Multi-scale heterogenous remote sensing data collaborative water quality inversion method, computer equipment and medium

Abstract

The invention discloses a multi-scale heterogenous remote sensing data collaborative water quality inversion method, computer equipment and a medium, and belongs to the technical field of water quality remote sensing monitoring. S3, an inversion model is constructed, and the step S3 comprises the steps of S31, wave band combination optimization and correlation analysis; s32, constructing a multi-scale heterogeneous remote sensing data collaborative fusion inversion model; and S33, multi-source data collaborative wave band selection and model determination. According to the method, high-frequency and high-spatial-resolution water quality inversion is realized, the spatial-temporal resolution and precision are improved, a more comprehensive and accurate water quality inversion model is constructed, the robustness of the model in a turbid water body and an eutrophicated water body is remarkably improved, the adaptability to a complex water body is enhanced, and the method is suitable for large-scale popularization and application. An efficient and economical technical means is provided for water quality inversion in a complex water body environment, and the method has remarkable low cost and universality.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of water quality remote sensing monitoring, and in particular relates to a multi-scale heterogeneous remote sensing data collaborative water quality inversion method, a computer device and a medium. BACKGROUND

[0002] The spectral characteristics of water bodies are determined by the light radiation scattering properties and absorption of various optically active substances in the water. When sunlight enters the water surface, part of it is absorbed by the atmosphere, part is directly reflected from the water surface, and part enters the water bottom after being absorbed and scattered by the substances in the water. The sensor receives the light absorbed and scattered by the substances in the water, and thus determines the type of the substances according to the spectral characteristics in a specific wavelength range.

[0003] The commonly used water quality remote sensing inversion models at present include empirical models, semi-empirical / semi-analytical models, analytical models, machine learning models and deep learning models, etc. The correlation between remote sensing data and water quality parameters is established to estimate the concentration of water quality parameters.

[0004] Due to the complex optical properties of inland water bodies, the spatial and temporal distribution and content of water quality indicators of different geographical regions or the same water body will change, various algorithms and models have their own advantages and disadvantages and applicability, and different sensor data have different time, space and spectral resolution, so that even in the same area using the same method, the estimation results will have certain differences. The current situation of using a single data source mainly depends on satellite data of different resolutions. Medium resolution data is widely used for regional water quality monitoring, and multi-spectral bands are used to invert water quality parameters; low resolution data provides high temporal resolution and is suitable for dynamic tracking of large-scale water bodies; high resolution data is used to improve the monitoring capability of small water bodies in recent years, but is limited by cloud cover and revisit period. However, single data source has many defects: (1) contradiction between time and space resolution, low resolution data is time-intensive but spatially vague, and high resolution data is spatially detailed but has poor time continuity; (2) spectral limitation, fixed band setting limits the comprehensive coverage of water quality parameters; (3) data blind area and poor consistency, cloud interference leads to data missing, and radiation difference between different sensors affects long-term analysis; (4) insufficient model generalization ability, unstable performance in cross-regional and cross-seasonal scenarios.

[0005] Based on the above reasons, the existing water quality inversion method has the problems of insufficient time and space resolution caused by single data source, limited model applicability, and dependence on hyperspectral data in complex water body environment: 1. The time and space resolution of single data source is insufficient, which is difficult to fully reflect the optical properties of water bodies.

[0006] Existing water quality inversion methods are usually based on a single data source (such as medium resolution or high resolution remote sensing images) for inversion, and it is difficult to simultaneously consider high temporal resolution and high spatial resolution. For example, although the medium resolution image has high temporal resolution, the spatial resolution is limited, and it is difficult to depict the small spatial changes of the water body; and although the high resolution image has high spatial resolution, the temporal resolution is low, and it is difficult to capture the rapid changes of the optical characteristics of the water body. The limitation of such single data source leads to the deficiency of the inversion results in time and space coverage and edge details.

[0007] 2. The model has limited applicability to complex water body environment.

[0008] For inland type II water body, the optical characteristics are complex and have significant temporal and spatial changes, and a single model is difficult to fully adapt to such complexity. Existing models (such as empirical models, machine learning models, etc.) are usually based on a single data source, which is difficult to capture more subtle changes in the optical characteristics of the water body, resulting in limited inversion accuracy.

[0009] 3. High dependence on hyperspectral remote sensing images, high data acquisition cost.

