A multi-sand river reservoir characteristic parameter dynamic monitoring technology based on remote sensing interpretation and multi-source data fusion

Abstract

This invention provides a dynamic monitoring technology for characteristic parameters of reservoirs in sediment-rich rivers based on remote sensing interpretation and multi-source data fusion, belonging to the field of water conservancy engineering monitoring technology. The technology includes the following steps: S1: Remote sensing image acquisition and preprocessing; S2: Water area extraction, fusing green band and shortwave infrared band, and calculating the improved normalized difference water index; S3: Constructing a water level inversion model, integrating preprocessed daily-scale measured water level data with the concurrent water area extracted in step S2, and dividing it into training and validation sets; S4: Calculating the dynamic water storage of the reservoir; S5: Constructing a sediment concentration inversion model, integrating preprocessed daily-scale measured sediment concentration data to construct a sediment concentration inversion model. This invention uses remote sensing interpretation to achieve automated, long-term, and large-scale continuous monitoring of water and sediment parameters in sediment-rich reservoirs, significantly improving monitoring efficiency; reducing manpower and material costs, and providing technological innovation for watershed-scale dynamic water and sediment perception.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering monitoring technology, and in particular to a dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion. Background Technology

[0002] Under the combined influence of global climate change and human activities, water resource management and monitoring have become important issues in the fields of water conservancy engineering and ecological protection. Reservoirs and dams are key infrastructure for ensuring national water security. High temporal resolution and high precision observation data of reservoir characteristic parameters are the core foundation for the accurate construction of hydrological models and the rational scheduling of reservoirs. Therefore, the monitoring and acquisition of various parameters of reservoirs has become a key research direction in the field of water resources and an important application scenario for new technologies.

[0003] Traditional reservoir characteristic parameter monitoring relies primarily on on-site measurements. This method suffers from significant drawbacks, including low efficiency, high manpower and resource consumption, poor timeliness, and limitations imposed by measurement conditions in remote and complex areas. It fails to achieve comprehensive, rapid, and dynamic control over reservoir operations. To address these issues, satellite remote sensing technology has been introduced. With its advantages of wide detection range, fast information acquisition speed, rich information content, and minimal susceptibility to ground conditions, satellite remote sensing has gradually become the mainstream technology for water resource surveys and ecological environment monitoring, providing a new approach for reservoir characteristic parameter monitoring. Existing methods for extracting water body information based on remote sensing spectral characteristics mainly include single-band methods, multi-band methods, and classification methods. Single-band methods utilize the difference in reflectance between water and non-water bodies in specific bands, combined with threshold values, to extract water body information. Multi-band methods enhance water body information and suppress background interference through band combination calculations; classification methods rely on algorithms such as random forests and support vector machines to classify and identify water bodies. However, these methods are affected by factors such as cloud cover, complex reservoir topography, and special spectral characteristics of sediment-laden water bodies in sediment-laden reservoir scenarios. Furthermore, the band method relies on manual experience to select thresholds and background noise is difficult to eliminate, while the classification method requires a large number of labeled samples and has strict requirements for spectral analysis and feature selection, making it difficult to guarantee rapid and accurate extraction of water bodies. Secondly, in terms of sediment concentration monitoring, although existing remote sensing interpretation methods can achieve continuous dynamic inversion of suspended sediment concentration, the research objects are mostly low-sediment water areas such as oceans, lakes, and estuaries. Research on high-turbidity, high-sediment-laden rivers and reservoirs is scarce, and the applicability of existing models is still unclear. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion, thereby solving the technical problems existing in the prior art.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0006] A dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion includes the following steps: S1: Remote sensing image acquisition and preprocessing, extracting long-term series remote sensing images covering the target reservoir area, and using the CFMask algorithm based on the impact quality assessment band for cloud removal, screening remote sensing images with cloud coverage below a preset threshold; simultaneously performing radiometric and atmospheric correction on the screened images; S2: Water area extraction, fusing green band and shortwave infrared band, calculating the improved normalized difference water index; automatically determining the optimal threshold using the Otsu method, binarizing the MNDWI image to generate an initial water mask; setting a connected region area threshold, removing isolated noise patches, and retaining continuous effective water areas; and counting the total number of pixels in the effective water area, calculating the reservoir water area based on the image spatial resolution; S3: Constructing a water level inversion model, integrating the preprocessed daily-scale measured water level data with the water area extracted in step S2. The data is fused and divided into training and validation sets. A water level inversion model is established using regression analysis and neural network algorithms, with water area as the independent variable and water level as the dependent variable. S4: Dynamic water storage of the reservoir is calculated. Based on the water level inversion model built in step S3, a definite integral method is used, with the lowest water level in the reservoir over the years as the lower limit of integration, and water level data from different dates are combined to calculate the dynamic water storage. S5: A sediment concentration inversion model is constructed. The pre-processed daily-scale measured sediment concentration data is integrated, and the water body range mask extracted in step S2 is combined to accurately extract the multi-band reflectance statistics of the corresponding water body area from the pre-processed image in step S1. The data is then spatiotemporally aligned with the water body area auxiliary parameter to construct a structured sample set. Pearson correlation analysis is used to quantify the correlation between reflectance and sediment concentration in each band, and significantly correlated bands are selected, while redundant and weakly correlated features are removed. A sediment concentration inversion model is constructed using multiple regression and machine learning methods, with reflectance of strongly correlated bands as the dependent variable and sediment concentration as the independent variable.

