A method for estimating the water storage of a blocked pit

Abstract

The application discloses a method for estimating water storage capacity of a pond with shelter. According to remote sensing images of a target area, a graph spot of a target pond is extracted, a graph spot area is determined, a remote sensing extraction water surface area and a remote sensing extraction vegetation area of the target pond are determined; at least based on the remote sensing extraction water surface area, a shelter degree of the target pond is determined; according to the graph spot area, a pond ridge area, the remote sensing extraction water surface area, the remote sensing extraction vegetation area and the shelter degree, a target pond water surface area is determined; based on the pond ridge area, an outer edge line is extracted; according to the shelter degree, a water depth of the target pond is determined; according to the water depth of the target pond and a bottom surface range, a pond bottom surface water depth model is determined; with the water depth of the outer edge line as zero and the pond bottom surface water depth model as a constraint condition, a pond water depth model is constructed; based on the target pond water surface area and the pond water depth model, a pond water storage capacity is determined. The application can improve the estimation precision of the water storage capacity of the pond with shelter.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of water storage capacity estimation, and particularly relates to a method for estimating water storage capacity of a pond with shelter. BACKGROUND

[0002] Ponds store water through rainwater collection, water diversion or groundwater, etc. They not only bear the function of drought and flood regulation, but also play an important role in flood regulation, ecological protection and water purification. Investigation of pond water storage capacity can provide theoretical support for rational development and utilization of regional water resources, and is crucial to realize joint utilization of multiple water sources in agriculture.

[0003] Common methods for investigating pond water storage capacity include the following: 1. A field investigation method for calculating pond water storage capacity by measuring the water surface area, water depth and terrain of the pond. This method is accurate, but is time-consuming and labor-intensive, and is not suitable for large-scale investigation. 2. A method for estimating pond water storage capacity by measuring water depth at a small number of sample points based on sampling statistical theory, and then estimating the water storage capacity of the pond by analogy with the grid. This method is relatively simple and low in cost, but its accuracy depends on the sample size and the rationality of sample design. 3. A remote sensing water depth inversion method for estimating pond water storage capacity by extracting water surface area, water depth and other water features from aerial images or satellite remote sensing images. This method is suitable for large-scale investigation, but due to the shelter of vegetation, shadow and hard surface (such as a breeding pond with a built roof), the remote sensing image cannot obtain the water depth information of the sheltered area, resulting in underestimation of the water storage capacity. SUMMARY

[0004] The technical purpose of the present application is to provide a method for estimating water storage capacity of a pond with shelter, which aims to solve the technical problem of underestimation of pond water storage capacity caused by the inability to estimate water depth in the sheltered area, and to realize accurate estimation of pond water storage capacity based on remote sensing means.

[0005] To achieve the above technical purpose, the present application adopts the following technical solution.

[0006] The present application provides a method for estimating water storage capacity of a pond with shelter, which comprises:

[0007] According to the remote sensing image of the target area, the map patch of the target pond is extracted and the patch area is determined;

[0008] According to the remote sensing image, the remote sensing extraction water surface area and the remote sensing extraction vegetation area of the target pond are determined; and at least based on the remote sensing extraction water surface area, the shelter degree of the target pond is determined.

[0009] determine a water surface area of the target pit pond according to the plot area, the dike area, the water surface area extracted by the remote sensing, the vegetation area extracted by the remote sensing, and the shielding degree;

[0010] extract an outer edge line of the target pit pond based on the dike area, determine a bottom surface range of the target pit pond, determine a water depth of the target pit pond according to the shielding degree, and determine a pit pond bottom surface water depth model according to the water depth of the target pit pond and the bottom surface range;

[0011] construct a pit pond water depth model with the water depth of the outer edge line being zero and the pit pond bottom surface water depth model as constraint conditions;

[0012] determine a pit pond water storage capacity based on the water surface area of the target pit pond and the pit pond water depth model.

[0013] Compared with the prior art, the application has the beneficial technical effects: the shielding degree of the target pit pond is analyzed by remote sensing image, the actual water surface area of the pit pond can be more accurately reflected, multiple data sources such as the plot area, the dike area, the water surface area extracted by the remote sensing, and the vegetation area are combined for comprehensive analysis, the water depth is determined according to the shielding degree, and the estimation accuracy of the water storage capacity of the pit pond with shielding can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] The drawings described herein are for illustrative purposes only, and are not intended to limit the scope of the present disclosure in any way. In addition, the shapes and scale sizes of the components in the drawings are only illustrative, and are used to help understand the present application, and are not specific limitations on the shapes and scale sizes of the components of the present application. Those skilled in the art can select various possible shapes and scale sizes to implement the present application according to specific circumstances under the teaching of the present application. In the drawings:

[0015] Figure 1 a flowchart of the method for estimating the water storage capacity of the pit pond with shielding provided by the embodiments of the present application;

[0016] Figure 2 a schematic diagram of the plot area, the dike area, the water surface area extracted by the remote sensing, and the vegetation area extracted by the remote sensing of the pit pond, the water surface area of the pit pond, and the bottom surface range of the pit pond for the embodiments of the present application;

