A method for determining a deformed bay of a multi-branch river

Abstract

The application discloses a kind of methods suitable for the determination of multi-branch river abnormal river bay, belong to the field of hydraulic engineering.The method includes the rough extraction of multi-branch river water body, the description of river multi-branch network structure, the batch calculation of river each branch plane geometric index, the identification of multi-branch river main flow path, the determination and output of multi-branch river abnormal river bay under variable space scale.The traditional quantification and discrimination method of abnormal river bay is extended to the multi-branch river with complex plane form.The determination method proposed in the application not only embodies the advantages of remote sensing and image processing technology in automation and computing efficiency, but also considers the computing accuracy close to visual interpretation, which not only saves a lot of cost, but also avoids the calculation error caused by data source and subjectivity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydraulic engineering, and particularly relates to a method for determining a deformed river bend suitable for a multi-branch river. BACKGROUND

[0002] The formation, development and cutting of river bends are one of the most basic self-organizing behaviors of alluvial rivers under the combined action of upstream water and sediment and local boundary conditions. However, when the water and sediment conditions and the local boundary conditions, especially the layout and shape parameters of river engineering, do not match, deformed river bends will occur. Deformed river bends usually have a large curvature and a relatively small bending radius, and present an "S" type, "Z" type, "Ω" type and the like in planar morphology. The formation and development of deformed river bends not only cause the river to deviate from the river, so that the local river control engineering cannot play a normal control role, but also may cause "horizontal" and "oblique" rivers, which may cause water flow to top and destroy the dikes or river regulation engineering, and seriously threaten the property and life safety of the residents along the river. Precise quantification and identification of deformed river bends can help to clarify the mechanism of the development and disappearance of deformed river bends and predict the formation of deformed river bends, so it has been one of the focuses of river scholars and water conservancy engineers for a long time.

[0003] The traditional quantification and identification method of deformed river bends relies on the estimation of a series of river bend related parameters, including the curvature of the river, the bending radius, the curvature coefficient, the amplitude and the river span, etc. Although the above deformed river bend parameters have clear definitions and calculation methods for rivers with single river channels, such as straight rivers and curved rivers, they still have the following three defects: first, large rivers in nature usually exhibit a multi-branch network structure with complex planar morphology, such as braided rivers and branch rivers, and the conventional deformed river bend parameter statistics and calculation method is no longer applicable; second, the traditional deformed river bend identification is mainly through visual interpretation, which not only consumes a lot of manpower and material resources for large multi-branch rivers, but also cannot avoid the estimation error caused by subjectivity; finally, the data sources (including remote sensing images, unmanned aerial vehicle pictures, prototype measurement data, etc.) relied on by the deformed river bend calculation are different, which makes the deformed river bend determination standards and quantitative parameter thresholds based on different data sources cannot be compared and learned from each other.

[0004] At the same time, the development of remote sensing and image processing technology has laid a technical foundation for the automatic quantitative calculation of river planar morphology, making it possible to automatically determine the deformed river bend of single river channel. However, how to extend the above deformed river network quantification indicators and calculation methods to multi-branch rivers with complex planar morphology still needs further consideration.

[0005] Therefore, we designed a method for determining a deformed river bend suitable for a multi-branch river to solve the above problems. SUMMARY

[0006] The purpose of the present application is to solve the problems in the prior art that the statistical and calculation methods of conventional and regular meandering river parameters are no longer applicable to the quantification and discrimination of meandering river, and consume a large amount of manpower and material resources, while also cannot avoid the estimation error caused by subjectivity, the meandering river discrimination criteria and quantification parameter thresholds obtained from different data sources cannot be compared and learned from each other, and the meandering river quantification index and its calculation method still need to be further considered when it is applied to a multi-branch river with complex plane morphology, and therefore a meandering river discrimination method suitable for multi-branch rivers is proposed, therefore, the research and proposal of an automatic meandering river discrimination method suitable for multi-branch rivers not only embody the advantages of remote sensing and image processing technology in automation and calculation efficiency, but also take into account the calculation accuracy close to visual interpretation, which not only saves a large amount of cost, but also avoids the calculation error caused by data sources and subjectivity.

[0007] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0008] A meandering river discrimination method suitable for multi-branch rivers, comprising the following steps:

[0009] Step S1, coarse extraction of multi-branch river water body: input a satellite remote sensing image, determine the initial value of the model parameter, calculate the improved normalized difference water index, and obtain the water body binary image of the satellite remote sensing image;

[0010] Step S2, description of river multi-branch network structure: extract the skeleton line of the river channel in the water body binary image, and record the pixel point coordinates of all skeleton lines, traverse the skeleton line pixel points, and determine the branch points of the river in the skeleton line one by one, and record the pixel point coordinates of the branch points;

[0011] Step S3, batch calculation of river each branch plane geometric index: execute a tracking algorithm on the determined river end point or river branch point, start from any branch point or end point, and traverse the pixel points of the skeleton line connected to the branch point or end point in turn, until other branch points or end points, during the tracking algorithm, based on the coordinate operation of the pixel points of the skeleton line, calculate and record the length and width of all single branches in the river in turn, and construct the length and width matrix of the single branches of the multi-branch river;

[0012] Step S4, identification of the main flow path of the multi-branch river: taking the ratio of the length and width of the branch as the weight, a characteristic path matrix and a length matrix of the multi-branch river are respectively constructed, the weighted characteristic path length between each river endpoint or branch point pair in the multi-branch river is calculated in turn, and the results of the characteristic path matrix and the length matrix are iterated to identify the weighted characteristic path from the uppermost branch point to the lowermost branch point of the multi-branch river as the main flow path of the river, and the pixel coordinates thereof are recorded in turn;

[0013] Step S5, determination and output of the deformed bay of the multi-branch river under variable spatial scales: for any pixel point on the main flow path of the river, a circular detection window with a spatial scale R is used to calculate the curvature of the main flow path of the river in the window, and it is determined whether the obtained river curvature value is greater than a given deformed bay curvature threshold value. If it is satisfied, it is considered that a deformed bay has been formed at the pixel point. All pixel points on the main flow path of the river are traversed, and the corresponding position and curvature of the deformed bay of the river under the spatial scale R are output.

