Pit-pond water depth calculation and storage capacity calculation method combined with multi-source data

Abstract

The invention discloses a multi-source data-combined pit-pond water depth calculation and storage capacity calculation method, which comprises the following steps of: acquiring actually measured water depth data, performing data preprocessing, and dividing the data into a training sample point set and a test sample point set; acquiring a multispectral remote sensing image of the reckoning area, and executing image preprocessing; unifying to the same coordinate system; fractal dimensions of each wave band and the combined wave band are calculated, and a water body region is segmented; establishing a mapping relation between the water depth values of the training sample points and the standardized multispectral apparent reflectance; establishing a water depth inversion model, and verifying the water depth inversion model by using the test sample point set; calculating discrete water depth points of the pit pond by using the optimal water depth inversion model; and establishing a minimum energy curved surface to obtain the water storage amount of a single pit pond, and accumulating to generate the total water storage amount. According to the method, actual measurement and remote sensing data are fused, the energy minimization water depth inversion model is constructed, accurate calculation of the water storage capacity of the pit pond is achieved, and the method has the advantages of being high in precision and high in applicability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydrology and water resources monitoring, and particularly relates to a pit pond water depth calculation and storage capacity calculation method combined with multi-source data. BACKGROUND

[0002] At present, small water bodies such as pits and ponds play an important role in agricultural irrigation, water resources scheduling and ecological environment management. The traditional methods for obtaining pit pond water depth and storage capacity mainly rely on manual measurement or single data source estimation methods, such as collecting water depth point data using a depth sounder or indirectly calculating based on water body range determination of remote sensing images. These methods have certain limitations in practical application: manual measurement is time-consuming and labor-intensive, and it is difficult to carry out efficiently in a large area; and single remote sensing data processing method is easily affected by light conditions, atmosphere and complex water body spectral characteristics, resulting in insufficient accuracy of the results.

[0003] The existing technology generally lacks the fusion and cooperation of multi-source data in the calculation of pit pond water depth spatial distribution and storage capacity, and it is difficult to simultaneously consider measurement accuracy and regional coverage. In addition, most of the existing methods focus on simple regression modeling or empirical formula calculation, ignoring the fine identification of water body boundaries, the system construction of water depth factors and the spatial continuity constraint, which easily causes water depth calculation deviation and storage capacity estimation error. Therefore, there is an urgent need for a comprehensive method that can fuse measured water depth data and remote sensing multi-source information to improve the accuracy and applicability of pit pond water depth inversion and storage capacity calculation. SUMMARY

[0004] One object of the present application is to provide a pit pond water depth calculation and storage capacity calculation method combined with multi-source data. The present application fuses measured and remote sensing data, constructs an energy minimization water depth inversion model, and realizes accurate calculation of pit pond storage capacity, with the advantages of high precision and strong applicability.

[0005] According to the pit pond water depth calculation and storage capacity calculation method combined with multi-source data according to the embodiment of the present application, the following steps are included: Collecting measured water depth data of pits and ponds in the calculation area, performing data preprocessing to generate standardized water depth data, and dividing them into a training sample point set and a test sample point set; Obtaining multi-spectral remote sensing images of the calculation area, and performing image preprocessing to generate standardized multi-spectral apparent reflectance images; Unifying the standardized water depth data and the standardized multi-spectral apparent reflectance images to the same coordinate system; Based on the standardized multi-spectral apparent reflectance images, calculating the fractal dimension of each band and combined band, determining the fractal threshold of water body and non-water body using the training sample point set, segmenting the water body area according to the fractal threshold, and generating a pit pond water surface boundary vector; A mapping relationship between the water depth values of the training sample points and the normalized multi-spectral apparent reflectance is established, and a water depth inversion factor vector is constructed; A water depth inversion model is established by taking the water depth inversion factor vector and the water depth values of the training sample points as inputs, and the water depth inversion model is verified by using the test sample point set to generate an optimal water depth inversion model; The optimal water depth inversion model is used to calculate the discrete water depth points of the pit pond to generate a discrete water depth point data set; The minimum energy surface is established by taking the pit pond water surface boundary vector as the boundary condition and the discrete water depth point data set as the inner point constraint, and the volume integral of the minimum energy surface is performed to obtain the water storage of a single pit pond, and the total water storage of the calculation area is generated by accumulating all the water storages of the pit ponds.

[0006] Optionally, the division of the training sample point set and the test sample point set specifically includes: The original measured water depth data in the pit pond of the calculation area are collected, the planar coordinates, collection time and water depth values of each sounding point are recorded, and an original measured water depth data set is formed; The original measured water depth data set is subjected to outlier rejection; The water depth data set after outlier rejection is subjected to duplicate value removal; The water depth data set after deduplication is subjected to standardization processing, and each record is standardized based on the average value and standard deviation of the water depth values in the set to generate a standardized water depth data set, which includes planar coordinates, collection time and standardized water depth values; The standardized water depth data set is divided into sample sets according to the training proportion to form the training sample point set and the test sample point set.

[0007] Optionally, the image preprocessing includes radiation correction, atmospheric correction, geometric correction, ortho correction, image fusion and noise removal.