[0017] Figure 3Figures showing the inversion results of the inversion factors in the examples, wherein (a) is a figure showing the inversion results of Ln(Rb / Rg); (b) is a figure showing the inversion results of the inversion factor Ln(Rb / Rnir); (c) is a figure showing the inversion results of the inversion factor Ln(Rb / Rr); (d) is a figure showing the inversion results of the inversion factor Ln(Rg / Rnir); (e) is a figure showing the inversion results of the inversion factor Ln(Rg / Rr); (f) is a figure showing the inversion results of the inversion factor Ln(Rr / Rnir); (g) is a figure showing the inversion results of the inversion factor Rb / Rg; (h) is a figure showing the inversion results of the inversion factor Rb / Rnir; (i) is a figure showing the inversion results of the inversion factor Rb / Rr; (j) is a figure showing the inversion results of the inversion factor Rg / Rnir; (k) is a figure showing the inversion results of the inversion factor Rg / Rr; (l) is a figure showing the inversion results of the inversion factor Rb / XLR; (m) is a figure showing the inversion results of the inversion factor Rg / XLR; (n) is a figure showing the inversion results of the inversion factor Rnir / XLR; (o) is a figure showing the inversion results of the inversion factor Rr / XLR; (p) is a figure showing the inversion results of the inversion factor Rb / XHR; (q) is a figure showing the inversion results of the inversion factor Rg / XHR; (r) is a figure showing the inversion results of the inversion factor Rnir / XHR; and (s) is a figure showing the inversion results of the inversion factor Rr / XHR;

[0018] Figure 4 Figures showing the water depth inversion effects obtained by different machine learning methods in the examples, wherein (a) is a figure showing the BP water depth inversion effect; (b) is a figure showing the CNN water depth inversion effect; (c) is a figure showing the SVM water depth inversion effect; (d) is a figure showing the RF water depth inversion effect; and (e) is a figure showing the water depth inversion effect based on ensemble learning;

[0019] Figure 5 Figures showing the pit pond bottom surface water depth models obtained by inversion based on original remote sensing data in the examples;

[0020] Figure 6 Figures showing the pit pond bottom surface water depth models obtained by inversion based on filtered remote sensing data in the examples;

[0021] Figure 7 Figures showing the pit pond bottom surface water depth models obtained by using the linear boundary gradient method in the examples;

[0022] Figure 8 Figures showing the pit pond bottom surface water depth models obtained by using the biharmonic spline method in the examples;

[0023] Figure 9 Figures showing the pit pond bottom surface water depth models obtained by using the nearest leader method in the examples;

[0024] Figure 10 Fig. 1 is a schematic diagram of a water depth model of a pit pond bottom surface obtained by using a thin-plate spline method in an embodiment;

[0025] Figure 11 Fig. 2 is a schematic diagram of a remote sensing image I of a non-occluded pit pond and extraction of a water surface and extraction of vegetation in an embodiment; wherein (a) is a schematic diagram of the remote sensing image I, (b) is a schematic diagram of a remote sensing extraction of a water surface obtained from the remote sensing image I (white color is water body), and (c) is a schematic diagram of a remote sensing extraction of vegetation obtained from the remote sensing image I (white color is vegetation);

[0026] Figure 12 Fig. 3 is a schematic diagram of water depth inversion based on the remote sensing image I of the non-occluded pit pond in an embodiment;

[0027] Figure 13 Fig. 4 is a schematic diagram of water depth inversion based on the remote sensing image I of the non-occluded pit pond in an embodiment after smoothing and denoising;

[0028] Figure 14 Fig. 5 is a schematic diagram of an unconstrained water depth model constructed based on the remote sensing image I of the non-occluded pit pond in an embodiment;

[0029] Figure 15 Fig. 6 is a schematic diagram of a water depth model with an outer edge line constraint constructed based on the remote sensing image I of the non-occluded pit pond in an embodiment;

[0030] Figure 16 Fig. 7 is a schematic diagram of a water depth model with a double constraint of an outer edge line constraint and a bottom surface range constraint constructed based on the remote sensing image I of the non-occluded pit pond in an embodiment;

[0031] Figure 17 Fig. 8 is a schematic diagram of a remote sensing image II of a non-occluded pit pond and extraction of a water surface and extraction of vegetation in an embodiment; wherein (a) is a schematic diagram of the remote sensing image II, (b) is a schematic diagram of a remote sensing extraction of a water surface obtained from the remote sensing image II (white color is water body), and (c) is a schematic diagram of a remote sensing extraction of vegetation obtained from the remote sensing image II (white color is vegetation);

[0032] Figure 18 Fig. 9 is a schematic diagram of water depth inversion based on the remote sensing image II of the non-occluded pit pond in an embodiment;

[0033] Figure 19 Fig. 10 is a schematic diagram of water depth inversion based on the remote sensing image II of the non-occluded pit pond in an embodiment after smoothing and denoising;

[0034] Figure 20 Fig. 11 is a schematic diagram of an unconstrained pit pond water depth model constructed based on the remote sensing image II of the non-occluded pit pond in an embodiment;

[0035] Figure 21Fig. 1 is a schematic diagram of a water depth model with only one constraint of the outer edge line for a non-full occlusion pit based on remote sensing images in an embodiment;

[0036] Figure 22 Fig. 2 is a schematic diagram of a water depth model with double constraints of the outer edge line constraint and the water depth model constraint of the pit bottom surface for a non-full occlusion pit based on remote sensing images in an embodiment;

