Classified processing pit-pond water storage capacity remote sensing statistics joint estimation method

Abstract

The invention discloses a classified processing pit-pond water storage amount remote sensing statistics joint estimation method, and relates to the technical field of water resource remote sensing monitoring, and the method comprises the steps: calculating the pit-pond density, and determining the grid width; calculating a normalized water body index, and extracting a pond water surface boundary; dividing the pit pond into an invertible pit pond and a non-invertible pit pond according to the water surface area and width of the pit pond; calculating the water storage amount of the invertible pit-pond by combining the area of the invertible pit-pond; calculating the water storage amount of the non-invertible pit pond by using the relationship between the water surface area of the non-invertible pit pond and the statistical model; and calculating the total water storage amount of the region, and generating a spatial distribution diagram of the water storage amount. According to the method, remote sensing inversion and a statistical model method are used for estimating the water storage amount, targeted processing can be carried out on different types of ponds, remote sensing images are combined with actually measured data, the water depth inversion precision of the invertible ponds can be effectively improved, and therefore a more reliable water storage amount estimation result is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing monitoring of water resources. Specifically, it relates to a remote sensing statistical joint estimation method for the storage volume of pond water with classified processing. Background Technique

[0002] As an important water resource on the earth, the accurate estimation of the water storage volume of ponds is of great significance for water resource management, ecological environment protection, cultivated land protection and other fields. The traditional method for estimating the water storage volume of water bodies mainly relies on observing water level data and measuring underwater terrain data. Although this method is accurate, it faces many limitations in practical applications. This method has high costs and limited coverage. Problems such as equipment failures and data transmission make it difficult to achieve large-area dynamic monitoring. The statistical model (area-storage relationship) is applicable to the estimation of the water storage volume of ponds in large areas, but it cannot reflect the spatial change trend of ponds.

[0003] With the development of remote sensing technology, satellite inversion bathymetry technology has gradually been applied to the fields of water depth and water storage volume estimation. However, the conventional remote sensing inversion method is affected by image resolution, aquatic plants and water quality interference, resulting in insufficient accuracy. Especially in shallow ponds, narrow ponds and small-area ponds, the error is significant. Moreover, most of the existing technologies are based on the research of water storage volume using statistical models and do not integrate multi-source remote sensing data. As a result, the estimation accuracy of the water storage volume of ponds in large areas under complex environments (diverse types, large quantities, wide areas, aquatic plant coverage, shadow interference) is low and the adaptability is poor.

[0004] In response to the problems in the related technologies, no effective solution has been proposed yet. Summary of the Invention

[0005] In response to the problems in the related technologies, the present invention proposes a remote sensing statistical joint estimation method for the storage volume of pond water with classified processing to overcome the above-mentioned technical problems existing in the existing related technologies.

[0006] To this end, the specific technical solution adopted by the present invention is as follows: A remote sensing statistical joint estimation method for the storage volume of pond water with classified processing, including: Based on the dynamic adjustable grid sampling method, sample and investigate the ponds in the test area, calculate the pond density and determine the grid width; Calculate the normalized water body index, determine the water body boundary through the threshold method, and perform masking processing on the survey data to extract the water surface boundary of the ponds; According to the water surface area and width of the ponds, divide the ponds into ponds that can be inverted and ponds that cannot be inverted; Use remote sensing image technology to calculate the water depth of the ponds that can be inverted, and combine with the area of the ponds that can be inverted to calculate the water storage volume of the ponds that can be inverted; Construct a statistical model to estimate the water storage capacity of non-invertible ponds, and use the relationship between the water surface area of non-invertible ponds and the statistical model to calculate the water storage capacity of non-invertible ponds; Fuse the water storage capacities of invertible ponds and non-invertible ponds, calculate the total regional water storage capacity, and generate a spatial distribution map of the water storage capacity.

[0007] Furthermore, based on the dynamic adjustable grid sampling method, sample the ponds in the test area, calculate the pond density, and determine the grid width, including: Classify the ponds in the test area into aquaculture ponds and natural ponds according to their types, preset the sampling ratio, determine the total sampling numbers of aquaculture ponds and natural ponds, as well as the overall total sampling number; Determine the width of the grid based on the density of the ponds in the test area, and allocate sample ponds to each grid; Sort the grids according to the number of sample ponds in the grids, and remove the grids with zero sample pond numbers; Traverse the valid grids, and sequentially extract one pond from each grid until the overall sample number requirement is met.

