Method for measuring underwater three-dimensional topography of sandy bed surface

Abstract

The method for measuring the underwater three-dimensional terrain of the sandy bed surface provided by the present disclosure comprises the following steps: building an image acquisition platform for obtaining an underwater image sequence of the sandy bed surface; arranging water and underwater control points and measuring their coordinates; determining the height, arrangement form and moving space step of the camera array; making all the cameras take pictures of the target water surface area at the same time based on the parameters determined in the previous step, and obtaining the underwater image sequence of the sandy bed surface; performing spatial calculation on the image sequence based on the motion recovery structure method, performing spatial registration only using the water control points, and then performing refraction correction, and then performing spatial registration only using the underwater control points without performing refraction correction, and then performing secondary correction on the point cloud obtained by using only the water control points and performing refraction correction through the average value of the difference between the two, and obtaining the final underwater three-dimensional terrain of the sandy bed surface. The time and space resolution and accuracy of the present disclosure are relatively high, and the observation object is not disturbed.

Description

Technical Field

[0001] This disclosure pertains to the fields of environmental and water conservancy engineering, river and lake geomorphology, waterway and nearshore engineering surveying, and particularly relates to a method for measuring the underwater three-dimensional topography of sandy beds. Background Technology

[0002] The alluvial riverbeds in the middle and lower reaches of large and medium-sized rivers are generally sandy, and sandy riverbeds are also widely distributed in nature along coastlines and lakeshores. The continuous topographic variations and sediment transport characteristics of sandy riverbeds are fundamental to studying key scientific issues such as the evolution of river and lake geomorphology, hydraulics, and ecological characteristics, as well as the migration and transformation of water pollution. They also serve as crucial basic data for engineering projects such as river management and flood control, coastal restoration and navigation, soil and water conservation, and ecological restoration. Sandy riverbeds develop various morphologies such as sand waves and sand dunes under different flow conditions, thus requiring three-dimensional topographic data for topographic surveying. However, due to the fine particle size of sandy riverbeds, sediment transport is easily facilitated, making three-dimensional topographic surveying quite challenging. On the one hand, invasive measurement methods (such as traditional needle surveying or RTK GPS measurement) can easily cause local topographic changes, introducing measurement errors. The same problem arises in laboratory measurements conducted without water, as the change in water flow alters the sandy bed surface topography, leading to discrepancies between the measured dry topography and the actual water-flowing topography. On the other hand, sandy bed topography changes rapidly under strong water flow conditions, and measurement methods with low temporal resolution cannot obtain detailed topographic data. For example, underwater laser scanning methods may continuously change the sandy bed morphology during the scanning process. Therefore, the topography of sandy beds should ideally be measured directly under water conditions, requiring a high degree of spatiotemporal resolution in the measurement methods.

[0003] Acoustic methods, based on the propagation and reflection characteristics of sound waves in water, are suitable for topographic surveying in deep water environments. However, regardless of whether it's single-point or multi-point measurement (such as multibeam bathymetry), the spatial resolution is low, and the large size of the equipment necessitates vessel transport, making it unsuitable for shallow water environments. Laboratory water depths are generally even smaller, thus significantly limiting the application of acoustic methods. Furthermore, acoustic measurement is a scanning method with limited temporal resolution. Photogrammetry, a widely used topographic surveying method based on computer vision technology, offers advantages such as low cost, speed, accuracy, and high spatial resolution. It can directly create 3D models of underwater topography in transparent water conditions. However, the refraction of light at the water surface leads to an overestimation of the measured elevation. Generally, photogrammetry relies on a single camera capturing images at different locations and reconstructing the 3D topography using Structure from Motion (SfM) methods. This process requires a long total time, resulting in low overall temporal resolution and making it unsuitable for direct application to sandy bed measurements. Moreover, photogrammetry cannot directly capture the fine sand transport characteristics of sandy beds. However, with proper modifications, the SfM photogrammetry method could significantly improve temporal resolution and resolve the sand transport zone on the sandy bed surface.

[0004] In response to the aforementioned measurement needs for underwater three-dimensional topography and sediment transport on sandy beds, and the problems existing in current technologies, it is necessary to develop a specialized photogrammetric method for measuring underwater three-dimensional topography on sandy beds. Summary of the Invention

[0005] This disclosure aims to address at least one of the technical problems existing in the prior art.

[0006] Therefore, the present disclosure provides a method for measuring the underwater three-dimensional topography of sandy beds with high spatial resolution and accuracy, comprising:

[0007] (1) Construct an image acquisition platform for acquiring underwater image sequences of sandy bed surface. The image acquisition platform includes a camera array consisting of J cameras arranged in the form of n×m, where n is the number of rows of the camera array, n≥1, m is the number of columns of the camera array, m≥3, and each row of cameras is required to capture the maximum width of the target water surface area. The rows of the camera array are defined as perpendicular to the water flow direction, and the columns are defined as in the direction of water flow.

[0008] (2) Arrange and measure control points, and obtain the spatial coordinates of the control points.

[0009] The control points include surface control points set on both sides of the target water surface area and underwater control points set on the bed surface of the target water surface area. Each control point includes a flat plate with a regular geometric pattern drawn on its upper surface and a horizontal slope. The geometric pattern has a clear center point position, which is defined as the coordinate of the control point. It is required that multiple control points can be captured in each image taken by a single camera.

[0010] The coordinates of each control point are measured using a measuring device. The position of the measuring device is defined as the origin. The horizontal direction with the water flow or the length direction along the target water surface area is defined as the X-axis direction, the horizontal direction perpendicular to the X-axis is defined as the Y-axis direction, and the direction perpendicular to the XY plane is defined as the Z-axis direction.

[0011] (3) Determine the height of the camera array above the bed surface, its arrangement, and the step size of its movement.

[0012] (3-1) Determine the height and arrangement of the camera array above the bed surface.

[0013] After confirming the number of columns m and focal length f of the camera array, the height of the camera array and the lateral spacing L of each camera in each row along the Y-axis can be initially determined using the following method. c :

[0014] Arrange the image sensors of all cameras along the Y-axis in length and the X-axis in width. Let H be the height of the camera array above the bed surface. Calculate the actual distance L corresponding to the length direction of each camera image sensor in the camera array using the following formula:

[0015]

[0016]

[0017] Where h is the image distance, l c The length of the camera's image sensor;

[0018] The overlap of individual images captured by two adjacent cameras in the Y-axis direction should be no less than 0. Y Then the lateral spacing L of the camera along the Y-axis c satisfy:

[0019] L c ≤(1-O Y )L

[0020] Choose the camera horizontal spacing L that satisfies the above formula. c Based on the calculation results, the following two conditions are used for verification. If both conditions cannot be met simultaneously, it is necessary to adjust the height H of the camera array above the bed surface and the lateral spacing L of the cameras. c Recalculate and verify until both conditions are met; the two conditions are:

[0021] Condition 1: If the total width of the water surface in the target area is required to be covered by the total shooting range of all cameras in each row except for the two outermost cameras, then:

[0022] (m-3)·L c +L≥(W w ) max

[0023] Among them, (W) w ) max It is the maximum width of the water surface in the target area;

[0024] Condition 2: The control points are clearly visible in every image. If the side length of a single control point pattern corresponds to at least T pixels in a single image, then:

[0025]

[0026] Where Res is the spatial resolution of a single image, and L GCP The actual side length of the control point pattern;

[0027] When n>1, the longitudinal spacing L of each row of cameras along the X-axis is determined according to the following formula. L :

[0028]

[0029] Among them, w c The width of the camera's image sensor;

[0030] (3-2) Determine the longitudinal movement spatial step size of the camera array

[0031] If the actual longitudinal distance S covered by the camera array is greater than or equal to the maximum longitudinal distance X of the target water surface area. max -X min X max and X min Let X and Y be the maximum and minimum coordinates of the target water surface area along the X-axis, respectively. Then, the longitudinal movement step size of the camera array is equal to 0. If the actual longitudinal distance S covered by the camera array is less than the maximum longitudinal distance X of the target water surface area... max -X min Then, the longitudinal spatial step size ΔL of the camera array is set according to the following formula:

[0032]

[0033] S=[n(1-O X )+O X W

[0034] ΔL=S-(1-OX W

[0035] Where W is the actual spatial dimension corresponding to the width direction of a single image, and the overlap of a single image in the X-axis direction should not be less than 0. X ;

[0036] (4) Based on the parameters determined in step (3), the target water surface area is photographed to obtain an underwater image sequence of the sandy bed surface.

