Machine learning ammonia nitrogen inversion method and device, terminal and storage medium

Abstract

The invention discloses a machine learning ammonia nitrogen inversion method and device, a terminal and a storage medium, and the method comprises the steps: obtaining actually-measured ammonia nitrogen data of a plurality of water quality samples and remote sensing image data, and carrying out the time matching, so as to extract the wave band feature information of each actually-measured point in a remote sensing image according to the geographical coordinate information of the water quality samples; performing wave band combination on the obtained actually measured wave band characteristic information of each point, sorting and filtering characteristic factors by using a Pearson correlation coefficient, and determining a plurality of candidate wave band characteristics; performing dimension reduction operation on the plurality of candidate wave band characteristics by using a principal component analysis method, and constructing a machine learning ammonia nitrogen remote sensing inversion model with optimized wave band characteristics by using the wave band characteristics after dimension reduction as independent variables and water quality parameter ammonia nitrogen as dependent variables; according to the method, rapid and accurate remote sensing inversion can be carried out on the ammonia nitrogen concentration of a large-range river network water system, and the time-consuming sampling and laboratory analysis process in a traditional water quality monitoring method is overcome.

Description

Technical Field

[0001] The present invention relates to water quality monitoring technology, and in particular to a machine learning ammonia nitrogen inversion method, device, terminal and storage medium. Background Art

[0002] Ammonia nitrogen (NH 3 -N refers to free ammonia (NH 3 ) and ammonium ions (NH 4 + ) in the form of nitrogen, which is one of the important indicators for measuring water quality categories and water blackness and odor. Accurate monitoring of ammonia nitrogen in urban rivers is an important step in effectively managing and controlling water pollution.

[0003] At present, water quality monitoring of urban river networks is still mainly based on on-site sampling and measurement, which is not only low in cost-effectiveness, but also difficult to fully reflect the overall status of water quality in a large area. With the continuous development of technology, multi-source remote sensing has been more widely used in efficient monitoring of water quality and color, and water environment protection and management due to its unique advantages such as fast speed, large-scale synchronization, and periodic repetition.

[0004] Urban rivers are fluid, narrow and long, and are polluted by tributaries along the way. Therefore, river water quality remote sensing is more complicated than that of marine water bodies and inland lakes. At present, domestic and foreign research on water quality pollution in urban river networks is basically concentrated on on-site tracking investigations of local river sections and key sections, causes of pollution, analysis of basic water quality characteristics, and treatment methods and plans. There are few related studies on remote sensing inversion of large-scale river networks and evaluation and analysis of treatment effectiveness.

[0005] For the inversion of inland water quality, there are few studies on remote sensing monitoring of non-optically active water quality parameters in water bodies, and the quantitative inversion model suitable for non-optically active water quality parameters in large-scale urban river networks needs further research.

[0006] Regarding the application of high-resolution remote sensing satellites in urban river network water quality monitoring, the core of the research is how to efficiently use high-resolution remote sensing satellites to achieve high-precision quantitative inversion of water quality parameters for large-scale urban river networks. So far, the problem of unclear characteristic bands of non-optically active water quality parameters and insufficient inversion accuracy still exists. Summary of the invention

[0007] The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art and to provide a machine learning ammonia nitrogen inversion method, device, terminal and storage medium to solve the problem of insufficient applicability and accuracy of remote sensing inversion of large-scale river networks and water systems.

[0008] To achieve the above object, the technical solution of the present invention is:

[0009] In a first aspect, the present invention provides a machine learning ammonia nitrogen inversion method, comprising: the method comprising:

[0010] Obtain the measured ammonia nitrogen data of several water quality samples and remote sensing image data for time matching, so as to extract the band characteristic information of each measured point in the remote sensing image according to the geographic coordinate information of the water quality samples;

[0011] The obtained band characteristic information of each measured point is combined, and the characteristic factors are sorted and filtered using the Pearson correlation coefficient to determine a number of candidate band characteristics; the principal component analysis method is used to perform a dimension reduction operation on the several candidate band characteristics to obtain the band characteristics after dimension reduction;

[0012] Using the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as the dependent variable, a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics was constructed;

[0013] The machine learning ammonia nitrogen remote sensing inversion model is used to perform remote sensing inversion of ammonia nitrogen concentration in river networks.