[0007] Furthermore, the remote sensing images used in step S1 are Landsat 5 / 7 / 8 and Sentinel-2A / B satellite images.

[0008] Furthermore, the formula for calculating the improved normalized difference water index in step S2 is as follows: In the formula, For green band reflection, This refers to the reflectivity in the shortwave infrared band.

[0009] Furthermore, the Otsu method calculation formula in step S2 is as follows: , In the formula, For inter-class variance, The segmentation threshold is... The global average gray level of the image. Foreground pixel ratio, The average gray level of the foreground. The percentage of background pixels. The average gray level of the background.

[0010] Furthermore, the formula for calculating the connectivity number in step S2 is as follows: In the formula, Let be the number of connected components. The connected area threshold. The image pixel resolution; the formula for calculating the water area is: In the formula, The water area of ​​the reservoir area For remote sensing image pixel resolution, This represents the number of pixels in the reservoir's water body.

[0011] Furthermore, the regression analysis method in step S3 includes one of linear regression, logarithmic regression, or exponential regression; the expression for constructing the inversion relationship model is as follows; In the formula, For the reservoir water level, The water area of ​​the reservoir area This is a regression function.

[0012] Furthermore, the calculation formulas for each evaluation index in step S3 are as follows: , , In the formula, To verify the number of data samples, For the first Inversion water level for each sample, For the first The actual water level of each sample This represents the average value of the actual water level.

[0013] Furthermore, the formula for calculating the dynamic water storage capacity in step S4 is as follows: In the formula, For water storage capacity, Water level The water surface area at that time This is the lowest water level the reservoir has ever recorded.

[0014] Furthermore, the machine learning methods in step S5 include support vector regression, extreme gradient boosting, K-nearest neighbor algorithm, and random forest.

[0015] Furthermore, the sediment content inversion method in step S5 includes a multiple regression model and a machine learning model, wherein the multiple regression model is as follows: The machine learning model is In the formula, For sand content, , All are undetermined coefficients in a multiple regression model. For reflectivity in the strongly correlated band, This is the regression function for the machine learning model.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] (i) This invention uses remote sensing interpretation to achieve automated, long-term, and large-scale continuous monitoring of water and sediment parameters in reservoirs with high sediment content, significantly improving monitoring efficiency; and reducing manpower and material costs, effectively overcoming practical constraints such as complex terrain and poor accessibility of monitoring in remote areas, providing technological innovation for watershed-scale dynamic sensing of water and sediment.