[0008] Furthermore, the density calculation formula of the ponds is: ; The calculation formula of the grid width is: ; In the formula, p represents the pond density in the test area; N represents the total number of ponds in the test area; S 1 represents the national land area in the test area; a represents the width of the grid; represents the number of sample ponds in the grid; k represents the sampling ratio.

[0009] Furthermore, calculate the normalized difference water index, determine the water body boundary through the threshold method, and perform masking processing on the survey data to extract the water surface boundary of the ponds, including the following steps: Calculate the normalized difference water index using remote sensing images, and obtain the initial information of the water bodies in the test area; Determine the threshold of the normalized difference water index through the threshold method. If the threshold is greater than or equal to the preset value, extract the water body boundary in the test area and perform boundary smoothing processing; Based on geographic information technology, perform masking processing on the non-pond water body types in the survey data, and extract the water surface boundary of the ponds.

[0010] Furthermore, calculating the water depth of retrievable ponds using remote sensing image technology and combining with the area of retrievable ponds to calculate the water storage volume of retrievable ponds includes: Obtain multispectral remote sensing images and use a custom geotransformation to convert the Universal Transverse Mercator projection coordinates of the multispectral remote sensing images into Gauss-Krüger projection coordinates; Calculate the Normalized Difference Vegetation Index (NDVI) and the Normalized Water Turbidity (NWT) using the multispectral remote sensing images, eliminate redundant points, thin out the measured points, and establish a one-to-one correspondence between the pixel grid and the measured points; Extract the water reflectance, generate band combination inversion factors through band calculations, and calculate the characteristic importance of each inversion factor in combination with the measured point data; Divide the measured point data into a training set and a validation set, construct a water depth inversion model to invert the water depth of retrievable ponds, and calculate the evaluation indices, the coefficient of determination and the root mean square error, by comparing the inverted water depth values with the actually measured water depths; According to the inversion results, extract the water depth data of each retrievable pond area, calculate the average water depth value of each retrievable pond, and calculate the water storage volume of retrievable ponds by multiplying the water surface area by the average water depth.

[0011] Furthermore, the calculation formula for the coefficient of determination of the evaluation index is: ; The calculation formula for the root mean square error is: ; In the formula, y i represents the i th true value; represents the i th estimated value; represents the average value of the true values; represents the total number of true values.

[0012] Furthermore, the formula for calculating the average water depth value of each retrievable pond is: ; In the formula, represents the i th average water depth value of the retrievable pond; represents the i th measured water depth value of the u th measurement point of the num th retrievable pond;

[0013] Furthermore, the formula for calculating the water storage volume of retrievable ponds is: ; In the formula, V1 represents the total water storage of the invertible ponds; represents the i average water depth value of the s i th i invertible pond; m represents the total number of invertible ponds.

[0014] Furthermore, the formula for calculating the water storage of non-invertible ponds is: ; ; In the formula, z represents the water storage; x represents the area; a , b , c all represent the fitting parameters of the quadratic polynomial function; V 2 represents the total water storage of non-invertible ponds; z j represents the j water storage of the n th non-invertible pond;

[0015] Furthermore, the water storage of invertible ponds and non-invertible ponds is fused to calculate the total water storage of the region, and the generation of the spatial distribution map of water storage includes: Using the average value method and the statistical model method to fuse the calculation results of the water storage of invertible ponds and non-invertible ponds to obtain the total water storage in the test area; Based on geographic information technology and the water surface boundary of the ponds, a new average water depth field is added, and the average water depth value of each pond is assigned to the vector pond surface to generate a visual regional pond water storage distribution map.

[0016] The beneficial effects of the present invention are: 1. By dividing the ponds into invertible ponds and non-invertible ponds, and applying remote sensing inversion and statistical model methods to estimate the water storage respectively, the present invention can perform targeted processing for different types of ponds. Combining remote sensing images with measured data can effectively improve the water depth inversion accuracy of invertible ponds, while the estimation of non-invertible ponds by the statistical model ensures the water storage estimation accuracy of the entire region, thus providing a more reliable water storage estimation result.