[0037] First, move the camera array center to its initial position. When the camera array's shooting range along the X-axis cannot completely cover the maximum length of the target water surface area, the initial position of the camera array center is located at the target water surface area X. min At the center of the cross section of the location, when the shooting range of the camera array along the X-axis can fully cover the maximum length of the target water surface area, the initial position of the center of the camera array is located directly above the center of the target water surface area and the entire water surface boundary of the target area can be seen in the camera's field of view.

[0038] Then, the camera array is moved from its initial position along the positive X-axis towards the other end of the target water surface area in a longitudinal spatial step of ΔL. After each longitudinal spatial step, all cameras in the camera array simultaneously capture the first image. Then, the array moves another longitudinal spatial step and stops capturing images, until it reaches the X-axis of the target water surface area. max Position; the first image captured by the j-th camera at the i-th position where the camera array stops shooting is denoted as (P1). ij When S≥X max -X min When, let i = 1, when S <X max -X min When i∈[1,p], j∈[1,J], J=n×m, The rounding function is derived from all images (P1). ij The underwater image sequence of the sandy bed surface includes the first image sequence P1, constituting the first image sequence P1.

[0039] (5) Obtain the underwater three-dimensional topography of the sandy bed surface based on the first image sequence P1.

[0040] (5-1) The underwater terrain modeling of the first image sequence P1 is performed based on the motion reconstruction structure method to obtain a three-dimensional point cloud. Spatial data registration is performed using control points. The initial encrypted three-dimensional point cloud and its corresponding digital orthophoto are denoted as C1 and DOM1, respectively, and the initial encrypted three-dimensional point cloud and its corresponding digital orthophoto are denoted as C1' and DOM1', respectively.

[0041] (5-2) Generate a water surface model WS1 containing the target water surface region based on the digital orthophoto DOM1, and rasterize the initial encrypted 3D point cloud C1 to obtain the rasterized initial 3D point cloud C1. g Rasterize the initial 3D point cloud C1 g The portion of the model below the water surface WS1 was selected to obtain the raster point cloud C1 located below the water surface. u For the grid point cloud C1 u Refraction correction is performed to obtain the refraction-corrected grid point cloud C1. c ;

[0042] (5-3) Generate a water surface model WS1' containing the target water surface area based on the digital orthophoto DOM1'. Apply the same rasterization method as the initial encrypted 3D point cloud C1' to the initial encrypted 3D point cloud C1' to obtain a rasterized initial 3D point cloud C1' using only underwater control points. g Rasterize the initial 3D point cloud C1' g The portion of the model below the water surface WS1' is selected to obtain the raster point cloud C1' located below the water surface. u Calculate the raster point cloud C1' u and the refraction-corrected grid point cloud C1 c The difference in average bed elevation ΔZ is used to adjust the refraction-corrected grid point cloud C1. c The bed elevation of each grid point is corrected to obtain the final underwater 3D grid point cloud C1. f This refers to the underwater three-dimensional topography of the target sandy bed surface.

[0043] In some embodiments, the image acquisition platform built in step (1) further includes:

[0044] The longitudinal module includes longitudinal slide rails placed on both sides of the water surface in the target area along the direction of water flow and longitudinal sliders disposed on the longitudinal slide rails and movable thereal;

[0045] A horizontal module includes a horizontal slide rail placed horizontally along the direction perpendicular to the water flow and a horizontal slider disposed on the horizontal slide rail and movable thereal. The end of the horizontal slide rail is connected to the vertical slider. The position of the horizontal slide rail in the horizontal direction along the water flow direction is changed by the vertical module.

[0046] The vertical module includes a vertical slide rail connected to and moving with the horizontal slider, a vertical slider disposed on the vertical slide rail and movable along it, and an L-shaped bracket connected to and moving with the vertical slider. The vertical section of the L-shaped bracket is parallel to the axis of the vertical slide rail, and the horizontal section of the L-shaped bracket is parallel to the axis of the longitudinal slide rail. The vertical module is used to adjust the height of the camera array from the bed surface.

[0047] There are n horizontal bars, each of which is connected to the horizontal section of the L-shaped bracket via a horizontal bar slider that can move along the horizontal section of the L-shaped bracket. The axial direction of all horizontal bars is parallel to the axial direction of the horizontal slide rail, and the height of all horizontal bars is consistent. Each horizontal bar is equipped with m gimbals for fixing the camera, and the angle and position of the camera are adjusted by the gimbals.

[0048] A synchronizer is used to send a single trigger signal to all cameras in a camera array so that all cameras can capture images simultaneously.

[0049] The control module is used to control the movement mode, movement speed and target position of the longitudinal module, the transverse module and the vertical module, control the synchronizer, and control the shooting parameters of all cameras and receive images captured by all cameras;

[0050] A power supply is used to power the various electronic devices in the image acquisition platform.

[0051] In some embodiments, the longitudinal module and the transverse module are respectively provided with a locator for determining the position of their respective sliders, and each locator uses a point laser rangefinder fixed to one end of its respective slide rail to realize slider distance measurement and positioning.

[0052] In some embodiments, the image acquisition platform further includes a light source system disposed on one or both sides of the crossbar.

[0053] In some embodiments, in step (2), the measuring device is an RTK GPS, a total station, or a laser scanner.

[0054] In some embodiments, step (5-1) specifically includes the following steps:

[0055] Spatial calculations are performed on the first image sequence P1 based on the structure-of-motion method to reconstruct the camera spatial position and lens orientation corresponding to each image. A sparse three-dimensional point cloud of the target water surface area is obtained through feature point matching. The center position of each control point appearing in the first image sequence P1 is identified and marked by manual or automatic methods to obtain the image coordinates of the control points. The spatial data registration of the sparse three-dimensional point cloud is performed using the correspondence between the real coordinates and image coordinates of the control points. Based on the registered three-dimensional point cloud, an initial encrypted three-dimensional point cloud is obtained using an encryption method, and the corresponding digital orthophoto is generated.

[0056] In some embodiments, step (5-2) specifically includes the following steps:

[0057] In the digital orthophoto DOM1, single-point calibration is used to identify water surface boundary points, obtaining a series of planar spatial coordinates for these points. These water surface boundary points are then projected onto the initial refined 3D point cloud C1, obtaining a series of 3D spatial coordinates for the water surface boundary points. Based on these water surface boundary points, a Delaunay triangulation mesh is generated as the water surface model WS1. The initial refined 3D point cloud C1 is then rasterized, with the bed elevation of the raster taken as the maximum elevation value of all point cloud points contained within the raster area, resulting in the rasterized initial 3D point cloud C1. g ; Calculate the initial rasterized 3D point cloud C1 g The distance from each grid point to the water surface model WS1 is used as the initial water depth d for each grid point. a ; Delete the rasterized initial 3D point cloud C1 g d a For grid points with values ​​less than 0, the resulting grid point cloud C1 is located below the water surface model WS1. u For each image corresponding to a camera position, the actual spatial range of the image captured at that camera position is calculated based on the camera orientation angle calculated using the structure-of-motion method, as well as the camera's focal length and sensor size. The grid point cloud C1 falling within this actual spatial range is then calculated. u The grid points in the image are considered visible at the camera's location;

[0058] For the selected raster point cloud C1 u Each grid point in the data is corrected for its initial water depth according to the following formula, resulting in the grid point cloud C1. u After correcting all grid points, the refraction-corrected grid point cloud C1 is obtained. c :

[0059]

[0060] d a =Z w -Z a

[0061]

[0062]

[0063]

[0064] Where, r ij X is the angle between the camera's optical axis and the water surface at that camera position. cij Y cij These are the X-axis and Y-axis coordinates of the camera position, respectively. wij X is the vertical distance between the camera position and the water surface model; a X a Za Let d be the X-axis, Y-axis, and Z-axis coordinates of a grid point located in the underwater region. a Let d be the initial water depth at a given grid point. ij This is the corrected water depth for a specific grid point calculated based on the camera position ij. The average corrected water depth is calculated for the camera positions corresponding to all images of a given grid point within the image range. p and q represent the number of camera array positions and camera numbers that stopped shooting, respectively, for all camera positions visible at that grid point. Z w Z represents the elevation of the water surface model corresponding to the horizontal position of a certain grid point. c is the corrected bed surface elevation for a given grid point; n1 and n2 are the refractive indices of water and air, respectively.