[0018] (ii) This invention improves the accuracy of water area extraction by progressively processing the water body information through the MNDWI index, the Otsu method to automatically determine the optimal threshold, and connectivity analysis to eliminate background noise. Based on the fusion of multi-source measured data and spectral features, and through correlation screening and multi-model cross-validation, a spectral inversion model for water level area and sediment content with a clear physical mechanism and strong generalization ability is constructed to ensure the accuracy and reliability of the calculation results of derived parameters such as water storage.

[0019] (III) The inversion model constructed by this invention is optimized for the special characteristics of sediment-laden river reservoirs. It can be extended to sediment-laden river reservoirs in different regions. In particular, it can be applied to small and medium-sized sediment-laden reservoirs with scarce data and remote geographical locations. It does not require the deployment of complex on-site monitoring equipment, thus lowering the monitoring threshold and having broad application prospects.

[0020] (iv) This invention can output the sequence of core parameters such as water area, water level, water storage, and sediment content at high frequency and with high precision, providing dynamic and quantitative data support for joint water and sediment scheduling, flood control and drought relief decision-making, ecological flow guarantee, siltation early warning and comprehensive benefit assessment, promoting the transformation of reservoir management from experience-driven to data-driven, effectively improving operational safety, and effectively ensuring the safe and stable operation of the reservoir and the realization of its comprehensive benefits. Attached Figure Description

[0021] Fig. 1 This is a framework diagram of a dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion disclosed in this invention;

[0022] Fig. 2 This is a schematic diagram of a dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion, as disclosed in this invention. Detailed Implementation

[0023] To make the content of this invention easier to understand, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Identical components are represented by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "up," and "down" used in the following description refer to directions in the accompanying drawings, while the terms "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0024] This embodiment uses the Xiaolangdi Reservoir area as the monitoring scope, such as... Figs. 1-2 As shown in the figure, this embodiment provides a dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion, including the following steps:

[0025] S1: Remote Sensing Image Acquisition and Preprocessing. Long-term, high-resolution remote sensing images of the Xiaolangdi Reservoir area from 2000 to 2025 were acquired using the Google Earth Engine cloud image processing platform. Landsat 5 / 7 / 8 and Sentinel-2A / B satellite images were prioritized, fully leveraging the historical archive advantage of the Landsat series (over thirty years) and the high revisit frequency and high spatial resolution of Sentinel-2. Through multi-source data synergy and complementarity, temporal discontinuities caused by single-sensor replacements, data gaps, or insufficient seasonal coverage were effectively avoided, constructing a complete and robust observation sequence to ensure the spatiotemporal continuity and reliability of the data. Integrated preprocessing was performed on the acquired remote sensing images, employing the CFMask algorithm based on impact quality assessment bands for cloud removal, specifically targeting clouds and cloud shadows. Refined identification and masking of snow / ice and high aerosol areas were performed to effectively remove cloud cover from affecting image quality. A cloud content threshold of 10% was set to screen remote sensing images with cloud coverage below 10% to ensure data quality and reduce cloud interference from polluting water spectral information. Radiometric and atmospheric corrections were simultaneously applied to the screened images to eliminate the influence of sensor errors, atmospheric scattering and absorption on image spectral information, resulting in a standardized remote sensing image dataset. The projection coordinate system, pixel resolution and band configuration were unified, and a standardized remote sensing image dataset with spatiotemporal alignment and spectral comparability was output.