[0014] As a further preferred embodiment of the present application, in step S1, the following steps are used for the rough extraction of the water body of the multi-branch river:

[0015] Step S11, inputting a satellite remote sensing image and determining the initial value of the model parameter: inputting a satellite remote sensing image and determining the resolution Re and the size of the input satellite remote sensing image, setting the value of the model parameter, the model parameter including: the improved normalized difference water index threshold MNDWI for distinguishing water body from non-water body in the remote sensing image T , the river width calculation parameter Δw, the maximum size R max and the minimum size R min of the circular window for detecting deformed river network under different scales, and the increase step ΔR, the curvature threshold S T for distinguishing normal bays from deformed bays;

[0016] Step S12, calculating the improved normalized difference water index to roughly extract the water body binary image from the satellite remote sensing image: calculating the improved normalized difference water index, and roughly extracting the water body binary image from the satellite remote sensing image based on the OSTU threshold segmentation method:

[0017] According to the improved normalized difference water index formula: The MNDWI index of each pixel point of the input satellite remote sensing image is calculated, and Green and MIR respectively represent the reflectivity of the green and mid-infrared bands in the satellite remote sensing image;

[0018] Each pixel point in the satellite remote sensing image is traversed, and the MNDWI index of each pixel point is compared with the given threshold MNDWI Tthe size of the MNDWI index is greater than a given threshold MNDWI T , the pixel point is determined as water body, otherwise, it is determined as non-water body, and finally the binary image of water body of the input satellite remote sensing image is obtained.

[0019] As further preferred in the present solution, in step S2, the delineation of the multi-branch network structure of the river includes the following steps:

[0020] In step S21, the skeleton line of the river channel in the binary image of water body is extracted by using a thinning algorithm, and the pixel point coordinates of all skeleton lines are recorded:

[0021] The water body with the largest connected area in the binary image of water body is retained, the water body is thinned into a set of continuous skeleton lines of the river channel with single resolution width by using Zhang-Suen thinning algorithm, and the coordinates of the pixel points belonging to the skeleton line are recorded in matrix Ske={X i ,Y i}n s , the pixel points of the skeleton line are sorted in the direction from the upstream to the downstream of the river, X i ,Y i represent the coordinates of the i-th pixel point on the skeleton line, and n s represents the total number of pixel points of the skeleton line;

[0022] In step S22, the pixel points of the skeleton line are traversed, the branch points of the river in the skeleton line are determined one by one, and the pixel point coordinates of the branch points are recorded:

[0023] The neighbor pixel points of any pixel point of the skeleton line are traversed, and whether each neighbor pixel point also belongs to the skeleton line is detected in turn, and the number n nei of neighbor pixel points that also belong to the skeleton line is counted.

[0024] According to the value of n nei , it is determined whether the pixel point of the skeleton line belongs to the branch point of the river: if n nei =1, the pixel point of the skeleton line is defined as the end point of the river, if n nei =2, the pixel point of the skeleton line is defined as a normal skeleton line pixel point, and if n nei >2, the pixel point of the skeleton line is defined as the branch point of the river.

[0025] In step S23, the pixel point coordinates of all the points determined as the end points of the river and the branch points of the river are recorded in matrix , the end points of the river and the branch points of the river are sorted in the order from the upstream to the downstream of the river, and n n represents the total number of the end points of the river or the branch points of the river.

[0026] As further preferred in the present solution, in step S3, the batch calculation of the planar geometric indicators of each branch of the river includes:

[0027] In step S31, a tracking algorithm is performed on the determined river endpoints or the branch points of the river, and from any branch point or endpoint, the pixel points of the skeleton lines connected to the branch point or endpoint are sequentially traversed until other branch points or endpoints:

[0028] The length matrix Len = {l i,j} and the width matrix Wid = {w i,j} of the river are defined, where i and j represent the serial numbers of the branch points or endpoints of the river, l i,j and w i,j represent the length and width of the single branch connected by the i-th branch point or endpoint and the j-th branch point or endpoint, respectively, and from the i-th branch point or endpoint of any river as the starting point, each pixel point of the skeleton line on the branch connected to the branch point is sequentially traversed until the j-th branch point or endpoint of the river;

[0029] In step S32, during the tracking algorithm, the length and width of the single branch in the river are sequentially calculated and recorded based on the coordinate operation of the pixel points of the skeleton line:

[0030] When traversing any pixel point k of the skeleton line of the branch connected by the i-th and j-th branch points or endpoints, first, the Euclidean distance between the pixel point k of the skeleton line and its adjacent previous pixel point k-1 is calculated based on the coordinate operation: Re represents the resolution of the input satellite remote sensing image, and the tangent slope of the branch at the pixel point k is calculated

[0031]

[0032] where X k-1 , Y k-1 and X k+1 , Y k+1 represent the coordinates of the previous pixel point k-1 and the next pixel point k+1 of the skeleton line directly connected to the pixel point k;

[0033] The length of the branch connected by the i-th and j-th branch points or endpoints of the river is equal in value to the cumulative value of the Euclidean distances of all adjacent pixel points of the skeleton line:

[0034]

[0035] where N K represents the total number of pixel points of the skeleton line included in the branch;

[0036] Taking the pixel point k of the skeleton line as the center, the pixel point k is taken as the center, and the pixel point k is taken as the center Extending to the non-water body pixel point on the left and right banks respectively to generate a single resolution section line perpendicular to the branch corresponding to the skeleton line. For the section line generated on the pixel point of any skeleton line, when the length difference between the left and right section lines is less than the river width calculation parameter Δw, the width of all single branches in the river is equal in value to the length of the section line, otherwise the river width at the pixel point of the skeleton line is equal in value to twice the shorter one of the left and right parts of the section line:

[0037]

[0038] Wherein,

[0039] In the formula, And Respectively represent the left and right section line lengths at the pixel point k;

[0040] Step S33, constructing the length and width matrix of the single branch of the multi-branch river:

[0041] After calculating the length and width of the branch connected by the river branch point or end point pair i and j, update the corresponding elements in the river length matrix Len={l i,j} and the width matrix Wid={w i,j}, and then in the clockwise direction, repeat step S32 for the remaining branches connected to the river branch point or end point i, until all branches connected to the river branch point or end point i are executed, and the length and width of all single branches in the river are calculated and recorded;

[0042] Based on the pixel point coordinate matrix of the river end point or the river branch point determined in step S23 Traverse all river end points or river branch points in the matrix Repeat steps S32 and S33 to finally matrix the length and width information of all single branches of the river, and construct the length matrix Len={l i,j} and the width matrix Wid={w i,j} of the single branch of the multi-branch river.