[0017] 2. Through refined classification estimation and generation of spatial distribution maps, the present invention provides specific spatial references and data support for regional water resource management. The water storage amounts of retrievable and non-retrievable ponds can be calculated separately, which helps to accurately grasp the water volume distribution of different types of ponds. Combined with the visual regional water storage amount distribution map, it can provide a more intuitive and detailed data basis for water resource managers, contributing to the implementation of more precise water resource scheduling and water area protection strategies.

[0018] 3. By combining dynamic adjustable grid sampling with remote sensing data, the present invention ensures adaptability under different geographical conditions and pond types. Through techniques such as grid sampling surveys, NDWI and water depth inversion, and statistical models, it can cover various complex terrains and water body changes, showing wide applicability. Whether it is aquaculture ponds or natural ponds, they can be flexibly processed and estimated according to specific situations, thus improving the overall efficiency and accuracy of water storage amount estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 is a flowchart of a remote sensing statistical joint estimation method for pond water storage amount with classified processing according to an embodiment of the present invention; Figure 2 is a flowchart of grid method sample extraction according to an embodiment of the present invention; Figure 3 is a schematic diagram of pond classification according to an embodiment of the present invention; Figure 4 is a schematic diagram of the principle of remote sensing water depth inversion according to an embodiment of the present invention; Figure 5 is a schematic diagram of comparison of the importance of inversion factor characteristics according to an embodiment of the present invention; Figure 6 is a schematic diagram of the fitting curve of the area - water storage amount statistical model according to an embodiment of the present invention; Figure 7 is a schematic diagram of comparison of the cross-sectional shapes of ponds in the longitudinal and transverse directions according to an embodiment of the present invention; Figure 8 is a schematic diagram of the inversion result according to an embodiment of the present invention; Figure 9 is a schematic diagram of comparison of accuracy verification according to an embodiment of the present invention; Figure 10Schematic diagram of the fitting curve of the training set area - water storage statistical model according to the embodiments of the present invention; Figure 11 Schematic diagram of the verification result of the water storage in the pond according to the embodiments of the present invention. Detailed implementation manners

[0021] To further illustrate the embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be combined with the relevant descriptions in the specification to explain the operating principle of the embodiments. With reference to these contents, those of ordinary skill in the art should be able to understand other possible implementation manners and the advantages of the present invention.

[0022] According to an embodiment of the present invention, a remote sensing statistical joint estimation method for the water storage in ponds with classification processing is provided.

[0023] Now, the present invention will be further described in combination with the accompanying drawings and detailed implementation manners. As Figure 1 shown, the remote sensing statistical joint estimation method for the water storage in ponds with classification processing according to the embodiments of the present invention includes: S1. Based on the dynamic adjustable grid sampling method, sample the ponds in the test area, calculate the pond density and determine the grid width; S2. Calculate the normalized difference water index, determine the water body boundary through the threshold method, and perform masking processing on the survey data to extract the water surface boundary of the ponds; S3. Divide the ponds into retrievable ponds and non - retrievable ponds according to the water surface area and width of the ponds; It should be explained that according to the water body morphological parameters (water surface area, width), the ponds are divided into two categories: The first category is retrievable ponds (pure water surface, area ≥ 2000 m², width ≥ 20 m), including two types: aquaculture ponds and natural ponds (as Figure 3 shown, the retrievable type at b); The second category is non - retrievable ponds (interfered by water plants, shadows, or small area, narrow width, covered by greenhouses), including narrow ponds, shallow ponds, small ponds, greenhouse - covered ponds, etc. (as Figure 3 shown, the non - retrievable type at a), providing a basis for subsequent differential modeling.

[0024] S4. Use remote sensing image technology to calculate the water depth of the retrievable ponds, and combine the area of the retrievable ponds to calculate the water storage of the retrievable ponds; S5. Construct a statistical model to estimate the water storage of the non - retrievable ponds, and use the relationship between the water surface area of the non - retrievable ponds and the statistical model to calculate the water storage of the non - retrievable ponds; S6. Integrate the water storage of the retrievable ponds and the non - retrievable ponds, calculate the total water storage of the region, and generate a spatial distribution map of the water storage.