[0037] Figure 23 Fig. 3 is a schematic diagram of a water surface patch and a water surface range for a full occlusion pit in an embodiment; wherein (a) is a schematic diagram of a water surface patch for a full occlusion pit, and (b) is a schematic diagram of a water surface range for a full occlusion pit;

[0038] Figure 24 Fig. 4 is a schematic diagram of an unconstrained water depth model for a full occlusion pit in an embodiment;

[0039] Figure 25 Fig. 5 is a schematic diagram of a double-constrained water depth model for a full occlusion pit in an embodiment;

[0040] Figure 26 Fig. 6 is a schematic diagram of a remote sensing image and a water surface range for an irregular pit in an embodiment; wherein (a) is a schematic diagram of a remote sensing image for an irregular pit, and (b) is a schematic diagram of a water surface range for an irregular pit;

[0041] Figure 27 Fig. 7 is a schematic diagram of a remote sensing inversion water depth for an irregular pit in an embodiment;

[0042] Figure 28 Fig. 8 is a schematic diagram of a remote sensing inversion water depth smoothing and denoising for an irregular pit in an embodiment;

[0043] Figure 29 Fig. 9 is a schematic diagram of an unconstrained water depth model for an irregular pit in an embodiment;

[0044] Figure 30 Fig. 10 is a schematic diagram of an outer edge line constrained water depth model for an irregular pit in an embodiment;

[0045] Figure 31 Fig. 11 is a schematic diagram of a double-constrained water depth model for an irregular pit in an embodiment;

[0046] Figure 32 Fig. 12 is a schematic diagram of a double-constrained water depth model for an irregular pit in an embodiment; DETAILED DESCRIPTION

[0047] In order for those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts should fall within the scope of protection of the present application.

[0048] In the description of the present application, the terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can be explicitly or implicitly included one or more features.

[0049] At present, the estimation method of pit pond water storage amount cannot obtain the water depth information of the blocked area in the remote sensing image due to the shielding of vegetation, shadow and hard surface (such as the cultivation pit pond with built roof), thereby leading to the underestimation of the water storage amount.

[0050] The embodiments of the present application provide a method for estimating the water storage amount of a blocked pit pond, which is used for estimating the water storage amount of a target pit pond with shielding, such as Figure 1 As shown in the figure, the method comprises the following steps:

[0051] Step 1, determining a target pit pond, extracting the graph spot of the target pit pond and determining the graph spot area according to the remote sensing image of the target area;

[0052] Step 2, comprising:

[0053] Step 2-1: determining the remote sensing extraction water surface area and the remote sensing extraction vegetation area of the target pit pond according to the remote sensing image;

[0054] Step 2-2: determining the shielding degree of the target pit pond based on at least the remote sensing extraction water surface area;

[0055] Step 3, comprising:

[0056] Step 3-1: determining the dike area of the target pit pond;

[0057] Step 3-2: determining the water surface area of the target pit pond according to the graph spot area, the dike area, the remote sensing extraction water surface area, the remote sensing extraction vegetation area and the shielding degree;

[0058] Step 4, comprising:

[0059] Step 4-1: extracting the outer edge line of the target pit pond based on the dike area;

[0060] Step 4-2: determining the bottom surface range of the target pit pond;

[0061] Step 4-3: Determine the water depth of the target pit pond according to the degree of occlusion;

[0062] Step 4-4: Determine the pit pond bottom water depth model according to the water depth of the target pit pond and the bottom surface range;

[0063] Step 5: Construct the pit pond water depth model with the water depth of the outer edge line as zero and the pit pond bottom water depth model as a constraint condition;

[0064] Step 6: Determine the pit pond water storage capacity based on the target pit pond water surface area and the pit pond water depth model.

[0065] It should be noted that the above steps are only a summary of the main steps of the method of the embodiments of the present application, and do not limit the order of the steps. In specific applications, the order of the steps is determined according to the actual situation, and the order of the corresponding steps can also be changed.

[0066] In some embodiments, before step 1, the near-year land change survey results and raw remote sensing images of a specific resolution (such as 2-meter resolution) and other data can be collected, the remote sensing image preprocessing can be completed, and the target pit pond polygon and polygon area can be extracted.

[0067] In some embodiments, the remote sensing image preprocessing process includes:

[0068] ① In ARCGIS (Geographic Information System software), the "1104 / 1104A / 1104k" pit pond polygon is extracted from the land class polygon field of the annual land change survey results. The complete pit pond polygon is formed by fusing according to the spatial position and attribute field;

[0069] ② In ENVI (The Environment for Visualizing Images, which is a complete remote sensing image processing platform), the remote sensing image is preprocessed by radiation calibration, atmospheric correction, and solar flare removal.

[0070] In some embodiments, step 2-1 can include: based on the spectral characteristics of the remote sensing image, for each pixel in the polygon, respectively determining the normalized water index NDWI and the normalized vegetation index NDVI, the expressions are as follows:

[0071] (1) ;

[0072] (2) ;

[0073] Wherein, Rg, Rr, Rnir respectively represent the reflectivity of the green wave band, the reflectivity of the red wave band and the reflectivity of the near-infrared wave band.