[0043] As a further preferred embodiment of the present application, in step S4, the identification of the main flow path of the multi-branch river comprises the following steps:

[0044] Step S41, taking the ratio of the length and width of the branch as the weight, respectively constructing the characteristic path matrix Pat and the length matrix Dis of the multi-branch river:

[0045] Define the characteristic path length d between any river end point or river branch point pair i and ji,j When i≠j and i is directly connected to j, the characteristic path length is equal in value to the ratio of length to width of the river channel between i and j at the river end point or the branching point of the river, i.e.:

[0046]

[0047] The characteristic path matrix Pat = {p i,j} of the multi-branch river and the length matrix Dis = {d i,j} of the multi-branch river are defined, the initial value of p i,j is p i,j = j, and the initial value of d i,j is equal to the characteristic path length between i and j at the river end point or the branching point of the river;

[0048] In step S42, each river end point or branching point pair of the multi-branch river is traversed, the weighted characteristic path length between each river end point or branching point pair is calculated in turn, and the characteristic path matrix Pat and the length matrix Dis of the multi-branch river in step S41 are iterated:

[0049] The river end point or branching point pair i and j in the river is traversed, the sum d i,k,j of the characteristic path lengths from the river end point or branching point to each river end point or branching point pair i and j is calculated in turn:

[0050] d i,k,j = d i,k + d k,j ,

[0051] In the formula, d i,k represents the characteristic path length from the river end point or branching point i to the river end point or branching point k, d k,j represents the characteristic path length from the river end point or branching point k to the river end point or branching point j, if d i,k,j < d i,j , the value of the element d i,j in the characteristic path length matrix Dis = {d i,j} of the multi-branch river is iterated as d i,k,j , and the value of the element p i,j in the characteristic path matrix Pat = {p i,j} is iterated as the value of p i,k ;

[0052] After the final characteristic path matrix Pat = {p i,jAfterwards, the main flow path of the river is identified from the matrix, i.e. the flow path starts from the uppermost branch point and connects to the downstream branch point along the branch with relatively large flow, until the lowermost branch point is reached;

[0053] In step S43, the weighted feature path from the uppermost branch point to the lowermost branch point of the multi-branch river is identified as the main flow path of the river based on the final results of the feature path matrix Pat and the length matrix Dis of the multi-branch river, and the pixel coordinates of the river channel skeleton line along the main flow path are recorded in sequence, including:

[0054] For any river endpoint or branch point i along the main flow path of the river, the last column element in the ith row of the feature path matrix Pat is found , and the value of the element is the serial number of the downstream river endpoint or branch point connected to the branch point i along the main flow path. The pixel coordinates of the river channel skeleton line connected to the two branch points are recorded in sequence from upstream to downstream.

[0055] From the first branch point in the main flow path to the lowermost river endpoint or branch point of the river, the serial numbers of the river endpoints or branch points along the main flow path are recorded in the point set , where m i represents the ith branch point in the main flow path, and n m is the total number of branch points included in the main flow path. The skeleton line coordinates along the main flow path are recorded in sequence from upstream to downstream in the matrix Cha_Path={X i ,Y i} nT , where X i ,Y i represent the ith pixel point coordinates along the skeleton line corresponding to the main flow path, and n T represents the total number of pixel points included in the skeleton line corresponding to the main flow path.

[0056] As a further preferred embodiment of the present scheme, the determination and output of the multi-branch river abnormal meander with variable spatial scale in step S5 include:

[0057] In step S51, for any pixel point along the main flow path of the river, a circular detection window with a spatial scale of R is used to calculate the curvature of the main flow path within the window, and the following method is adopted:

[0058] A circular detection window with a variable radius and a spatial scale of R is set with the pixel point i along the skeleton line of the main flow path as the center, and the window can move downstream along the main flow path. The curvature value SI R of the main flow path within the window is calculated:

[0059]

[0060] In the formula, represents the bending length of the main flow path of the river within the circular detection window, which is equal in value to the cumulative value of the pixel point distances of all adjacent skeleton lines on the main flow path in the window:

[0061]

[0062] n j represents the total number of pixel points of the skeleton line on the main flow path of the river within the circular detection window, X j , Y j represents the coordinate of the jth pixel point on the skeleton line, X j+1 , Y j+1 represents the coordinate of the j+1th pixel point on the skeleton line;

[0063] represents the straight line length of the main flow path of the river within the circular detection window, which is equal in value to the straight line distance between the pixel point of the most upstream skeleton line and the pixel point of the most downstream skeleton line on the main flow path in the window:

[0064]

[0065] X1,Y1represents the coordinate of the pixel point of the most upstream skeleton line on the main flow path, represents the coordinate of the pixel point of the most downstream skeleton line on the main flow path;

[0066] The calculated main flow path bending value SI R is taken as the river bending value at the pixel point i of the skeleton line at the spatial scale R, and the larger the value, the more likely it is to form a deformed meander at this point;

[0067] The obtained result is recorded in the river bending matrix , R max and R min respectively represent the maximum and minimum scales of the given circular detection window, and ΔR is the increasing step of the circular detection window;

[0068] Step S52, it is judged whether the main flow path bending value obtained in step S51 is greater than a given deformed meander bending threshold value, and if it is satisfied, it is considered that a deformed meander has been formed at the pixel point:

[0069] The spatial scale R of the circular detection window is gradually increased from R min in steps of ΔR until R max, repeat step S51 to obtain the calculation results of the sinuosity of the river along the main flow path from upstream to downstream at different spatial scales, and output a river sinuosity matrix SI, and compare each element in the river sinuosity matrix SI with the sinuosity threshold value S T for detecting the abnormal meander if the size of the element is greater than the size of the threshold value S , it is determined that there is an abnormal meander at the spatial scale R at the pixel point on the skeleton line of the river;

[0070] Step S53: traversing all the pixel points on the skeleton line of the main flow path of the river, repeating steps S51 and S52, and outputting the corresponding position of the abnormal meander of the river at the spatial scale R and the sinuosity result.