[0025] In this optional embodiment, based on the dynamically adjustable grid sampling method, a sampling survey is conducted on the ponds in the test area, and the pond density is calculated and the grid width is determined, including: The ponds in the test area are divided into aquaculture ponds and natural ponds by type, and the sampling ratio is preset to determine the total number of samples of aquaculture ponds and natural ponds and the total number of samples as a whole; The width of the grid was determined by the density of ponds in the experimental area, and sample ponds were assigned to each grid; The grids are sorted according to the number of sample ponds in the grids, and the grids with zero sample ponds are removed; Traverse the valid grids and select a pond from each grid in turn until the overall sample quantity requirement is met.

[0026] It should be explained that, based on the principles of comprehensiveness, typicality and quality-efficiency, a certain proportion of sample ponds were selected in the test area, and a single-beam instrument was used to conduct field sampling surveys. The grid method was used to select sample ponds. First, the pond density was calculated, then the grid width was determined, and finally samples were extracted from each grid, so that the samples were evenly distributed in space.

[0027] It needs to be explained that Figure 2 As shown in the figure, the grid method for extracting sample ponds includes: first, dividing the ponds in the area into two categories: aquaculture ponds and natural ponds; second, setting the sampling ratio to determine the total number of aquaculture ponds, the total number of natural ponds and the total number of overall samples; then, determining the grid width according to the density of ponds in the jurisdiction; third, sorting the grids according to the number of ponds in the grid, and removing the grids with 0 ponds; then, looping through the grids, extracting a pond from each grid in turn until the quantity requirement is met.

[0028] In this optional embodiment, the density calculation formula of the pond is: ; The calculation formula for grid width is: ; In the formula, p represents the density of ponds in the experimental area; N represents the total number of ponds in the experimental area; S 1 It represents the land area within the experimental area, in square kilometers; a Indicates the width of the grid in kilometers; Indicates the number of sample ponds in the grid; n Set to 2, which means that 2 sample ponds are selected from each grid; k Represents the sampling ratio, which is a constant.

[0029] In this optional embodiment, the normalized difference water index (NDWI) is calculated, the water body boundary is determined by the threshold method, and the survey data is masked. Extracting the pond water surface boundary includes the following steps: Use remote sensing images (high-resolution optical remote sensing images) to calculate the normalized water index and obtain initial information on the water bodies in the test area; The threshold value of the normalized water index is determined by the threshold method. If the threshold value is greater than or equal to the preset value, the water body boundary in the test area is extracted and the boundary smoothing is performed; Based on geographic information technology (geographic information software), the non-pond water body types (reservoirs, lakes, rivers, and ditch vector layers) in the survey data (the latest annual land change survey data) are masked and the pond water surface boundaries are extracted.

[0030] It should be explained that based on high-resolution optical remote sensing images (such as high-resolution series, resource series, Sentinel-2, etc.), after radiation calibration and atmospheric correction processing, the normalized water index NDWI is calculated, and the threshold value of NDWI is determined by the threshold method. When the threshold value is ≥ 0.2 (preset value), it is identified as the boundary of all water bodies in the study area, and the boundary is smoothed to remove the jagged edge. The vector layers of reservoirs, lakes, rivers, and ditches in the latest annual land change survey data are collected, and these non-pond water body types are masked using geographic information software to accurately extract the water surface boundaries of regional pond types. The formula for calculating the normalized water index NDWI is: ; In the formula, ρ represents the spectral reflectance; Green Indicates the green light band; NIR Represents the near-infrared band.

[0031] In this optional embodiment, the water depth of the invertible pond is calculated using remote sensing imaging technology, and the water storage capacity of the invertible pond is calculated in combination with the area of ​​the invertible pond, including: Acquire multispectral remote sensing images (Sentinel-2 images) and use custom geographic transformation to convert the Universal Transverse Mercator projection coordinates of the multispectral remote sensing images into Gauss-Krüger projection coordinates; The normalized vegetation index and normalized water turbidity were calculated using multispectral remote sensing images, redundant points were eliminated, measured points were thinned out, and pixel grids and measured points were matched one by one. Extract water body reflectance, generate band combination inversion factors through band calculation, and calculate the characteristic importance of each inversion factor in combination with measured point data; Divide the measured point data into a training set and a validation set, construct a water depth inversion model to invert the water depth of the ponds that can be inverted, and calculate the evaluation indexes of the coefficient of determination and the root mean square error by comparing the inverted water depth value with the actually measured water depth; According to the inversion results, extract the water depth data of each area of the ponds that can be inverted, calculate the average water depth value of each pond that can be inverted, and calculate the water storage capacity of the ponds that can be inverted by combining the water surface area and the average water depth.