[0010] Although the hyperspectral remote sensing image has rich spectral information and can provide more fine water quality parameter inversion capability, its acquisition cost is high, and the data acquisition frequency is limited, which is difficult to meet the demand of large-scale and high-frequency water quality monitoring.

[0011] The information disclosed in this Background section is intended only to increase an understanding of the general background of the application, and is not admitted to be prior art against the present application. SUMMARY

[0012] The purpose of the present application is to solve the problems faced by the existing water quality inversion method, and to provide a multi-scale heterogeneous remote sensing data collaborative water quality inversion method, a computer device and a medium.

[0013] The first aspect of the present application provides a multi-scale heterogeneous remote sensing data collaborative water quality inversion method, which comprises the steps of: S2, data preprocessing; S3, inversion model construction, wherein, The step S3 comprises the steps of: S31, band combination optimization and correlation analysis; S32, construction of multi-scale heterogeneous remote sensing data collaborative fusion inversion model; S33, multi-source data collaborative band selection and model determination.

[0014] In an embodiment of the present application, the step S2 comprises: Radiometric calibration, atmospheric correction and orthorectification, wherein the geographical position registration of two scenes of medium and high resolution images is required, that is, by adjusting the spatial position difference between the images, it is ensured that they are accurately aligned in the same geographical reference frame.

[0015] In an embodiment of the present application, comprising: In the modeling, the reflectivity values are extracted on the corresponding satellite remote sensing images based on the measured point data, that is, according to the longitude and latitude coordinates of the sorted measured points, the points are spread on the two images to determine the pixel positions corresponding to the points, and all band values in the pixels are extracted respectively to obtain the medium and high resolution image reflectivity values, that is, to obtain the optical image bands.

[0016] In an embodiment of the present application, the step S31 comprises: The optical image bands are combined and optimized, and reasonable band combinations are expanded, including two-band, three-band and four-band combinations, the remote sensing image reflectivity under each band combination is extracted, and the correlation analysis is performed with the measured values of each index of the corresponding measured points, and the correlation is sorted according to the correlation size, and the N highest correlation combinations are selected for subsequent processing, wherein N is an integer and takes a value of 6-15.

[0017] In an embodiment of the present application, the correlation analysis uses Pearson correlation coefficient r to measure the correlation size, and for any combination, the calculation formula of r is as follows: , In the above formula, represents the Pearson correlation coefficient, represents the number of measured points, represents the i-th measured data, represents the band combination calculation value of the corresponding position of the i-th measured point, represents the mean value of the measured data, represents the mean value of the band combination calculation value of the corresponding position of the measured point.

[0018] In an embodiment of the present application, the step S32 comprises: The formula of the model is expressed as: , Wherein, X1 and X2 are the band combination calculation values of the medium resolution remote sensing image and the high resolution remote sensing image respectively; Y is the inversion result, that is, the target result of inversion; α and β are spectral sensitivity adjustment indexes, and satisfy α+β=1; γ is a characteristic normalization coefficient; λ is a gradient coupling factor; is a gradient operator, and x and y represent the horizontal coordinate and the vertical coordinate in the horizontal coordinate system in the horizontal plane, For each of the selected N highly coherent bands, i.e. the N combinations with the highest correlation, the final fitting model Y is obtained by referring to the above formula, based on the medium-high resolution band combination values X1 and X2 and the monitoring site measured data to obtain α, β, γ, λ, From the N combinations with the highest correlation, N sets of parameters α, β, γ, λ are obtained, i.e. N models are obtained.

[0019] In an embodiment of the present application, the step S33 comprises: The accuracy of the model corresponding to different sets of parameters is verified by using the monitoring point measured value and the corresponding remote sensing image reflectance combination value, from the root mean square error RMSE, the mean absolute error MAPE and the determination coefficient R 2 After comprehensive consideration, the optimal parameters X1 and X2 are obtained, wherein, , , In the above two formulas, Y represents the result calculated by the model, is the corresponding site measured value, and m is 20% of the number of measured points.

[0020] In an embodiment of the present application, N=10.

[0021] The second aspect of the present application provides a computer device, which comprises a memory, a first processor and a first computer program stored on the memory and executable on the first processor, wherein the first computer program is executed by the first processor to implement the multiscale heterogeneous remote sensing data collaborative water quality inversion method described above.