[0071] Compared with the prior art, the present application has the following advantages: the present application is composed of five parts, i.e., coarse extraction of water body of a multi-branch river, description of network structure of multi-branch channels of the river, batch calculation of planar geometric indexes of each branch channel of the river, identification of the main flow path of the multi-branch river, and determination and output of abnormal meanders of the multi-branch river at variable spatial scales, and the traditional quantitative and discriminant method of abnormal meanders is extended to the multi-branch river with complex planar morphology; the determination method proposed in the present application not only embodies the advantages of remote sensing and image processing technology in automation and calculation efficiency, but also takes into account the calculation accuracy close to visual interpretation, thereby saving a large amount of cost and avoiding calculation errors caused by data sources and subjectivity. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 A determination method for abnormal meanders of a multi-branch river according to the present application;

[0073] Figure 2 A schematic diagram of a binary image of water body extracted in an embodiment of the present application;

[0074] Figure 3 A schematic diagram of a single branch channel skeleton line of a river and branch points and end points thereof extracted in an embodiment of the present application;

[0075] Figure 4 A schematic diagram of a hydrological section generated in the process of calculating the river width of a single branch channel in an embodiment of the present application;

[0076] Figure 5 A schematic diagram of a main flow path of a multi-branch river and a variable size circular window for detecting abnormal meanders in an embodiment of the present application. DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present application will be clearly and completely described below 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, but not all the embodiments.

[0078] Referring to Figure 1 , the embodiment proposes a method for determining the abnormal river bay of multi-branch river, which comprises the following steps:

[0079] Step S1, coarse extraction of water body of multi-branch river: input satellite remote sensing image, determine the initial value of model parameter, calculate improved normalized difference water index, obtain water body binary image of satellite remote sensing image, and adopt the following steps:

[0080] Step S11, input satellite remote sensing image and determine the initial value of model parameter: input satellite remote sensing image and determine the resolution Re and size MxN (wherein M and N are respectively the row number and column number of image pixel points) of the input satellite remote sensing image; set the value of model parameter, which includes: improved normalized difference water index threshold MNDWI for distinguishing water body from non-water body in remote sensing image, river width calculation parameter Aw, maximum size R and minimum size R of circular window for detecting abnormal river network at different scales, and its increasing step size AR, and bending threshold S for distinguishing normal river bay from abnormal river bay. T max min T .

[0081] Step S12, calculate improved normalized difference water index, and coarsely extract water body binary image from satellite remote sensing image: calculate improved normalized difference water index, and coarsely extract water body binary image from satellite remote sensing image based on OSTU threshold segmentation method:

[0082] According to the formula of improved normalized difference water index: Calculate the MNDWI index of each pixel point of the input satellite remote sensing image, wherein Green and MIR respectively refer to the reflectivity of green and mid-infrared band in satellite remote sensing image.

[0083] Traverse each pixel point in the satellite remote sensing image, and compare the size of the MNDWI index of each pixel point with the given threshold MNDWI T , if the MNDWI index is greater than the given threshold MNDWI T , it is determined that the pixel point is water body, otherwise it is non-water body, and finally the water body binary image (shown in 1) of the input satellite remote sensing image is obtained. Figure 2

[0084] Step S2, description of multi-branch river network structure: extract the skeleton line of river channel in water body binary image, record the pixel point coordinates of all skeleton lines, traverse the skeleton line pixel points, determine the branch point of river in the skeleton line one by one, and record the pixel point coordinates of the branch point, including the following steps:

[0085] ​​​​Step S21: Use a thinning algorithm to extract the skeletal lines of the river channel in the binary image of the water body, and record the pixel coordinates of all skeletal lines:

[0086] First, retain the water bodies with the largest connected area in the binary water image obtained in step S12 to eliminate noise caused by the presence of lakes and ponds. Figure 2 As shown in Figure 2), secondly, the Zhang-Suen thinning algorithm is used to refine the water body into a set of continuous, single-resolution width channel skeleton lines (as shown in Figure 2). Figure 3 As shown in Figure 3), the coordinates of the pixels belonging to the skeleton line are recorded in the matrix Ske = {X} i ,Y i}n s In the diagram, the pixels of the skeletal lines are sorted from upstream to downstream of the river. X i ,Y i Represents the coordinates of the i-th pixel on the skeletal line, n s This represents the total number of pixels in the skeleton line.

[0087] Step S22: Traverse the pixels of the skeleton line, determine the branching points of the river in the skeleton line one by one, and record the pixel coordinates of the branching points:

[0088] Iterate through the neighboring pixels of any skeletal line pixel, checking each neighboring pixel in turn whether it also belongs to a skeletal line, and count the neighboring pixels that also belong to a skeletal line. Figure 3 The number n (as shown in Figure 4) nei ;

[0089] According to n nei The value of n determines whether a pixel of the skeleton line belongs to a river branching point: if n nei =1, then the pixel of the skeleton line is defined as the river endpoint ( Figure 3 As shown in Figure 5), if n nei =2, then the pixel of this skeleton line is defined as a normal skeleton line pixel. Figure 3 As shown in Figure 6), if n nei If the value is greater than 2, then the pixel of the skeleton line is defined as the branching point of the river. Figure 3 (As shown in Figure 7).

[0090] Step S23: Iterate through the pixels of each skeletal line obtained in step S21, repeat step S22, and record the coordinates of all pixels identified as river endpoints and river branching points in a matrix. In the diagram, the river's endpoints and its bifurcation points are ordered from upstream to downstream, n. n This indicates the total number of river endpoints or river branching points.

[0091] Step S3, batch calculation of planar geometric indexes of each branch of the river: a tracking algorithm is performed on the determined river endpoints or branch points of the river, and each pixel point of the skeleton line connected to the branch point or endpoint is traversed in sequence from any branch point or endpoint until the other branch point or endpoint. During the tracking algorithm, the length and width of all single branches in the river are calculated and recorded in sequence based on coordinate calculation of the pixel points of the skeleton line, and a length and width matrix of single branches of the multi-branch river is constructed. Specifically, the following steps are included:

[0092] Step S31, a tracking algorithm is performed on the determined river endpoints or branch points of the river, and each pixel point of the skeleton line connected to the branch point or endpoint is traversed in sequence from any branch point or endpoint until the other branch point or endpoint.