[0032] It should be noted that for the ponds that can be inverted, carry out the water depth inversion experiment according to the inversion process (as Figure 4 shown). Obtain Sentinel-2 images with close dates, create a custom geotransformation (transform from WGS1984 geographic coordinate system to CGCS2000 geographic coordinate system), use the geographic information raster projection tool, load the custom geotransformation, and convert the UTM projection coordinates (Universal Transverse Mercator projection coordinates) to the Gauss-Kruger projection coordinates of the 3-degree zone of CGCS2000 (Gauss-Kruger projection coordinates), so as to ensure that the converted projection is consistent with the projection of the measured data, and generate the input data set by combining the measured data and the inversion factors.

[0033] It should be noted that for the preprocessing of the input data. Eliminate redundant points, thin out the measured points, and make the pixel raster correspond to the measured points one by one to avoid data redundancy and overfitting. Calculate the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Turbidity Index (NDTI) using the Sentinel-2 images, preprocess the input data, and eliminate the points in the vegetation-covered areas (NDVI greater than 0.1), high turbidity areas (NDTI>0), and mixed pixel areas (points within 10 meters from the shore), eliminate the points with relatively shallow water depth (below 0.6 meters), and eliminate the points with large changes in the water body state of the ponds due to the inconsistency between the measured time and the image time, so as to improve the reliability of the inversion model. The formulas for calculating the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Turbidity Index (NDTI) are respectively: ; ; In the formula, ρ represents the spectral reflectance; Green represents the green light band; NIR represents the near-infrared band; Red represents the red light band.

[0034] It should be noted that according to the spatial position of the measured water depth sample points, extract the water body reflectance of the preprocessed image as a single-band inversion factor, and the reflectance after band calculation (logarithm, ratio, log ratio, etc.) as a band combination inversion factor. A total of 20 inversion factors are formed (as shown in Table 1). Combine the measured point data and calculate the feature importance of each inversion factor (asFigure 5 As shown in the figure, the higher the feature importance value, the higher the correlation between the factor and water depth.

[0035] Table 1. Comparison of calculation methods and feature importance of each inversion factor Serial number Inversion factor Feature importance Serial number Inversion factor Feature importance 1 B1 0.09 11 Ln(B1) / Ln(B2) 0.08 2 B2 0.04 12 Ln(B1) / Ln(B3) 0.04 3 B3 0.04 13 Ln(B1) / Ln(B4) 0.02 4 B4 0.05 14 Ln(B2) / Ln(B3) 0.05 5 B1 / B2 0.08 15 Ln(B2) / Ln(B4) 0.04 6 B1 / B3 0.04 16 Ln(B3) / Ln(B4) 0.03 7 B1 / B4 0.02 17 B1 + B2 0.03 8 B2 / B3 0.05 18 B1 + B3 0.04 9 B2 / B4 0.04 19 B2 + B3 0.06 10 B3 / B4 0.03 20 B1 + B4 0.12 Note: B1, B2, B3, and B4 represent the reflectance of blue, green, red, and near-infrared bands respectively.

[0036] It should be explained that the processed measured data is divided into a training set and a validation set. Combining the results of data preprocessing, a water depth inversion model controlled and constrained by multiple factors such as NDVI, NDTI, water depth range, and spatial distance is constructed using the training set. According to the correlation between water depth and reflectance, the water depth of the pond is inverted. The inversion models include random forest, GBDT, XGBoost, BP neural network model, etc. Other machine learning and deep learning models can also be tried. The accuracy of the inversion results is evaluated using the validation set, and the coefficient of determination (R 2 ) and root mean square error (RMSE) of the evaluation index are calculated by the inversion water depth value and the actual water depth. The calculation formula for the coefficient of determination of the evaluation index is: ; The calculation formula for the root mean square error is: ; In the formula, y i represents the i th true value ( i = 1, 2,..., n ); represents the i th estimated value; represents the average value of the true values; represents the total number of true values.