[0008] Optionally, the generation of the pit pond water surface boundary vector specifically includes: A band set and a combined band set are established on the normalized multi-spectral apparent reflectance image, the apparent reflectance of each single-band pixel is directly recorded, and the weighted average of the apparent reflectance of each single band is taken as the apparent reflectance of the combined band, and the band set includes blue light band, green light band, red light band and near-infrared band; A multi-scale window sequence is constructed around the pixel, the covered box number is calculated at each scale by using the difference box counting method, and the relationship between the covered box number and the scale is recorded, and the fractal dimension of the pixel is generated by fitting the slope of the straight line between the logarithm of the covered box number and the logarithm of the scale; The fractal dimensions of all single bands and combined bands are summarized to construct the fractal feature vector of the pixel. The fractal feature vector is then mapped to the training sample point set at the same coordinate position. Based on the water body label and non-water body label of the training sample point set, the fractal feature vector is divided into a water body set and a non-water body set. The mean and variance of the water body set and the non-water body set are calculated for the fractal characteristic component of each pixel respectively. The ratio of the mean difference to the sum of the variances is taken as the discrimination criterion, and the value with the largest discrimination criterion is selected as the fractal threshold of the fractal characteristic component to obtain the fractal threshold set. Based on the fractal threshold set, the fusion coefficient corresponding to each fractal feature component is used to construct the pixel comprehensive judgment value, and the water body area mask is generated based on the comparison result of the pixel comprehensive judgment value and the judgment threshold. If the pixel comprehensive judgment value is greater than or equal to the judgment threshold, it is judged as a water body pixel. If the pixel comprehensive judgment value is less than the judgment threshold, it is judged as a non-water body pixel. In the water body area mask, the water body pixel is marked as 1 and the non-water body pixel is marked as 0; Connected domain screening and morphological processing are performed on the water area mask. Connected domains with an area smaller than the minimum area threshold are deleted and closed operations are performed to obtain water areas with satisfactory connectivity and smooth boundaries. Boundary tracing is performed on the outer boundary of the water area to obtain a boundary polygon sequence. The boundary polyline is simplified with a simplification tolerance to generate vertex coordinates. An ordered vector boundary is constructed according to the vertex sequence to generate the pond water surface boundary vector.

[0009] Optionally, the construction of the water depth inversion factor vector specifically includes: In a unified coordinate system, the standardized water depth data and the standardized multispectral apparent reflectance image are spatially aligned to determine the corresponding pixel position of each training sample point on the standardized multispectral apparent reflectance image. At the corresponding pixel position of the training sample point, the apparent reflectance of each single band, the apparent reflectance of the combined band, the difference factor between the two bands, and the ratio factor between the two bands are extracted, and the corresponding values ​​are matched with the water depth value of the training sample point to establish a mapping relationship between the water depth value of the training sample point and the apparent reflectance; The mapping results of each training sample point are sorted out, and for each training sample point, a factor set consisting of a single-band apparent reflectance, a combined band apparent reflectance, a difference factor between two bands, and a ratio factor between two bands is formed; For each training sample point, its factor set is arranged in a fixed order to form the water depth inversion factor vector of the training sample point. The water depth inversion factor vectors of all training sample points are summarized to form a water depth inversion factor vector set.

[0010] Optionally, the generation of the optimal water depth inversion model specifically includes: Based on the water depth inversion factor vector set and the water depth values ​​of the training sample points, the training sample point set and the test sample point set are distinguished. For each training sample point in the training sample point set, the water depth inversion factor vector is recorded as the input variable, and the water depth value is recorded as the output variable. The water depth inversion factor vectors of all training sample points are arranged in a fixed order to form a training sample point design matrix, and the water depth values ​​of the training sample points are arranged in the same order to form a training sample point target vector; The water depth inversion model is set as an energy minimization water depth inversion model based on variational constraints. The energy functional of the energy minimization water depth inversion model consists of three parts. Specifically, the first part is the weighted sum of the absolute values ​​of the differences between the predicted water depth values ​​of the training sample points and the water depth values ​​of the training sample points, which serves as a data consistency term; the second part is the sum of the squares of the differences between the predicted water depth values ​​of adjacent training sample points, which serves as a spatial smoothness constraint term; the third part is the product of the sum of the squares of the parameter vector corresponding to the water depth inversion model and the regularization coefficient, which serves as a regularization constraint term. The three parts are added together to form the energy functional; An iterative optimization method based on gradient descent is used to determine the parameter vector of the water depth inversion model by minimizing the energy functional, and the parameter estimation results under different regularization coefficients are obtained. The parameter estimation results are verified using a set of test sample points. For each test sample point, its water depth inversion factor vector is input, and the predicted water depth value is output. This is then compared with the water depth value of the test sample point. The sum of the absolute values ​​of the differences between the predicted values ​​and the water depth values ​​of all test sample points is calculated as the verification error metric. In a set of preset regularization coefficient candidate sets, the verification error metric corresponding to each regularization coefficient is calculated one by one, and the regularization coefficient with the smallest verification error metric is selected. The corresponding parameter estimation is used as the parameter of the optimal water depth inversion model to generate the optimal water depth inversion model.

[0011] Optionally, the generation of the total water storage capacity of the estimated area specifically includes: In a unified coordinate system, the calculation area of ​​a single pond is defined by the pond surface boundary vector, and the calculation area is plane triangulated to obtain a grid set consisting of the pond surface boundary vector and the discrete water depth point dataset. Boundary conditions are imposed on the boundary vector of the pond water surface, and the value of the continuous water depth function on the boundary is fixed to zero, so that the pond water surface boundary serves as a constraint boundary; An interior point constraint is imposed on each discrete water depth point located inside the calculation area in the discrete water depth point data set, so that the value of the continuous water depth function at the discrete water depth point is equal to the predicted water depth value of the discrete water depth point; Constructing an objective function for the minimum energy surface, the objective function consists of a membrane energy term that measures the change in the first-order partial derivative of the continuous water depth function within the calculation region, and a thin plate bending energy term that measures the change in the second-order partial derivative of the continuous water depth function within the calculation region. The objective function is a weighted sum of the two components. Taking the minimization of the objective function as the criterion and satisfying the boundary conditions and interior point constraints at the same time, the minimum energy surface of a single pond is obtained; The finite element method is used to discretize the minimum energy surface on the grid set. The continuous water depth function is represented by a linear combination of node basis functions, and the discrete equations are assembled. The water depth values ​​of the grid nodes are obtained by solving the equations. After obtaining the water depth value of the grid node, the volume integration is performed on the calculation area, and the vertex average method of the grid unit is used for discrete integration to obtain the water storage capacity of a single pond. Calculate the pond water storage capacity of all ponds within the estimated area and add them up to generate the total water storage capacity of the estimated area.

