A method for constructing an RCBAM model for remotely sensing river flow

Abstract

The application discloses a kind of RCBAM model construction methods of river flow remote sensing inversion, which comprises the following steps:1) collecting remote sensing image data;2) high-precision DEM production;3) river center line acquisition;4) river width and slope acquisition;5) river section shape extraction;6) flow area calculation;7) river flow estimation.The application obtains the section shape of river channel based on high-precision digital elevation model DEM, combines the instantaneous water surface information obtained from remote sensing image, dynamically calculates the river section area, effectively solves the problem that the flow area in the traditional BAM flow model relies on empirical formula to obtain, realizes the river flow estimation considering the shape characteristics of river channel, and can realize efficient and accurate river flow remote sensing inversion.

Description

TECHNICAL FIELD

[0001] The present application relates to river flow remote sensing inversion technology, belonging to the technical field of geographic information processing. In particular, it relates to a river flow inversion method in an ungauged area based on river channel shape and Bayesian method, and specifically relates to a RCBAM (River Channel Bayesian AMHG-Manning) model construction method for remote sensing inversion of river flow, which can realize efficient and accurate remote sensing inversion of river flow. BACKGROUND

[0002] River flow, as a key element in the hydrological cycle, is a basic strategic data resource and has a profound impact on many fields. For example, in the aspect of flood control and disaster reduction, the monitoring of river flow is extremely important for the early warning and prevention and control of disasters such as floods and droughts. Timely and accurate flow data can provide decision support for management departments and reduce disaster losses. In the aspect of water resources management, accurate river flow data is an important basis for rational allocation of water resources and planning of water conservancy facility construction.

[0003] Traditional river flow monitoring mainly relies on hydrological monitoring stations. However, due to factors such as geographical environment and economic development level, the spatial distribution of hydrological monitoring stations is uneven. In particular, in remote areas, mountainous areas or economically underdeveloped areas, the number of stations is small, and these areas are widely known as areas lacking hydrological data.

[0004] Remote sensing, as a remote and non-contact information acquisition technology, has the advantages of wide coverage, high observation frequency and no geographical restrictions, and can obtain near-real-time data of rivers worldwide, providing new possibilities for river flow monitoring. In early studies, the area-flow relationship or the width-flow relationship was established according to the river area or river width extracted from remote sensing and the measured flow to estimate the river flow. Such methods have simple models and are highly dependent on the empirical relationship between remote sensing monitoring variables and field measured flow, and the precision is limited. Moreover, such a relationship is only applicable to areas with field monitoring data and cannot be applied to areas lacking data.

[0005] In addition, some scholars use remote sensing observation data as input parameters for existing hydrodynamic equations (such as the Manning equation) or combine them with numerical hydrodynamic models to estimate river flow. The Bayesian AMHG-Manning (BAM) model, as a typical representative, uses satellite remote sensing to obtain river width, slope and height data, and uses Bayesian methods to estimate river flow and quantify the uncertainty of the estimation results. BAM has shown certain advantages in river flow estimation, but the method of obtaining the important parameter of the over-flow area (A it =A 0i +δA it ) in the algorithm has defects, and the over-flow area Ait The reference cross-sectional area A 0i And the change of flow area delta A it Composed of, respectively, by means of empirical formula and satellite altimetry data calculation, there are the following defects: (1) the inaccuracy of the reference cross-sectional area: using the empirical formula related to the river width to set the reference cross-sectional area, the sample of the river width database is limited, and it is difficult to reflect the real characteristics of the complex and diverse rivers in the world. The empirical formula has inaccuracy, and it is difficult to correct uniformly; (2) the satellite altimetry data is limited, and only the data of large rivers can be captured, and it is not used for small and medium-sized rivers, and the data quantity is less. SUMMARY

[0006] In view of the defects of the flow area in the traditional BAM model, the application provides a RCBAM model construction method for remote sensing inversion of river flow considering the shape information of the river channel. The model obtains the cross-sectional shape of the river channel based on the independently produced high-precision (5m) digital elevation model (DEM), and dynamically calculates the flow area of the river channel combined with the instantaneous water surface information obtained from the remote sensing image, so as to realize the estimation of the river flow considering the shape characteristics of the river channel.