[0026] S2: Water area extraction, fusion of green band and shortwave infrared band, calculation of improved normalized difference water index, enhancing the spectral difference between water body and background features, and improving the ability to identify water bodies in sandy scenarios; the calculation formula of the improved normalized difference water index is: In the formula, For green band reflection, The reflectivity is for the shortwave infrared band, and the center wavelength of the green band is 0.56. The center wavelength of the shortwave infrared band is 0.161 nm. The optimal threshold is automatically determined using the Otsu method. The MNDWI image is binarized to generate an initial water body mask, specifically through automatic threshold separation. This process requires no manual intervention. The method iterates through all possible grayscale thresholds, calculates the inter-class variance between the water foreground and non-water background at different thresholds, and uses the grayscale value corresponding to the maximum inter-class variance as the optimal water body recognition threshold. Based on this threshold, a binary image of the water body is generated, with water body pixels labeled as 1 and non-water body pixels labeled as 0, thus achieving preliminary extraction of water body information. The Otsu method calculation formula is as follows: , In the formula, For inter-class variance, The segmentation threshold is... The global average gray level of the image. Foreground pixel ratio, The average gray level of the foreground. The percentage of background pixels. Set the average gray level of the background; set the area threshold for connected regions to 30000. Given an image pixel resolution of 30m, calculate the connected number. Regions with a connectivity number greater than 34 are selected, isolated noise patches are removed, and continuous effective water areas are retained; the formula for calculating the connectivity number is: In the formula, Let be the number of connected components. The connected area threshold. The image pixel resolution; the formula for calculating the water area is: In the formula, The water area of ​​the reservoir area For remote sensing image pixel resolution, The number of pixels in the reservoir's water body is calculated; the total number of pixels in the effective water body area is also counted. The water area of ​​the reservoir is calculated based on the image spatial resolution. If the number of water body pixels on a certain date is 100,000, then the water area is calculated as follows: ;

[0027] S3: Construct a water level inversion model by integrating the preprocessed daily-scale measured water level data with the contemporaneous water area extracted in step S2, and dividing it into a training set and a validation set. The preprocessing involves collecting daily-scale measured water level series data from the monitoring reservoir area and performing quality control on the data, including outlier removal and missing value completion. The preprocessed measured water level data is then spatiotemporally matched with the corresponding date and corresponding reservoir area water area data extracted in step S2 to construct a water area-water level fusion dataset. The fusion dataset is randomly divided into a model training set and a validation set according to a 7:3 or 8:2 ratio. The training set is used for model parameter fitting, and the validation set is used for model accuracy evaluation. Using the water area in the fusion dataset as the independent variable and the water level as the dependent variable, regression analysis and neural network algorithms are used to establish a water level inversion model. The expression for constructing the inversion model is as follows: In the formula, For the reservoir water level, The water area of ​​the reservoir area The regression function has different forms for different models; the water area data from the validation set are substituted into each candidate model to calculate the inverted water level; the coefficient of determination is selected. Mean absolute error Root mean square error As an indicator for evaluating model accuracy, the model with the smallest error and highest accuracy is selected as the optimal water level inversion model by comprehensively comparing the evaluation index results of various models; the calculation formulas for each evaluation index are as follows: , , In the formula, To verify the number of data samples, For the first Inversion water level for each sample, For the first The actual water level of each sample The average actual water level; the regression analysis method includes one of linear regression, logarithmic regression, or exponential regression, all of which are existing technologies and will not be elaborated upon here; in this embodiment, daily measured water level data of Xiaolangdi Reservoir from 2000 to 2025 are collected, and after removing outliers, missing values ​​are filled in using linear interpolation; the measured water level data are matched with the water area data extracted by remote sensing on the corresponding dates to form a fused dataset, which is divided into a training set (2000-2021) and a validation set (2021-2025) at an 8:2 ratio; through , , Evaluation accuracy, and finally selection =0.924、 =3.66m The linear regression model with a water level of 5.87m is used as the optimal water level inversion model, and the model expression is: ,in, Water level, unit: meters (m). For water body area, unit ;