[0012] The beneficial effects of the present invention are: By integrating measured water depth data with multispectral remote sensing image data, the present invention establishes a method for estimating pond water depth and calculating storage capacity by combining multi-source data, overcoming the limitations of traditional methods that rely on a single data source or manual measurement. The method fully utilizes the complementary advantages of high precision of measured data and wide coverage of remote sensing data in terms of technical path. Through data preprocessing, standardization and coordinate unification, data from different sources are incorporated into the same analysis framework, ensuring the scientific nature and consistency of the estimation results. At the same time, the fractal dimension discrimination and boundary vector generation mechanism are introduced in the water body extraction process, realizing the fine identification of pond water body boundaries and accurate expression of geometric shapes, effectively improving the accuracy and reliability of water body area division.

[0013] Secondly, in terms of water depth inversion, the present invention constructs a systematic water depth inversion factor vector, which not only includes the apparent reflectivity of single bands and combined bands, but also introduces differential and ratio factors, thereby forming a richer and more comprehensive feature set. Combined with the energy minimization modeling method of variational constraints, data consistency, spatial smoothing and regularization constraints are introduced, so that the model can take into account spatial continuity and complexity control while ensuring accuracy, avoiding the overfitting and spatial mutation problems of traditional regression models. Through gradient descent iterative optimization and test sample verification, the optimal regularization coefficient can be automatically selected to obtain the optimal water depth inversion model, which significantly improves the accuracy and stability of water depth prediction.

[0014] Finally, in the water storage calculation process, the minimum energy surface and the finite element method are introduced to organically combine the water body boundary conditions and the discrete water depth point constraints, and a continuous water depth distribution function is established in the pit pond range. The surface not only meets the physical constraint of zero water depth on the boundary, but also strictly approximates the predicted point value, and realizes smooth transition between the boundary and the inner point, effectively reproducing the spatial variation of the water depth inside the pit pond. On this basis, the volume integral is used to accurately calculate the water storage of a single pit pond, and the results of all pit ponds in the region are accumulated to obtain the total storage of the calculation area. This method avoids the errors caused by boundary simplification or inaccurate water depth interpolation in traditional volume estimation, and improves the scientificity and reliability of the calculation. BRIEF DESCRIPTION OF DRAWINGS

[0015] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings: Fig. 1 A flow chart of a pit pond water depth calculation and storage calculation method combining multi-source data according to the present application; Fig. 2 A schematic diagram of a pit pond water depth calculation and storage calculation method combining multi-source data according to the present application, which uses the finite element method to discretely solve the minimum energy surface and calculate the pit pond water storage. DETAILED DESCRIPTION

[0016] The application will now be described in further detail with reference to the drawings. These drawings are simplified schematic diagrams that only schematically illustrate the basic structure of the application, and therefore only show the components related to the application.

[0017] Reference Figs. 1-2 A pit pond water depth calculation and storage calculation method combining multi-source data, comprising the following steps: Collecting measured water depth data of pit ponds in the calculation area, performing data preprocessing, generating standardized water depth data, and dividing into a training sample point set and a test sample point set; Obtaining multispectral remote sensing images of the calculation area, and performing image preprocessing to generate standardized multispectral apparent reflectance images; Unifying the standardized water depth data and the standardized multispectral apparent reflectance images to the same coordinate system; Based on the standardized multispectral apparent reflectance images, calculating the fractal dimension of each band and combined band, determining the fractal threshold of water body and non-water body using the training sample point set, segmenting the water body area according to the fractal threshold, and generating pit pond water surface boundary vectors; Establishing a mapping relationship between the training sample point water depth values and the standardized multispectral apparent reflectance, and constructing a water depth inversion factor vector; Taking the water depth inversion factor vector and the water depth values ​​of the training sample points as input, a water depth inversion model is established, and the water depth inversion model is verified using the test sample point set to generate the optimal water depth inversion model; The optimal water depth inversion model is used to estimate the discrete water depth points of the pond and generate a discrete water depth point dataset; The minimum energy surface was established with the pond water surface boundary vector as the boundary condition and the discrete water depth point dataset as the interior point constraint. The volume integral of the minimum energy surface was performed within the pond water surface boundary to obtain the water storage capacity of each pond. The water storage capacity of all ponds was then accumulated to generate the total water storage capacity of the estimated area.

[0018] In this embodiment, the division of the training sample point set and the test sample point set specifically includes: Collect the original measured water depth data in the ponds in the estimation area, record the plane coordinates, collection time and water depth value of each sounding point, and form the original measured water depth data set; Perform outlier removal on the original measured water depth data set; Perform duplicate value removal on the water depth data set after outliers are removed; Performing standardization on the deduplicated water depth data set, standardizing each record based on the mean and standard deviation of the water depth values ​​in the set to generate a standardized water depth data set, which includes plane coordinates, acquisition time, and standardized water depth values; The standardized water depth data set is divided into samples according to the training ratio to form a training sample point set and a test sample point set.

[0019] In this embodiment, the image preprocessing includes radiation correction, atmospheric correction, geometric correction, orthorectification, image fusion and noise removal.