[0014] Optionally, before using the machine learning ammonia nitrogen remote sensing inversion model to perform remote sensing inversion of ammonia nitrogen concentration in a river network, it also includes: performing accuracy testing on the machine learning ammonia nitrogen remote sensing inversion model.

[0015] Optionally, the accuracy test of the machine learning ammonia nitrogen remote sensing inversion model includes:

[0016] Using the machine learning ammonia nitrogen remote sensing inversion model, the measured values ​​of the detection points are compared with the model prediction values, the determination coefficient, root mean square error, and mean absolute error of each detection point are calculated, and the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model is verified;

[0017] The machine learning ammonia nitrogen remote sensing inversion model is used to invert the ammonia nitrogen concentration in the water body in other time periods in the target area, and the determination coefficient, root mean square error, and mean absolute error of each detection point are calculated to verify the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model.

[0018] Optionally, the obtained measured band feature information of each point is combined, and the feature factors are sorted and filtered using the Pearson correlation coefficient to determine a number of candidate band features; and the principal component analysis method is used to perform a dimension reduction operation on the several candidate band features to obtain the band features after dimension reduction, including:

[0019] Construct multiple band transformation forms such as single band, band sum, difference, ratio and square root transformation;

[0020] The Pearson correlation coefficient is used to sort and filter the characteristic factors, and the band or band combination that is most closely related to the parameter concentration of the measured points in each area is selected to determine m candidate features, which are recorded as v i,i＝1,2,…,m ,v i =n×1 vector, the calculation formula is shown in formula (1):

[0021]

[0022] In the formula, r xy is the correlation coefficient, n is the number of samples, x is the water quality inversion factor, and y is the water quality parameter ammonia nitrogen;

[0023] The principal component analysis method is used to reduce the dimension of m candidate features to k, denoted as w i,i＝1,2,…,k ,w i =n×1 vector.

[0024] Optionally, the band characteristics after dimensionality reduction are used as independent variables, and the water quality parameter ammonia nitrogen is used as a dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics, including:

[0025] After dimension reduction, the optimized feature matrix w n×k =[w 1 w 2 …w k ] as input, and the measured ammonia nitrogen data of water quality samples as output. Combined with the extreme gradient boosting model, a machine learning ammonia nitrogen remote sensing inversion model with band feature optimization is constructed.

[0026] Optionally, the acquisition of ammonia nitrogen data measured from a plurality of water quality samples and remote sensing image data for time matching, so as to extract band characteristic information of each point measured in the remote sensing image according to geographic coordinate information of the water quality samples, includes:

[0027] Determine whether the data level of the remote sensing image data corresponding to the sentinel image is the specified level. If so, determine that the remote sensing image data of the specified level has completed the calibration correction operation. If not, perform atmospheric correction on the remote sensing image data of the non-specified level.

[0028] The average value of the satellite remote sensing reflectance within a 3×3 pixel window centered on the sampling point is extracted as the remote sensing reflectance of the sampling point.

[0029] In a second aspect, the present invention provides a machine learning ammonia nitrogen inversion device, comprising:

[0030] The water quality data collection and processing module is used to obtain the measured ammonia nitrogen data of several water quality samples and the remote sensing image data for time matching, so as to extract the band characteristic information of each point measured in the remote sensing image according to the geographic coordinate information of the water quality samples;

[0031] The band feature optimization module is used to combine the band feature information of each measured point, sort and filter the feature factors using the Pearson correlation coefficient, and determine a number of candidate band features; use the principal component analysis method to perform dimensionality reduction operation on the several candidate band features to obtain the band features after dimensionality reduction;

[0032] The inversion model construction module is used to use the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as the dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics;

[0033] The remote sensing inversion module is used to perform remote sensing inversion of the ammonia nitrogen concentration in the river network by using the machine learning ammonia nitrogen remote sensing inversion model.