[0022] The third aspect of the present application provides a computer readable storage medium for storing a second computer program, wherein the second computer program is executable by at least one second processor to make the at least one second processor execute the multiscale heterogeneous remote sensing data collaborative water quality inversion method described above.

[0023] Compared with the prior art, the present application achieves the following technical effects: 1. By comprehensively using medium resolution and high resolution remote sensing images, the spatiotemporal complementarity and information complementarity of the two are fully utilized, the spectral bands of multi-source data are jointly extracted, the water quality inversion with high frequency and high spatial resolution is realized, the spatiotemporal resolution and accuracy are improved, and a more comprehensive and accurate water quality inversion model is constructed; 2. Based on the multispectral feature fusion of sensitive weight optimization, the robustness of the model in turbid water and eutrophic water is significantly improved, and the adaptability to complex water bodies is enhanced; 3、Not only weaken the limitations of single data source, but also reduce the dependence on hyperspectral data, and replace hyperspectral image with medium-high resolution data to reduce monitoring cost and support large-scale high-frequency application, but also can significantly improve the applicability, stability and spatio-temporal resolution of the model, provide an efficient and economical technical means for water quality inversion in complex water environment, and has significant low cost and universality. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is a flow chart of a multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] Unless otherwise explicitly indicated, throughout the specification and claims, the term "comprise" or variations such as "comprises" or "comprising" will be understood to imply the inclusion of a stated element or group of elements but not the exclusion of any other element or group of elements.

[0026] The technical solutions of the present application are described below through specific examples. It should be understood that the one or more steps mentioned in the present application do not exclude other methods and steps before and after the combination steps, or other methods and steps can be inserted between these explicitly mentioned steps. It should also be understood that these examples are only used to illustrate the present application and not to limit the scope of the present application. Unless otherwise specified, the numbering of each method step is only for the purpose of identifying each method step, and is not limited to the arrangement order of each method or the scope of the implementation of the present application. Changes or adjustments of the relative relationship, without substantial technical content changes, can also be considered as the implementation scope of the present application.

[0027] The raw materials and instruments used in the examples are not specifically limited in source, and can be purchased on the market or prepared according to conventional methods well known to those skilled in the art.

[0028] As shown in Figure 1 , taking the inversion process of the suspended matter concentration index in Taihu Lake in Jiangsu Province as an example, the multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to the preferred embodiment of the present application includes the following steps.

[0029] S1, data acquisition.

[0030] Obtain high-resolution remote sensing image data and water quality index measured data of the water area to be monitored, obtain two kinds of resolution remote sensing image data obtained by a wide field of view (WFV, Wide Field of View, 16m resolution, m represents meter) camera and a high-resolution (PMS, Panchromatic / Multispectral Sensor, 2m resolution) camera respectively on the same day, and 10-20 measured point data, preferably 15 measured point data, including suspended solids concentration and coordinate information. Among them, the WFV camera obtained is a medium-resolution remote sensing image data, and the PMS camera obtained is a high-resolution remote sensing image data, which can be downloaded from the land observation system, and the specific download address is: Land observation satellite data service. The remote sensing image data is divided according to the scene, and one scene image is a compressed package of original data, the unit of the original data is digital (Digital Number, DN) value, it should be noted that the DN value is the brightness value of the remote sensing image element, which records the gray value of the ground object, and has no unit, which is an integer value, the value is related to the radiation resolution of the sensor, the emissivity of the ground object, the atmospheric transmittance and the scattering rate. The suspended solids concentration can be obtained by collecting the water sample at the corresponding point from the measured point, such as a measured point in Taihu Lake, and obtaining the concentration through laboratory test, and the coordinate information can be the position information of the collected water sample obtained by global positioning system (Global Positioning System, GPS) positioning, including longitude and latitude.

[0031] S2, data preprocessing.

[0032] In order to meet the high-precision requirement of water quality inversion modeling, the data obtained in step S1 needs to be further preprocessed to convert the image DN value in the original data into reflectivity value. This step mainly includes radiation calibration, atmospheric correction and orthographic correction, and since multi-scale heterogeneous remote sensing data is used, the geographical position of the two scenes of medium and high resolution images needs to be registered, that is, by adjusting the spatial position difference between the images, it is ensured that they are accurately aligned under the same geographical reference frame, so as to meet the requirement of the same point of medium and high resolution remote sensing images.

