic coordinate information;

[0007] Based on the target information of the sampling points, generate the target information of a second preset number of random points through spatial interpolation technology;

[0008] Extract the spectral values of each band at the sampling points and the random points in the remote sensing image;

[0009] Perform correlation analysis based on the spectral values and the target parameters to determine the optimal band combination;

[0010] Based on the spectral values of the optimal band combination and the target parameters, construct a data set, train and verify a target machine learning algorithm, and obtain an inversion model for determining the target parameters based on the spectral values;

[0011] Based on the inversion model and the remote sensing image to be inverted, perform inversion of the target parameters.

[0012] Further, in some embodiments of the present application, generating the target information of the second preset number of random points through spatial interpolation technology based on the target information of the sampling points includes:

[0013] Generating the surface distribution data of the target parameters through spatial interpolation technology based on the target information of the sampling points;

[0014] Randomly sampling to generate the second preset number of random points, and extracting the corresponding target parameters from the surface distribution data.

[0015] Further, in some embodiments of the present application, after extracting the spectral values of each band at the sampling points and the random points in the remote sensing image, it further includes: performing standardization and normalization processing on the extracted spectral values.

[0016] Further, in some embodiments of the present application, after extracting the spectral values of each band at the sampling points and the random points in the remote sensing image, it further includes: dividing two-thirds of the sampling points and all the random points into a training set; dividing one-third of the sampling points into a validation set.

[0017] Further, in some embodiments of the present application, the correlation analysis based on the spectral values and the target parameters to determine the optimal band combination includes:

[0018] For the sampling points and random points divided into the training set, performing correlation analysis on the target parameters and the spectral values of each band of the corresponding sampling points or random points to determine the sensitive bands with high correlation;

[0019] Combining the sensitive bands to obtain a sensitive band combination;

[0020] Performing correlation analysis on the sensitive band combination and the target parameters of the corresponding sampling points or random points to determine the sensitive band combination with high correlation as the optimal band combination.

[0021] Further, in some embodiments of the present application, constructing a data set based on the spectral values and the target parameters of the optimal band combination, training and validating a target machine learning algorithm, and obtaining an inversion model for determining the target parameters based on the spectral values includes:

[0022] Constructing a training data set based on the spectral values and target parameters of the optimal band combination corresponding to the sampling points or random points divided into the training set;

[0023] Training the target machine learning algorithm based on the training data set to obtain the inversion model.

[0024] Further, in some embodiments of the present application, the target machine learning algorithm includes at least one of a support vector machine, a random forest, and a neural network.

[0025] Further, in some embodiments of the present application, the target parameter includes a water quality parameter, and the water quality parameter includes a pH value, a dissolved oxygen, and a total phosphorus.

[0026] In a second aspect, the present application provides a remote sensing image inversion device based on spatial interpolation and machine learning, including:

[0027] An acquisition module, configured to select a first preset number of sampling points in a to-be-measured area and obtain target information of the sampling points, where the target information includes a target parameter and geographic coordinate information;

[0028] A spatial interpolation module, configured to generate target information of a second preset number of random points through a spatial interpolation technique based on the target information of the sampling points;

[0029] An extraction module, configured to extract spectral values of each band at the sampling points and the random points in a remote sensing image;

[0030] An analysis module, configured to perform a correlation analysis based on the spectral values and the target parameter to determine an optimal band combination;

[0031] A construction and application module, configured to construct a data set based on the spectral values of the optimal band combination and the target parameter, train and verify a target machine learning algorithm, and obtain an inversion model for determining the target parameter based on the spectral values; and perform inversion of the target parameter based on the inversion model and a to-be-inverted remote sensing image.

[0032] In a third aspect, the present application provides a remote sensing image inversion device based on spatial interpolation and machine learning, including a processor and a memory, where the processor is connected to the memory:

[0033] Wherein, the processor is configured to call and execute a program stored in the memory;

[0034] The memory is configured to store the program, and the program is at least used to execute the above-mentioned remote sensing image inversion method based on spatial interpolation and machine learning.