[0034] In a third aspect, the present invention provides an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0035] Memory, used to store computer programs;

[0036] A processor is used to implement the steps of machine learning ammonia nitrogen inversion as described in any one of claims 1-6 when executing a program stored in a memory.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of machine learning ammonia nitrogen inversion as described in any one of the above items are as follows:

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention can quickly and accurately perform remote sensing inversion of ammonia nitrogen concentrations in a large range of river networks by constructing a machine learning ammonia nitrogen inversion model based on band feature optimization, overcoming the time-consuming sampling and laboratory analysis process in traditional water quality monitoring methods, and significantly improving the efficiency and real-time performance of water quality monitoring. The machine learning ammonia nitrogen inversion model optimizes the band features of the model input, improves the applicability and accuracy of the model, can provide more comprehensive and timely data support for water quality monitoring, and provide a scientific basis for relevant departments to formulate water environment governance and protection measures, thereby helping to improve the effectiveness of water environment management, promote the sustainable use of water resources and improve the regional ecological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0042] Figure 1 A flow chart of the machine learning ammonia nitrogen inversion method provided in an embodiment of the present application;

[0043] Figure 2 Flowchart for prediction of ammonia nitrogen remote sensing inversion model using machine learning;

[0044] Figure 3 A schematic diagram of the composition of a machine learning ammonia nitrogen inversion device provided in an embodiment of the present application;

[0045] Figure 4 A schematic diagram of the composition of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0047] To facilitate the understanding of the embodiments of the present application, further explanation will be given below in conjunction with the drawings and specific embodiments. The embodiments do not constitute a limitation on the embodiments of the present application.

[0048] See also Figure 1 As shown, the machine learning ammonia nitrogen inversion method provided in this embodiment mainly includes the following steps:

[0049] 110. Obtain the measured ammonia nitrogen data of several water quality samples and remote sensing image data for time matching, so as to extract the band characteristic information of each measured point in the remote sensing image according to the geographic coordinate information of the water quality samples;

[0050] 120. Combining the obtained measured band feature information of each point, and sorting and filtering the feature factors using the Pearson correlation coefficient to determine a number of candidate band features; performing dimensionality reduction operation on the several candidate band features using the principal component analysis method to obtain the band features after dimensionality reduction;

[0051] In this step, the band features of the remote sensing images were screened and sorted using the Pearson correlation coefficient, and the principal component analysis (PCA) was used for dimensionality reduction, which effectively reduced redundant features, optimized the selection of model input features, and significantly improved the efficiency and accuracy of the inversion model.

[0052] 130. Using the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as the dependent variable, a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics was constructed;

[0053] In this step, the band characteristics after dimensionality reduction are used as independent variables, and the water quality parameter ammonia nitrogen is used as the dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics, which can improve the applicability and accuracy of remote sensing inversion of large-scale river networks and water systems, and solve the problem of insufficient accuracy of remote sensing inversion of ammonia nitrogen in large-scale river networks and water systems in existing models.

[0054] 140. Use the machine learning ammonia nitrogen remote sensing inversion model to perform remote sensing inversion of ammonia nitrogen concentration in river networks.

[0055] In this step, the machine learning ammonia nitrogen remote sensing inversion model can be widely used in the remote sensing inversion of ammonia nitrogen concentration in large-area water bodies at the urban scale, providing an efficient and accurate technical means for water quality monitoring and water environment management.

[0056] It can be seen that this method can quickly and accurately perform remote sensing inversion of ammonia nitrogen concentrations in a large range of river networks by constructing a machine learning ammonia nitrogen inversion model based on band feature optimization, overcoming the time-consuming sampling and laboratory analysis process in traditional water quality monitoring methods, and significantly improving the efficiency and real-time performance of water quality monitoring. This model optimizes the band characteristics of the model input, improves the applicability and accuracy of the model, and can provide more comprehensive and timely data support for water quality monitoring, and provide a scientific basis for relevant departments to formulate water environment governance and protection measures, thereby helping to improve the effectiveness of water environment management, promote the sustainable use of water resources and improve the regional ecological environment.

[0057] As a preferred method for machine learning ammonia nitrogen inversion provided in this embodiment, Figure 2 As shown, before using the machine learning ammonia nitrogen remote sensing inversion model to perform remote sensing inversion of ammonia nitrogen concentration in the river network, it also includes:

[0058] The machine learning ammonia nitrogen remote sensing inversion model is subjected to an accuracy test to ensure the accuracy of the machine learning ammonia nitrogen remote sensing inversion model, including:

[0059] The machine learning ammonia nitrogen remote sensing inversion model was used to compare the measured values ​​of the detection points with the model predicted values, and the coefficient of determination (R2 ), root mean square error (RMSE), and mean absolute error (MAE) to verify the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model;

[0060] The machine learning ammonia nitrogen remote sensing inversion model is used to invert the ammonia nitrogen concentration in the water body in other time periods in the target area, and the determination coefficient, root mean square error, and mean absolute error of each detection point are calculated to verify the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model.