[0020] In this embodiment, the generation of the pond water surface boundary vector specifically includes: A band set and a combined band set are established on the standardized multispectral apparent reflectance image, the pixel apparent reflectance of each single band is directly recorded, and for each combined band, the weighted average of the apparent reflectances of each single band is used as the combined band apparent reflectance, wherein the band set includes a blue band, a green band, a red band, and a near-infrared band; A multi-scale window sequence is constructed with the pixel as the center. The number of covered boxes is calculated using the differential box counting method at each scale, and the relationship between the number of covered boxes and the scale is recorded. The fractal dimension of the pixel is generated by fitting the slope of the straight line between the logarithm of the number of covered boxes and the logarithm of the scale. The generation of the fractal dimension of the pixel specifically includes: In the standardized multi-spectral apparent reflectance image, a sequence of multi-scale windows centered on the pixel is established for each pixel, and the side length of the window is taken as an odd number of pixels and increases gradually; In each scale window, the differential box counting method is performed on the single-band apparent reflectance and the combined-band apparent reflectance respectively, the apparent reflectance in the scale window is linearly normalized to a fixed gray level range according to the minimum apparent reflectance and the maximum apparent reflectance, and is divided into several intensity layers; The scale window is divided into equal sub-grids, the maximum layer number and the minimum layer number of the intensity layer are read in each sub-grid, and the difference between the two plus one is taken as the cover box number of the sub-grid, and the cover box numbers of all sub-grids under the scale window are summed to obtain the cover box number; The corresponding relationship between the cover box number and the scale is recorded for all scales, and the least square linear fitting is performed with the logarithm of the cover box number as the vertical coordinate and the logarithm of the scale as the horizontal coordinate, and the slope of the fitting straight line is taken as the fractal dimension of the target pixel under the single-band apparent reflectance or the combined-band apparent reflectance; When the multi-scale window is out of bounds, the neighborhood padding method is used to fill in, when the window contains invalid pixels of cloud or shadow mask, the current scale calculation of the pixel is skipped, and when the number of valid scales is insufficient, the fractal dimension of the pixel is not output; Finally, the fractal dimension is obtained under all single-band apparent reflectance and combined-band apparent reflectance of each pixel respectively; The fractal dimensions of all single-band and combined-band are summarized to construct the fractal feature vector of the pixel, and the fractal feature vector is divided into water body set and non-water body set according to the water body label and non-water body label of the training sample point set, the water body label and non-water body label of the training sample point set represent that the sample points are classified based on the measured water depth value, and whether the sample point belongs to the water area is marked, if the measured water depth value of the sample point is greater than zero, the sample point is in the water area, and is marked as water body label, if the measured water depth value of the sample point is equal to zero or the sample point is in the non-water area, it is marked as non-water body label; The mean and variance of the water body set and the non-water body set are calculated for each fractal feature component respectively, the ratio of the mean difference to the sum of the variances is taken as the discrimination criterion, and the value with the maximum discrimination criterion is selected as the fractal threshold of the fractal feature component, and the fractal threshold set is obtained, the fractal feature component is an element of the fractal feature vector; Based on the fractal threshold set, the fusion coefficient corresponding to each fractal feature component is used to construct the pixel comprehensive judgment quantity, and the water body area mask is generated according to the comparison result of the pixel comprehensive judgment quantity and the judgment threshold, if the pixel comprehensive judgment quantity is greater than or equal to the judgment threshold, it is determined as a water body pixel, if the pixel comprehensive judgment quantity is less than the judgment threshold, it is determined as a non-water body pixel, the water body pixel in the water body area mask is marked as 1, and the non-water body pixel is marked as 0; The connected domain screening and morphological processing are performed on the water area mask, the connected domains with an area less than a minimum area threshold are deleted, and a close operation is performed, so that the water area with required connectivity and smooth boundary is obtained; The boundary tracking is performed on the outer boundary of the water area, a sequence of boundary polylines is obtained, the vertices are simplified according to a simplification tolerance, the vertex coordinates are generated, the ordered vector boundary is constructed according to the order of the vertices, and the pit pond water surface boundary vector is generated. Specifically, the boundary tracking operation is performed on the outer boundary of the water area, the boundary positions of adjacent water pixels are recorded in sequence, and a continuous sequence of boundary polylines is obtained. On the basis of the sequence of boundary polylines, the simplification tolerance is set, the boundary polylines are simplified, the redundant vertices are reduced under the premise of ensuring the accurate expression of the boundary geometry, the vertex coordinates of the simplified boundary polylines are generated, the ordered vector boundary is constructed according to the order of the vertices, and finally the ordered vector boundaries of the water areas are collected to generate the pit pond water surface boundary vector. The simplification tolerance represents the maximum geometric deviation allowed in the simplification process of the water boundary vector polylines, and is used to balance between accuracy and data complexity.

[0021] In the embodiment, the construction of the water depth inversion factor vector specifically includes: In the unified coordinate system, the normalized water depth data is spatially corresponded to the normalized multi-spectral apparent reflectance image, and the corresponding pixel position of each training sample point on the normalized multi-spectral apparent reflectance image is determined; At the corresponding pixel position of the training sample point, the apparent reflectance of each single band, the apparent reflectance of the combined band, the difference factor between the two bands and the ratio factor between the two bands are extracted, and are corresponded to the water depth value of the training sample point, so as to establish the mapping relationship between the water depth value and the apparent reflectance of the training sample point; The mapping results of each training sample point are arranged, and for each training sample point, a factor set composed of the apparent reflectance of the single band, the apparent reflectance of the combined band, the difference factor between the two bands and the ratio factor between the two bands is formed; For each training sample point, the factor set is arranged in a fixed order to form the water depth inversion factor vector of the training sample point, and the water depth inversion factor vectors of all training sample points are collected to form a water depth inversion factor vector set.