[0035] The present invention relates to the technical field of data processing, and particularly to a remote sensing image inversion method and related devices based on spatial interpolation and machine learning. The method includes: selecting a first preset number of sampling points within the area to be measured, and obtaining the target information of the sampling points; based on the target information of the sampling points, generating the target information of a second preset number of random points through spatial interpolation technology; extracting the spectral values of each band in the remote sensing image; performing correlation analysis based on the spectral values and target parameters to determine the optimal band combination; constructing a data set based on the spectral values of the optimal band combination and the target parameters, training and validating the target machine learning algorithm to obtain an inversion model for determining the target parameters based on the spectral values; and performing inversion of the target parameters based on the inversion model and the remote sensing image to be inverted. In this way, on the basis of ensuring the inversion accuracy of the remote sensing image, the data requirements can be greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 It is a flowchart of the remote sensing image inversion method based on spatial interpolation and machine learning provided by an embodiment of the present invention.

[0038] Figure 2 It is a flowchart of the remote sensing image inversion method based on spatial interpolation and machine learning provided by another embodiment of the present invention.

[0039] Figure 3 It is a structural schematic diagram of the remote sensing image inversion device based on spatial interpolation and machine learning provided by an embodiment of the present invention.

[0040] Figure 4 It is a structural schematic diagram of the remote sensing image inversion device based on spatial interpolation and machine learning provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope protected by the present invention.

[0042] Figure 1It is a schematic flowchart of a remote sensing image inversion method based on spatial interpolation and machine learning provided by an embodiment of the present invention. Please refer to Figure 1 , this embodiment may include the following steps:

[0043] S101. Select a first preset number of sampling points within the area to be measured, and obtain the target information of the sampling points.

[0044] Among them, the target information includes target parameters and geographic coordinate information. For example, for water quality monitoring, the target parameters are water quality parameters such as pH value, dissolved oxygen, and total phosphorus.

[0045] S102. Based on the target information of the sampling points, generate the target information of a second preset number of random points through spatial interpolation technology.

[0046] Specifically, first, based on the target information of the sampling points, generate the surface distribution data of target parameters such as water quality parameters through spatial interpolation technology, then randomly sample to generate a second preset number of random points, and extract the corresponding target parameters from the surface distribution data.

[0047] S103. Extract the spectral values of each band at the sampling points and random points in the remote sensing image.

[0048] Specifically, the remote sensing image mentioned in this application may be a hyperspectral remote sensing image. After preprocessing the hyperspectral remote sensing image, based on the pixel information thereon, the spectral values of each band corresponding to the actual geographical location can be obtained (that is, the spectral reflectance values of each band).

[0049] S104. Conduct a correlation analysis based on the spectral values and target parameters to determine the optimal band combination.

[0050] Specifically, first, through correlation analysis, obtain the sensitive bands with high correlation with target parameters such as water quality parameters, then perform various combinations on the sensitive bands to obtain sensitive band combinations, and then conduct a correlation analysis on the sensitive band combinations with target parameters such as water quality parameters to obtain the optimal band combination.

[0051] S105. Based on the spectral values and target parameters of the optimal band combination, construct a data set, train and verify the target machine learning algorithm, and obtain an inversion model for determining the target parameters based on the spectral values.

[0052] S106. Based on the inversion model and the remote sensing image to be inverted, perform the inversion of the target parameters.

[0053] Based on the spectral values and target parameters of the optimal band combination, a dataset is constructed for model training and learning, enabling the model to learn the mapping relationship from spectral bands to water quality parameters, and validating the performance of the model to obtain an inversion model. In practical applications, the inversion model is used to invert the remote sensing image to be inverted.

[0054] Figure 2 FIG. is a schematic flowchart of a remote sensing image inversion method based on spatial interpolation and machine learning provided by another embodiment of the present invention. Combining Figure 2 , the following takes water quality parameters as target parameters to describe each step in the present application in detail:

[0055] First, on-site sampling is carried out, that is, a first preset number of sampling points are selected in the area to be measured, and the target information of the sampling points is obtained, where the target information includes target parameters and geographic coordinate information.