[0007] The purpose of the application is realized as follows:

[0008] A RCBAM model construction method for remote sensing inversion of river flow, comprising the following steps:

[0009] 1) Collect remote sensing image data:

[0010] Including satellite front and rear view image pairs and time series of optical remote sensing images.

[0011] 2) High-precision DEM production:

[0012] The satellite front and rear view image pairs are taken as input, and the digital surface model (DSM) is obtained through directional adjustment and stereo matching algorithm, and then the high-precision numerical elevation model (DEM) (5m) data is obtained by applying filtering to remove artificial features and vegetation.

[0013] The satellite front and rear view images are used to produce the digital elevation model (DEM) of the image coverage area by using the binocular stereo vision principle. Binocular stereo vision is an important form of machine vision, which is based on the principle of parallax and uses imaging equipment to obtain two images of the measured object from different positions. By calculating the position deviation between the corresponding points of the image, the three-dimensional geometric information of the object can be obtained.

[0014] 3) River centerline acquisition:

[0015] The optical remote sensing image is taken as input, a normalized water body index NDWI is calculated to obtain a water body binary image, and an erosion operation is performed on the water body binary image to obtain a river center line.

[0016] 4) River width and slope acquisition:

[0017] Based on the river center line, the vertical distance method is used to acquire the river width; and based on the DEM data, the slope is acquired.

[0018] 5) River section shape extraction:

[0019] In the DEM data, a profile is cut perpendicularly to the acquired river center line to obtain river section elevation data, and based on the river section elevation data, a river section shape curve is acquired by performing linear interpolation on discrete elevation points.

[0020] 6) Flow area calculation: The river width and the river section shape curve are corresponded, Gaussian-Kronrod integration is performed, and the flow area formed by the river section under different river width conditions is obtained.

[0021] 7) The Manning formula and the multi-station hydraulic geometry method AMHG are used as the basic equation for flow estimation, the Manning formula and the AMHG equation are logarithmically transformed and a likelihood function is constructed by using the Bayesian method; the Monte Carlo sampling method of Bayesian inference is used to extract samples from the specified distribution and calculate statistics to approximate the posterior distribution; the parameters used in the above formula: the reference cross-sectional area is set as the flow area corresponding to the minimum river width; the flow area change part is the difference between the instantaneous river width, the minimum river width and the area formed by the river section shape curve; the obtained river width, slope and flow area are calculated according to the Bayesian inference to obtain the flow estimation result.

[0022] In a further scheme, the time of the front and rear view image pair data used for DEM production in step 2) should be in the dry season of the river flow, so as to maximize the inclusion of river channel shape information; the precision of the high-precision DEM is 5m. The specific steps of step 2) are as follows:

[0023] 2-1 Utilize satellite image RPC parameters or ground control points to perform directional adjustment on the image to obtain accurate orientation parameters, use accurate positioning models and orientation parameters to perform epipolar sampling on the front and rear view images of the image pair respectively to obtain approximate epipolar images;

[0024] 2-2 Use a semi-global stereo matching algorithm (SGM) to perform stereo matching and point intersection on the approximate epipolar images to obtain high-density three-dimensional point cloud;

[0025] 2-3 Register the point cloud data to generate a digital surface model (DSM) containing ground buildings and vegetation, and then apply filtering to the DSM to remove artificial features and vegetation to obtain the DEM.

[0026] In a further aspect, in step 3), a normalized difference water index (NDWI) of the remote sensing image is calculated, and an OTSU adaptive threshold method is used to obtain a water body binary image, wherein a water body pixel value is 1 and a non-water body pixel value is 0.

[0027]

[0028] wherein ρ Green is the reflectivity of the green band, and ρ NIR is the reflectivity of the near-infrared band.

[0029] Based on the water body binary image, a morphological thinning algorithm is used to continuously erode the water body edge while avoiding complete deletion of the water body, and finally a center line with a width of 1 pixel is left.

[0030] For an image pixel coordinate (x, y) in the binary water body image I (the pixel value is 0 or 1, and 1 represents water body), a structure element B(m, n) (which is a small matrix, and m and n are its coordinates) slides on the image. The erosion operation formula is:

[0031]

[0032] The erosion operation is continuously repeated, and when the water body is only left with a line with a width of 1 pixel after erosion, this line is the river center line.