[0065] In some embodiments, in steps (5-2) and (5-3),

[0066] When the initial 3D point cloud C1 is rasterized g or C1' g If, after deleting grid points with a water depth less than 0, the resulting grid point cloud located below the water surface model WS1 or WS1' still contains some grid points actually located above the water surface model WS1 or WS1', then it is necessary to generate a spatial fitting plane FWS1 or FWS1' for the water surface boundary points and calculate the initial rasterized 3D point cloud C1. g or C1' g The distance to the spatial fitting plane FWS1 or FWS1' is used to perform secondary screening of the three-dimensional points on the water surface. Different screening thresholds are tested until all grid points located above the water surface model WS1 or WS1' are removed, thus finally obtaining the grid point cloud located below the water surface model WS1 or WS1'.

[0067] In some embodiments, the measurement method further includes obtaining the corresponding underwater three-dimensional topography of the target sandy bed surface by taking the first image sequence P1 obtained in step (4) at different times according to steps (5-1) to (5-3), thereby obtaining a sequence of the three-dimensional topography of the target sandy bed surface changing over time.

[0068] In some embodiments, the measurement method further includes identifying sand transport zones in the target sandy bed, with the following specific steps:

[0069] S1, Obtain the second image sequence P2

[0070] In step (4), let the distance be closer to Y. min The camera position is set as the first camera, close to Y. max The position setting is for the m-th column of cameras, Y max and ymin Let X and Y represent the maximum and minimum coordinates of the target water surface area along the Y-axis, respectively. After all cameras in the camera array simultaneously capture the first image at position i, the position of the camera array remains unchanged, and the cameras in the remaining columns (excluding the first column) capture the second image sequentially at equal time intervals ΔT. After all cameras in the camera array have captured the second image at position i, the camera array moves one vertical spatial step ΔL to position i+1, and the above shooting method is repeated until the camera array moves to X. max The camera array stops shooting at position i, and the second image obtained by the j-th camera at that position is denoted as (P2). ij (from all images P2) ij This constitutes the second image sequence P2;

[0071] S2. Based on the second image sequence P2 and the target area sandy bed surface obtained in step (5), underwater three-dimensional topography identification of the sediment transport area.

[0072] Following step (5-1), underwater terrain modeling is performed on the second image sequence P2 to obtain a 3D point cloud. Spatial data registration is performed using surface control points to obtain the initial encrypted 3D point cloud C2 and its corresponding digital orthophoto DOM2. Following step (5-2), refraction correction is performed on the initial encrypted 3D point cloud C2 to obtain the refraction-corrected raster point cloud, which is used as the final underwater 3D raster point cloud C2 obtained based on the second image sequence P2. f ;C1 f With C2 f Subtracting the values ​​yields the raster point cloud ΔC, where each raster point is assigned the value C1. f With C2 f The elevation difference of the grid points on the bed surface is used to identify grid points in the grid point cloud ΔC whose grid point values ​​exceed the error threshold as sand transport areas in the target sandy bed surface.

[0073] The beneficial effects of this disclosure are as follows:

[0074] 1. The underwater 3D topography of sandy beds exhibits high temporal and spatial resolution and accuracy. Due to the use of multiple cameras for simultaneous imaging, continuous image sequences can be obtained at a high frequency for underwater topographic measurement, resulting in high temporal resolution. The fine sand transport and resulting topographic changes between images within a synchronously acquired image sequence are negligible, ensuring the accuracy of topographic modeling, reaching up to the millimeter level. This method achieves high spatial resolution based on visible light and images, and precise control of the spatial resolution is achieved through fine-grained control of the image acquisition platform (such as height, camera spacing, and angle), reaching a spatial resolution up to the sub-millimeter level.

[0075] 2. It can accurately determine the fine sand transport area. By comparing the modeling results of synchronously captured image sequences and image sequences captured at certain intervals, it can accurately quantify and determine the rapidly moving and inconspicuous fine sand transport area. The method is simple and fast.

[0076] 3. It can measure underwater three-dimensional topography directly without stopping the water supply, so the test process does not require stopping the water supply for measurement, and field measurements do not require dredging the river channel.

[0077] 4. The positions of each camera in the camera array are easy to adjust.

[0078] 5. It is a non-invasive measurement method that will not interfere with the object being observed. Attached Figure Description

[0079] Figure 1 This is a flowchart of a method for measuring the underwater three-dimensional topography of a sandy bed provided in an embodiment of this disclosure.

[0080] Figure 2 This is a schematic diagram of the structure of the image acquisition platform built in the measurement method provided in this embodiment.

[0081] Figure 3 (a) to (d) are Figure 2 The diagram shows the structure of the water control point in the image acquisition platform.

[0082] Figure 4 (a) to (c) are Figure 2 The diagram shows the structure of the underwater control point in the image acquisition platform.

[0083] Figure 5 The underwater three-dimensional topography and sediment transport area of ​​the sandy bed surface obtained by the measurement method provided in the embodiments of this disclosure are shown in (a) underwater three-dimensional topography of the sandy bed surface and (b) sediment transport area of ​​the sandy bed surface.

[0084] In the diagram, 1 is the camera array, 1-1 is the camera, 1-2 is the altitude positioner, 1-3 is the crossbar, 1-4 is the crossbar slider, 2-1 is the longitudinal slide rail, 2-2 is the longitudinal slider, 3-1 is the transverse slide rail, 3-2 is the transverse slider, 4-1 is the vertical slide rail, 4-2 is the vertical slider, 4-3 is the L-shaped bracket, 5 is the synchronizer, 6 is the control module, 7 is the power supply, 8-1 is the surface control point, 8-2 is the underwater control point, 8-a is the frame, 8-b is the flat plate, 8-c is the ground stake, 8-d is the pivot, 8-e is the support rod, and 9 is the light source system. Detailed Implementation

[0085] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining this application and are not intended to limit this application.

[0086] Conversely, this application covers any alternatives, modifications, equivalent methods, and schemes made within the spirit and scope of this application as defined by the claims. Furthermore, to provide the public with a better understanding of this application, certain specific details are described in detail below. However, this application can be fully understood by those skilled in the art even without these detailed descriptions.

[0087] In the description of this disclosure, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing this disclosure and simplifying the description, and do not indicate or imply that the foundation or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this disclosure, "a plurality of" means two or more, unless otherwise explicitly specified.

[0088] In the description of this disclosure, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure according to the specific circumstances.

[0089] In this disclosure, unless otherwise expressly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0090] See Figure 1 The method for measuring the underwater three-dimensional topography of a sandy bed provided in the first aspect of this disclosure includes:

[0091] (1) Building an image acquisition platform

[0092] See Figure 2 The image acquisition platform that was built includes:

[0093] Camera array 1 consists of J cameras 1-1 with fixed focal lengths that support remote control, arranged in an n×m pattern. It is used to capture underwater image sequences of the sandy bed surface. n is the number of rows in camera array 1, n≥1, and m is the number of columns in camera array 1, corresponding to the number of cameras 1-1 in each row, m≥3. Each row of cameras is required to completely cover the maximum width of the target water surface area. The rows of the camera array are perpendicular to the water flow direction, and the columns are in the same direction as the water flow. All cameras 1-1 in camera array 1 have the same focal length, and all cameras 1-1 support remote control shooting. Preferably, cameras that support remote control of focusing, setting shooting parameters such as aperture, shutter speed, and ISO, as well as data transmission are used.

[0094] The longitudinal module includes longitudinal slide rails 2-1 placed on both sides of the target area along the direction of water flow (one longitudinal slide rail 2-1 is set on each side), and longitudinal sliders 2-2 set on the longitudinal slide rails 2-1 and movable along them. The longitudinal module is used to adjust the position of the camera array 1 as a whole along the direction of water flow.

[0095] The horizontal module includes a horizontal slide rail 3-1 that is generally placed horizontally along the direction of water flow, and a horizontal slider 3-2 that is set on the horizontal slide rail 3-1 and can move along it. The end of the horizontal slide rail 3-1 is connected to the vertical slider 2-2. The horizontal module changes the horizontal position of the horizontal slide rail 3-1 along the direction of water flow. The horizontal module is used to adjust the position of the entire camera array 1 along the direction of water flow.

[0096] The vertical module includes a vertical slide rail 4-1 connected to and moving with the horizontal slider 3-2, a vertical slider 4-2 mounted on the vertical slide rail 4-1 and movable along it, and an L-shaped bracket 4-3 connected to and moving with the vertical slider 4-2. The vertical section of the L-shaped bracket 4-3 is parallel to the axis of the vertical slide rail 4-1, and the horizontal section of the L-shaped bracket 4-3 is parallel to the axis of the longitudinal slide rail 2-1. The vertical module is used to adjust the overall height of the camera array 1 from the bed surface.