[0061] R 2 The closer it is to 1, the better the model fit is. RMSE is used to measure the deviation between the predicted value and the measured value. The smaller the value, the higher the prediction accuracy of the model. MAE represents the average value of the absolute value of the error between the predicted value and the measured value. The specific formula is as follows:

[0062]

[0063] Among them, y i and are the measured value and predicted value of ammonia nitrogen content respectively; For NH 3 -The average of the N content measurements; n is the number of samples.

[0064] In a specific embodiment, the step 120 includes:

[0065] Construct multiple band transformation forms such as single band, band sum, difference, ratio and square root transformation;

[0066] The Pearson correlation coefficient is used to sort and filter the characteristic factors, and the band or band combination that is most closely related to the parameter concentration of the measured points in each area is selected to determine m candidate features, which are recorded as v i,i＝1,2,…,m ,v i =n×1 vector, the calculation formula is shown in formula (1):

[0067]

[0068] In the formula, r xy is the correlation coefficient, n is the number of samples, x is the water quality inversion factor, and y is the water quality parameter ammonia nitrogen;

[0069] The principal component analysis method is used to reduce the dimension of m candidate features to k, denoted as w i,i＝1,2,…,k ,w i =n×1 vector.

[0070] Thus, through the above steps, redundant features are effectively reduced, the selection of model input features is optimized, and the efficiency and accuracy of the inversion model are significantly improved.

[0071] In a specific embodiment, the above step 130 includes:

[0072] After dimension reduction, the optimized feature matrix w n×k =[w 1 w 2 …w k ] as input and the water quality parameter ammonia nitrogen as output, and combined with the extreme gradient boosting model, a machine learning ammonia nitrogen remote sensing inversion model with band feature optimization is constructed.

[0073] In a specific embodiment, the above step 110 includes:

[0074] Determine whether the data level of the remote sensing image data corresponding to the sentinel image is the specified level. If so, determine that the remote sensing image data of the specified level has completed the calibration correction operation. If not, perform atmospheric correction on the remote sensing image data of the non-specified level. Since the signal from the water body (water-off-radiance) in the total radiation received by the remote sensing sensor is very small, more than 90% comes from atmospheric Rayleigh scattering, aerosol scattering and solar reflection. Atmospheric correction can eliminate these radiation errors caused by atmospheric influences and invert the true surface reflectivity of the ground object.

[0075] The average value of satellite remote sensing reflectance within a 3×3 pixel window centered on the sampling point is extracted as the remote sensing reflectance of the sampling point. The purpose of extracting the average value of satellite remote sensing reflectance within a 3×3 pixel window centered on the sampling point is to improve the representativeness and reliability of remote sensing reflectance values ​​through local averaging of spatial scales and reduce the influence of pixel positioning errors and accidental outliers.

[0076] Accordingly, if Figure 3 As shown, this embodiment also provides a machine learning ammonia nitrogen inversion device 300, comprising:

[0077] The water quality data collection and processing module 310 is used to obtain the measured ammonia nitrogen data of several water quality samples and the remote sensing image data for time matching, so as to extract the band characteristic information of each point measured in the remote sensing image according to the geographic coordinate information of the water quality samples;

[0078] The band feature optimization module 320 is used to combine the obtained measured band feature information of each point, and use the Pearson correlation coefficient to sort and filter the feature factors to determine a number of candidate band features; use the principal component analysis method to perform dimensionality reduction operation on the several candidate band features to obtain the band features after dimensionality reduction;

[0079] The inversion model construction module 330 is used to use the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as a dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics;

[0080] The remote sensing inversion module 340 is used to perform remote sensing inversion of the ammonia nitrogen concentration in the river network by using the machine learning ammonia nitrogen remote sensing inversion model.

[0081] As a preferred embodiment of the machine learning ammonia nitrogen inversion device of this embodiment, the device also includes:

[0082] The accuracy verification module is used to verify the accuracy of the machine learning ammonia nitrogen remote sensing inversion model.