[0022] In the embodiment, the generation of the optimal water depth inversion model specifically includes: Based on the water depth inversion factor vector set and the water depth values of the training sample points, the training sample point set and the test sample point set are distinguished, for each training sample point in the training sample point set, the water depth inversion factor vector is recorded as an input variable, and the water depth value is recorded as an output variable; The water depth inversion factor vectors of all training sample points are arranged in a fixed order to form a training sample point design matrix, and the water depth values ​​of the training sample points are arranged in the same order to form a training sample point target vector; The water depth inversion model is set as an energy minimization water depth inversion model based on variational constraints. The energy functional of the energy minimization water depth inversion model consists of three parts. Specifically, the first part is the weighted sum of the absolute values ​​of the differences between the predicted water depth values ​​of the training sample points and the water depth values ​​of the training sample points, which serves as a data consistency term; the second part is the sum of the squares of the differences between the predicted water depth values ​​of adjacent training sample points, which serves as a spatial smoothness constraint term; the third part is the product of the sum of the squares of the parameter vector corresponding to the water depth inversion model and the regularization coefficient, which serves as a regularization constraint term. The three parts are added together to form the energy functional; The energy minimization water depth inversion model based on variational constraints refers to a process in which, in the water depth inversion process, the water depth inversion factor vector of the training sample point set and the water depth value of the training sample point are used as the input and output relationship to construct an energy functional model including data consistency constraints, spatial smoothness constraints and regularization constraints, and the parameter vector of the water depth inversion model is determined by minimizing the energy functional. The data consistency constraint is used to measure the difference between the water depth value predicted by the model and the water depth value of the training sample point. The spatial smoothness constraint is used to maintain the continuity of the predicted water depth values ​​of adjacent training sample points. The regularization constraint is used to prevent the model parameter vector from being too large or overfitting. By minimizing the energy functional, it is possible to maintain the smoothness of the spatial distribution and take into account the control of model complexity while ensuring the consistency of the predicted water depth value and the water depth value of the training sample point, thereby obtaining the optimal water depth inversion model. The predicted water depth value of the training sample point is the output value obtained by inputting the water depth inversion factor vector of the training sample point into the current water depth inversion model and performing model parameter calculation; An iterative optimization method based on gradient descent is used to determine the parameter vector of the water depth inversion model by minimizing the energy functional, and the parameter estimation results under different regularization coefficients are obtained. The parameter estimation results are verified using a set of test sample points. For each test sample point, its water depth inversion factor vector is input, and the predicted water depth value is output. This is then compared with the water depth value of the test sample point. The sum of the absolute values ​​of the differences between the predicted values ​​and the water depth values ​​of all test sample points is calculated as the verification error metric. In a set of preset regularization coefficient candidate sets, the verification error metric corresponding to each regularization coefficient is calculated one by one, and the regularization coefficient with the smallest verification error metric is selected. The corresponding parameter estimation is used as the parameter of the optimal water depth inversion model to generate the optimal water depth inversion model.

[0023] In this embodiment, the generation of the total water storage capacity of the estimated area specifically includes: In a unified coordinate system, a calculation area of a single pit pond is defined by pit pond water surface boundary vectors, and the calculation area is planarly triangulated to obtain a grid set composed of pit pond water surface boundary vectors and a discrete water depth point data set; A boundary condition is applied to the pit pond water surface boundary vectors, and the value of the continuous water depth function on the boundary is fixed to zero, so that the pit pond water surface boundary serves as a constraint boundary; The continuous water depth function refers to a two-dimensional function covering the range of the pit pond water surface boundary in a unified coordinate system, the input is a plane coordinate, and the output is a water depth value corresponding to the position. At the discrete water depth points, the value of the continuous water depth function is consistent with the predicted water depth value. On the pit pond water surface boundary vectors, the value of the continuous water depth function is zero. Between the boundary and the discrete points, the continuous water depth function forms a smooth transition through an energy minimization method, which is used to describe the spatial distribution of the water depth inside the pit pond. An inner point constraint is applied to each discrete water depth point in the discrete water depth point data set located inside the calculation area, so that the value of the continuous water depth function at the discrete water depth point is equal to the predicted water depth value of the discrete water depth point; A target function of the minimum energy surface is constructed, which is composed of a membrane energy term and a thin plate bending energy term. The membrane energy term measures the first-order partial derivative change of the continuous water depth function in the calculation area. The thin plate bending energy term measures the second-order partial derivative change of the continuous water depth function in the calculation area. The target function is a weighted sum of the two parts: ; Wherein, The target function of the minimum energy surface is represented by The calculation area defined by the pit pond water surface boundary vectors is represented by The continuous water depth function is represented by , The first-order partial derivative of the continuous water depth function in the and directions is represented by , , The second-order partial derivative of the continuous water depth function is represented by , The weight coefficient is represented by The importance of the membrane energy term and the thin plate bending energy term is balanced. The minimum energy surface of a single pit pond is obtained by taking the minimum of the target function as the criterion and simultaneously satisfying the boundary condition and the inner point constraint. In the minimum energy surface construction process, the boundary condition refers to the value of the continuous water depth function at the position corresponding to the pit pond water surface boundary vector must be equal to zero, to ensure that the water depth at the water surface boundary is zero, the inner point constraint refers to the value of the continuous water depth function at the position of the discrete water depth point data set must be equal to the predicted water depth value of the discrete water depth point, to ensure that the model is strictly consistent with the prediction result at these positions, the boundary condition and the inner point constraint are satisfied simultaneously, which means that when solving the minimum energy surface, the optimization process not only requires the energy minimization of the continuous water depth function in the whole region, but also requires the boundary condition of zero water depth at the boundary and the inner point constraint of consistent with the predicted water depth value at all discrete water depth points, so that the final continuous water depth function not only meets the physical characteristics of the boundary, but also maintains the consistency with the discrete point data; The minimum energy surface is a specific solution that simultaneously satisfies the boundary condition and the inner point constraint and minimizes the objective function among all continuous water depth functions, which is strictly zero at the boundary, strictly equal to the predicted water depth value at the discrete water depth points, and filled by smooth transition at other positions, thereby generating a water depth distribution surface that meets the physical boundary characteristics and the observation point constraints; The minimum energy surface is discretely solved on the grid set by using the finite element method, the continuous water depth function is represented by a linear combination of node basis functions, and a discrete equation set is assembled to obtain the water depth values of the grid nodes by solving the equation set; In a unified coordinate system, a grid set is constructed in the calculation region defined by the pit pond water surface boundary vector, the continuous water depth function is represented by node basis functions at the nodes of each grid element, specifically, nodes are selected in each grid element and node basis functions corresponding to the nodes are defined, the continuous water depth function is represented by a linear combination of node basis functions, on this basis, the objective function of the minimum energy surface is discretized on the grid set to form algebraic equations for the water depth values of each node, and all algebraic equations corresponding to the grid elements are assembled in turn to obtain a discrete equation set covering the entire calculation region, the boundary conditions corresponding to the pit pond water surface boundary vector and the inner point constraints corresponding to the discrete water depth point data set are applied to the discrete equation set, the water depth values of the boundary nodes and the inner point nodes are fixed, and finally the numerical solving method is used to solve the discrete equation set to obtain the water depth value distribution of all nodes in the grid set; The node basis function is a local function defined on the grid set element in the finite element method, which takes value 1 at the node and value 0 at other nodes, and is used to represent the continuous water depth function as a linear combination of node water depth values; Within the calculation area defined by the pond water surface boundary vector, the continuous water depth function is represented on the grid set as a linear combination of nodal basis functions. That is, within each grid cell, the nodal water depth value is used as a coefficient and the nodal basis function is used to represent the continuous water depth function. On this basis, the integral term involving the continuous water depth function in the objective function of the minimum energy surface is replaced by a linear combination of nodal basis functions, and the integral operation is converted into a finite summation of each grid cell in the grid set. For each grid cell, the local stiffness matrix and local load vector are calculated separately, and the local stiffness matrices and local load vectors of all grid cells are assembled into a discrete equation system covering the entire grid set. Finally, a system of algebraic equations that depends only on the nodal water depth value is obtained, thereby realizing the discretization of the minimum energy surface objective function. The local stiffness matrix is ​​a coefficient matrix obtained by discretizing the membrane energy term and the thin plate bending energy term in the objective function within a single grid unit, and then taking the derivative and integrating the node basis function; The local load vector is a vector obtained by integrating the node basis function after discretizing the terms related to the external constraints or boundary conditions in the objective function within a single grid unit; After obtaining the water depth value of the grid node, the volume integration is performed on the calculation area, and the vertex average method of the grid unit is used for discrete integration to obtain the water storage capacity of a single pond. Calculate the pond water storage capacity of all ponds within the estimated area and add them up to generate the total water storage capacity of the estimated area.