[0056] Specifically, a first preset number, such as N, of sampling points are selected in the area to be measured, and relevant parameters are measured on-site. Taking water quality as an example, it includes pH value, dissolved oxygen, total phosphorus, etc. At the same time, a GPS device is used to record the geographic coordinate information of each sampling point.

[0057] It should be noted that in the present application, it also includes using software such as GIS to extract the inversion area in combination with remote sensing image data, determine the boundary, shape, and area of the area, and prepare the data basis required for inversion. The above-mentioned area to be measured and the area to be inverted subsequently can both be carried out in the extracted inversion area.

[0058] Further, spatial interpolation, random point generation, and information extraction are carried out, that is, based on the target information of the sampling points, the target information of a second preset number of random points is generated through spatial interpolation technology.

[0059] Specifically, first, the water quality parameter distribution is generated by spatial interpolation. For example, the spatial interpolation can be performed on the data of the N sample points that have been obtained to generate the surface distribution data of the water quality parameters. Among them, in practical applications, interpolation methods such as Kriging interpolation and inverse distance weighting can be selected to ensure the continuity of the data within the inversion area. Then, random points are generated and parameters are extracted. For example, based on the above-obtained surface distribution data, M random points (generally M > N) are generated by random sampling, and the water quality parameter values generated by spatial interpolation and their corresponding geographic coordinates are extracted at these random points. Both the random points and the sampling points are used for subsequent processing, thereby achieving an expansion of the data volume and geographical coverage.

[0060] At the same time, first, the remote sensing image is preprocessed and spectral extraction is carried out.

[0061] Specifically, first, the original remote sensing images such as hyperspectral remote sensing images can be radiometrically calibrated, atmospherically corrected, and geographically corrected through ENVI or other remote sensing image processing software.

[0062] Then, the spectral curve data at the above-mentioned sampling points and random points are extracted from the corrected image, and the spectral values of each band of the hyperspectrum at these points are obtained.

[0063] In some embodiments of the present application, after obtaining the above spectral values, the spectral values of different bands can also be standardized or normalized to ensure the scale consistency between different bands. The standardization formula is as follows:

[0064]

[0065] where μ is the mean of the data, σ is the standard deviation, a is the spectral value before standardization, and a′ is the spectral value after standardization.

[0066] Furthermore, sample division is performed.

[0067] Specifically, the above-mentioned N sampling points (i.e., samples) are divided into a training set and a validation set. For example, 2N / 3 is divided into the training set and N / 3 is divided into the validation set. The 2N / 3 sampling points in the training set and M random points are used for the training of the inversion model, and the N / 3 sampling points in the validation set are used for the validation of the inversion model.

[0068] Furthermore, sensitive band screening is performed.

[0069] Specifically, a correlation analysis is performed on the spectral values of each band of the target parameter and the corresponding sampling point or random point, and several bands with higher correlations are selected as sensitive bands to reduce the model calculation amount and improve the inversion accuracy.

[0070] Furthermore, the best band combination screening is performed, including combining the sensitive bands to obtain sensitive band combinations; performing a correlation analysis on the sensitive band combinations and the target parameters of the corresponding sampling points or random points to determine the best band combination with high correlation.

[0071] Specifically, for the above sensitive bands, any two sensitive bands are combined in ways such as addition, subtraction, and ratio. For more than two sensitive bands, different combinations such as multi-band combination addition and (Ra + Rb) / Rc and (Ra + Rb) / (Ra - Rb) are adopted to obtain sensitive band combinations. Then, a correlation analysis is performed on each sensitive band combination and the target parameter such as the water quality parameter respectively, and the sensitive band combination with the highest correlation is selected as the final best band combination for subsequent model construction (of course, in some embodiments of the present application, some sensitive bands can also be directly selected for subsequent model construction).

[0072] Further, model training and evaluation are carried out.