[0074] An adaptive image thresholding method using local first-order statistics in Matlab, based on the mean intensity of local normalized difference water index (NDWI) and normalized difference vegetation index (NDVI) of each pixel neighborhood, dynamically selects water detection threshold T NDWI and vegetation detection threshold T NDVI ; if the normalized water index of a pixel is greater than the water detection threshold, the pixel is determined to belong to the unobstructed area; if the mean normalized vegetation index of a pixel is greater than the vegetation detection threshold, the pixel is determined to belong to the obstructed area; the sum of the areas of all pixels in the unobstructed area is determined as the water surface area S NDWI extracted by remote sensing; the sum of the areas of all pixels in the obstructed area is determined as the vegetation area S NDVI extracted by remote sensing.

[0075] In some embodiments, the obstruction degree of the pit pond can include full obstruction and non-full obstruction, and the non-full obstruction pit pond can refer to a pit pond with most of the water surface exposed, and the full obstruction pit pond can refer to a pit pond completely covered by vegetation, shadow or hard surface.

[0076] As an example, the pit pond with the water surface area S NDWI extracted by remote sensing being zero in step 2-2 is divided into a full obstruction pit pond; the rest of the pit ponds can be divided into non-full obstruction pit ponds. In other embodiments, the obstruction degree of the pit pond can be divided according to actual needs.

[0077] In the embodiments, the schematic diagram of the pit pond plot area, the dike area, the water surface area S Figure 2 extracted by remote sensing, the vegetation area S extracted by remote sensing and the water surface area of the pit pond can be seen in FIG. 1.

[0078] In some embodiments, step 3-1 can specifically include: obtaining the resolution of the remote sensing image; comparing the plot area with a preset plot area threshold, and presetting a dike width parameter according to the comparison result; using a circle as a first structure element, determining the first structure element radius according to the dike width parameter and the resolution; and using the first structure element radius to perform morphological operation on the plot to extract the dike area.

[0079] The structure element is a basic element in morphological transformation, and the larger the structure element radius, the more obvious the shrinkage.

[0080] The inventors found through analysis of existing terrain data that the larger the pit pond plot area, the wider the dike width. As an example, the dike width parameter is about 1-6 meters, and the pixels occupied by the dike are equal to the dike width k meters divided by the resolution p meters of the remote sensing image. The dike area can be extracted in Matlab using morphological operation

[0081] As an example, the parameters are set as follows:

[0082] ① Pit pond plot area < first plot area threshold (such as 400 square meters, can be set according to actual needs), the first structural element radius is 0 pixel, and the pond ridge width is not removed, that is, the pond ridge area is zero;

[0083] ② The second plot area threshold (such as 1000 square meters, can be set according to actual needs) ≥ pit pond plot area ≥ first plot area threshold, the first structural element radius is k / p, k=2 pixels;

[0084] ③ The third plot area threshold (such as 4000 square meters, can be set according to actual needs) ≥ pit pond plot area > second plot area threshold, the first structural element radius is k / p, k=4 pixels;

[0085] ④ Pit pond plot area > third plot area threshold, the first structural element radius is k / p, k=6 pixels.

[0086] In some embodiments, step 3-2 determines the target pit pond water surface area as one of the water storage calculation parameters, which can specifically include: the shielding degree includes full shielding and non-full shielding. For the target pit pond with non-full shielding, the target pit pond water surface area is the union of the remote sensing extracted water surface area and the remote sensing extracted vegetation area, and the maximum value of the target pit pond water surface area is the complement set of the pond ridge area in the plot area. The water surface area of the non-full shielding pit pond is shown in the range of the blue dashed line in FIG. 3, and the calculation expression is shown in formula (3) and formula (4): Figure 2

[0087] (3) ;

[0088] (4) ;

[0089] Wherein, is the non-full shielding target pit pond water surface area, is the remote sensing extracted water surface area, is the remote sensing extracted vegetation area, is the plot area of the pit pond, is the pond ridge area.

[0090] For the target pit pond with full shielding, the target pit pond water surface area is the complement set of the pond ridge area in the plot area, and the calculation expression is shown in formula (5):

[0091] (5) ;

[0092] Wherein, is the full shielding target pit pond water surface area.

[0093] ​The embodiments of the present application distinguish between full occlusion and non-full occlusion, and respectively adopt different calculation methods for the target pond water surface area. In the case of full occlusion, the part of the area of the graph spot excluding the pond ridge area is directly considered. In the case of non-full occlusion, the water surface and the vegetation area extracted by remote sensing are combined, because these vegetation areas may be partially composed of aquatic vegetation on the water surface, and thus the present application can more accurately reflect the actual area of the pond water surface.

[0094] In some embodiments, step 4-1 extracts the outer edge line (the outer edge line is the intersection line of the highest water level of the pond and the bank) based on the pond ridge area, which can be seen in the red line in Figure 2 Based on the analysis of the existing terrain data, the underwater terrain of the pond is usually an inverted trapezoidal platform, and the outer edge line is located at the top of the slope, and the water depth of the outer edge line can be considered as 0, which is the first constraint condition for constructing the three-dimensional pond water depth model.