[0033] Step 3): Since the river width at different positions is inconsistent, erosion judgment is performed when erosion is performed, and the judgment is specifically divided into 8 cases: 1 data point around the judgment pixel, the center point gradient is 7; 2 data points around the judgment pixel, the center point gradient is 6; 3 data points around the judgment pixel, the center point gradient is 5; 4 data points around the judgment pixel, the center point gradient is 4; 5 data points around the judgment pixel, the center point gradient is 3; 6 data points around the judgment pixel, the center point gradient is 2; 7 data points around the judgment pixel, the center point gradient is 1.

[0034] In a further aspect, based on the vertical distance method described in step 4), the river width is obtained: starting from each point on the center line, searching in a direction perpendicular to the center line, finding two corresponding points of the water body boundary in the vertical direction, and calculating the distance between the two points to obtain the river width at that position.

[0035] Suppose the river center line is composed of a series of coordinate points (x i , y i ), and the distance between any two adjacent points is less than 1 pixel, then the river width at any position can be obtained by searching in a direction perpendicular to the center line. i, i ​) can be determined according to the tangent slope of the centerline. If the tangent slope of the centerline at the point is k, then the slope of the vertical line is -1 / k, and the intersection coordinates (x j1 ,y j1 ) and (x j2 ,y j2 ) of the line equation with the water boundary can be calculated by the distance formula between two points: .

[0036] In step 4), the slope is obtained based on the DEM data: the slope is obtained by calculating the ratio of the elevation difference and the horizontal distance between pixels within or at a certain distance from the river channel;

[0037] The coordinates of the two pixels are (i, j) and (m, n), and their elevation values are Z i,j and Z m,n , respectively. Then the elevation difference between them is ΔZ = Z i,j - Z m,n , and the horizontal distance . Then the slope is:

[0038]

[0039] In step 4), according to the long time series of optical images, the river width and slope data of the selected section are obtained, and the section is not less than 4.

[0040] Further, in step 5), the obtained river section shape includes the entire river within the dike on both sides of the river. Assuming that the river monitoring section position coordinates are (x0, y0), the profile is cut in the DEM data obtained in step 2) perpendicular to the river centerline obtained in step 3). Assuming that the profile width is W, for each pixel (x, y) in the profile, its elevation value Z(x, y) is directly obtained from the DEM data. In actual calculation, by traversing all pixels in the profile, a series of elevation values Z1, Z2, …, Z m are obtained.

[0041] Based on the profile elevation data, the discrete elevation points are processed by linear interpolation method to draw a smooth section contour line. Assuming that the adjacent two elevation points (x i , Z i ) and (x i+1 , Z i+1 ) have a interval Δx in x direction, for any point x between x i and x i+1 , its elevation Z can be calculated by linear interpolation formula:

[0042]

[0043] Through such interpolation calculation on the points in the whole profile, enough dense elevation points are obtained, and then the river cross-section shape curve Z=f(x) is drawn.

[0044] Further, in step 6), the original flow area calculation formula is as follows:

[0045] A it = A 0i + δAit

[0046] Wherein, A it is the flow area, m 2 ; A 0i is the reference section area, m 2 ; δA it is the flow area change, calculated as follows:

[0047] First, the minimum river width length obtained by remote sensing monitoring in step 4) is positionally corresponding to the river cross-section shape line Z=f(x) drawn in step 5) with the long time series river width. The corresponding principle is: according to the river width length, the water surface is found from the lowest part of the river f(x) upwards, and when the river width length intersects with the curve on both sides of f(x), it is considered that the water surface is at this time. Due to the influence of environmental factors, the two sides of the water surface may not be completely horizontal, so the intersection point of the river width length and the river shape on both sides is allowed to have a certain inclination angle and height difference. That is, the following conditions are met:

[0048]

[0049] Wherein, h is the vertical height from the bottom to the top to find the satisfying condition, m; w is the river width, m; ∈ is the error, ∈=0.1m.