[0073] First, a dataset is constructed.

[0074] Specifically, after obtaining the above-mentioned optimal band combination, first, for the points divided into the training set, the data of the corresponding bands are combined with the target parameters such as water quality parameter values to form a training dataset. In practical applications, this training dataset can be in the form of a matrix, where each row represents a sample (such as the sampling points or random points mentioned above), each column represents the spectral value of a band (or band combination), and the last column is the target variable (water quality parameter).

[0075] For example, let X be the feature matrix, which contains the spectral values of the optimal band combination (each sample has m bands),

[0076] and Y be the target variable matrix, which contains the corresponding parameter values (such as pH, dissolved oxygen, total nitrogen, etc.) and is represented as follows:

[0077] X = [x 1 x 2 …x m , Y = [y 1 y 2 …y n (2)

[0078] where x i is the spectral value of a certain band or combination, and y i is the value of the corresponding water quality parameter.

[0079] Then, the target machine learning algorithm is determined.

[0080] Specifically, in this application, the target machine learning algorithms include support vector machine, random forest, and neural network. The following is a brief introduction to the training processes and model prediction forms of support vector machine, random forest, and neural network:

[0081] For the support vector machine (Support Vector Machine, SVM), for the regression task, support vector regression is a commonly used choice, which fits the model by minimizing the loss function. The training process can be simply expressed by the following formula:

[0082]

[0083] where w is the parameter of the model, ε i is the slack variable, and C is the regularization parameter.

[0084] The model prediction can be expressed as:

[0085]

[0086] For Random Forest, it performs regression by integrating multiple decision trees. During training, each tree learns a random subset of the data and generates a prediction result. The final result is the average of the output results of each tree. The training process can be simply expressed by the following formula:

[0087]

[0088] where f t (x) is the prediction result of the i-th tree, and T is the number of trees.

[0089] For Neural Networks, it is especially suitable for dealing with complex non-linear relationships and can be trained through a Multi-Layer Perceptron (MLP). The training process involves adjusting the weights to make the model output as close as possible to the true value. The training process uses the backpropagation algorithm to minimize the loss function (such as Mean Squared Error, MSE), which can be simply expressed by the following formula:

[0090]

[0091] The model prediction can be expressed as:

[0092]

[0093] where f is the activation function, W L and b L are the weights and biases of the output layer, and a L-1 is the output of the previous layer.

[0094] It should be noted that in this application, no improvement is made to the principles of the above machine learning algorithms. Their basic principles of use are the same as those of the corresponding machine learning algorithms in the prior art. The specific principles and parameters not explained can be understood by referring to the machine learning algorithm models in the prior art, and no more detailed introduction will be given here.

[0095] Then training is carried out.

[0096] Specifically, input the prepared training dataset into the target machine learning algorithm for training, enabling the model to learn the mapping relationship from the spectral values of spectral bands to target parameters such as water quality parameters. During the training process, the model optimizes its internal parameters according to the target function (such as minimizing the loss function). In practical applications, the corresponding models can be obtained by training through all the above-mentioned target machine learning algorithms respectively, and then select the model with the best performance as the final inversion model, or use multiple models suitable for different target parameters as the final inversion models, or only use one of them for training to directly obtain the inversion model.

[0097] Finally, conduct model verification and evaluation.

[0098] Specifically, for example, after training through all the above-mentioned target machine learning algorithms respectively to obtain the corresponding models, use the models to invert the parameters of the inversion verification points (i.e., the sampling points divided into the validation set above) (in practical applications, the validation dataset can also be generated through the data generation method in the above training dataset for this validation step), and verify and evaluate the performance of the trained models through the predicted values and the measured values of the original sampling points, so as to select the most suitable model as the inversion model. Specifically, the goodness of fit of the model can be verified and measured through the coefficient of determination of the model, the prediction error can be verified and measured through the root mean square error, and the accuracy of the model prediction can be verified and measured through the mean absolute error, etc.