[0097] There are n horizontal bars 1-3. Each horizontal bar 1-3 is connected to the horizontal section of the L-shaped bracket 4-3 via a horizontal bar slider 1-4 that can move along the horizontal section of the L-shaped bracket 4-3. The axial direction of all horizontal bars 1-3 is parallel to the axial direction of the horizontal slide rail 3-1, and the height of all horizontal bars 1-3 is consistent. Each horizontal bar 1-3 is equipped with m gimbals for fixing the camera 1-1. The angle of the camera 1-1 can be adjusted through the gimbals. The distance between two adjacent horizontal bars 1-3 and the distance between two adjacent gimbals on the same horizontal bar 1-3 can be adjusted.

[0098] The control module 6 is used to control the movement mode (including stepping and constant speed) of the longitudinal module, the transverse module and the vertical module, as well as the movement speed and target position. It can also remotely control the shooting parameters of all cameras 1-1 and receive the underwater image sequence captured by all cameras 1-1.

[0099] Power supply 7 is used to power the various electronic devices in this image acquisition platform;

[0100] The control points are divided into surface control points 8-1 and underwater control points 8-2. Surface control points 8-1 are fixedly positioned on both sides of the target water surface area, while underwater control points 8-2 are fixedly positioned on the bed surface (such as a riverbed) of the target water surface area, near the water's edge. Both types of control points consist of a frame 8-a and a plate 8-b connected to the frame 8-a. The upper surface of plate 8-b is decorated with a pattern composed of white and dark colors. The pattern is generally a regular geometric shape with a clearly defined center point, which is defined as the coordinate of the control point. The side length of the plate at each control point is related to the height of camera array 1, ensuring that the center point of the control point can be clearly identified in the image from camera 1-1. The main body of the control points is installed on both sides of the target area or on the bed surface using ground stakes 8-c. The difference between the two types of control points is that the frame and plate of surface control point 8-1 are rotatably connected to facilitate adjustment of the slope of the plate in surface control point 8-1, ensuring that the slope of the plate in surface control point 8-1 remains horizontal. See [link to relevant documentation]. Figure 3(a) shows the form of the surface control point 8-1 when placed horizontally; (b) shows the form of the underwater control point 8-2 when placed on a slope; (c) is a bottom view of the surface control point 8-1; and (d) shows the ground stake 8-c of the surface control point 8-1, which is fixed to the outer frame of the surface control point 8-1 by a screw section. The frame of the underwater control point 8-2 is fixedly connected to the plate. To ensure that the underwater control point 8-2 does not significantly obstruct water flow and is not easily covered by sediment transport, the plate of the underwater control point 8-2 is 2mm to 3mm higher than the bed surface. See [reference needed]. Figure 4 (a) shows the shape of underwater control point 8-2 when its placement position is consistent with the slope of the bed surface, (b) is a bottom view of underwater control point 8-2, and (c) is the ground nail 8-c of underwater control point 8-2, which is fixed to the outer frame of underwater control point 8-2 by a screw section.

[0101] Furthermore, since the particles on the fine sand bed are easily agitated, it is necessary to use an array of multiple cameras to take pictures simultaneously in order to obtain high temporal resolution measurement results. The image resolution of camera 1-1 should be as high as possible, as the image resolution directly determines the spatial resolution of the measurement results.

[0102] Furthermore, when the camera array consists of n=1 rows of cameras, the number of cameras m ≥ 5. The lenses of all cameras 1-1 can be set to face the same direction, perpendicular to the water surface. Alternatively, the lenses of the outermost cameras on both sides can be tilted 5°–10° towards the central camera in the camera array 1, with the lenses of the remaining cameras facing perpendicular to the water surface. The length direction of all camera sensors is parallel to the axis of the horizontal bar 1-3 where the single row of cameras is located, and the width direction of all camera sensors is perpendicular to the axis of the horizontal bar 1-3 where the single row of cameras is located.

[0103] Furthermore, when camera array 1 consists of n>1, i.e., multiple rows of cameras, the number of cameras in each row remains consistent, which is m. The arrangement of cameras in each row (camera spacing and orientation) is exactly the same, and the horizontal bars 1-3 of each row of cameras remain parallel and located on the same plane. The horizontal bars 1-3 are installed on the horizontal section of the L-shaped bracket 4-3 of the vertical module via horizontal bar sliders 1-4. The spacing between adjacent horizontal bars 1-3 is adjusted by changing the position of the horizontal bar sliders 1-4 on the L-shaped bracket 4-3.

[0104] Furthermore, to ensure the stability of the camera array 1, multiple vertical modules can be installed, such as... Figure 2 The image acquisition platform of the embodiment shown has two vertical modules. During the measurement process, the distance between the two vertical modules remains fixed, that is, the two vertical modules move synchronously along the transverse slide rail 3-1.

[0105] Furthermore, the camera array is equipped with a height locator 1-2, which is a point laser rangefinder, fixed at the center of any horizontal bar of the camera array, facing perpendicular to the plane where the camera array is located or facing the same direction as the camera lens at the center position (when there is a single row of cameras), which can remotely control and measure the vertical distance of the camera array from the water surface or the bed surface.

[0106] Furthermore, the longitudinal module and the transverse module are each equipped with a positioner for determining the position of their respective sliders. Figure 2 (The individual locators are not shown in the diagram.) Each locator uses a point laser rangefinder fixed to one end of its respective slide rail to achieve slider distance measurement and positioning functions.

[0107] Furthermore, both types of control points are solid rectangular (mostly square) plastic plates with a thickness of 5mm-10mm. The side length is determined by the height of camera array 1 from the bed surface, and the principle is that the center point of the control point can be clearly identified in the image of camera 1-1. The plate 8-b of the water control point 8-1 is connected to one side of the frame 8-a via a pivot 8-d, and the plate 8-b of the water control point 8-1 is connected to the other side of the frame 8-a via a support rod 8-e. The frame 8-a of the water control point 8-1 has screw holes for installing ground stakes 8-c. The water control point 8-1 is generally installed on the shore, and the ground stakes 8-c are embedded into the bed surface to fix the water control point 8-1. Then, the slope of the plate 8-b is adjusted to a basic level using the pivot 8-d, thus facilitating clear shooting from the top by camera 1-1. In the absence of frame processing conditions, other methods can be used to set up a base to support the plastic plate (such as a cement base) to make it basically level. The plastic plate of underwater control point 8-2 is the same size as that of surface control point 8-1, but it is directly connected to the frame and cannot be rotated. The ground stake is also connected to the screw hole on the back of the frame.

[0108] Furthermore, the image acquisition platform in this embodiment also includes a light source system 9, which can provide supplementary lighting. When natural light cannot guarantee the camera shutter speed, the light source system 9 can be activated. The light source system 9 consists of one or more strip LED lights. To minimize reflections from the water surface, a diffuser is arranged on the outside of each strip LED light. The strip LED lights are fixed between adjacent horizontal bars and placed parallel to each row of cameras.

[0109] Furthermore, the image acquisition platform in this embodiment also includes a synchronizer 5, used to synchronize all cameras in the camera array 1, so that all cameras can take pictures simultaneously using a single trigger signal. The synchronizer 5 is controlled by the control module 6. The control module transmits image acquisition commands to the synchronizer 5, and the synchronizer 5 then distributes trigger signals to each camera for image acquisition.

[0110] Furthermore, power supply 7 is used to power various electronic devices in this image acquisition platform (such as camera array 1, light source system 9, synchronizer 5, various modules and control modules 6); if it is a laboratory setting, power supply 7 is preferably a power supply facility that can provide a stable AC power environment; if it is in the field, an outdoor power supply or generator can be used.

[0111] (2) Arrange and measure control points, and obtain the spatial coordinates of the control points.

[0112] Both above-water and underwater control points are required. Above-water control points 8-1 should be placed as close to the water surface as possible to maximize control effectiveness, and their distribution along the shoreline of the target water area should be as even as possible. The spacing is related to the height and focal length of camera 1-1. The principle is that at least two above-water control points should be captured in each camera image, preferably more than four simultaneously. The spacing between two adjacent above-water control points 8-1 on the same shore is initially determined by the following formula, regardless of whether the water flow direction in the target area is clearly defined. It should be noted that after the initial arrangement of above-water control points 8-1, after camera 1-1 is installed, it is necessary to test whether the aforementioned standard of at least two above-water control points 8-1 being captured in each camera image can be met. If not, the number and spacing of above-water control points 8-1 need to be adjusted until the requirement is met. The number of underwater control points 8-2 is generally the same as that of surface control points 8-1. The X-axis position of the underwater control points can be the same as that of the surface control points, while the Y-axis position is determined by moving 5-10cm from the toe of the slope towards the center of the water surface. In addition, it should be avoided as much as possible to place them in areas with strong water flow and severe sediment transport, otherwise sediment transport may bury the underwater control points 8-2.