[0095] As an example, step 4-1 extracts the outer edge line of the target pond based on the pond ridge area, which can specifically include: binarizing the pond ridge area using the imbinarize function of the Matlab software (the imbinarize function is a tool in the MATLAB image processing toolbox for binarizing gray-scale images or volume data), and further extracting the boundary line using the bwboundaries function (the bwboundaries function is a function in MATLAB that is specifically used to track the outer boundaries of objects in a binary image and the boundaries of holes inside these objects), and taking the inner boundary line of the pond ridge area as the outer edge line.

[0096] In some embodiments, step 4-2 of determining the bottom surface range of the target pond includes: comparing the graph spot area with a preset graph spot area threshold, presetting the interval distance j meters between the bottom of the target pond and the edge of the graph spot according to the comparison result; obtaining the resolution p meters of the remote sensing image; according to the interval distance and the resolution, a circle is used as a second structure element to determine the second structure element radius; using the second structure element radius, morphological operation is performed on the graph spot to shrink and crop inward from the periphery of the graph spot to obtain the bottom surface range.

[0097] As an example, the second structure element radius for calculating the bottom surface range of the pond is as follows:

[0098] ① If the pond graph spot area is less than 400 square meters, the second structure element radius r=j / p, j=2 pixels;

[0099] ② If 1000 square meters is greater than or equal to the pond graph spot area and is greater than 400 square meters, the second structure element radius r=j / p, j=4 pixels;

[0100] ③ If 4000 square meters is greater than or equal to the pond graph spot area and is greater than 1000 square meters, the second structure element radius r=j / p, j=6 pixels;

[0101] (iv) the area of the pit pond pattern is greater than 4000 square meters, the second structural element radius r = j / p, j = 8 pixels.

[0102] In some embodiments, step 4-3 can specifically include:

[0103] Step 4-31: For a non-fully occluded target pit pond, based on the remote sensing image, the reflectivity of each band is mathematically transformed to determine a plurality of water depth inversion factors, the mathematical transformation including ratio transformation and logarithmic transformation; based on the water depth inversion factor, the water depth of the non-occluded area (i.e. remote sensing inversion water depth) is fitted by using a machine learning model;

[0104] Step 4-32: Based on the water depth of the non-occluded area, the water depth of the occluded area is fitted by using a thin plate spline function.

[0105] As an example, the water depth inversion factor in step 4-31 includes any one or more of a first logarithmic ratio factor, a second logarithmic ratio factor, a third logarithmic ratio factor, a fourth logarithmic ratio factor, a fifth logarithmic ratio factor, a sixth logarithmic ratio factor, a first ratio factor, a second ratio factor, a third ratio factor, a fourth ratio factor, a fifth ratio factor, a first suspended sediment factor ratio, a second suspended sediment factor ratio, a third suspended sediment factor ratio, a fourth suspended sediment factor ratio, a fifth suspended sediment factor ratio, a sixth suspended sediment factor ratio, a seventh suspended sediment factor ratio, and an eighth suspended sediment factor ratio.

[0106] As shown in Table 1, the first logarithmic ratio factor is represented as: Ln(Rb / Rg); the second logarithmic ratio factor is represented as: Ln(Rb / Rnir); the third logarithmic ratio factor is represented as: Ln(Rb / Rr); the fourth logarithmic ratio factor is represented as: Ln(Rg / Rnir); the fifth logarithmic ratio factor is represented as: Ln(Rg / Rr); the sixth logarithmic ratio factor is represented as: Ln(Rr / Rnir); the first ratio factor is represented as: Rb / Rg; the second ratio factor is represented as: Rb / Rnir; the third ratio factor is represented as: Rb / Rr; the fourth ratio factor is represented as: Rg / Rnir; the fifth ratio factor is represented as: Rg / Rr; the first suspended sediment factor ratio is represented as: Rb / XLR; the second suspended sediment factor ratio is represented as: Rg / XLR; the third suspended sediment factor ratio is represented as: Rnir / XLR; the fourth suspended sediment factor ratio is represented as: Rr / XLR; the fifth suspended sediment factor ratio is represented as: Rb / XHR; the sixth suspended sediment factor ratio is represented as: Rg / XHR; the seventh suspended sediment factor ratio is represented as: Rn / XHR; the eighth suspended sediment factor ratio is represented as: Rr / XHR; XLR is a first concentration sediment factor, represented as: XLR=(Rg+Rnir) / (Rr / Rnir); XHR is a second concentration sediment factor, represented as: XHR=(Rg+Rr) / (Rg / Rr). Wherein, Rb, Rg, Rr, Rnir represent reflectivity of blue band, reflectivity of green band, reflectivity of red band and reflectivity of near-infrared band respectively.

[0107] The embodiment of the present application is aimed at a turbid pit pond, and uses the ratio of single band and low and high concentration sediment factors to obtain the water depth of the non-full-shading pit pond, to weaken the influence of suspended sediment on remote sensing depth measurement, and to make the result more accurate. The inversion result using the inversion factor can be seen from Figure 3 .