[0099] For example, conduct verification and evaluation based on the following formula:

[0100]

[0101] where, y i is the actual value, is the predicted value, is the mean of the actual values. R 2 is the coefficient of determination, which is used to measure the goodness of fit of the model. RMSE is the root mean square error, which is used to measure the prediction error. MAE is the mean absolute error, which is used to measure the accuracy of the model prediction. It should be noted that the coefficient of determination, root mean square error, and mean absolute error mentioned above are all indicators in the prior art, and this application does not redefine them. The specific application principles and parameters not explained can be understood by referring to the corresponding applications in the prior art, and no more detailed description will be provided here.

[0102] Furthermore, conduct remote sensing image inversion.

[0103] Specifically, invert remote sensing images such as hyperspectral images through the inversion model to obtain inversion results such as water quality parameter surface distribution data.

[0104] The remote sensing image inversion method based on spatial interpolation and machine learning provided by this application effectively reduces the requirement for the amount of original data in remote sensing image inversion and reduces the sampling workload by combining spatial interpolation and a small number of sampling points. At the same time, the screening and optimization of sensitive bands improve the applicability and prediction accuracy of the model, enabling high-precision inversion analysis of water quality, soil quality, etc. in complex terrain environments, and providing a more economical and effective solution for the application of remote sensing images, especially hyperspectral remote sensing images, in environmental monitoring.

[0105] Based on the same inventive concept, this application also provides a remote sensing image inversion device based on spatial interpolation and machine learning. Figure 3 It is a schematic structural diagram of the remote sensing image inversion device based on spatial interpolation and machine learning provided by an embodiment of the present invention. As Figure 3 shown, the device includes:

[0106] An acquisition module 11, configured to select a first preset number of sampling points in the area to be measured and obtain the target information of the sampling points, where the target information includes target parameters and geographic coordinate information;

[0107] A spatial interpolation module 12, configured to generate the target information of a second preset number of random points through spatial interpolation technology based on the target information of the sampling points;

[0108] An extraction module 13, configured to extract the spectral values of each band at the sampling points and the random points in the remote sensing image.

[0109] An analysis module 14, configured to perform a correlation analysis on the spectral values and the target parameters to determine the optimal band combination.

[0110] A construction and application module 15, configured to construct a data set based on the spectral values and the target parameters of the optimal band combination, train and verify a target machine learning algorithm, and obtain an inversion model for determining the target parameters based on the spectral values; and perform inversion of the target parameters based on the inversion model and the remote sensing image to be inverted.

[0111] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0112] Based on the same inventive concept, this application also provides a remote sensing image inversion device based on spatial interpolation and machine learning. Figure 4 It is a schematic structural diagram of the remote sensing image inversion device based on spatial interpolation and machine learning provided by an embodiment of the present invention. As Figure 4As shown in the figure, the device includes: a processor 21 and a memory 22, and the processor 21 is connected to the memory 22. Among them, the processor 21 is used to call and execute the program stored in the memory 22; the memory 22 is used to store the program, and this program is at least used to execute the remote sensing image inversion method based on spatial interpolation and machine learning in the above embodiments.

[0113] For the specific implementation of the remote sensing image inversion device based on spatial interpolation and machine learning provided by the embodiments of the present application, reference can be made to the implementation manners of the remote sensing image inversion method based on spatial interpolation and machine learning in any of the above embodiments, and details are not described herein again.

[0114] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be seen in the same or similar content in other embodiments.

[0115] It should be noted that in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise stated, the meaning of "a plurality of" refers to at least two.

[0116] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of the code of executable instructions including one or more steps for implementing a specific logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the functions involved, rather than in the order shown or discussed, and this should be understood by those skilled in the technical field of the embodiments of the present invention.

[0117] It should be understood that each part of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple 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, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0118] Those of ordinary skill in the technical field of the present invention can understand that all or part of the steps carried by the methods in the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0119] In addition, each functional unit in various embodiments of the present invention may be integrated into a processing module, may exist separately physically for each unit, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0120] The above-mentioned storage medium may be a read-only memory, a magnetic disk, an optical disc, etc.