[0050] Using numerical iteration method to solve, starting from h min =min(f(x)), gradually increasing h; for each h, solving the equation f(x)=h, getting the left intersection point x left (h) and the right intersection point x right (h); judge the following equation:

[0051]

[0052] When the above minimum value ≤ ∈, accept the solution. That is, the corresponding position of the river width and the river is found.

[0053] After finding the corresponding relationship, the flow area formed by the river width and the cross-section shape is Gauss-Kronrod integrated, and the formula is as follows:

[0054]

[0055] Wherein, A is the calculated flow area, m2 ; W is river width, m; W / 2 is the scaling factor of the integral interval; m is the total number of nodes; W k is the weight corresponding to the integral node k; x k is the position of the integral node in the original interval [0, w].

[0056] Finally, the difference between the long time series river width and the minimum river width corresponding to the area is obtained, that is, the flow area change amount:

[0057] δA it = A i -A min

[0058] Where, δA it is the flow area change amount, m 2 ; A i is the flow area corresponding to time i, m 2 ; A min is the flow area corresponding to the minimum river width in all time series, m 2 .

[0059] In step 6), the length of the long time series river width is matched with the river cross section shape, the elevation difference of the intersection points on the left and right sides is not more than 10 cm, the flow area formed by the river width and the cross section shape is calculated by Gauss-Kronrod integral, and the integral rule is G7-K15.

[0060] In further schemes, in step 7), the Manning formula and the multi-station hydraulic geometry method AMHG are used as the basic equation for flow estimation, the Manning formula and the AMHG equation are logarithmically transformed and a likelihood function is constructed by using the Bayesian method, and the Manning equation and the logarithmically transformed equation are respectively:

[0061]

[0062] Where, Q t is the flow, m 3 / s; t is the time; n is the Manning coefficient, m -1 / 3 ·s; A it is the cross-sectional area: A 0i is the reference cross-sectional area, δA it is the flow area change amount, m 2 ; W it is the river width, m; S it is the slope, dimensionless.

[0063] The AMHG equation and the logarithmically transformed equation are respectively:

[0064]

[0065] log Wit = b i (log Q t - log Q c ) + log W c + ∈ g

[0066] where Wit is the river width, m; Q t is the discharge, m 3 / s; W c , Q c and b i are AMHG parameters: Q c is the characteristic discharge, m 3 / s; W c is the global characteristic river width corresponding to Q c , m; b i is the width-discharge exponent, dimensionless; ∈ g is the AMHG error term, assumed to be normally distributed (∈g ~ N(0, σg)), dimensionless.

[0067] Prior distribution settings:

[0068] A 0i is the cross-sectional area corresponding to all minimum river widths of step 6), A 0i = A min ; logbi is defined by the regression model: logbi = 0.0216 + 0.4578 · SD(logWit) + ∈b; Qt is set as a truncated normal distribution based on the global hydrological model results and physical constraints (W min × d min × V min ≤ Qt ≤ W max × d max × V max ), where W min and W max are the minimum / maximum observed river widths, m; d min = 0.5 m, d max = 10 m: minimum / maximum water depths; V min = 0.5 m / s, V max = 5 m / s: minimum / maximum flow velocities.

[0069] Joint posterior distribution:

[0070] p(Θ | x) ∝ f(x | θ) π(Θ)

[0071] Θ is the set of unknown parameters, including Q t , n, A 0i , W c , Q c , b i; x is the observed data (W it , S it , δA it ); f(x|Θ) is the joint likelihood function, which is composed of the Manning equation and the AMHG equation error term; π(Θ) is the prior distribution of the parameters.

[0072] The posterior distribution is approximated by Bayesian inference Monte Carlo sampling, which draws samples from the specified distribution and calculates statistics (mean, standard deviation, quantile) to approximate the posterior distribution. The long time series of river width and slope obtained in step 4) and the long time series of flow area obtained in step 6) are used to obtain the flow estimation results according to the Bayesian inference Monte Carlo sampling method.