[0113] L G ≤0.5W

[0114] Among them, L G The distance between two adjacent water control points on the same bank is denoted by W, which is the actual spatial size corresponding to the width direction of a single camera image (the smaller of the two side lengths of the image).

[0115] After the control points are set up, the coordinates of the center points of each control point are measured using high-precision measuring equipment such as RTK GPS, total station, and laser scanner. Using the position of the RTK GPS, total station, or laser scanner as the origin, the horizontal direction along the water flow direction is defined as the positive X-axis, the horizontal direction perpendicular to it to the left is defined as the positive Y-axis, and the vertical upward direction is defined as the positive Z-axis. When the water flow direction is not clearly defined, the longer direction of the water surface area is defined as the X-axis, and the horizontal direction perpendicular to it is defined as the Y-axis. The positive X-axis and Y-axis directions are determined according to the measurement direction. Therefore, in this embodiment, the horizontal sections of the longitudinal slide rail 2-1, the L-shaped bracket 4-3, and the width direction of all camera sensors are arranged along the X-axis, while the length directions of the transverse slide rail 3-1, all crossbars 1-3, and all camera sensors are arranged along the Y-axis.

[0116] (3) Determine the height of the camera array above the bed surface, its arrangement, and the step size of its movement.

[0117] (3-1) Determine the height and arrangement of the camera array above the bed surface.

[0118] After confirming the total number of cameras (m) and focal length (f) in each row of the camera array based on the actual equipment conditions, the height of the camera array and the lateral (Y-direction) spacing (L) between cameras in each row are initially determined using the following method. c :

[0119] First, arrange the image sensors of all cameras along the Y-axis in length and the X-axis in width. Then, assume that the height of camera array 1 above the bed surface is H, and calculate the actual distance L corresponding to the length direction of the image sensor of each camera 1-1 in camera array 1 using the following formula:

[0120]

[0121]

[0122] Where h is the image distance, l c This refers to the length of the camera's image sensor.

[0123] The overlap of individual images captured by two adjacent cameras in the Y-axis direction should be no less than 0. Y (O in this embodiment) Y If we take 80%, then the lateral spacing L of the camera is... c satisfy:

[0124] L c ≤(1-O Y )L

[0125] Choose the camera horizontal spacing L that satisfies the above formula. c Based on the calculation results, the following two conditions are used for verification. If both conditions cannot be met simultaneously, it is necessary to adjust the height H of the camera array above the bed surface and the lateral spacing L of the cameras. c Recalculate and verify until both conditions are met. The two conditions are:

[0126] Condition 1: If the total width of the water surface in the target area is required to be covered by the total shooting range of all cameras in each row except for the two outermost cameras, then:

[0127] (m-3)·L c +L≥(W w ) max

[0128] Among them, (W) w ) maxIt is the maximum width of the water surface in the target area (the direction of the water surface width is parallel to the axis of the crossbar 1-3).

[0129] Condition 2: Control points should be clearly visible in every image. Generally, the side length of a single control point pattern should correspond to at least T pixels (in this embodiment, T is 20) in a single image. Therefore:

[0130]

[0131] Where Res is the spatial resolution of a single image, and L GCP This refers to the actual side length of the control point pattern.

[0132] Furthermore, when n>1, that is, when camera array 1 consists of multiple rows of cameras, the longitudinal (X-direction) spacing L of each row of cameras also needs to be determined according to the following formula. L :

[0133]

[0134] Among them, w c This refers to the width of the camera's image sensor.

[0135] (3-2) Determine the longitudinal movement spatial step size of the camera array

[0136] The longitudinal movement spatial step size of the camera array is determined based on the relationship between the actual longitudinal distance covered by the camera array and the maximum longitudinal distance to the water surface in the target area. Specifically:

[0137] If the actual longitudinal distance S covered by the camera array is greater than or equal to the maximum longitudinal distance X of the target area's water surface, then the longitudinal movement step size of the camera array is 0. This means that the camera array does not need to be moved during the measurement process; only the position of the camera array needs to be adjusted before measurement to ensure the total shooting area covers the target area. If the actual longitudinal distance S covered by the camera array is less than the maximum longitudinal distance X of the target area's water surface... max -X min (X max and X min Let be the maximum and minimum coordinates of the water surface in the target area along the X-axis, respectively. Then, during the measurement process, the longitudinal spatial step size ΔL of the camera array needs to be moved along the X-axis according to the following formula:

[0138]

[0139] S=[1+(n-1)(1-O X )]W=[n(1-O X )+O X W

[0140] ΔL=S-(1-OX W

[0141] Where W is the actual spatial dimension corresponding to the width direction of a single underwater image, and the overlap of a single image in the X-axis direction should not be less than 0. X .

[0142] (4) Based on the parameters determined in step (3), the target area is photographed to obtain an underwater image sequence of the sandy bed surface. First, the center of the camera array is moved to the initial position. When using a single-row camera or a multi-row camera whose shooting range cannot fully cover the target area, the initial position of the camera array center is located at the upstream position of the target area (i.e., X). min In the middle of the cross-section of the location, when there is no obvious direction of water flow, it is located at the end with a smaller X value in the length direction of the water surface (i.e., X). min (Location). If it is a single-row camera array, then the X-axis of the water surface boundary of the target area... min The location should be on the side with the larger X-value of the camera's field of view. If it is a multi-row camera array, the X-value of the target area's water surface boundary should be considered. min The position should be within the field of view of the camera in the row with the largest X value. If a multi-row camera array is used and step (3) shows that its measurement range can cover the entire target area, then move the center of the camera array directly above the center of the target area so that the entire boundary of the target water surface area can be seen in the camera's field of view.

[0143] During the measurement process, regardless of whether it's a single-row camera or a camera array composed of multiple rows of cameras, the camera moves from its initial position along the positive X-axis towards the other end of the target area in a longitudinal spatial step of ΔL. After each longitudinal spatial step, all cameras in the array simultaneously capture the first image, then move another longitudinal spatial step before stopping to capture the next image, until the camera reaches the downstream position of the target area (or the end of the target area with the largest X-value). A 1-second pause is generally maintained between two adjacent spatial intervals to ensure the camera captures a clear image. If the total shooting range of the multi-row camera array encompasses the entire target area (i.e., S ≥ (X...)...) max -X min If the camera array is stationary, then no movement of the camera array is required during the measurement process. The first image obtained by the j-th camera at the i-th position where the camera array stops shooting is denoted as (P1). ij When S≥X max -X min When, let i = 1, when S <X max -X min When i∈[1,p], j∈[1,J], J=n×m, The rounding function is derived from all images (P1). ij This constitutes the first image sequence P1.

[0144] (5) Obtain the underwater three-dimensional topography of the sandy bed surface based on image sequence P1

[0145] (5-1) Based on the Structure from Motion (SfM) method, underwater terrain modeling is performed on the first image sequence P1 to obtain a 3D point cloud. Spatial data registration is performed using control points. The initial encrypted 3D point cloud and its corresponding digital orthophoto obtained using only the location information of the surface control points are denoted as C1 and DOM1, respectively. The initial encrypted 3D point cloud and its corresponding digital orthophoto obtained using only the location information of the underwater control points are denoted as C1' and DOM1', respectively. The specific steps are as follows:

[0146] Spatial computation is performed on the first image sequence P1 using the SfM method to reconstruct each image (P1). ij The corresponding camera spatial position and lens orientation are used to obtain a sparse 3D point cloud of the target area through feature point matching. The center positions of each control point appearing in the first image sequence P1 (water control points are used for C1 and DOM1; underwater control points are used for C1' and DOM1', and so on) are identified and marked using manual or automatic methods to obtain the image coordinates of the control points. The manual method involves visually identifying the center of the control points in each image (P1). ij The location within the image; automatic methods identify images through image processing (such as edge detection) or by training convolutional neural networks (CNNs) (P1). ij The system identifies the control point marker regions and then approximates these regions using predefined geometric shapes (such as circles, ellipses, rectangles, etc., depending on the actual shape of the control point pattern). The centroid of these approximate regions is then calculated as the image coordinates of the control point centers. Spatial data registration is performed on the sparse 3D point cloud using the correspondence between the actual and image coordinates of the control points. Based on the registered 3D point cloud, an initial encrypted 3D point cloud is obtained using encryption methods (such as multi-view stereo, MVS method). Finally, a digital orthophoto is generated based on this initial encrypted 3D point cloud.