[0121] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0122] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A remote sensing image inversion method based on spatial interpolation and machine learning, characterized in that: include: Selecting a first preset number of sampling points in the area to be measured, and acquiring target information of the sampling points, wherein the target information includes target parameters and geographic coordinate information; Based on the target information of the sampling points, generating target information of a second preset number of random points by using a spatial interpolation technique; Extracting spectral values ​​of each band at the sampling point and the random point in the remote sensing image; Performing correlation analysis based on the spectral value and the target parameter to determine the optimal band combination; Based on the spectral value of the optimal band combination and the target parameter, a data set is constructed, and a target machine learning algorithm is trained and verified to obtain an inversion model for determining the target parameter based on the spectral value; Based on the inversion model and the remote sensing image to be inverted, the target parameters are inverted.

2. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 1, characterized in that: The step of generating target information of a second preset number of random points based on the target information of the sampling points by using a spatial interpolation technique comprises: Based on the target information of the sampling points, generating surface distribution data of the target parameters by using a spatial interpolation technique; A second preset number of random points are generated by random sampling, and corresponding target parameters are extracted from the surface distribution data.

3. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 1, characterized in that: After extracting the spectral values ​​of each band at the sampling point and the random point in the remote sensing image, the method further includes: standardizing and normalizing the extracted spectral values.

4. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 1, characterized in that: After extracting the spectral values ​​of each band at the sampling points and the random points in the remote sensing image, the method further includes: dividing two-thirds of the sampling points and all the random points into a training set; and dividing one-third of the sampling points into a verification set.

5. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 4, characterized in that: The performing correlation analysis based on the spectral value and the target parameter to determine the optimal band combination includes: For the sampling points and random points divided into the training set, correlation analysis is performed between the target parameter and the spectral values ​​of each band of the corresponding sampling points or random points to determine sensitive bands with high correlation; Combining the sensitive bands to obtain a sensitive band combination; A correlation analysis is performed on the sensitive band combination and the target parameters of the corresponding sampling points or random points, and a sensitive band combination with a high correlation is determined as the optimal band combination.

6. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 5, characterized in that: The method of constructing a data set based on the spectral value of the optimal band combination and the target parameter, training and verifying the target machine learning algorithm, and obtaining an inversion model for determining the target parameter based on the spectral value includes: Constructing a training data set based on the spectral values ​​and target parameters of the optimal band combination corresponding to the sampling points or random points divided into the training set; The target machine learning algorithm is trained based on the training data set to obtain the inversion model.

7. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 1, characterized in that: The target machine learning algorithm includes at least one of a support vector machine, a random forest, and a neural network.

8. The remote sensing image inversion method based on spatial interpolation and machine learning according to claim 1, characterized in that: The target parameters include water quality parameters, and the water quality parameters include pH value, dissolved oxygen and total phosphorus.

9. A remote sensing image inversion device based on spatial interpolation and machine learning, characterized in that: include: A collection module, used to select a first preset number of sampling points in the area to be measured, and obtain target information of the sampling points, wherein the target information includes target parameters and geographic coordinate information; A spatial interpolation module, used to generate target information of a second preset number of random points by a spatial interpolation technique based on the target information of the sampling points; An extraction module, used for extracting the spectral values ​​of each band at the sampling point and the random point in the remote sensing image; An analysis module, used for performing correlation analysis based on the spectral value and the target parameter to determine an optimal band combination; An application module is constructed to construct a data set based on the spectral values ​​of the optimal band combination and the target parameters, train and verify the target machine learning algorithm, and obtain an inversion model for determining the target parameters based on the spectral values; and invert the target parameters based on the inversion model and the remote sensing image to be inverted.

10. A remote sensing image inversion device based on spatial interpolation and machine learning, characterized in that: The invention comprises a processor and a memory, wherein the processor is connected to the memory: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the remote sensing image inversion method based on spatial interpolation and machine learning as described in any one of claims 1-8.

Images