[0083] In a large-scale urban river network, water quality may change rapidly due to seasonal changes, human activities and other factors. Traditional single models may find it difficult to adapt to such dynamic changes, resulting in insufficient timeliness and accuracy of the inversion results. In view of the problem of insufficient model generalization ability caused by the spatiotemporal heterogeneity of water quality in a large-scale river network, as a priority of the machine learning ammonia nitrogen inversion device of this embodiment, the device also includes:

[0084] Multi-model building module, used to build a variety of machine and deep learning models. For example, in addition to XGBoost, add random forest (RF), support vector regression (SVR), long short-term memory network (LSTM) and other machine learning and deep learning models, and design independent hyperparameter optimizers for each machine and deep learning model, such as Bayesian optimization or grid search

[0085] The spatiotemporal partitioning module is used to divide the study area into multiple spatiotemporal sub-regions based on geographic location and time information, and automatically determine the optimal number and boundaries of partitions using clustering algorithms (such as K-means or DBSCAN);

[0086] Dynamic weight allocation module, which is used for dynamic weight allocation algorithm based on the performance of machine and deep learning models, and introduces online learning mechanism to update the weights of each machine and deep learning model in different spatiotemporal sub-areas in real time

[0087] Model integration and prediction module, which is used to implement the model integration method based on weighted voting, and design a sliding time window mechanism to dynamically adjust the machine and learning deep model integration strategy

[0088] Anomaly detection and feedback module, which is used to introduce outlier detection algorithms, such as Isolation Forest or Local Outlier Factor (LOF), and design a feedback mechanism to feed back detected anomalies to the machine and learning deep model training and weight assignment modules

[0089] The spatiotemporal partitioning module first divides the study area into multiple spatiotemporal sub-regions. The multi-model construction module trains multiple different types of models in each sub-region. The dynamic weight allocation module calculates and updates the model weights in real time based on the historical performance of each model in different sub-regions. The model integration and prediction module selects the appropriate model combination according to the current spatiotemporal position and uses the weighted voting method for the final prediction. The anomaly detection and feedback module monitors the prediction results in real time and triggers model retraining or weight adjustment when anomalies are found. Through this dynamic adaptive multi-model integration, the problem of insufficient model generalization ability caused by the spatiotemporal heterogeneity of water quality conditions in a large range of river networks can be effectively solved. It can also dynamically select the optimal model combination according to the characteristics of different spatiotemporal sub-regions, and continuously optimize the model performance through a real-time feedback mechanism, thereby improving the accuracy, stability and adaptability of ammonia nitrogen concentration inversion.

[0090] like Figure 3 As shown, an embodiment of the present application provides an electronic device, including a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114; the memory 113 is used to store computer programs; the processor 111 is used to implement the steps of machine learning ammonia nitrogen inversion provided by any of the above method embodiments when executing the program stored in the memory 113. Exemplarily, the steps of machine learning ammonia nitrogen inversion may include the following steps: obtaining a target video stream containing a target object, and extracting a frame sequence from the target video stream; performing feature analysis and fusion on the frame sequence through a preset feature analysis model group to obtain a spatiotemporal feature fusion result; performing behavior analysis through a preset prediction model group based on the spatiotemporal feature fusion result and the frame sequence to obtain a frame sequence analysis result; performing abnormal behavior detection based on the spatiotemporal feature fusion result and the frame sequence analysis result to obtain an abnormal behavior detection result of the target object.

[0091] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of machine learning ammonia nitrogen inversion provided in any of the aforementioned method embodiments are implemented.

[0092] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A machine learning ammonia nitrogen inversion method, characterized in that: include: The method comprises: Obtain the measured ammonia nitrogen data of several water quality samples and remote sensing image data for time matching, so as to extract the band characteristic information of each measured point in the remote sensing image according to the geographic coordinate information of the water quality samples; The obtained band characteristic information of each measured point is combined, and the characteristic factors are sorted and filtered using the Pearson correlation coefficient to determine a number of candidate band characteristics; the principal component analysis method is used to perform a dimension reduction operation on the several candidate band characteristics to obtain the band characteristics after dimension reduction; Using the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as the dependent variable, a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics was constructed; The machine learning ammonia nitrogen remote sensing inversion model is used to perform remote sensing inversion of ammonia nitrogen concentration in river networks.

2. The machine learning ammonia nitrogen inversion method according to claim 1, characterized in that: Before using the machine learning ammonia nitrogen remote sensing inversion model to perform remote sensing inversion of the ammonia nitrogen concentration in the river network, it also includes: performing accuracy testing on the machine learning ammonia nitrogen remote sensing inversion model.