[0093] First, define the length matrix Len = {l i,j} and the width matrix Wid = {w i,j} of the river, where i and j represent the serial numbers of the branch points or endpoints of the river, l i,j and w i,j represent the length and width of the single branch connected by the i k-1 th branch point or endpoint and the j k-1 th branch point or endpoint. Second, take any branch point or endpoint i of the river as the starting point, and traverse each pixel point of the skeleton line on the branch connected to the branch point until another branch point or endpoint j of the river.

[0094] Step S32, during the tracking algorithm, the length and width of the single branch in the river are calculated and recorded in sequence based on coordinate calculation of the pixel points of the skeleton line:

[0095] When traversing any pixel point k of the skeleton line of the branch connected by i and j,

[0096] First, the Euclidean distance between the pixel point k of the skeleton line and the pixel point k-1 of the adjacent previous skeleton line is calculated based on coordinate calculation:

[0097] Re represents the resolution of the input satellite remote sensing image, and the tangent slope of the branch at the pixel point k is calculated

[0098]

[0099] In the formula, X k-1 , Y k-1 , X k+1 , and Y k+1 represent the coordinates of the pixel point k-1 of the previous skeleton line and the pixel point k+1 of the next skeleton line directly connected to the pixel point k of the skeleton line;

[0100] The length of the branch connected by the bifurcation point or the end point i and j is equal to the sum of the Euclidean distance of all the adjacent pixel points of the skeleton line on the branch:

[0101]

[0102] where N K represents the total number of the pixel points of the skeleton line contained in the branch.

[0103] Secondly, the pixel point k of the skeleton line is taken as the center, and the left and right sides are extended to the non-water pixel points, respectively, to generate a single-resolution cross-section line (shown in Fig. 8) perpendicular to the branch corresponding to the skeleton line. Figure 4 The width of the branch is equal to the average value of the lengths of the cross-section lines perpendicular to the branch.

[0104] It should be noted that not all the cross-section lines generated at the pixel points of the skeleton lines in the branch can be directly used for the calculation of the river width, especially the pixel points of the skeleton lines near the bifurcation point or the end point of the river. The lengths of the left and right cross-section lines (shown in Fig. 9) at these pixel points usually have a large difference, which will cause an error in the calculation of the river width. Therefore, the river width calculation parameter Δw is set in the method. For the cross-section line generated at the pixel point of any skeleton line, when the difference between the lengths of the left and right cross-section lines is less than the river width calculation parameter Δw, the width of all the single branches in the river is equal to the length of the cross-section line. Otherwise, the river width at the pixel point of the skeleton line is equal to twice the shorter one of the left and right parts of the cross-section line: Figure 4

[0105] wherein,

[0106]

[0107] wherein, and represent the lengths of the left and right cross-section lines (shown in Figs. 10 and 11) at the pixel point k, respectively. Figure 4

[0108] Step S33: Constructing the length and width matrix of the single branch of the multi-branch river:

[0109] After the lengths and widths of the branches connected by the bifurcation point or the end point i and j are calculated, the length matrix Len = {l i,j} and the width matrix Wid = {w i,j ​​​Then, for the corresponding elements in}, repeat step S32 in a clockwise direction for the remaining channels connected to the river branching point or endpoint i, until all channels connected to the river branching point or endpoint i have been processed, thus completing the calculation and recording of the length and width of all individual channels in the river.

[0110] The pixel coordinate matrix based on the river endpoints or river bifurcation points determined in step S23. Traversing the matrix For all river endpoints or river bifurcation points, repeat steps S32 and S33 to finally matrix the length and width information of all individual channels of the river, constructing the length matrix Len={l i,j} and the width matrix Wid={w i,j}

[0111] Step S4: Identification of the main flow path of the multi-branched river: Using the ratio of the length to the width of the branch channel as the weight, construct the feature path matrix and length matrix of the multi-branched river respectively. Traverse each river endpoint or branch point pair in the multi-branched river, calculate the weighted feature path length between each river endpoint or branch point pair in turn, and iterate the results of the feature path matrix and length matrix. Identify the river skeleton line traversed by the weighted feature path from the upstream branch point to the downstream branch point of the multi-branched river as the main flow path of the river, and record its pixel coordinates in turn.

[0112] It is worth mentioning that, in this embodiment, the identification of the main flow path of a multi-branched river includes the following steps:

[0113] Step S41: Using the ratio of the length to the width of the channel as weights, construct the feature path matrix Pat and the length matrix Dis for the multi-branched river.

[0114] First, define the characteristic path length d between any pair of river endpoints or river bifurcation points i and j. i,j When i≠j and i and j are directly connected, the characteristic path length is numerically equal to the ratio of the length to the width of the channel connecting the river endpoints or river bifurcation points i and j; that is, the wider and shorter the channel, the smaller the characteristic path length. When i≠j and i and j are not directly connected, d i,j =∞; when i = j, d i,j =0, that is:

[0115]

[0116] Secondly, define the characteristic path matrix Pat = {p} of the multi-branched river. i,j} and the length matrix Dis={d of the multi-branched river i,j The row and column indices of these two matrices correspond to the indices of the river's bifurcation points or endpoints, respectively. The element p...i,j The initial value is: p i,j = j, d i,j The initial value is equal to the characteristic path length between the river endpoints or the bifurcation points of the river for i and j.

[0117] Step S42, traverse each pair of river endpoints or river bifurcation points in the multi-branched river, calculate the weighted characteristic path length between each pair of river endpoints or river bifurcation points in turn, and iterate the results of the characteristic path matrix Pat and the length matrix Dis of the multi-branched river in step S41:

[0118] Traverse all pairs of river endpoints or river bifurcation points i and j in the river, and calculate the sum d of the characteristic path lengths from the river endpoint or river bifurcation point to each pair of river endpoints or river bifurcation points i and j in turn i,k,j :

[0119] d i,k,j = d i,k + d k,j ,

[0120] In the formula, d i,k represents the characteristic path length from the river endpoint or river bifurcation point i to the river endpoint or river bifurcation point k, and d k,j represents the characteristic path length from the river endpoint or river bifurcation point k to the river endpoint or river bifurcation point j (assuming i < k < j). Compare d i,k , j with d i,j : If d i,k,j < d i,j , then iterate the value of the element d i,j in the characteristic path length matrix Dis = {d i,j} of the multi-branched river as d​​​​​​​​​​​​​​​​​​

[0123] For any river endpoint or river bifurcation point i that passes through the main flow path of a river, find the last column element p in the i-th row of the feature path matrix Pat. i,nn The value is the sequence number of the downstream river endpoint or river branch point that passes through the main flow path and is directly connected to the river endpoint or river branch point i (the river branch point sorting here is the same as in the matrix Node). The pixel coordinates of the river channel skeleton line connected to the above two branch points are recorded in order from upstream to downstream.