[0073] The advantages and beneficial effects of the present application are:

[0074] The present application uses self-produced high-precision digital elevation model DEM, fully considers the reality and spatial variability of river shape, and combines real-time water surface to invent a RCBAM model construction method for remote sensing inversion of river flow, which solves the defects of parameter acquisition uncertainty in traditional BAM model and improves the accuracy of river flow estimation. Compared with the Pearson correlation coefficient analysis of rainfall, the correlation coefficients of RCBAM and BAM are 0.67 and 0.59 respectively, and RCBAM is improved by 0.08 compared with BAM. It can realize efficient and accurate remote sensing inversion of river flow. BRIEF DESCRIPTION OF DRAWINGS

[0075] The present application will be further described below in conjunction with the drawings and examples.

[0076] Figure 1 It is a DEM image of Xibaxiaqu River Basin based on GF-7 production in the example.

[0077] Figure 2 It is a center line of a certain section of Xibaxiaqu River in December 2020 in the example.

[0078] Figure 3 It is an example of river surface width at four cross sections of Xibaxiaqu River in December 2020 in the example.

[0079] Figure 4 It is an example of slope calculation at one cross section of Xibaxiaqu River in December 2020 in the example.

[0080] Figure 5 It is an example of river cross section shape at one cross section of Xibaxiaqu River in December 2020 in the example.

[0081] Figure 6 It is an example of flow area (green shaded area) corresponding to river width at one cross section of Xibaxiaqu River in December 2020 in the example.

[0082] Figure 7 River flow graph of Xibaxiaqu in 2003-2022 for the example.

[0083] Figure 8 Persion correlation analysis graph of rainfall data and remote sensing monitoring Xibaxiaqu runoff results for the example. DETAILED DESCRIPTION

[0084] Example 1:

[0085] A method for constructing an RCBAM model for remotely sensing river flow, comprising the following steps:

[0086] 1) Collect remote sensing image data; including satellite front and rear view image pairs and time series of optical remote sensing images. In this embodiment, DEM production data comes from the China Resource Satellite Application Center's GF-7, with a spatial resolution of sub-meter; water surface remote sensing data comes from the European Space Agency's Sentinel-2 satellite, with a spatial resolution of 10m; the Xibaxiaqu river is selected as an implementation case, which is located in the Brahmaputra River basin and belongs to an area with no data, and is relatively representative.

[0087] 2) High-precision DEM generation: First, use the satellite front and rear view images of high-resolution images as input images, and use the satellite RPC parameters or ground control points to perform directional adjustment on the images, and perform stereo matching on the stereo image pairs to obtain a digital surface model (DSM) containing ground buildings and vegetation, and then apply filtering to the DSM to remove artificial features and vegetation to obtain a DEM.

[0088] In this embodiment, GF-7 is used as input image, SGM is used to obtain DSM, and IMF is used to remove features and vegetation to obtain 5m high-precision digital elevation model DEM data. As shown in Figure 1 Xibaxiaqu basin DEM image produced based on GF-7.

[0089] 3) River centerline acquisition: based on optical images, calculate the normalized water index (NDWI), and use the OTSU adaptive threshold method to obtain a water binary image, where the water pixel value is 1 and the non-water pixel value is 0. Repeat the erosion operation, and when the erosion is only left with a line with a width of 1 pixel, this line is the river centerline.

[0090] In this embodiment, NDWI is calculated by Sentinel-2 optical image and a binary water image is obtained, and the sliding window is set to 3x3. Figure 2 River centerline of Xibaxiaqu in December 2020.

[0091] 4) River width and slope acquisition: Based on the water body image and river centerline obtained in step 3), the river width is obtained by using the perpendicular distance method. Starting from each point on the centerline, search in the direction perpendicular to the centerline, and find the two corresponding points of the water body boundary in the vertical direction. By measuring the distance between the two points, the river width at that position can be obtained.

[0092] In this embodiment, four river sections are selected, which are perpendicular to the river centerline and extend outward until they intersect with the boundaries of the river water body. The river width is the distance between the two intersection points. Figure 3 River surface width of four sections of Xibaxia River in December 2020.

[0093] Based on the DEM obtained in step 2), the slope is obtained by calculating the ratio of the elevation difference between pixels within a certain distance and the horizontal distance. In this embodiment, the slope is calculated by extending 500m along the upstream and downstream of the section, i.e. S=(127-110) / 1000=0.027, Figure 4 Slope calculation example image of one section of Xibaxia River in December 2020.