[0024] Example 1: In order to verify the feasibility of the present invention in implementation, the present invention was applied to a contiguous pond water system in the suburbs of a certain city. There are many natural and artificial water bodies of various shapes and areas distributed in the area. The local water resources management department has long faced the problem of lacking a precise grasp of the distribution of pond water depth and actual water storage capacity. Previous work mostly used manual measurement of water depth combined with empirical methods to estimate water storage capacity, which is time-consuming, labor-intensive, and difficult to cover a large area. When the water level fluctuates frequently during the rainy season, the timeliness and accuracy of the data are even more difficult to guarantee. At the same time, the simple area estimation method based on remote sensing water body identification is affected by lighting conditions, changes in water body reflectivity, and obstruction by shoreline vegetation. The errors in the calculation of water body boundaries and reserves are large, which cannot meet the current needs of high-precision water resource management and scheduling.

[0025] In this scenario, field measurements were conducted on multiple ponds in the area. Portable depth sounders and high-precision locators were used to obtain the original water depth data of several sampling points in the ponds. The coordinates and time information of the measurement points were accurately recorded. The collected measured data were preliminarily screened, and abnormal records were eliminated. After removing duplicate measurement points, standardized methods were used for processing. All sounding point data were unified to the same benchmark and randomly divided into training and test sets according to a predetermined ratio to provide support for subsequent model construction and verification.

[0026] At the same time, data collection was conducted across the entire region using high-resolution multispectral remote sensing imagery from the cloudless day. The images were corrected for radiation, atmosphere, geometry, and orthorectification to ensure that the reflectance of each pixel accurately reflected the actual surface conditions. Further noise removal and image fusion were performed to create standardized multispectral apparent reflectance images corresponding to the measured points. To facilitate subsequent analysis, all data were projected into a unified coordinate system, achieving a seamless integration of remote sensing and measured data.

[0027] In remote sensing images, based on the distribution characteristics of ponds, a multi-scale window sequence including blue light, green light, red light and near-infrared bands was constructed. The number of covering boxes at each scale was calculated by the differential box counting method, and the fractal dimension of each pixel was fitted. The fractal dimensions of all single bands and combined bands were sorted out. Combined with the water body and non-water body markers of the measured water depth point set, the differences in fractal characteristics between water bodies and non-water bodies were systematically analyzed, and the most discriminative fractal threshold set was extracted. Based on the fractal threshold, a comprehensive water body judgment value was constructed, and a water body mask was generated for the remote sensing image. After connected domain screening and morphological processing, noise and small area clutter were effectively removed, and finally a water surface boundary vector that conforms to the actual pond shoreline morphology was formed. Through boundary tracking and simplified tolerance processing, the complex boundary was converted into ordered vector data with clear structure and refined vertices, laying a solid foundation for subsequent spatial modeling.

[0028] In the water depth inversion stage, the standardized water depth data is accurately spatially aligned with the multispectral image. With each training sample point as the center, the single-band and combined-band apparent reflectance of the corresponding pixel is extracted, the difference and ratio factors are calculated, and a multidimensional water depth inversion factor vector set is constructed. These factors fully reflect the potential correlation between different spectral information and water depth, providing a rich representation for the model input layer. Based on the training set, a water depth inversion model based on energy minimization variational constraints is established. This model not only constrains the difference between the predicted water depth and the measured water depth, but also combines the spatial smoothness of adjacent measurement points and parameter regularization requirements to effectively prevent overfitting and extreme value fluctuations. The model is verified for multiple rounds using the test set data. By adjusting the regularization coefficient, the model achieves the optimal balance between accuracy and generalization ability, and ultimately determines the optimal water depth inversion model parameters in the region.

[0029] Applying the optimal water depth inversion model, the water depth values ​​of all pixels within the water body range of the remote sensing image are batch-calculated to obtain a discrete water depth point dataset covering the entire pond area. These estimated values ​​not only retain the accuracy of the measured points, but also can achieve high-precision extrapolation in non-measured areas based on remote sensing characteristics. With the water surface boundary vector as the boundary condition and the discrete water depth points as the interior point constraints, the finite element method is used to construct the minimum energy surface within the pond range. The continuous water depth function is discretized through the node basis function, and the stiffness matrix and load vector are assembled for each grid unit to form a discrete set of equations. After imposing the zero water depth condition of the boundary nodes and the water depth constraint of the interior point nodes, the numerical method is used to efficiently solve the water depth distribution of all nodes.