[0124] Starting from the first bifurcation point in the main flow path of the river, that is, the upstream bifurcation point, and continuing to the downstream endpoint or bifurcation point, record the serial numbers of the river endpoints or bifurcation points through which the main flow path passes in the point set. In the diagram, mi represents the i-th branch point in the main flow path, and n... m This represents the total number of branch points included in the main flow path. Simultaneously, the coordinates of the skeletal lines along the main flow path are recorded sequentially from upstream to downstream in the matrix Cha_Path = {X...} i ,Y i} nT X i ,Y i This represents the coordinates of the i-th pixel on the skeletal line corresponding to the main flow path, where n is the coordinate of the i-th pixel. T This represents the total number of pixels contained in the skeleton line corresponding to the main flow path.

[0125] Step S5: Determination and output of deformed river bends in multi-branched rivers under varying spatial scales: For any pixel on the main flow path of a river, a circular detection window with spatial scale R is used to calculate the curvature of the main flow path within the window. It is determined whether the obtained river curvature value is greater than the given deformed river bend curvature threshold. If it is satisfied, it is considered that a deformed river bend has been formed at that pixel. All pixels on the main flow path of the river are traversed, and the position and curvature results of the deformed river bends under spatial scale R are output.

[0126] Step S51: For any pixel along the main flow path of a river, a circular detection window with a spatial scale of R is used to calculate the curvature of the main flow path within the window, in the following manner:

[0127] Using pixel i on the skeleton line along the main flow path of any river as the center, set a circular detection window with a variable radius and spatial scale R that can move downstream along the main flow path of the river. Figure 5 As shown in Figure 13, calculate the tortuosity value SI of the main flow paths contained within this window. R :

[0128]

[0129] wherein, represents the length of the bend of the main flow path of the river within the circular detection window, which is equal in value to the cumulative value of the pixel point distances of all the adjacent skeleton lines on the main flow path in the window:

[0130]

[0131] n j represents the total number of pixel points of the skeleton lines on the main flow path of the river within the circular detection window, X j , Y j represents the coordinate of the jth pixel point on the skeleton line, X j+1 , Y j+1 represents the coordinate of the j+1th pixel point on the skeleton line;

[0132] represents the length of the straight line of the main flow path of the river within the circular detection window, which is equal in value to the straight line distance between the pixel point of the most upstream skeleton line and the pixel point of the most downstream skeleton line on the main flow path in the window:

[0133]

[0134] X1,Y1represents the coordinate of the pixel point of the most upstream skeleton line on the main flow path, represents the coordinate of the pixel point of the most downstream skeleton line on the main flow path;

[0135] with the calculated main flow path bend value SI R as the river bend value at the pixel point i of the skeleton line at the spatial scale R, the greater the value, the more likely it is to form a deformed meander at this point;

[0136] record the results in the river bend matrix , R max and R min respectively represent the maximum and minimum scales of the given circular detection window, and ΔR is the increasing step of the circular detection window;

[0137] Step S52, determine whether the main flow path bend value obtained in step S51 is greater than a given deformed meander bend threshold value, if it is satisfied, it is considered that a deformed meander has been formed at the pixel point:

[0138] the spatial scale R of the circular detection window is gradually increased from R min in steps of ΔR until R max, the step S51 is repeated to obtain the calculation results of the sinuosity of the river along the main flow path from the upstream to the downstream at different spatial scales, and a river sinuosity matrix SI is output, and each element in the river sinuosity matrix SI is compared with the sinuosity threshold S T in size, if , it is determined that there is a meandering river bend at the spatial scale R at the pixel point on the skeleton line.

[0139] In step S53, all pixel points on the main flow path skeleton line of the river are traversed, the step S51 and the step S52 are repeated, and the corresponding position of the meandering river bend and the sinuosity result at the spatial scale R are output.

[0140] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art in the technical range disclosed by the present application, according to the technical scheme and the inventive concept of the present application, equivalent replacement or change, should be covered in the protection scope of the present application.

Claims

1. A method for determining a meander of a multi-branch river, characterized by, The method comprises the following steps: Step S1, coarse extraction of the water body of the multi-branch river: inputting a satellite remote sensing image, determining initial values of model parameters, calculating an improved normalized difference water body index, and obtaining a water body binary image of the satellite remote sensing image; Step S2, description of the network structure of the multi-branch river: extracting the skeleton lines of the river in the water body binary image, recording the pixel point coordinates of all the skeleton lines, traversing the pixel points of the skeleton lines, determining the branch points of the river in the skeleton lines one by one, and recording the pixel point coordinates of the branch points; Step S3, batch calculation of the planar geometric indexes of each branch of the river: performing a tracking algorithm on the determined end points of the river or the branch points of the river, sequentially traversing the pixel points of the skeleton lines connected to the branch point or the end point from any branch point or end point, until the other branch point or end point, sequentially calculating and recording the length and width of all single branches in the river based on the coordinate calculation of the pixel points of the skeleton lines, and constructing a length and width matrix of the single branches of the multi-branch river; Step S4, identification of the main flow path of the multi-branch river: taking the ratio of the length and width of the branch as the weight, respectively constructing a characteristic path matrix and a length matrix of the multi-branch river, sequentially calculating the weighted characteristic path length between each pair of end points of the river or branch points of the river, and iteratively calculating the results of the characteristic path matrix and the length matrix to identify the main flow path of the multi-branch river, and sequentially recording the pixel point coordinates of the main flow path of the multi-branch river from the uppermost branch point to the lowermost branch point; Step S5, determination and output of the deformed river bend of the multi-branch river under variable spatial scales: for any pixel point on the main flow path of the river, a circular detection window with a spatial scale R is used to calculate the curvature of the main flow path of the river in the window, and it is judged whether the obtained river curvature value is greater than a given deformed river bend curvature threshold value, if yes, it is considered that a deformed river bend is formed at the pixel point, all pixel points on the main flow path of the river are traversed, and the corresponding position and curvature of the deformed river bend of the river under the spatial scale R are output.