[0147] (5-2) Perform refraction correction on the initial encrypted 3D point cloud C1 to obtain the refraction-corrected raster point cloud C1. c

[0148] In the digital orthophoto DOM1, single-point manual calibration of water surface boundary points is used to obtain a series of planar spatial coordinates of these points. These points are then projected onto the initial refined 3D point cloud C1 to obtain a series of 3D spatial coordinates of the water surface boundary points. Based on these water surface boundary points, a Delaunay triangulation mesh is generated as the water surface model WS1. The initial refined 3D point cloud C1 is then rasterized. The raster size is determined by the spatial resolution requirements of the data analysis and should be smaller than the required spatial resolution. The bed elevation of the raster is taken as the maximum elevation value of all point cloud points contained within the raster range, resulting in the rasterized initial 3D point cloud C1. g Calculate the initial rasterized 3D point cloud C1 g The distance from each grid point to the water surface model WS1 is used as the initial water depth d for each grid point. a Delete the rasterized initial 3D point cloud C1 g d a Grid points with a value less than 0 are selected, meaning grid points located below the water surface model. If the grid points selected based on the above selection criteria still include some grid points actually located above the water surface model, then it is necessary to generate the spatial fitting plane FWS1 for the water surface boundary points and calculate the initial rasterized 3D point cloud C1. g The distance to the spatial fitting plane FWS1 is used to rasterize the initial 3D point cloud C1. g A secondary screening process is performed, testing different screening thresholds until all grid points above the water surface are deleted, thus obtaining the grid point cloud C1 below the water surface. u For each image corresponding to a camera position, the actual spatial range of the image captured at that camera position is calculated based on the camera orientation angles (pitch, roll, and yaw) calculated using the SfM method, as well as the camera's focal length and sensor size. The grid point cloud C1 falling within this range is then calculated. u In this context, grid points are considered visible at that camera location, and each grid point is typically visible at multiple camera locations.

[0149] For the selected raster point cloud C1 u Each grid point in the data is corrected for its initial water depth according to the following formula, resulting in the grid point cloud C1. u After correcting all grid points, the refraction-corrected grid point cloud C1 is obtained. c :

[0150]

[0151] d a =Z w -Z a

[0152]

[0153]

[0154]

[0155] Where, r ij X is the angle between the camera's optical axis and the water surface at that camera position. cij Y cij These are the X-axis and Y-axis coordinates of the camera position, respectively. wij X is the vertical distance between the camera position and the water surface model; a X a , z a Let d be the X-axis, Y-axis, and Z-axis coordinates of a grid point located in the underwater region. a Let d be the initial water depth at a given grid point. ij This is the corrected water depth for a specific grid point calculated based on the camera position ij. The average corrected water depth is calculated for the camera positions corresponding to all images of a given grid point within the image range. p and q represent the number of camera array positions and camera numbers that stopped shooting, respectively, for all camera positions visible at that grid point. Z w Z represents the elevation of the water surface model corresponding to the horizontal position of a certain grid point. c is the corrected bed surface elevation for a given grid point; n1 and n2 are the refractive indices of water and air, respectively.

[0156] (5-3) Using the initial encrypted 3D point cloud C1' to refine the refraction-corrected raster point cloud C1 c Secondary corrections were performed to obtain the underwater three-dimensional topographic measurement results of the sandy bed surface in the target area.

[0157] A water surface model WS1' containing the target water surface area is generated based on the digital orthophoto DOM1'. The initial encrypted 3D point cloud C1' is processed using the same rasterization method as the initial encrypted 3D point cloud C1 (i.e., the raster size is the same, and the bed elevation of the raster is the maximum elevation value of all point cloud points contained within the raster range) to obtain a rasterized initial 3D point cloud C1' using only underwater control points. g For the rasterized initial 3D point cloud C1' g Using the rasterized initial 3D point cloud C1 g The same filtering method is used to filter the raster point cloud C1' located below the water surface model WS1'. u (i.e., the filtering method provided in step (5-2)), and calculate the raster point cloud C1'. u Average elevation of bed surface The refraction-corrected grid point cloud C1 c Calculate the average elevation of the bed surface Calculate the raster point cloud C1' u and the refraction-corrected grid point cloud C1 c The difference in average bed surface elevation:

[0158] The refraction-corrected grid point cloud C1 c Bed elevation Z of each grid point c We need to add ΔZ, i.e., Z. c =Z c +ΔZ, thus obtaining the final underwater 3D grid point cloud C1 f This refers to the underwater three-dimensional topography of the target sandy bed surface.

[0159] In this embodiment, using underwater control points for spatial registration and refraction correction of the 3D point cloud has opposite effects on the elevation error of the 3D topography of the sandy bed. Furthermore, the underwater 3D topography obtained using underwater control points without refraction correction has a smaller overall elevation error and a more uniform spatial distribution compared to the underwater 3D topography obtained using surface control points with refraction correction, although the latter is slightly lower than the true value overall. However, the average elevation difference between the underwater 3D topography obtained using underwater control points without refraction correction and the underwater 3D topography obtained using surface control points with refraction correction is close to the elevation error of the latter. Therefore, this elevation difference can be used to further correct the underwater 3D topography obtained using surface control points with refraction correction.

[0160] Furthermore, the first image sequence P1 obtained through step (4) at different times can be used to obtain the corresponding underwater three-dimensional topography of the target sandy bed surface according to steps (5-1)-(5-3), thereby obtaining the sequence of the three-dimensional topography of the target sandy bed surface changing over time.

[0161] Furthermore, the measurement method of this disclosure embodiment also includes identifying sediment transport zones in the target area, with the following specific steps:

[0162] S1, Obtain the second image sequence P2

[0163] In step (4), let Y be the boundary of the water surface near the target area. min The camera position is set as the first camera, located near the water surface boundary of the target area in the Y direction. maxThe positions of the cameras in the array are set as follows: After all cameras in the array simultaneously capture the first image at position i, the positions of the camera array remain unchanged. The cameras in the remaining columns (excluding the first column) are then controlled to capture the second image sequentially at equal time intervals ΔT. Specifically, after ΔT of time following the simultaneous capture of the first image by all cameras, the second camera (in a single-row camera array) or the second column of cameras (in a multi-row camera array) is controlled to capture the second image. After 2ΔT of time following the simultaneous capture of the first image by all cameras, the third camera... The kth camera (single-row camera array) or the 3rd column camera (multi-row camera array) takes the second image, and so on. After all cameras have taken the first image, at a time interval of (k-1)·ΔT, the kth camera (single-row camera array) or the kth column camera (multi-row camera array) takes the second image (k∈[2,m]). After all cameras in the camera array have taken the second image at position i, the camera array moves one vertical spatial step ΔL to position i+1, and the above shooting method is repeated until the camera array moves to the X-axis of the water surface boundary of the target area. max The camera array stops shooting at position i, and the second image obtained by the j-th camera at that position is denoted as (P2). ij , composed of all images (P2) ij The second image sequence P2 is constructed; ΔT generally ranges from 2s to 5s and can be appropriately selected based on the water flow intensity. When the water flow intensity is high, a smaller ΔT is selected, and when the water flow intensity is low, a larger ΔT is selected. The second image sequence allows for fine sand transport between each frame, thus containing the fine sand transport characteristics of the sandy bed surface, which can be used to analyze the sand transport region of the sandy bed surface.

[0164] S2. Based on the second image sequence P2 and the target area sandy bed surface obtained in step (5), underwater three-dimensional topography identification of the sediment transport area.

[0165] Following step (5-1), underwater terrain modeling is performed on the second image sequence P2 using the SfM method to obtain a 3D point cloud. Spatial data registration is performed using surface control points to obtain the initial encrypted 3D point cloud C2 and its corresponding digital orthophoto DOM2. Following step (5-2), refraction correction is performed on the initial encrypted 3D point cloud C2 to obtain the refraction-corrected raster point cloud, which is used as the final underwater 3D raster point cloud C2 obtained based on the second image sequence P2. f C1 f With C2 f Since the rasterization methods are exactly the same, they can be directly subtracted to obtain the raster point cloud ΔC, where each raster point is assigned the value C1. f With C2 fThe elevation difference of the grid point bed surface is used to identify grid points in the grid point cloud ΔC whose grid point values ​​exceed the error threshold as sand transport areas in the target area. Considering the limited thickness of the fine sand transport layer, the error threshold can generally be selected as 1cm to 2cm or the value corresponding to 95%-99% of the elevation probability density distribution of the grid point cloud ΔC.