[0108] Table 1: Water depth remote sensing inversion factor

[0109]

[0110] As an example, in step 4-31, based on the water depth inversion factor, a machine learning model is used to fit the water depth of the unshaded area, including: if a plurality of water depth inversion factors are determined, a Spearman correlation coefficient analysis method is used to analyze the correlation coefficient between all water depth inversion factors and pit pond water depth values, and a set number (such as 8) of water depth inversion factors with the highest correlation coefficient are selected to form an optimal water depth inversion factor combination; the machine learning model includes a plurality of machine learning sub-models, and the water depth inversion factor combination is input into each machine learning sub-model; the final fitted water depth of the machine learning model is obtained by weighted summation according to the weight and fitting value of each machine learning sub-model, and the expression is as follows:

[0111] (6);

[0112] wherein n is the number of machine learning sub-models, wi is the weight of the i-th machine learning sub-model, yi is the fitting value of the inversion result corresponding to the i-th machine learning sub-model.

[0113] In some embodiments, the R2 of the machine learning sub-models can be used as the weight, and the better the fitting effect of the machine learning sub-model, the greater the weight. The value range of R2 is between 0 and 1. R2 (R-squared), also known as the coefficient of determination, is a statistical quantity used to evaluate the fitting degree of the model. As an example, the R2, RMSE and bias of each machine learning sub-model can be calculated, and if the R2 is greater than a preset value (such as 0.6), it is taken as a candidate machine learning sub-model, as shown in equation (7). 2 2

[0114] (7);

[0115] wherein R2i is the coefficient of determination of the i-th machine learning sub-model.

[0116] In specific embodiments, the machine learning sub-models can be a back propagation neural network model (BP), a deep learning model (CNN), a support vector machine regression model (SVR) and a random forest regression model (RF), etc. The inversion effects of the machine learning sub-models are shown in Figs. (a)-(d) of Figure 4 , and the integrated learning water depth inversion effect is shown in Fig. (e) of Figure 4 .

[0117] The embodiments of the present application can further reduce redundant information by selecting the water depth inversion factors most related to the pit pond water depth value, and can more accurately reflect the actual situation of the water depth. The integrated learning of multiple machine learning sub-models can make full use of the advantages of different models and avoid the overfitting or underfitting problems that may exist in a single model. By combining the prediction results of each sub-model through weighted summation, a more stable and reliable fitting result can be obtained.

[0118] In some embodiments, step 4-31 further includes filtering to remove outliers of the remotely sensed water depth. White noise caused by water surface sunlight reflection and mixed pixels at the water-land boundary of the pit pond can both cause outliers in the remotely sensed water depth. Wiener filtering is used to smooth and remove water depth outliers, and as an example, the kernel function size is 5x5. As shown in Fig. Figures 5-6 ​​As shown, the filtered water depth surface is smooth and has no obvious noise points, and can express the undulation of the underwater terrain, which is conducive to the construction of the subsequent pit and pond bottom surface water depth model.

[0119] In the embodiments, four methods for constructing the bottom surface water depth model under no constraints are compared, such as Figures 7-10 Figure 7 Figure 9 It can be seen that the pit and pond bottom surface water depth model constructed based on the linear boundary gradient and the nearest neighbor point has a large difference from the actual value, and Figure 8 Figure 10 The pit and pond bottom surface water depth model constructed based on the double harmonic thin plate spline function is easy to lose the trend of the original water depth, and the pit and pond bottom surface water depth model constructed based on the thin plate spline function can better maintain the trend of the original water depth.

[0120] In some embodiments, in step 4-32, the filtered remote sensing inversion water depth is fitted by using the thin plate spline function under no constraints, and the water depth of the blocked area is filled according to the spatial distribution and trend of the water depth of the unblocked area, so that each pixel in the pit and pond bottom surface range has a water depth value.

[0121] In some embodiments, the water depth of the target pit and pond in step 4-3 is determined according to the blocking degree, and further includes step 4-33: for the target pit and pond that is fully blocked, the geometric center point of the pit and pond is generated in ARCGIS; the minimum distance from the geometric center point of each non-full-blocked pit and pond in a preset range around the target pit and pond to the geometric center point of the target pit and pond is determined, and the average water depth of the non-full-blocked pit and pond corresponding to the minimum distance is taken as the water depth of the fully blocked target pit and pond.

[0122] As an example, the measured water depth data can be collected in advance, and the measured water depth data at least includes the water depth value and the two-dimensional plane coordinates of the sampling point. The average water depth of each non-full-blocked pit and pond around the target pit and pond can refer to the measured water depth data.

[0123] As an example, the pit and pond bottom surface water depth model constructed in step 4-4 is different from the non-full-blocked pit and pond bottom surface water depth model, and the full-blocked pit and pond bottom surface water depth model is a plane with uniform water depth value, such as Figures 23-25 .

[0124] For the fully blocked pit and pond, the pit and pond bottom surface water depth model is a plane, the water depth value of the plane is derived from the average water depth of the nearest non-full-blocked pit and pond to the target pit and pond, and the size of the plane is equal to the pit and pond bottom surface range. For the non-full-blocked pit and pond, the water depth of the blocked area is filled by using the remote sensing inversion water depth, and the size of the plane is equal to the pit and pond bottom surface range.

[0125] ​​​In some embodiments, step 5 constructs the pit pond water depth model, and the water depth of the outer edge line is 0, which is the first constraint condition for constructing the pit pond water depth model; the pit pond bottom surface water depth model is the second constraint condition, and the pit pond water depth model under the double constraints is constructed, which is one of the parameters for water storage calculation.