[0037] It should be explained that using the mask extraction tool of the geographic information software and the obtained vector map patches of ponds that can be inverted, the mask extraction of the inversion results is carried out, and the formula for calculating the average water depth of the pond is: ; In the formula, represents the average water depth value of the i th pond that can be inverted; represents the measured water depth value of the i th pond that can be inverted at the u th measurement point, u = 1, 2,..., num ; num represents the total number of measurement points.

[0038] It should be noted that the water storage volume of the pond can be inverted by multiplying the water surface area by the average water depth. The formula for calculating the water storage volume of the pond that can be inverted is: ; In the formula, V 1 represents the total water storage volume of the pond that can be inverted; represents the average water depth value of the i th pond that can be inverted; s i represents the area of the i th pond that can be inverted, i = 1, 2,..., m ; m represents the total number of ponds that can be inverted.

[0039] It should be noted that for ponds that cannot be inverted, first use all the measured ponds as samples to establish a quadratic polynomial function relationship between area and water storage volume, and obtain the parameters a , b , c through regression fitting. Input the water surface area of the non-invertible pond class extracted into the pre-trained statistical model (as shown in Figure 6 ), and the formula for calculating the water storage volume of the non-invertible pond is: ; ; In the formula, z represents the water storage volume; x represents the area; a , b , c all represent the fitting parameters of the quadratic polynomial function; V 2 represents the total water storage volume of the non-invertible pond; z j represents the water storage volume of the j th non-invertible pond, j = 1, 2,..., n ; n represents the total number of non-invertible ponds.

[0040] It should be noted that in the method for estimating the water storage volume of ponds based on a statistical model, the cross-sectional shape of the pond has a certain impact on the construction and accuracy of the model. If the pond shapes are similar (for example, most aquaculture ponds are regular rectangles), then the relationship between area and water storage volume is more stable. Shallow flat ponds with a depth of less than 2.5 meters generally have uniform water depth and a strong linear relationship between area and storage. According to field research, most artificial aquaculture ponds are regular geometric shapes (rectangles, circles), with flat or gently sloping bottoms and neat slopes. Common types of natural ponds include dish-shaped, pot-shaped, and long-strip-shaped, asFigure 7 As shown, three typical pond cross-sections are presented, indicating a positive correlation between the visible area and water storage volume.

[0041] In this alternative embodiment, the water storage volumes of the invertible ponds and non-invertible ponds are integrated to calculate the total regional water storage volume, and the generation of the spatial distribution map of the water storage volume includes: Using the average value method and the statistical model method to integrate the calculation results of the water storage volumes of the invertible ponds and non-invertible ponds to obtain the total water storage volume within the test area; Based on geographic information technology and the pond water surface boundary, a new average water depth field is added, and the average water depth value of each pond is assigned to the vector pond surface to generate a visual regional pond water storage volume distribution map.

[0042] It should be explained that by integrating the results of the average value method (for invertible ponds) and the statistical model method (for non-invertible ponds), the total regional water storage volume value is calculated. Using geographic information software and the pond surface (pond water surface boundary), a new "average water depth" field is added, and the average water depth value of each pond is assigned to the vector pond surface, thereby generating a visual regional pond water storage volume distribution map. The formula for calculating the total water storage volume value of the region (test area) is: ; In the formula, V represents the total water storage volume within the working area, V 1 and V 2 are the water storage volumes of the invertible category (invertible ponds) and non-invertible category ponds (non-invertible ponds) respectively.

[0043] It should be explained that one cloud-free Sentinel-2 image with a resolution of 10 meters is obtained, and 71 pond water depth data are collected in the field. The pond water surface data in the study area are extracted, and the pond patches in the study area are divided into two categories: invertible and non-invertible. For the invertible ponds, four inversion models such as the random forest, GBDT, XGboost, and BP neural network models are used. The verification results show (as Figure 8 shown) that the XGboost model has the best effect, and the inversion result is relatively consistent with the measured water depth (R 2 = 0.88). For the non-invertible ponds, typical samples are extracted, and the water depth and bottom elevation data are measured in the field. The cross-sectional profiles of each type of pond are explored, the functional relationship between the water surface area and water storage volume is calculated, and the statistical model quadratic polynomial function is fitted. Comparing with the measured data for verification (as Figure 9 shown), the root mean square error (RMSE) of the inversion result of the XGBoost model is 0.12 m.