[0166] In one embodiment of the measurement method provided in this disclosure, the underwater three-dimensional topography of a sandy bed and its sediment transport zone are measured in an indoor water tank. The maximum length of the water surface area is 4m, and the maximum width is 2.5m. The median particle size of the fine sand particles constituting the bed surface is 0.36mm. The bed surface width is 2m, the measurement distance is 4m, and the slope is 2‰. The test uses a single-row camera array consisting of 5 Canon 80D SLR cameras, all facing vertically downwards, essentially perpendicular to the water surface, with a fixed focal length of 18mm and a sensor size of 22.3×14.9mm. 2 The photo resolution is 6000×4000 pixels. 2 The spatial resolution of the acquired images must be no less than 1 mm / pixel. A point laser rangefinder is installed in the middle of the camera array, with the array height from the bed surface approximately 3.4 m. The horizontal spacing between the cameras in the camera array is 0.5 m. The calculated actual spatial dimension corresponding to the length direction of a single image from the camera array is 4.19 m. The total shooting range of the three middle cameras (excluding the outermost ones) is 5.19 m, exceeding the maximum width of the water surface in the target area by 2.5 m. The actual resolution of the acquired underwater image is 0.7 mm / pixel, simultaneously satisfying both conditions in step (3). A strip LED light is arranged on one side of the camera array.

[0167] Twelve control points were established in the target area, with each surface control point supported by a corresponding concrete base on each bank. There were six control points in total: three surface control points and three underwater control points on each bank. All control points were constructed from plastic sheets with sides of 10cm. The three surface control points on the same bank were generally aligned in a straight line along the downstream direction (X-direction), and the three underwater control points on the same bank were also generally aligned in a straight line along the X-direction. The X-direction spacing between the surface and underwater control points on the same bank was 1.6m. The coordinates of the control points were measured using a total station with an accuracy of 3mm. This example illustrates the measurement and data processing flow for the underwater three-dimensional topography of the sandy bed and the sediment transport area.

[0168] The measurement process is as follows:

[0169] 1. Turn on the water tank circulation system and set the inflow rate to 100 L / s. After the flow rate stabilizes, begin measurement. The longitudinal movement interval of the camera array is 0.5 m. The camera array is set to move one spatial interval along the X direction and then remain stationary for 18 seconds. The five cameras first simultaneously capture the first image. Then, from left to right, the first to fifth cameras capture their second images at time intervals of 3 seconds, 6 seconds, 9 seconds, 12 seconds, and 15 seconds, respectively. During the shooting process, all cameras are set to aperture 4.0, shutter speed 1 / 100 s, and ISO 1600. A total of 70 images (first image) and 70 images (second image) of the target area are acquired, constructing image sequences P1 and P2, respectively.

[0170] 2. The SfM method is used to reconstruct the first image sequence P1 in 3D to obtain a sparse point cloud. The center position of the underwater control point in each image is manually marked. The coordinates of the underwater control point are used to spatially register the sparse point cloud to generate a dense point cloud C1, and the corresponding digital orthophoto DOM1 is generated with a spatial resolution of 0.62 mm / pixel. The SfM method is used again to reconstruct the first image sequence P1 in 3D, but this time the center position of the underwater control point is used for spatial registration to generate a dense point cloud C1' and a digital orthophoto DOM1'.

[0171] 3. In the digital orthophoto DOM1, the water surface boundary points were manually calibrated using single points, obtaining and exporting the spatial coordinates of 226 water surface boundary points. The Delaunay triangulation mesh was generated based on these water surface boundary points using Cloud Compare software to serve as the water surface model. Both C1 and C1' were rasterized to 0.005m. The bed elevation of the raster was taken as the maximum elevation value of all point cloud points contained within the raster range, resulting in the initial rasterized 3D point cloud C1. g and C1' g Calculate C1 g The distance to the water surface is used as the initial water depth d. a According to d a <0 Deletes grid points located above the water surface model. After this filtering, grid points above and below the water surface model can be completely distinguished, thus obtaining the grid point cloud C1 located in the underwater region. u Therefore, no further filtering is needed. At this point, the raster point cloud data includes the raster point coordinates X... a Y a Z a water surface elevation Z w Initial water depth d a Based on the grid point coordinates, camera position (three-dimensional coordinates and three angles representing orientation) data derived from the SfM method, and the camera's own sensor data, refraction correction is performed to obtain the corrected water depth for each underwater grid point. Compared with the corrected bed surface elevation Z cThus, the refraction-corrected grating point cloud C1 is obtained. c Then, using the same method, C1' g The refraction-corrected raster point cloud C1' was obtained by screening. u Calculate C1' u and C1 c The difference in average bed elevation ΔZ, for C1 c Bed elevation Z of each grid point c We need to add ΔZ to obtain the final corrected bed elevation Z. c Generate underwater 3D raster point cloud C1 f See Figure 5 (a)

[0172] 4. Process the first image sequence P1 obtained at different times using steps 2 to 3 above to obtain the underwater three-dimensional topography of the target sandy bed surface at different times, thereby obtaining the sequence of the three-dimensional topography of the target sandy bed surface changing over time.

[0173] 5. Following the processing in step S2 above, the bed elevation difference raster point cloud ΔC is obtained. Areas in ΔC with values ​​exceeding the error threshold of 2 cm are identified as fine sand transport zones and marked accordingly. Figure 5 As shown in (b).