3. The machine learning ammonia nitrogen inversion method according to claim 1, characterized in that: The accuracy test of the machine learning ammonia nitrogen remote sensing inversion model includes: Using the machine learning ammonia nitrogen remote sensing inversion model, the measured values ​​of the detection points are compared with the model prediction values, the determination coefficient, root mean square error, and mean absolute error of each detection point are calculated, and the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model is verified; The machine learning ammonia nitrogen remote sensing inversion model is used to invert the ammonia nitrogen concentration in the water body in other time periods in the target area, and the determination coefficient, root mean square error, and mean absolute error of each detection point are calculated to verify the prediction effect of the machine learning ammonia nitrogen remote sensing inversion model.

4. The machine learning ammonia nitrogen inversion method according to claim 1, characterized in that: The obtained measured band feature information of each point is combined, and the feature factors are sorted and filtered using the Pearson correlation coefficient to determine a number of candidate band features; the principal component analysis method is used to reduce the dimension of the candidate band features to obtain the band features after the dimension reduction, including: Construct multiple band transformation forms such as single band, band sum, difference, ratio and square root transformation; The Pearson correlation coefficient is used to sort and filter the characteristic factors, and the band or band combination that is most closely related to the parameter concentration of the measured points in each area is selected to determine m candidate features, which are recorded as v i,i＝1,2,…,m ,v i =n×1 vector, the calculation formula is shown in formula (1): In the formula, r xy is the correlation coefficient, n is the number of samples, x is the water quality inversion factor, and y is the water quality parameter ammonia nitrogen; The principal component analysis method is used to reduce the dimension of m candidate features to k, denoted as w i,i＝1,2,…,k ,w i =n×1 vector.

5. The machine learning ammonia nitrogen inversion method according to claim 4, characterized in that: The band characteristics after dimensionality reduction are used as independent variables, and the water quality parameter ammonia nitrogen is used as the dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics, including: After dimension reduction, the optimized feature matrix w n×k =[w1 w2…w k ] as input, and the measured ammonia nitrogen data of water quality samples as output. Combined with the extreme gradient boosting model, a machine learning ammonia nitrogen remote sensing inversion model with band feature optimization is constructed.

6. The machine learning ammonia nitrogen inversion method according to claim 1, characterized in that: The method of obtaining ammonia nitrogen data measured by a plurality of water quality samples and remote sensing image data for time matching, and extracting band characteristic information of each point measured in the remote sensing image according to the geographic coordinate information of the water quality samples, includes: Determine whether the data level of the remote sensing image data corresponding to the sentinel image is the specified level. If so, determine that the remote sensing image data of the specified level has completed the calibration correction operation. If not, perform atmospheric correction on the remote sensing image data of the non-specified level. The average value of the satellite remote sensing reflectance within a 3×3 pixel window centered on the sampling point is extracted as the remote sensing reflectance of the sampling point.

7. A machine learning ammonia nitrogen inversion device, characterized in that: include: The water quality data collection and processing module is used to obtain the measured ammonia nitrogen data of several water quality samples and the remote sensing image data for time matching, so as to extract the band characteristic information of each point measured in the remote sensing image according to the geographic coordinate information of the water quality samples; The band feature optimization module is used to combine the band feature information of each measured point, sort and filter the feature factors using the Pearson correlation coefficient, and determine a number of candidate band features; use the principal component analysis method to perform dimensionality reduction operation on the several candidate band features to obtain the band features after dimensionality reduction; The inversion model construction module is used to use the band characteristics after dimensionality reduction as independent variables and the water quality parameter ammonia nitrogen as the dependent variable to construct a machine learning ammonia nitrogen remote sensing inversion model with optimized band characteristics; The remote sensing inversion module is used to perform remote sensing inversion of the ammonia nitrogen concentration in the river network by using the machine learning ammonia nitrogen remote sensing inversion model.

8. The machine learning ammonia nitrogen inversion device according to claim 7, characterized in that: Also includes: The accuracy verification module is used to verify the accuracy of the machine learning ammonia nitrogen remote sensing inversion model.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor is used to implement the steps of machine learning ammonia nitrogen inversion as described in any one of claims 1-6 when executing a program stored in a memory.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of machine learning ammonia nitrogen inversion according to any one of claims 1 to 6 are implemented.

Images