[0044] It should be noted that in order to verify the effect of the remote sensing inversion and statistical model fusion method (hereinafter referred to as the "fusion method") in estimating water storage, two experiments were conducted. Experiment 1: The single statistical model method was used to estimate the water storage in regional ponds. Assuming that the water storage in all ponds in the working area was calculated using the statistical model, the sample patches were randomly divided into a training set and a validation set. As Figure 10 shown, the statistical model fitting curve of the training set area - water storage. Figure 10 Figure (a) is the model fitting curve of the training set of the single statistical model method. Experiment 2: The fusion method was used to estimate the water storage in regional ponds. Assuming that the ponds in the working area were divided into two categories: "invertible class" and "non - invertible class", the sample patches were divided into a training set and a validation set. The validation set only included the "non - invertible class", and the patches outside the validation set were the training set. Figure 10 Figure (b) is the model fitting curve of the training set of the fusion method. As Figure 11 shown, the verification results of the water storage in ponds. Figure 11 Figures (a), (c), and (e) are the verification scatter plots of the three models: the single statistical model method, the statistical model in the fusion method, and the inversion model in the fusion method, respectively. Figure 11 Figures (b), (d), and (f) are the comparison curves of the measured and estimated water storage of the three models: the single statistical model method, the statistical model in the fusion method, and the inversion model in the fusion method, respectively. It can be found that the root mean square error of the single statistical model method is relatively large, which is 50355.06 m 3 . The water storage estimated by the fusion method is in good agreement with the measured value, and the RMSEs are 6507.98 and 10604.76 m 3 respectively. Since the ponds using the statistical model in the fusion method are generally patches with smaller areas, the obvious water storage estimation error is smaller. In addition, the remote sensing inversion method based on dynamically adjustable grid sampling can not only improve the estimation accuracy of water storage, but also depict the details of the water depth change and the trend of the underwater terrain change of each pond.

[0045] In summary, by means of the above technical solutions of the present invention, through refined classification estimation and spatial distribution map generation, the present invention provides specific spatial references and data supports for regional water resources management. The water storage volumes of retrievable and non-retrievable ponds can be calculated separately, which helps to accurately grasp the water volume distribution of different types of ponds. Combined with the visualized regional water storage volume distribution map, it can provide more intuitive and detailed data basis for water resources managers, and contribute to the implementation of more accurate water resources scheduling and water area protection strategies. By combining dynamic adjustable grid sampling with remote sensing data, the present invention ensures adaptability under different geographical conditions and pond types. Through technologies such as grid sampling surveys, NDWI and water depth inversion, and statistical models, it can cover various complex terrains and water body changes, and has wide applicability. Whether it is a breeding pond or a natural pond, it can be flexibly processed and estimated according to specific circumstances, thus improving the overall efficiency and accuracy of water storage volume estimation.

[0046] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A remote sensing statistical joint estimation method for pond water storage capacity based on classification, characterized in that: include: Based on the dynamic adjustable grid sampling method, a sampling survey was conducted on the ponds in the test area, the pond density was calculated and the grid width was determined; Calculate the normalized water index, determine the water body boundary by threshold method, and perform mask processing on the survey data to extract the pond water surface boundary; According to the water surface area and width of the pond, the ponds are divided into invertible ponds and non-invertible ponds; The water depth of the invertible pond is calculated using remote sensing imaging technology, and the water storage capacity of the invertible pond is calculated based on the area of ​​the invertible pond; Construct a statistical model to estimate the water storage capacity of the irreversible pond, and calculate the water storage capacity of the irreversible pond by using the relationship between the water surface area of ​​the irreversible pond and the statistical model; The water storage capacity of invertible ponds and non-invertible ponds is merged to calculate the total water storage capacity of the region and generate a spatial distribution map of the water storage capacity.

2. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 1 is characterized in that: The method of sampling ponds in the test area based on the dynamic adjustable grid sampling method, calculating the density of ponds and determining the grid width includes: The ponds in the test area are divided into aquaculture ponds and natural ponds by type, and the sampling ratio is preset to determine the total number of samples of aquaculture ponds and natural ponds and the total number of samples as a whole; The width of the grid was determined by the density of ponds in the experimental area, and sample ponds were assigned to each grid; The grids are sorted according to the number of sample ponds in the grids, and the grids with zero sample ponds are removed; Traverse the valid grids and select a pond from each grid in turn until the overall sample quantity requirement is met.

3. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 2 is characterized in that: The density calculation formula of the pond is: ； The calculation formula of the grid width is: ； In the formula, p represents the density of ponds in the experimental area; N represents the total number of ponds in the experimental area; S 1 indicates the land area within the experimental area; a Indicates the width of the grid; Indicates the number of sample ponds in the grid; k Indicates the sampling ratio.

4. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 1 is characterized in that: The calculation of the normalized water body index, determination of the water body boundary by the threshold method, and masking of the survey data to extract the pond water surface boundary include the following steps: Use remote sensing images to calculate the normalized water index and obtain initial information on the water bodies in the test area; The threshold value of the normalized water index is determined by the threshold method. If the threshold value is greater than or equal to the preset value, the water body boundary in the test area is extracted and the boundary smoothing is performed; Based on geographic information technology, the non-pond water body types in the survey data were masked and the pond water surface boundaries were extracted.

5. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 1 is characterized in that: The method of calculating the water depth of the invertible pond by using remote sensing imaging technology and calculating the water storage capacity of the invertible pond in combination with the area of ​​the invertible pond includes: Obtain multispectral remote sensing images and use custom geographic transformation to convert the universal transverse Mercator projection coordinates of the multispectral remote sensing images into Gauss-Krüger projection coordinates; The normalized vegetation index and normalized water turbidity were calculated using multispectral remote sensing images, redundant points were eliminated, measured points were thinned out, and pixel grids and measured points were matched one by one. Extract water body reflectance, generate band combination inversion factors through band calculation, and calculate the characteristic importance of each inversion factor in combination with measured point data; The measured point data were divided into training set and validation set, and a water depth inversion model was constructed to invert the water depth of the invertible ponds. The inverted water depth value was compared with the actual measured water depth to calculate the determination coefficient and root mean square error of the evaluation index. According to the inversion results, the water depth data of each invertible pond area is extracted, the average water depth of each invertible pond is calculated, and the water storage capacity of the invertible pond is calculated by combining the water surface area and the average water depth.

6. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 5 is characterized in that: The calculation formula of the determination coefficient of the evaluation index is: ； The calculation formula of the root mean square error is: ； In the formula, y i Indicates i a truth value; Indicates i An estimated value; represents the average of the true values; Represents the total number of true values.

7. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 5 is characterized in that: The formula for calculating the average water depth of each invertible pond is: ； In the formula, Indicates i The average water depth of the invertible ponds; Indicates i Reversible ponds u The measured water depth value of each measuring point; num Indicates the total number of measurement points.

8. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 5 is characterized in that: The formula for calculating the water storage capacity of the invertible pond is: ； In the formula, V 1 indicates the total water storage capacity of the invertible pond; Indicates i The average water depth of the invertible ponds; s i Indicates i The area of ​​the invertible pond; m Represents the total number of invertible ponds.

9. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 1 is characterized in that: The formula for calculating the water storage capacity of the non-reversible pond is: ； ； In the formula, z Indicates water storage capacity; x Indicates area; a , b , c All represent the fitting parameters of the quadratic polynomial function; V 2 indicates the total water storage capacity of the non-invertible pond; z j Indicates j The water storage capacity of the irreversible ponds; n Represents the total number of non-invertible ponds.

10. The method for remote sensing statistics combined estimation of pond water storage volume by type processing according to claim 1 is characterized in that: The method of fusing the water storage of the invertible pond and the non-invertible pond, calculating the total water storage of the region, and generating a spatial distribution map of the water storage includes: The calculation results of water storage capacity of invertible ponds and non-invertible ponds were merged by using the average value method and statistical model method to obtain the total water storage capacity in the test area. Based on geographic information technology and pond water surface boundaries, a new average water depth field is added, and the average water depth value of each pond is assigned to the vector pond surface to generate a visual regional pond water storage distribution map.

Images