[0126] As Figures 11-22 , the pit pond water depth model constructed without constraints is obtained by interpolating the water depth of the blank area using the known water depth trend and other information. The known water depth trend is difficult to reflect the topographic characteristics that the water depth of the edge area gradually decreases and the water bottom elevation gradually rises.

[0127] The pit pond water depth model under the constraint of the outer edge line can reflect the characteristics that the water depth of the slope gradually decreases. However, this method has poor construction effect on the pit pond water depth model with high vegetation sheltering degree. As Figures 11-22 It can be seen that in the sheltered area, the pit pond water depth constructed under only one constraint of the outer edge line is low, and the bottom surface water depth model is distorted.

[0128] Therefore, when constructing the water depth model by using the thin plate spline function, adding the two constraint conditions of the outer edge line and the pit pond bottom surface water depth model can better estimate the water depth of the pit pond under different sheltering degrees, as shown in Figure 16 and Figure 22 .

[0129] The water depth model construction method provided in the embodiments has good effect on the construction of the water depth model of an irregularly shaped pit pond, as shown in Figures 26-32 .

[0130] In some embodiments, step 6 calculates the pit pond water storage based on the pit pond water surface area and the water depth model, including: dividing the pit pond water surface area into pixels to determine the pixel area corresponding to each pixel; determining the water depth value corresponding to each pixel according to the pit pond water depth model; and multiplying the pixel area of each pixel by the water depth value corresponding to the pixel and summing up to determine the pit pond water storage, and the calculation formula is as follows:

[0131] (8) ;

[0132] wherein n represents the number of pixels of the water surface area, H i represents the water depth value corresponding to the i th pixel, S i represents the pixel area corresponding to the i th pixel. In general cases, S i is equal to the square of the resolution, which is a fixed value. In the embodiments, a 2-meter resolution remote sensing image is used, and S i = 4 square meters.

[0133] For a single non-full sheltered pit pond water storage calculation, . For a single full sheltered pit pond water storage calculation, In some embodiments, the total area pond water storage amount is calculated .

[0134] The above describes in detail the method for estimating the sheltered pond water storage amount provided by the present application, and the principles and implementation manners of the present application are described by using specific examples. The above descriptions of the embodiments are only used to help understand the concept of the present application, and should not be understood as a limitation on the protection scope of the present application.

Claims

1. A method for estimating a water storage volume of a water body with obstruction, characterized by, The method comprises the following steps: According to the remote sensing image of the target area, the graph spot of the target pit pond is extracted and the area of the graph spot is determined; According to the remote sensing image, the remote sensing extraction water surface area and the remote sensing extraction vegetation area of the target pit pond are determined; and at least based on the remote sensing extraction water surface area, the shielding degree of the target pit pond is determined; The dike area of the target pit pond is determined; and according to the graph spot area, the dike area, the remote sensing extraction water surface area, the remote sensing extraction vegetation area and the shielding degree, the water surface area of the target pit pond is determined; Based on the dike area, the outer edge line of the target pit pond is extracted; the bottom surface range of the target pit pond is determined; according to the shielding degree, the water depth of the target pit pond is determined; and according to the water depth and the bottom surface range, the pit pond bottom surface water depth model is determined; The pit pond water depth model is constructed with the outer edge line water depth as zero and the pit pond bottom surface water depth model as a constraint condition; Based on the target pit pond water surface area and the pit pond water depth model, the pit pond water storage capacity is determined.

2. The obstructed pit and pond water storage volume estimation method of claim 1, wherein, According to the remote sensing image, the remote sensing extraction water surface area and the remote sensing extraction vegetation area of the target pit pond are determined, which comprises the following steps: Based on the spectral characteristics of the remote sensing image, for each pixel in the graph spot, the normalized water body index and the normalized vegetation index are determined respectively; According to the average value of the normalized water body index and the average value of the normalized vegetation index in the local neighborhood of each pixel, the water body detection threshold and the vegetation detection threshold are determined; If the normalized water body index of a pixel is greater than the water body detection threshold, the pixel is determined to belong to an unshielded area; if the normalized vegetation index of a pixel is greater than the vegetation detection threshold, the pixel is determined to belong to a shielded area; The sum of the areas of all pixels in the unshielded area is determined as the remote sensing extraction water surface area; and the sum of the areas of all pixels in the shielded area is determined as the remote sensing extraction vegetation area.

3. The obstructed pit and pond water storage volume estimation method of claim 1, wherein, The dike area of the target pit pond is determined, which comprises the following steps: The resolution of the remote sensing image is obtained; According to the comparison between the graph spot area and the preset graph spot area threshold, the preset dike width parameter is determined according to the comparison result; According to the dike width parameter and the resolution, the first structure element radius is determined; The first structure element radius is used for morphological operation on the graph spot to extract the dike area.

4. The obstructed pit and pond water storage volume estimation method of claim 1, wherein, According to the graph spot area, the dike area, the remote sensing extraction water surface area, the remote sensing extraction vegetation area and the shielding degree, the target pit pond water surface area is determined, which comprises the following steps: The shielding degree comprises full shielding and non-full shielding; For the target pit pond with non-full shielding, the target pit pond water surface area is the union of the remote sensing extraction water surface area and the remote sensing extraction vegetation area, and the maximum value of the target pit pond water surface area is the complement of the dike area in the graph spot area; For the target pit pond with full shielding, the target pit pond water surface area is the complement of the dike area in the graph spot area.