2. The method according to claim 1, wherein, In step S1, the following steps are used for coarse extraction of the water body of the multi-branch river: Step S11, inputting a satellite remote sensing image and determining initial values of model parameters: inputting a satellite remote sensing image, determining a resolution Re and a size of the inputted satellite remote sensing image, and setting values of model parameters, the model parameters including: an improved normalized difference water index threshold MNDWI for distinguishing water bodies from non-water bodies in the remote sensing image T , a river width calculation parameter Aw, a maximum size R of a circular window for detecting a deformed river network at different scales max , a minimum size R min , and an increase step size AR thereof, a sinuosity threshold S for distinguishing normal river bends from deformed river bends T ; Step S12, calculation of the improved normalized difference water body index, and coarse extraction of the water body binary image from the satellite remote sensing image: the improved normalized difference water body index is calculated, and the water body binary image is coarsely extracted from the satellite remote sensing image based on the OSTU threshold segmentation method: According to the improved normalized difference water index formula: The MNDWI index of each pixel point of the input satellite remote sensing image is calculated, and Green and MIR respectively refer to the reflectivity of the green and mid-infrared bands in the satellite remote sensing image. Traverse each pixel point in the satellite remote sensing image, compare the MNDWI index of each pixel point with the size of the given threshold MNDWI in turn T If the MNDWI index is greater than the given threshold MNDWI T , it is determined that the pixel point is water body, otherwise it is non-water body, and finally the water body binary image of the input satellite remote sensing image is obtained.

3. The method according to claim 1, wherein, In step S2, the description of the network structure of the multi-branch river comprises the following steps: Step S21, extracting the skeleton lines of the river in the water body binary image by using a thinning algorithm, and recording the pixel point coordinates of all the skeleton lines: The water body with the largest connected area in the binary image of the water body is reserved, the water body is thinned into a group of continuous and single-resolution width skeleton lines of the river channel by using the Zhang-Suen thinning algorithm, and the coordinates of the pixel points belonging to the skeleton lines are recorded in a matrix The pixel points of the skeleton lines are sorted in the direction from the upstream to the downstream of the river i X i Y s i represents the i-th pixel point coordinate on the skeleton line, and n represents the total number of the pixel points of the skeleton line. Step S22, traversing the pixel points of the skeleton lines, determining the branch points of the river in the skeleton lines one by one, and recording the pixel point coordinates of the branch points: The neighbor pixel points of the pixel points of any bone line are traversed, whether each neighbor pixel point also belongs to the bone line is detected in turn, and the number n of neighbor pixel points that also belong to the bone line is counted nei ; According to the value of n nei , it is determined whether the pixel point of the skeleton line belongs to the branch point of the river: if n nei =1, the pixel point of the skeleton line is defined as the end point of the river, if n nei =2, the pixel point of the skeleton line is defined as a normal skeleton line pixel point, and if n nei >2, the pixel point of the skeleton line is defined as the branch point of the river. Step S23, record all pixel point coordinates judged as river end points and river branch points in matrix In the matrix, the river end points and the river branch points are sorted in order from upstream to downstream of the river, n n represents the total number of river end points or river branch points.

4. The method according to claim 3, wherein, In step S3, the batch calculation of the planar geometric indexes of each branch of the river comprises: Step S31, performing a tracking algorithm on the determined end points of the river or the branch points of the river, sequentially traversing the pixel points of the skeleton lines connected to the branch point or the end point from any branch point or end point, until the other branch point or end point: Define the length matrix Len = {l i,j} and the width matrix Wid = {w i,j} of the river, where i and j represent the serial number of the bifurcation point or the end point of the river, l i,j and w i,j represent the length and width of the single branch connected by the i-th bifurcation point or end point and the j-th bifurcation point or end point, respectively. Starting from the bifurcation point or end point i of any river, traverse each pixel point of the skeleton line on the branch connected with the bifurcation point until the other bifurcation point or end point j of the river. Step S32, during the tracking algorithm, the length and width of a single branch of the river are calculated and recorded in sequence based on the coordinate calculation of the pixel points of the skeleton line: When traversing the pixel point k of any skeleton line of the connected branch of the river endpoint or the bifurcation point i and j, first, the Euclidean distance between the pixel point k of the skeleton line and the pixel point k-1 of the previous skeleton line adjacent thereto is calculated based on coordinate operation: Re represents the resolution of the input satellite remote sensing image, and the tangent slope of the branch at the pixel point k is calculated In the formula, X k-1 , Y k-1 respectively represent the coordinates of the pixel point k-1 of the previous skeleton line and the pixel point k+1 of the next skeleton line directly connected with the pixel point k of the skeleton line. k+1 , Y k+1 respectively represent the coordinates of the pixel point k-1 of the previous skeleton line and the pixel point k+1 of the next skeleton line directly connected with the pixel point k of the skeleton line. The length of the branch connected by the river bifurcation point or endpoint i and j is equal to the cumulative value of the Euclidean distance of all adjacent pixel points on the skeleton line: In the formula, N K represents the total number of pixel points of the skeleton line contained on the branch. With the pixel point k of the skeleton line as the center, a vertical line is generated according to the local tangent slope of the pixel point k The vertical line is extended to the non-water body pixel points on the left and right banks to generate a single-resolution cross-section line perpendicular to the branch corresponding to the skeleton line. For the cross-section line generated on the pixel point of any skeleton line, when the length difference between the left and right cross-section lines is less than the river width calculation parameter Δw, the width of all single branches in the river is equal to the length of the cross-section line in value. Otherwise, the river width at the pixel point of the skeleton line is equal to twice the shorter one of the left and right parts of the cross-section line in value. wherein wherein, and respectively represent the left and right section line lengths at the pixel point k. Step S33, construct the length and width matrix of a single branch of a multi-branch river: After the length and width of the branches connected with the river branch point or end point i and j are calculated, the corresponding elements in the river length matrix Len = {l i,j} and the width matrix Wid = {w i,j} are updated, and then the remaining branches connected with the river branch point or end point i are sequentially processed in a clockwise direction, and step S32 is repeated until all branches connected with the river branch point or end point i are processed, and the length and width of all single branches in the river are calculated and recorded. The pixel point coordinate matrix of the river end point or the bifurcation point of the river determined in step S23 The matrix is traversed All the river end points or the bifurcation points of the river in the matrix are repeated with step S32 and step S33, the length and width information of all the single branches of the river are finally matrixed, and the length matrix Len={l i,j} and the width matrix Wid={w i,j} of the single branches of the multi-branch river are constructed.