[0030] Finally, for each pond, within the bounded calculation area, the vertex averaging method was used to perform a volume integration based on the obtained node water depths to estimate its respective water storage capacity. The water storage capacity of all ponds in the entire region was then accumulated to determine the total water storage capacity of the target area, providing a scientific basis for practical applications such as water resource scheduling, irrigation management, and drought early warning.

[0031] In order to verify the performance of the present invention in practice, a comparison was made with the traditional method, and the results are shown in Table 1.

[0032] Table 1 Comparison of the effects of the method of the present invention and the traditional method in estimating the water depth and water storage capacity of the pond

[0033] It can be seen intuitively from Table 1 that the multi-source data fusion and energy minimization model method proposed in the present invention is superior to traditional manual measurement estimation and single remote sensing image estimation in the task of pond water depth and water storage capacity estimation.

[0034] In terms of water depth error, traditional manual methods have an average error of up to 0.32 meters due to sparse measurement points and human error. Single remote sensing methods, limited by water spectral variations and resolution, can only reduce the error to 0.28 meters. However, this invention combines measured water depth points with remote sensing multispectral information, combining fractal feature boundary recognition, spatial smoothing, and energy minimization inversion to reduce the water depth estimation error to 0.10 meters, achieving higher estimation accuracy.

[0035] In terms of the relative error of water storage capacity, the traditional method has a calculation error of 14.7% due to the difficulty in accurately obtaining the boundary; the single remote sensing method has an error of 11.3% due to the limited extrapolation of water depth. The method of the present invention reduces the cumulative error through minimum energy surface, finite element discrete solution and high-density discrete point extrapolation under spatial constraints. The relative error of water storage capacity is only 3.8%, which can provide more reliable data support for actual scheduling and ecological management.

[0036] The work efficiency and the spatial resolution also embody the advantages of the present application. The work time of the traditional manual method is extremely long, 22 hours are needed for measuring 10 pits and ponds at a time, and the spatial resolution is difficult to cover all the areas. Although the speed of the single remote sensing method is improved, the resolution is limited by the sensor, and the processing of the boundaries of some small water bodies is still insufficient. The present application uses automatic remote sensing interpretation, mathematical modeling and limited manual assistance, greatly shortens the work time to 2.5 hours, improves the spatial resolution to 2 meters, and realizes the comprehensive and fine coverage of small pits and ponds.

[0037] In terms of human participation, the present application minimizes the on-site requirements, only one person is needed to complete data collection and verification, which reduces the labor intensity of the management department, reduces the error caused by human interference, and improves the automation and intelligent level of the process.

[0038] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for calculating pond water depth and storage capacity by combining multi-source data, characterized in that: The steps include: Collect measured water depth data of ponds in the estimation area, perform data preprocessing, generate standardized water depth data, and divide it into training sample point set and test sample point set; Obtain multispectral remote sensing images of the estimated area and perform image preprocessing to generate standardized multispectral apparent reflectance images; Unify the standardized water depth data and the standardized multispectral apparent reflectance image into the same coordinate system; Based on the standardized multispectral apparent reflectance image, the fractal dimension of each band and the combined band is calculated. The fractal threshold between water and non-water bodies is determined using the training sample point set. The water area is segmented based on the fractal threshold to generate the pond water surface boundary vector. Establish a mapping relationship between the water depth values ​​of the training sample points and the standardized multispectral apparent reflectance, and construct a water depth inversion factor vector; Taking the water depth inversion factor vector and the water depth values ​​of the training sample points as input, a water depth inversion model is established, and the water depth inversion model is verified using the test sample point set to generate the optimal water depth inversion model; The optimal water depth inversion model is used to estimate the discrete water depth points of the pond and generate a discrete water depth point dataset; The minimum energy surface was established with the pond water surface boundary vector as the boundary condition and the discrete water depth point dataset as the interior point constraint. The volume integral of the minimum energy surface was performed to obtain the water storage capacity of a single pond. The water storage capacity of all ponds was then accumulated to generate the total water storage capacity of the estimated area.

2. The method for calculating pond water depth and storage capacity by combining multi-source data according to claim 1 is characterized in that: The division of the training sample point set and the test sample point set specifically includes: Collect the original measured water depth data in the ponds in the estimation area, record the plane coordinates, collection time and water depth value of each sounding point, and form the original measured water depth data set; Perform outlier removal on the original measured water depth data set; Perform duplicate value removal on the water depth data set after outliers are removed; Performing standardization on the deduplicated water depth data set, standardizing each record based on the mean and standard deviation of the water depth values ​​in the set to generate a standardized water depth data set, which includes plane coordinates, acquisition time, and standardized water depth values; The standardized water depth data set is divided into samples according to the training ratio to form a training sample point set and a test sample point set.

3. The method for calculating pond water depth and storage capacity by combining multi-source data according to claim 1, characterized in that: The image preprocessing includes radiation correction, atmospheric correction, geometric correction, orthorectification, image fusion and noise removal.