[0033] S3, construction of inversion model.

[0034] The modeling process is mainly constructed around the principle of "infinite fitting of remote sensing spectral values to measured data". During modeling, the reflectivity value is extracted from the corresponding satellite remote sensing image based on the measured point data, that is, according to the longitude and latitude coordinates of the measured point, the corresponding pixel position of the point is determined on the medium and high resolution remote sensing image, and all the band values in the pixel are extracted, so that the reflectivity value of the medium and high resolution image is obtained, that is, the optical image band is obtained. Among them, 80% of the extracted data is used for model inversion construction, and the remaining 20% is used for accuracy verification. The inversion model construction principles of different water area indicators are consistent, and the corresponding band combination modes are different. The construction includes the following steps.

[0035] S31, band combination optimization and correlation analysis.

[0036] The obtained optical image bands are combined and optimized, reasonable band combinations are expanded and explored, including two-band, three-band and four-band combination modes, the reflectivity of the remote sensing image under each band combination is extracted, and the correlation analysis is performed with the measured values of each index of the corresponding measured points respectively, and the correlation is sorted according to the size, and the N highest correlation combinations are selected for subsequent processing. N is an integer and takes a value of 6-15, preferably 10.

[0037] The correlation analysis uses Pearson correlation coefficient r to measure the correlation size, the larger r is, the greater the correlation is, and for any band combination mode, the calculation formula of r is as follows: , Wherein represents the Pearson correlation coefficient, represents the number of measured points, represents the i-th measured data, represents the band combination calculation value of the i-th measured point corresponding position, represents the mean value of the measured data, represents the mean value of the band combination calculation value of the measured point corresponding position. The band combination calculation value is the result obtained by bringing the reflectivity of each band into the combination mode.

[0038] S32, constructing a multi-scale heterogeneous remote sensing data collaborative fusion inversion model.

[0039] The present application proposes a multi-scale heterogeneous remote sensing data fusion inversion model based on spectral-spatial collaborative constraint, which realizes the information complementarity expression of multi-source data, and the formula of the model is: , wherein, X1 and X2 are respectively the band combination calculation values of the medium resolution remote sensing image and the high resolution remote sensing image; Y is the inversion result, i.e. the inversion target result, such as the inversion result of the suspended substance concentration of the Taihu Lake, at this time, the data unit is milligrams / liter; a and b are spectral sensitivity adjustment indexes, and satisfy a+b=1; g is a characteristic normalization coefficient and has no unit; l is a gradient coupling factor (used for controlling the interaction intensity of multi-scale spatial characteristics) and has no unit; is a gradient operator, and x and y represent the horizontal coordinate and the vertical coordinate in the horizontal coordinate system.

[0040] In the above formula, is a spectral complement term, which introduces a power-law ratio form Breaking the limitation of traditional linear combination, the non-linear coupling of spectral response is realized through the exponential parameter; is a spatial gradient interaction term, which establishes a gradient tensor product across resolutions Captures the synergistic effect of multi-scale spatial characteristics.

[0041] Then, for each of the selected N highly coherent bands, i.e. the N combinations with the highest correlation, the final fitting model Y is obtained by referring to the above formula, based on the medium-high resolution band combination values (i.e. X1 and X2) and the monitoring station measured data to obtain a, b, g, l. From the N combinations with the highest correlation, N sets of parameters a, b, g, l are obtained, i.e. N fitting models Y are obtained.

[0042] S33, multi-source data collaborative band selection and model determination.

[0043] The remaining 20% of the monitoring point measured values and the corresponding remote sensing image reflectance combination values are used to verify the accuracy of the inversion model corresponding to different sets of the above parameters, from the root mean square error (RMSE), the mean absolute error (MAPE) and the determination coefficient R 2 After comprehensive consideration, the best band combination method under medium-high resolution is selected, i.e. the best parameters and X1 and X2 are obtained. Among them, , , How to calculate R 2 is known to the skilled person and will not be repeated here. In the above two formulas, represents the result calculated by the model, is the corresponding station measured value, and m is 20% of the number of measured points. The smaller the root mean square error (RMSE) and the mean absolute error (MAPE) values, the higher the model inversion accuracy. The determination coefficient R 2The determination degree of the independent variable in the model to the dependent variable, that is, the fitting degree between the predicted value of the model and the actual observation value, and the value is between 0 and 1, and the closer to 1 indicates that the fitting effect of the model is better.