5. The method according to claim 1, wherein, In step S4, the identification of the main flow path of a multi-branch river includes the following steps: Step S41, construct the feature path matrix Pat and the length matrix Dis of the multi-branch river by taking the length-to-width ratio of the branch as the weight: d = length of the river course between the river end or bifurcation point pair i and j i,j d = length of the river course between the river end or bifurcation point pair i and j define a characteristic path matrix Pat = {p i,j} of the multi-branch river and a length matrix Dis = {d i,j} of the multi-branch river, the initial value of element p i,j is p i,j = j, and the initial value of d i,j is equal to the length of the characteristic path between the river endpoints or the branching points of the river for i and j; Step S42, iterate through each river endpoint or river bifurcation point pair in the multi-branch river, calculate the weighted feature path length between each river endpoint or river bifurcation point pair, and iterate the results of the feature path matrix Pat and the length matrix Dis of the multi-branch river in step S41: The sum d of the characteristic path lengths from the river endpoint or river bifurcation point i to each river endpoint or river bifurcation point j is calculated in turn for all river endpoints or river bifurcation points i and j in the river i,k,j : d i,k,j = d i,k + d k,j , wherein d i,k represents the characteristic path length from river endpoint or river branching point i to river endpoint or river branching point k, d k,j represents the characteristic path length from river endpoint or river branching point k to river endpoint or river branching point j, d i,k,j <d i,j , the value of the element d i,j in the characteristic path length matrix Dis = {d i,j} of the multi-branch river is iterated to d i,k,j , and the value of the element p i,j in the characteristic path matrix Pat = {p i,j} is iterated to the value of p i,k . After the final multi-branch river characteristic path matrix Pat = {p i,j} is obtained through iteration, the main flow path of the river is identified, i.e. the flow path starts from the uppermost branch point and always connects to the branch with larger flow along the river downstream until the lowermost branch point. Step S43, based on the final results of the feature path matrix Pat and the length matrix Dis of the multi-branch river, identify the weighted feature path from the uppermost bifurcation point to the lowermost bifurcation point of the multi-branch river as the main flow path of the river, and record the pixel point coordinates in sequence, including: For any river endpoint or river bifurcation point i passing through the main flow path of the river, find the last column element of the i-th row in the characteristic path matrix Pat The value is the serial number of the downstream river endpoint or river bifurcation point connected directly to the river endpoint or river bifurcation point i passing through the main flow path, and the pixel point coordinates of the river skeleton line connected by the above two bifurcation points are recorded in the order from upstream to downstream. Starting from the first bifurcation point in the main river channel, and continuing down to the downstream endpoint or bifurcation point, record the serial numbers of the river endpoints or bifurcation points traversed by the main channel in a point set. In, where m i This represents the i-th branch point in the main flow path, n. m This represents the total number of branch points included in the main flow path. Simultaneously, the coordinates of the skeletal lines along the main flow path are recorded sequentially from upstream to downstream in the matrix Cha_Path = {X...} i ,Y i } nT X i ,Y i This represents the coordinates of the i-th pixel on the skeletal line corresponding to the main flow path, where n is the coordinate of the i-th pixel. T This represents the total number of pixels contained in the skeleton line corresponding to the main flow path.

6. The method according to claim 1, wherein, In step S5, the determination and output of the abnormal meander of the multi-branch river under the spatial scale include: Step S51, for any pixel point on the main flow path of the river, use a circular detection window with a spatial scale of R to calculate the curvature of the main flow path within the window, and use the following method: A pixel point i on the skeleton line on the main flow path of any river is taken as the center of a circle, a circular detection window with a variable radius and a space scale R is set, and the window can move downstream along the main flow path of the river. The main flow path curvature value SI contained in the window is calculated R : wherein represents the length of the bend of the main flow path of the river within the circular detection window, which is equal in value to the cumulative value of the pixel distances of all adjacent skeleton lines on the main flow path in the window: n j represents the total number of pixel points of the skeleton line on the main flow path of the river within the circular detection window, X j , Y j represents the coordinate of the jth pixel point on the skeleton line, X j+1 , Y j+1 represents the coordinate of the j+1th pixel point on the skeleton line; represents the length of the straight line representing the main flow path of the river within the circular detection window, which is equal in value to the straight-line distance between the pixel point of the most upstream skeleton line and the pixel point of the most downstream skeleton line on the main flow path in the window: X1, Y1 represents the pixel point coordinate of the most upstream skeleton line on the main flow path, X2, Y2 represents the pixel point coordinate of the most downstream skeleton line on the main flow path; with the computed main flow path tortuosity value SI R As the river tortuosity value at pixel point i of the skeleton line under spatial scale R, the larger the value, the more likely it is to form a deformed meander at this point. The resulting results are recorded in the river sinuosity matrix n T In the formula, R max and R min respectively represent the maximum and minimum scales of the given circular detection window, and ΔR is the increasing step of the circular detection window. Step S52, determine whether the curvature value of the main flow path obtained in step S51 is greater than the given abnormal meander curvature threshold, if it is satisfied, it is considered that an abnormal meander has been formed at the pixel point: The spatial scale R of the circular detection window is gradually increased from R min in steps of ΔR until R max Step S51 is repeated to obtain the calculation results of the sinuosity of the river along the main flow path from the upstream to the downstream at different spatial scales, and a river sinuosity matrix SI is output. Each element in the river sinuosity matrix SI is compared with the sinuosity threshold S T for detecting a meandering bay in sequence. If S T , it is determined that a meandering bay at the spatial scale R exists at the pixel point on the skeleton line. Step S53, iterate through all pixel points on the skeleton line of the main flow path of the river, repeat steps S51 and S52, and output the corresponding position and curvature of the abnormal meander of the river under the spatial scale R.

Images