5. The obstructed pit and pond water storage volume estimation method of claim 1, wherein, The bottom surface range of the target pit pond is determined, which comprises the following steps: According to the comparison between the graph spot area and the preset graph spot area threshold, the interval distance between the bottom of the target pit pond and the edge of the graph spot is preset according to the comparison result; acquiring a resolution of the remote sensing image; determining a second structural element radius according to the interval distance and the resolution; performing morphological operation on the polygon by using the second structural element radius to obtain a bottom surface range by inwardly shrinking and cropping from the polygon periphery.

6. The obstructed pit and pond water storage volume estimation method of claim 1, wherein, determining the water depth of the target pit pond according to the shielding degree, including: the shielding degree includes non-full shielding; for the target pit pond with non-full shielding, performing mathematical transformation on reflectivity of each band based on the remote sensing image to determine a plurality of water depth inversion factors, the mathematical transformation including ratio transformation and logarithmic transformation; fitting the water depth of the non-shielding area by using a machine learning model based on the water depth inversion factors; fitting the water depth of the shielding area by using a thin plate spline function based on the water depth of the non-shielding area.

7. The obstructed pit and pond water storage volume estimation method of claim 6, wherein, the water depth inversion factors include any one or more of a first logarithmic ratio factor, a second logarithmic ratio factor, a third logarithmic ratio factor, a fourth logarithmic ratio factor, a fifth logarithmic ratio factor, a sixth logarithmic ratio factor, a first ratio factor, a second ratio factor, a third ratio factor, a fourth ratio factor, a fifth ratio factor, a first suspended sediment factor ratio, a second suspended sediment factor ratio, a third suspended sediment factor ratio, a fourth suspended sediment factor ratio, a fifth suspended sediment factor ratio, a sixth suspended sediment factor ratio, a seventh suspended sediment factor ratio, and an eighth suspended sediment factor ratio; wherein the first log-ratio factor is expressed as: Ln(Rb / Rg); The second log-ratio factor is expressed as: Ln(Rb / Rnir); The third log-ratio factor is expressed as: Ln(Rb / Rr); The fourth log-ratio ratio factor is expressed as: Ln(Rg / Rnir); The fifth logarithmic ratio factor is expressed as: Ln(Rg / Rr); The sixth logarithmic ratio factor is expressed as: Ln(Rr / Rnir); The first ratio factor is expressed as: Rb / Rg; The second ratio factor is expressed as: Rb / Rnir; The third ratio factor is expressed as: Rb / Rr; The fourth ratio factor is expressed as: Rg / Rnir; The fifth ratio factor is expressed as: Rg / Rr; The first suspended sediment factor ratio is expressed as: Rb / XLR; The second suspended sediment factor ratio is expressed as: Rg / XLR; The third suspended sediment factor ratio is expressed as: Rr / XLR; The fourth suspended sediment factor ratio is expressed as: Rnir / XLR; The fifth suspended sediment factor ratio is expressed as: Rb / XHR; The sixth suspended sediment factor ratio is expressed as: Rg / XHR; The seventh suspended sediment factor ratio is expressed as: Rnir / XHR; The eighth suspended sediment factor ratio is expressed as: Rr / XHR XLR is the ratio of the first concentration of sediment factor, expressed as: XLR=(Rg+Rnir)(Rr / Rnir); XHR is the ratio of the second concentration of sediment factor, expressed as: XHR=(Rg+Rr)(Rg / Rr); Rb , Rg , Rr , Rnir respectively represent reflectivity of a blue band, reflectivity of a green band, reflectivity of a red band, and reflectivity of a near-red band.

8. The obstructed pit and pond water storage volume estimation method of claim 6, wherein, fitting the water depth of the non-shielding area by using a machine learning model based on the water depth inversion factors, including: if a plurality of the water depth inversion factors are determined, analyzing the correlation coefficients between all the water depth inversion factors and the pit pond water depth values by using a Spearman correlation coefficient analysis method, selecting a set number of the water depth inversion factors with the highest correlation coefficients to form a water depth inversion factor combination; the machine learning model includes a plurality of machine learning sub-models, and the water depth inversion factor combination is input into each machine learning sub-model; performing weighted summation according to the weights and fitting values of each machine learning sub-model to obtain the water depth of the non-shielding area fitted by the machine learning model.

9. The obscured impoundment water storage volume estimation method of claim 1, wherein, determining the water depth of the target pit pond according to the shielding degree, including: the shielding degree includes full shielding; For a target pit pond with full occlusion, the minimum distance from the geometric center point of each non-full occlusion pit pond in a preset range around the target pit pond to the geometric center point of the target pit pond is determined, and the average water depth of the non-full occlusion pit pond corresponding to the minimum distance is taken as the water depth of the target pit pond with full occlusion.

10. The obscured impoundment water storage volume estimation method of claim 1, wherein, Based on the pit pond water surface area and the pit pond water depth model, the pit pond water storage capacity is determined, including: Dividing the pixels based on the pit pond water surface area, determining the pixel area corresponding to each pixel; According to the pit pond water depth model, the water depth value corresponding to each pixel is determined; The pixel area of each pixel is multiplied by the water depth value corresponding to the pixel, and then summed, thereby determining the pit pond water storage capacity.

Images