[0044] Example: X1 can be taken as the reflectance combination value of remote sensing image taken by a 16m resolution high-resolution camera, X2 can be taken as the reflectance combination value of remote sensing image taken by a 2m resolution high-resolution camera, and the inversion model formula Y is obtained by the multiscale heterogeneous remote sensing data collaborative fusion inversion model formula, that is, the reflectance value under the medium-high resolution of Taihu Lake and the measured results of the measured point position of the suspended matter concentration are brought into operation, after band combination optimization and correlation analysis, the best X1 and X2 are screened out, and finally the inversion model Y of the suspended matter concentration of Taihu Lake is obtained, and b1 is the blue spectrum reflectance value, b2 is the green spectrum reflectance value, b3 is the red spectrum reflectance value, and b4 is the near-infrared spectrum reflectance value. The following results can be obtained: For WFV, , For PMS, , .

[0045] In the above formula, is the model finally determined in this example.

[0046] S34, result and precision evaluation.

[0047] From the root mean square error (RMSE), mean absolute error (MAPE) and determination coefficient R 2 The final inversion model is comprehensively evaluated to determine whether the result meets the requirements, if not, the spectral sensitivity adjustment index is adjusted again, and γ and λ are adjusted to determine the model again, if it meets the requirements, the model is applied to the remote sensing image to obtain the inversion result map of the monitoring index.

[0048] The application also provides a computer device, which comprises a memory, a first processor and a first computer program stored on the memory and executable on the first processor, and the first computer program is executed by the first processor to realize the multiscale heterogeneous remote sensing data collaborative water quality inversion method.

[0049] The application also provides a computer readable storage medium for storing a second computer program, and the second computer program can be executed by at least one second processor to make the at least one second processor execute the multiscale heterogeneous remote sensing data collaborative water quality inversion method.

[0050] The present application solves the deficiencies of the existing water quality inversion method in terms of spatial and temporal resolution, applicability to complex water bodies, and data dependency, by using multi-scale heterogeneous remote sensing data inversion technology, combined with optimal band combination and sensitive weight optimization. 1. A multi-scale heterogeneous remote sensing data collaborative inversion method is proposed, which fully utilizes the complementarity of medium resolution and high resolution remote sensing images in time and space, extracts multi-scale features based on medium and high resolution image collaboration technology, and enhances the spatio-temporal consistency and universality of the model.

[0051] 2. A multi-scale heterogeneous remote sensing data collaborative band selection and dynamic sensitive weight distribution mechanism is proposed, which breaks through the spectral, spatial and temporal resolution limitations of a single data source through data-driven optimization, significantly improving the accuracy and universality of water quality inversion. According to the spectral characteristics of medium resolution and high resolution remote sensing data, the optimal band combination sensitive to water quality parameters is selected; and a nonlinear interactive sensitive weight distribution formula is proposed, which adaptively adjusts the contribution proportion of medium and high resolution data through dynamic parameter ω.

[0052] 3. By selecting the optimal band combination of medium resolution and high resolution images, the key spectral features of water bodies are extracted, without relying on hyperspectral data, significantly reducing the data acquisition cost, and improving the convenience of data acquisition.

[0053] 4. The present application further embodies the spectral feature transferable mechanism and multi-index coupled modeling framework, making the inversion model not only applicable to suspended solids concentration, but also extendable to dynamic monitoring of key water quality parameters such as total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH3-N), permanganate index (CODMn), and chlorophyll a (Chl-a).

[0054] In summary, the present application uses medium resolution and high resolution remote sensing images collaboratively, fully utilizes the high temporal resolution of medium resolution images and the high spatial resolution of high resolution images, constructs a multi-scale heterogeneous remote sensing data collaborative inversion model, simultaneously meets the needs of high frequency monitoring and high spatial detail expression, significantly improves the spatial and temporal resolution of water quality inversion; selects the optimal band combination of medium resolution and high resolution images, and combines the proportion of the two combination results selected by the sensitive weight, to construct an exponential regression model; fully utilizes the spectral information of multi-scale heterogeneous remote sensing data, more comprehensively extracts the optical characteristics of water bodies, thereby improving the applicability and inversion accuracy of the model in complex water body environment. At the same time, the present application does not rely on hyperspectral remote sensing images, but selects the optimal band combination of medium resolution and high resolution images to extract the key spectral features of water bodies. The present application not only reduces the data acquisition cost, but also improves the convenience of data acquisition, and is suitable for large-scale water quality monitoring.