[0028] S4: Calculation of dynamic reservoir storage capacity. Based on the water level inversion model constructed in step S3, a definite integral method is used, with the lowest water level in the reservoir over the years as the lower limit of integration. Combined with water level data from different dates, the dynamic storage capacity is calculated to achieve spatiotemporal dynamic active monitoring of the storage capacity. The reservoir storage capacity is the integral of the water area corresponding to the water level from the lowest value to the current value. The formula for calculating dynamic storage capacity is: In the formula, For water storage capacity, Water level The water surface area at that time This is the lowest water level the reservoir has ever recorded. The current monitoring water level is used as the basis for the operation. First, the daily water level data for each year within the monitoring period is extracted, including the inverted water level and the measured water level, to determine the minimum water level for each year. Second, the optimal water level area inversion model obtained in step S2 is substituted into the above integral formula, and combined with the water level data for each day of the year, the dynamic water storage for each day of the year is calculated. Finally, the arithmetic mean of the daily water storage for each year is calculated to obtain the annual average dynamic water storage, realizing the annual dynamic change analysis of the reservoir's water storage. In this embodiment, the daily water level data for the Xiaolangdi Reservoir area from 2000 to 2025 is extracted to determine the minimum water level for each year, such as 2005 = 219.47m, 2006 = 221.04m, and so on. The optimal water level area inversion model is then used to calculate the dynamic water storage for each day of the year. Substituting into the integral formula for water storage, calculate the daily dynamic water storage. For example, if the water level on a certain day in 2005 is Z=257.55m, then the water storage is... 100 million The average daily water storage for each year was taken, resulting in an average daily dynamic water storage of 3.769 billion cubic meters in 2005. The same principle applies to other years;

[0029] S5: Construct a sediment concentration inversion model, integrate preprocessed daily-scale measured sediment concentration data, and combine the water body range mask extracted in step S2 to accurately extract multi-band reflectance statistics of the corresponding water body area from the preprocessed image in step S1. Align these statistics with the water body area auxiliary parameters spatiotemporally to construct a structured sample set. Specific steps are as follows: S5.1: Spectral feature analysis and strong correlation band selection: Based on the fundamental principle that changes in water sediment concentration cause changes in spectral reflectance, and that these changes can be captured by satellite sensors, analyze the multi-band spectral characteristics of the preprocessed remote sensing image. Use Pearson correlation analysis to quantify the correlation between reflectance and sediment concentration in each band, select significantly correlated bands, eliminate redundant and weakly correlated features, reduce the model input dimension, and improve the model's inversion efficiency. S5.2: Collect daily-scale measured sediment concentration series data of the monitoring reservoir area and preprocess the data; perform spatiotemporal matching between the preprocessed measured sediment concentration data and the remote sensing strongly correlated band reflectance data of the reservoir water body range obtained on the corresponding date and in the feedback, and construct a fusion dataset of band reflectance and sediment concentration; divide the fusion dataset into a model training set and a validation set according to a preset ratio of 7:3; S5.3: Using strongly correlated band reflectance as the dependent variable and sediment concentration as the independent variable, construct a sediment concentration inversion relationship model using multiple regression and machine learning methods; the machine learning methods include support vector regression, extreme gradient boosting, K-nearest neighbor algorithm, and random forest; the two core forms of the sediment concentration inversion are: multiple regression model: Machine learning models In the formula, For sand content, , All are undetermined coefficients in a multiple regression model. For reflectivity in the strongly correlated band, The regression function for the machine learning model is used; the strongly correlated band reflectance data of the validation set are substituted into each candidate sediment concentration inversion model to calculate the inverted sediment concentration; the same accuracy evaluation index as the water level inversion model is used to evaluate the accuracy of each model, and the model with the best overall performance is selected as the final sediment concentration inversion model to ensure that the model is applicable to sediment concentration monitoring of different sediment-rich reservoirs; this embodiment collects daily measured sediment concentration data of Xiaolangdi Reservoir from 2000 to 2021, and after preprocessing, compares it with the blue and green bands of the corresponding date remote sensing images. Reflectance data from the red, near-infrared, and red bands were matched. Pearson correlation analysis was used to select the red and near-infrared bands with the largest absolute values ​​of correlation coefficients with sediment concentration as strongly correlated bands. The fused dataset was divided into training and validation sets in an 8:2 ratio. With red and near-infrared reflectance as dependent variables and sediment concentration as independent variables, a multivariate regression model and a support vector regression model were constructed. The validation results showed that the support vector regression model had better accuracy and was determined to be the optimal sediment concentration inversion model, which can achieve accurate sediment concentration inversion.