[0094] 5) River section shape extraction: At the river section, based on the DEM data produced in step 2), the river profile elevation data is cut perpendicular to the river centerline obtained in step 3). Based on the river profile elevation data, the river shape section shape diagram is drawn. Figure 5 River section shape example diagram of one section of Xibaxia River in December 2020.

[0095] 6) Flow area calculation: The river width obtained in step 4) and the river shape section shape line obtained in step 5) are integrated by Gauss-Kronrod integration to calculate the section area corresponding to the river width, and the flow area formed by the river section under different river width conditions is obtained. Figure 6 Flow area corresponding to river width (green shaded area) example diagram of one section of Xibaxia River in December 2020.

[0096] 7) Reasoning estimation flow: Mannings formula and multi-station hydraulic geometry AMHG are used as the basic equation for flow estimation. The Mannings formula and AMHG equation are logarithmically transformed and a likelihood function is constructed using the Bayesian method. Samples are drawn from the specified distribution and statistical quantities are calculated to approximate the posterior distribution through the Bayesian inference Monte Carlo sampling method. The parameters used in the above methods and formulas are as follows: the reference cross-sectional area is set to the flow area corresponding to the minimum river width; the flow area change part is the difference between the instantaneous river width, the minimum river width and the area formed by the river section shape curve; based on the river width and slope obtained in step 4) and the flow area with river shape information obtained in step 6), the flow is estimated. In this embodiment, the river flow of Xibaxiaqu River for 20 years (2003-2022) is calculated. Figure 7 River flow graph of Xibaxiaqu for 2003-2022.

[0097] Results verification:

[0098] Since the Xibaxiaqu River basin is an ungauged area, Pearson correlation analysis is performed on the rainfall data and the results of remote sensing monitoring of Xibaxiaqu River runoff (RCBAM) and BAM. Figure 8 The results show that the remote sensing monitored river flow has a significant positive correlation with rainfall, and the correlation coefficients of RCBAM and BAM are 0.67 and 0.59, respectively. The monitoring accuracy of RCBAM based on river shape for river flow is improved by 0.08.

[0099] Finally, it should be noted that the above is only used to illustrate the technical solutions of the present application and not to limit. Although the present application has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for constructing an RCBAM model for remote sensing inversion of river flow, characterized in that: include: 1) Acquire remote sensing image data; This includes satellite forward and backward image pairs and time-series optical remote sensing images; 2) High-precision DEM generation: Using satellite forward and backward image pairs as input, the digital surface model (DSM) is obtained through directional adjustment and stereo matching algorithms. The DSM is then filtered to remove artificial features and vegetation to obtain numerical elevation model (DEM) data. 3) Obtaining the river centerline: Using optical remote sensing images as input, the Normalized Difference Water Index (NDWI) is calculated to obtain a binary image of the water body. The binary image of the water body is then subjected to an erosion operation to obtain the river centerline. 4) Obtaining river width and slope: The river width is obtained using the vertical distance method based on the river centerline; the slope is obtained based on DEM data. 5) Extraction of river cross-section shape: Obtain the river cross-section elevation data by cutting a profile perpendicular to the acquired river centerline in the DEM data. Based on the river cross-section elevation data, obtain the river cross-section shape curve by linear interpolation of discrete elevation points. 6) Calculation of flow area: Correspond the river width to the river channel cross-sectional shape curve, and perform Gauss-Cronrod integral to obtain the flow area formed by the river channel cross-section under different river width conditions; The Gauss-Cronrod integral is given by the following formula: Where A is the calculated flow area, m 2 W is the river width, in meters; W / 2 is the scaling factor for the integration interval; m is the total number of nodes; W k It is the weight corresponding to the integration node k; x k It is the position of the integration node in the original interval [0, w]; The change in flow area is obtained by subtracting the area corresponding to the minimum river width from the area corresponding to the maximum river width over a long time series. δA it = Yes i -IN min Where, δA it The change in flow area is m 2 A i Let m be the flow area corresponding to time i. 2 A min Let m be the flow area corresponding to the minimum river width across all time series. 2 ; 7) River Flow Estimation: The Manning formula and AMHG are used as the basic equations for flow estimation. Bayesian methods are used to perform logarithmic transformations on the Manning formula and AMHG equations and construct likelihood functions. A Monte Carlo sampling method based on Bayesian inference is used to draw samples from a specified distribution and calculate statistics to approximate the posterior distribution. The parameters used in the above formulas are: the reference cross-sectional area is set as the flow area corresponding to the minimum river width; the change in flow area is the difference between the instantaneous river width, the minimum river width, and the area formed by the river cross-sectional shape curve; the obtained river width, slope, and flow area are calculated according to Bayesian inference to obtain the flow estimation results.

2. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: In step 2), the data for the front and rear view images used for DEM production must be taken during the dry season of the river to maximize the inclusion of river channel shape information; the accuracy of the high-precision DEM is 5m.

3. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: The specific steps for step 2) are as follows: 2-1 Use satellite imagery RPC parameters or ground control points to perform orientation adjustment on the imagery to obtain accurate orientation parameters. Use the accurate positioning model and orientation parameters to perform epipolar sampling on the forward and backward images of the image pair to obtain approximate epipolar images. 2-2 A semi-global stereo matching algorithm is used to perform stereo matching and point intersection on approximate epipolar images to obtain high-density three-dimensional point clouds; 2-3 point cloud data registration generates a DSM containing ground buildings and vegetation, and then the DSM is filtered to remove man-made features and vegetation to obtain a DEM.

4. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: In step 3), NDWI is calculated, and the OTSU adaptive thresholding method is used to obtain a binary image of the water body, where the value of water body pixels is 1 and the value of non-water body pixels is 0. The erosion operation is repeated continuously until only a line with a width of 1 pixel remains in the water body. This line is the center line of the river channel.

5. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: In step 3), during the erosion process, an erosion determination is performed, specifically divided into 8 cases: 1 data point around the pixel, with a center point gradient of 7; 2 data points around the pixel, with a center point gradient of 6; 3 data points around the pixel, with a center point gradient of 5; 4 data points around the pixel, with a center point gradient of 4; 5 data points around the pixel, with a center point gradient of 3; 6 data points around the pixel, with a center point gradient of 2; and 7 data points around the pixel, with a center point gradient of 1.

6. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: Based on the vertical distance method used in step 4) to obtain the specific river width: starting from each point on the center line, search in the direction perpendicular to the center line, find two corresponding points of the water body boundary in the vertical direction, and calculate the distance between these two points to obtain the river width at that location; Step 4) Obtain slope based on DEM data: Obtain slope by calculating the ratio of the elevation difference to the horizontal distance between pixels within or along the river channel; Step 4) Obtain river width and slope data for selected cross-sections based on long-term optical images, with no fewer than 4 cross-sections.

7. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: The cross-sectional shape of the river obtained in step 5) should include the entire river channel within the dikes on both sides of the river.

8. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: In step 6), the river width length of the obtained long-term sequence is matched with the shape of the river channel cross section, and the elevation difference between the intersection points on the left and right sides does not exceed 10 cm.

9. The method for constructing an RCBAM model for remote sensing inversion of river flow according to claim 1, characterized in that: In step 7), the Manning equation and the equation after the logarithmic transformation are as follows: Among them, Q t For flow rate, m 3 / s; t is time; n is the Manning coefficient, m -1 / 3 ·s;A it Cross-sectional area: A 0i Let δA be the reference cross-sectional area. it The change in flow area is m. 2 W it River width, m; S it Slope, dimensionless; The AMHG equation and its logarithmic transformation are as follows: log W it =b i (log Q t -log Q c )+log W c +∈ g Among them, W it The width of the river is m; Q t For flow rate, m 3 / s;W c Q c With b i For AMHG parameters; Q c For characteristic flow, m 3 / s;W c To Q c The corresponding global feature river width, m; b i Width-flow index; ∈ g For AMHG error terms; Where A0 is the flow area corresponding to the minimum river width of all cross sections; δA it The value is the difference between the instantaneous river width, the minimum river width, and the area formed by the river channel cross-sectional shape curve at that cross-section.

Images