4. The method for calculating pond water depth and storage capacity by combining multi-source data according to claim 1, characterized in that: The generation of the pond water surface boundary vector specifically includes: A band set and a combined band set are established on the standardized multispectral apparent reflectance image, the pixel apparent reflectance of each single band is directly recorded, and for each combined band, the weighted average of the apparent reflectances of each single band is used as the combined band apparent reflectance, wherein the band set includes a blue band, a green band, a red band, and a near-infrared band; A multi-scale window sequence is constructed with the pixel as the center. The number of covered boxes is calculated using the differential box counting method at each scale, and the relationship between the number of covered boxes and the scale is recorded. The fractal dimension of the pixel is generated by fitting the slope of the straight line between the logarithm of the number of covered boxes and the logarithm of the scale. The fractal dimensions of all single bands and combined bands are summarized to construct the fractal feature vector of the pixel. The fractal feature vector is then mapped to the training sample point set at the same coordinate position. Based on the water body label and non-water body label of the training sample point set, the fractal feature vector is divided into a water body set and a non-water body set. The mean and variance of the water body set and the non-water body set are calculated for the fractal characteristic component of each pixel respectively. The ratio of the mean difference to the sum of the variances is taken as the discrimination criterion, and the value with the largest discrimination criterion is selected as the fractal threshold of the fractal characteristic component to obtain the fractal threshold set. Based on the fractal threshold set, the fusion coefficient corresponding to each fractal feature component is used to construct the pixel comprehensive judgment value, and the water body area mask is generated based on the comparison result of the pixel comprehensive judgment value and the judgment threshold. If the pixel comprehensive judgment value is greater than or equal to the judgment threshold, it is judged as a water body pixel. If the pixel comprehensive judgment value is less than the judgment threshold, it is judged as a non-water body pixel. In the water body area mask, the water body pixel is marked as 1 and the non-water body pixel is marked as 0; Connected domain screening and morphological processing are performed on the water area mask. Connected domains with an area smaller than the minimum area threshold are deleted and closed operations are performed to obtain water areas with satisfactory connectivity and smooth boundaries. Boundary tracing is performed on the outer boundary of the water area to obtain a boundary polygon sequence. The boundary polyline is simplified with a simplification tolerance to generate vertex coordinates. An ordered vector boundary is constructed according to the vertex sequence to generate the pond water surface boundary vector.

5. The method for calculating pond water depth and storage capacity by combining multi-source data according to claim 1, characterized in that: The construction of the water depth inversion factor vector specifically includes: In a unified coordinate system, the standardized water depth data and the standardized multispectral apparent reflectance image are spatially aligned to determine the corresponding pixel position of each training sample point on the standardized multispectral apparent reflectance image. At the corresponding pixel position of the training sample point, the apparent reflectance of each single band, the apparent reflectance of the combined band, the difference factor between the two bands, and the ratio factor between the two bands are extracted, and the corresponding values ​​are matched with the water depth value of the training sample point to establish a mapping relationship between the water depth value of the training sample point and the apparent reflectance; The mapping results of each training sample point are sorted out, and for each training sample point, a factor set consisting of a single-band apparent reflectance, a combined band apparent reflectance, a difference factor between two bands, and a ratio factor between two bands is formed; For each training sample point, its factor set is arranged in a fixed order to form the water depth inversion factor vector of the training sample point. The water depth inversion factor vectors of all training sample points are summarized to form a water depth inversion factor vector set.

6. The method for calculating pond water depth and storage capacity by combining multi-source data according to claim 1, characterized in that: The generation of the optimal water depth inversion model specifically includes: Based on the water depth inversion factor vector set and the water depth values ​​of the training sample points, the training sample point set and the test sample point set are distinguished. For each training sample point in the training sample point set, the water depth inversion factor vector is recorded as the input variable, and the water depth value is recorded as the output variable. The water depth inversion factor vectors of all training sample points are arranged in a fixed order to form a training sample point design matrix, and the water depth values ​​of the training sample points are arranged in the same order to form a training sample point target vector; The water depth inversion model is set as an energy minimization water depth inversion model based on variational constraints. The energy functional of the energy minimization water depth inversion model consists of three parts. Specifically, the first part is the weighted sum of the absolute values ​​of the differences between the predicted water depth values ​​of the training sample points and the water depth values ​​of the training sample points, which serves as a data consistency term; the second part is the sum of the squares of the differences between the predicted water depth values ​​of adjacent training sample points, which serves as a spatial smoothness constraint term; the third part is the product of the sum of the squares of the parameter vector corresponding to the water depth inversion model and the regularization coefficient, which serves as a regularization constraint term. The three parts are added together to form the energy functional; An iterative optimization method based on gradient descent is used to determine the parameter vector of the water depth inversion model by minimizing the energy functional, and the parameter estimation results under different regularization coefficients are obtained. The parameter estimation results are verified using a set of test sample points. For each test sample point, its water depth inversion factor vector is input, and the predicted water depth value is output. This is then compared with the water depth value of the test sample point. The sum of the absolute values ​​of the differences between the predicted values ​​and the water depth values ​​of all test sample points is calculated as the verification error metric. In a set of preset regularization coefficient candidate sets, the verification error metric corresponding to each regularization coefficient is calculated one by one, and the regularization coefficient with the smallest verification error metric is selected. The corresponding parameter estimation is used as the parameter of the optimal water depth inversion model to generate the optimal water depth inversion model.

7. The method for estimating pond water depth and calculating storage capacity by combining multi-source data according to claim 1 is characterized in that: The generation of the total water storage capacity of the estimated area specifically includes: In a unified coordinate system, the calculation area of ​​a single pond is defined by the pond surface boundary vector, and the calculation area is plane triangulated to obtain a grid set consisting of the pond surface boundary vector and the discrete water depth point dataset. Boundary conditions are imposed on the boundary vector of the pond water surface, and the value of the continuous water depth function on the boundary is fixed to zero, so that the pond water surface boundary serves as a constraint boundary; An interior point constraint is imposed on each discrete water depth point located inside the calculation area in the discrete water depth point data set, so that the value of the continuous water depth function at the discrete water depth point is equal to the predicted water depth value of the discrete water depth point; Constructing an objective function for the minimum energy surface, the objective function consists of a membrane energy term that measures the change in the first-order partial derivative of the continuous water depth function within the calculation region, and a thin plate bending energy term that measures the change in the second-order partial derivative of the continuous water depth function within the calculation region. The objective function is a weighted sum of the two components. Taking the minimization of the objective function as the criterion and satisfying the boundary conditions and interior point constraints at the same time, the minimum energy surface of a single pond is obtained; The finite element method is used to discretize the minimum energy surface on the grid set. The continuous water depth function is represented by a linear combination of node basis functions, and the discrete equations are assembled. The water depth values ​​of the grid nodes are obtained by solving the equations. After obtaining the water depth value of the grid node, the volume integration is performed on the calculation area, and the vertex average method of the grid unit is used for discrete integration to obtain the water storage capacity of a single pond. Calculate the pond water storage capacity of all ponds within the estimated area and add them up to generate the total water storage capacity of the estimated area.

Images