[0030] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion, characterized in that: Includes the following steps: S1: Remote sensing image acquisition and preprocessing, extracting long-term remote sensing images covering the target reservoir area, and using the CFMask algorithm based on the impact quality assessment band for cloud removal, screening remote sensing images with cloud coverage below a preset threshold; radiometric correction and atmospheric correction are simultaneously applied to the screened images. S2: Water area extraction, fusion of green band and shortwave infrared band, calculation of improved normalized differential water index; automatic determination of optimal threshold using Otsu method, binarization of MNDWI image to generate initial water mask; setting connected region area threshold, removing isolated noise patches, retaining continuous effective water area; and counting the total number of pixels in the effective water area, combined with image spatial resolution to calculate reservoir water area. S3: Construct a water level inversion model by integrating the pre-processed daily measured water level data with the water area extracted in step S2, and dividing it into a training set and a validation set; using water area as the independent variable and water level as the dependent variable, establish a water level inversion model using regression analysis and neural network algorithms. S4: Calculation of dynamic water storage in the reservoir. Based on the water level inversion relationship model constructed in step S3, the dynamic water storage is calculated by using the lowest water level in the reservoir over the years as the lower limit of integration and combining water level data from different dates. S5: Construct a sediment concentration inversion model, integrate preprocessed daily-scale measured sediment concentration data, combine with the water body range mask extracted in step S2, accurately extract multi-band reflectance statistics of the corresponding water body area from the preprocessed image in step S1, and align them with the water body area auxiliary parameters in time and space to construct a structured sample set; use Pearson correlation analysis to quantify the correlation between reflectance of each band and sediment concentration, screen significantly correlated bands, and remove redundant and weakly correlated features; Using the reflectance of strongly correlated bands as the dependent variable and sediment content as the independent variable, a sediment content inversion relationship model was constructed using multiple regression and machine learning methods.

2. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The remote sensing images used in step S1 are Landsat 5 / 7 / 8 and Sentinel-2A / B satellite images.

3. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The formula for calculating the improved normalized differential water index in step S2 is as follows: In the formula, For green band reflection, This refers to the reflectivity in the shortwave infrared band.

4. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The Otsu method calculation formula in step S2 is as follows: , In the formula, For inter-class variance, The segmentation threshold is... The global average gray level of the image. Foreground pixel ratio, Foreground average gray level, The percentage of background pixels. The average gray level of the background.

5. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The formula for calculating the connectivity number in step S2 is as follows: In the formula, Let be the number of connected components. The connected area threshold. The image pixel resolution; the formula for calculating the water area is: In the formula, The water area of ​​the reservoir area For remote sensing image pixel resolution, This represents the number of pixels in the reservoir's water body.

6. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The regression analysis method in step S3 includes one of linear regression, logarithmic regression, or exponential regression; the expression for constructing the inversion relationship model is: In the formula, For the reservoir water level, The water area of ​​the reservoir area This is a regression function.

7. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1 or 6, characterized in that: The calculation formulas for each evaluation index in step S3 are as follows: , , In the formula, To verify the number of data samples, For the first Inversion water level for each sample, For the first The actual water level of each sample This represents the average value of the actual water level.

8. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The formula for calculating the dynamic water storage capacity in step S4 is as follows: In the formula, For water storage capacity, Water level The water surface area at that time This is the lowest water level the reservoir has ever recorded.

9. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1, characterized in that: The machine learning methods in step S5 include support vector regression, extreme gradient boosting, K-nearest neighbor algorithm, and random forest.

10. The dynamic monitoring technology for characteristic parameters of reservoirs in sandy rivers based on remote sensing interpretation and multi-source data fusion as described in claim 1 or 9, characterized in that: The sediment content inversion method in step S5 includes a multiple regression model and a machine learning model. The multiple regression model is as follows: The machine learning model is In the formula, For sand content, , All are undetermined coefficients in a multiple regression model. For reflectivity in the strongly correlated band, This is the regression function for the machine learning model.

Images