[0055] The foregoing description of specific exemplary embodiments of the application has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the application to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teaching. It is intended that the scope of the application be limited not with this detailed description, but rather by the claims appended hereto.

Claims

1. A multi-scale heterogeneous remote sensing data collaborative water quality inversion method, characterized in that, Including the following steps: S2, Data Preprocessing; S3, Inversion Model Construction in, Step S3 includes the following steps: S31. Band combination optimization and correlation analysis; S32. Construct a multi-scale heterogeneous remote sensing data collaborative fusion and inversion model; S33. Multi-source data collaborative band selection and model determination.

2. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 1, characterized in that, Step S2 includes: Radiometric calibration, atmospheric correction, and orthorectification are required. Among these, it is necessary to perform geographic registration on two images of medium and high resolution, that is, to ensure that they are accurately aligned under the same geographic reference frame by adjusting the spatial differences between the images.

3. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 2, characterized in that, include: During modeling, reflectance values ​​are extracted from the corresponding satellite remote sensing images based on the measured point data. That is, according to the latitude and longitude coordinates of the measured points, the points are plotted on the two images to determine the pixel positions corresponding to the points, and all band values ​​in the pixels are extracted to obtain the medium-to-high resolution image reflectance values, i.e., the optical image bands.

4. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 3, characterized in that, Step S31 includes: The optical image bands are combined and optimized to explore reasonable band combinations, including two-band, three-band, and four-band combinations. The reflectance of remote sensing images under each band combination is extracted and correlated with the measured values ​​of each index at the corresponding measured points. The images are sorted according to the magnitude of the correlation, and the N combinations with the highest correlation are selected for subsequent processing, where N is an integer and ranges from 6 to 15.

5. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 4, characterized in that, The correlation analysis uses the Pearson correlation coefficient r to measure the magnitude of the correlation. For any of the aforementioned combinations, the formula for calculating r is as follows: ， In the above formula, This represents the Pearson correlation coefficient. Indicates the number of measured points. This represents the i-th measured data point. This represents the calculated value of the band combination corresponding to the i-th measured point. This represents the mean of the measured data. This represents the mean of the calculated values ​​of the band combination at the location corresponding to the measured point.

6. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 5, characterized in that, Step S32 includes: The model's formula is expressed as follows: ， Where X1 and X2 are the calculated values ​​of the band combination of medium-resolution and high-resolution remote sensing images, respectively; Y is the inversion result, i.e. the target result of the inversion; α and β are spectral sensitivity adjustment indices, and satisfy α+β=1; γ is the feature normalization coefficient; λ is the gradient coupling factor. This is the gradient operator, where x and y represent the x-coordinate and y-coordinate in a Cartesian coordinate system on the horizontal plane. For each of the N selected high-coherence bands, i.e., the N combinations with the highest correlation, the above formula is used as a reference. Based on the band combination values ​​X1 and X2 at medium and high resolution and the measured data from the monitoring stations, α, β, γ, and λ are calculated to obtain the final fitting model Y. From the N combinations with the highest correlation, N sets of parameters α, β, γ, and λ are obtained, thus yielding N models.

7. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to claim 6, characterized in that, Step S33 includes: By combining measured values ​​from monitoring points with corresponding remote sensing image reflectance values, the accuracy of the models corresponding to different sets of parameters is verified, using root mean square error (RMSE), mean absolute error (MAPE), and coefficient of determination (R²). 2 Taking all factors into consideration, the optimal parameters and X1 and X2 are obtained, where, ， ， In the two equations above, This represents the result calculated using the model. The values ​​are the actual measured values ​​for the corresponding stations, where m is 20% of the number of measured points.

8. The multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to any one of claims 5-7, characterized in that, N=10。 9. A computer device, characterized in that, The system includes a memory, a first processor, and a first computer program stored in the memory and executable on the first processor. When the first computer program is executed by the first processor, it implements the multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a second computer program, which can be executed by at least one second processor to enable the at least one second processor to perform the multi-scale heterogeneous remote sensing data collaborative water quality inversion method according to any one of claims 1-8.

Images