[0174] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0175] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for measuring the three-dimensional underwater topography of a sandy bed, characterized in that, include: (1) Construct an image acquisition platform for acquiring underwater image sequences of sandy bed surface. The image acquisition platform includes a camera array consisting of J cameras arranged in the form of n×m, where n is the number of rows of the camera array, n≥1, m is the number of columns of the camera array, m≥3, and each row of cameras is required to capture the maximum width of the target water surface area. The rows of the camera array are defined as perpendicular to the water flow direction, and the columns are defined as in the direction of water flow. (2) Arrange and measure control points, and obtain the spatial coordinates of the control points. The control points include surface control points set on both sides of the target water surface area and underwater control points set on the bed surface of the target water surface area. Each control point includes a flat plate with a regular geometric pattern drawn on its upper surface and a horizontal slope. The geometric pattern has a clear center point position, which is defined as the coordinate of the control point. It is required that multiple control points can be captured in each image taken by a single camera. The coordinates of each control point are measured using a measuring device. The position of the measuring device is defined as the origin. The horizontal direction with the water flow or the length direction along the target water surface area is defined as the X-axis direction, the horizontal direction perpendicular to the X-axis is defined as the Y-axis direction, and the direction perpendicular to the XY plane is defined as the Z-axis direction. (3) Determine the height of the camera array above the bed surface, its arrangement, and the step size of its movement. (3-1) Determine the height and arrangement of the camera array above the bed surface. After confirming the number of columns m and focal length f of the camera array, the height of the camera array and the lateral spacing L of each camera in each row along the Y-axis can be initially determined using the following method. c : Arrange the image sensors of all cameras along the Y-axis in length and the X-axis in width. Let H be the height of the camera array above the bed surface. Calculate the actual distance L corresponding to the length direction of each camera image sensor in the camera array using the following formula: Where h is the image distance, l c The length of the camera's image sensor; The overlap of individual images captured by two adjacent cameras in the Y-axis direction should be no less than 0. Y Then the lateral spacing L of the camera along the Y-axis c satisfy: L c ≤(1-O Y )L Choose the camera horizontal spacing L that satisfies the above formula. c Based on the calculation results, the following two conditions are used for verification. If both conditions cannot be met simultaneously, it is necessary to adjust the height H of the camera array above the bed surface and the lateral spacing L of the cameras. c Recalculate and verify until both conditions are met; the two conditions are: Condition 1: If the total width of the water surface in the target area is required to be covered by the total shooting range of all cameras in each row except for the two outermost cameras, then: (m-3)·L c +L≥(W w ) max Among them, (W) w ) max It is the maximum width of the water surface in the target area; Condition 2: The control points are clearly visible in every image. If the side length of a single control point pattern corresponds to at least T pixels in a single image, then: Where Res is the spatial resolution of a single image, and L GCP The actual side length of the control point pattern; When n>1, the longitudinal spacing L of each row of cameras along the X-axis is determined according to the following formula. L : Among them, w c The width of the camera's image sensor; (3-2) Determine the longitudinal movement spatial step size of the camera array If the actual longitudinal distance S covered by the camera array is greater than or equal to the maximum longitudinal distance X of the target water surface area. max -X min X max and X min Let X and Y be the maximum and minimum coordinates of the target water surface area along the X-axis, respectively. Then, the longitudinal movement step size of the camera array is equal to 0. If the actual longitudinal distance S covered by the camera array is less than the maximum longitudinal distance X of the target water surface area... max -X min Then, the longitudinal spatial step size ΔL of the camera array is set according to the following formula: S=[n(1-O X )+O X ]W ΔL=S-(1-O X )W Where W is the actual spatial dimension corresponding to the width direction of a single image, and the overlap of a single image in the X-axis direction should not be less than 0. X ; (4) Based on the parameters determined in step (3), the target water surface area is photographed to obtain an underwater image sequence of the sandy bed surface. First, move the camera array center to its initial position. When the camera array's shooting range along the X-axis cannot completely cover the maximum length of the target water surface area, the initial position of the camera array center is located at the target water surface area X. min At the center of the cross section of the location, when the shooting range of the camera array along the X-axis can fully cover the maximum length of the target water surface area, the initial position of the center of the camera array is located directly above the center of the target water surface area and the entire water surface boundary of the target area can be seen in the camera's field of view. Then, the camera array is moved from its initial position along the positive X-axis towards the other end of the target water surface area in a longitudinal spatial step of ΔL. After each longitudinal spatial step, all cameras in the camera array simultaneously capture the first image. Then, the array moves another longitudinal spatial step and stops capturing images, until it reaches the X-axis of the target water surface area. max Position; the first image captured by the j-th camera at the i-th position where the camera array stops shooting is denoted as (P1). ij When S≥X max -X min When, let i = 1, when S <X max -X min When i∈[1,p], The rounding function is used to round up all images (P1). ij The underwater image sequence of the sandy bed surface includes the first image sequence P1, constituting the first image sequence P1. (5) Obtain the underwater three-dimensional topography of the sandy bed surface based on the first image sequence P1. (5-1) The underwater terrain modeling of the first image sequence P1 is performed based on the motion reconstruction structure method to obtain a three-dimensional point cloud. Spatial data registration is performed using control points. The initial encrypted three-dimensional point cloud and its corresponding digital orthophoto are denoted as C1 and DOM1, respectively, and the initial encrypted three-dimensional point cloud and its corresponding digital orthophoto are denoted as C1' and DOM1', respectively. (5-2) Generate a water surface model WS1 containing the target water surface region based on the digital orthophoto DOM1, and rasterize the initial encrypted 3D point cloud C1 to obtain the rasterized initial 3D point cloud C1. g Rasterize the initial 3D point cloud C1 g The portion of the model below the water surface WS1 was selected to obtain the raster point cloud C1 located below the water surface. u For the grid point cloud C1 u Refraction correction is performed to obtain the refraction-corrected grid point cloud C1. c ; (5-3) Generate a water surface model WS1' containing the target water surface area based on the digital orthophoto DOM1'. Apply the same rasterization method as the initial encrypted 3D point cloud C1' to the initial encrypted 3D point cloud C1' to obtain a rasterized initial 3D point cloud C1' using only underwater control points. g Rasterize the initial 3D point cloud C1' g The portion of the model below the water surface WS1' is selected to obtain the raster point cloud C1' located below the water surface. u Calculate the raster point cloud C1' u and the refraction-corrected grid point cloud C1 c The difference in average bed elevation ΔZ is used to adjust the refraction-corrected grid point cloud C1. c The bed elevation of each grid point is corrected to obtain the final underwater 3D grid point cloud C1. f This refers to the underwater three-dimensional topography of the target sandy bed surface.

2. The measurement method according to claim 1, characterized in that, The image acquisition platform built in step (1) also includes: The longitudinal module includes longitudinal slide rails placed on both sides of the water surface in the target area along the direction of water flow and longitudinal sliders disposed on the longitudinal slide rails and movable thereal. A horizontal module includes a horizontal slide rail placed horizontally along the direction perpendicular to the water flow and a horizontal slider disposed on the horizontal slide rail and movable thereal. The end of the horizontal slide rail is connected to the vertical slider. The position of the horizontal slide rail in the horizontal direction along the water flow direction is changed by the vertical module. The vertical module includes a vertical slide rail connected to and moving with the horizontal slider, a vertical slider disposed on the vertical slide rail and movable along it, and an L-shaped bracket connected to and moving with the vertical slider. The vertical section of the L-shaped bracket is parallel to the axis of the vertical slide rail, and the horizontal section of the L-shaped bracket is parallel to the axis of the longitudinal slide rail. The vertical module is used to adjust the height of the camera array from the bed surface. There are n horizontal bars, each of which is connected to the horizontal section of the L-shaped bracket via a horizontal bar slider that can move along the horizontal section of the L-shaped bracket. The axial direction of all horizontal bars is parallel to the axial direction of the horizontal slide rail, and the height of all horizontal bars is consistent. Each horizontal bar is equipped with m gimbals for fixing the camera, and the angle and position of the camera are adjusted by the gimbals. A synchronizer is used to send a single trigger signal to all cameras in a camera array so that all cameras can capture images simultaneously. The control module is used to control the movement mode, movement speed and target position of the longitudinal module, the transverse module and the vertical module, control the synchronizer, and control the shooting parameters of all cameras and receive images captured by all cameras; A power supply is used to power the various electronic devices in the image acquisition platform.

3. The measurement method according to claim 2, characterized in that, The longitudinal module and the transverse module are respectively equipped with a locator for determining the position of their respective sliders. Each locator uses a point laser rangefinder fixed to one end of its respective slide rail to measure the distance and position the slider.

4. The measurement method according to claim 1, characterized in that, The image acquisition platform also includes a light source system deployed on one or both sides of the crossbar.

5. The measurement method according to claim 1, characterized in that, In step (2), the measuring equipment used is an RTK GPS, a total station, or a laser scanner.

6. The measurement method according to claim 1, characterized in that, Step (5-1) specifically includes the following steps: Spatial calculations are performed on the first image sequence P1 based on the structure-of-motion method to reconstruct the camera spatial position and lens orientation corresponding to each image. A sparse three-dimensional point cloud of the target water surface area is obtained through feature point matching. The center position of each control point appearing in the first image sequence P1 is identified and marked by manual or automatic methods to obtain the image coordinates of the control points. The spatial data registration of the sparse three-dimensional point cloud is performed using the correspondence between the real coordinates and image coordinates of the control points. Based on the registered three-dimensional point cloud, an initial encrypted three-dimensional point cloud is obtained using an encryption method, and the corresponding digital orthophoto is generated.

7. The measurement method according to claim 1, characterized in that, Step (5-2) specifically includes the following steps: In the digital orthophoto DOM1, single-point calibration is used to identify water surface boundary points, obtaining a series of planar spatial coordinates for these points. These water surface boundary points are then projected onto the initial refined 3D point cloud C1, obtaining a series of 3D spatial coordinates for the water surface boundary points. Based on these water surface boundary points, a Delaunay triangulation mesh is generated as the water surface model WS1. The initial refined 3D point cloud C1 is then rasterized, with the bed elevation of the raster taken as the maximum elevation value of all point cloud points contained within the raster area, resulting in the rasterized initial 3D point cloud C1. g ; Calculate the initial rasterized 3D point cloud C1 g The distance from each grid point to the water surface model WS1 is used as the initial water depth d for each grid point. a ; Delete the rasterized initial 3D point cloud C1 g d a For grid points with values ​​less than 0, the resulting grid point cloud C1 is located below the water surface model WS1. u For each image corresponding to a camera position, the actual spatial range of the image captured at that camera position is calculated based on the camera orientation angle calculated using the structure-of-motion method, as well as the camera's focal length and sensor size. The grid point cloud C1 falling within this actual spatial range is then calculated. u The grid points in the image are considered visible at the camera's location; For the selected raster point cloud C1 u Each grid point in the data is corrected for its initial water depth according to the following formula, resulting in the grid point cloud C1. u After correcting all grid points, the refraction-corrected grid point cloud C1 is obtained. c : d a =Z w -Z a Where, r ij X is the angle between the camera's optical axis and the water surface at that camera position. cij Y cij These are the X-axis and Y-axis coordinates of the camera position, respectively. wij X is the vertical distance between the camera position and the water surface model; a X a Z a Let d be the X-axis, Y-axis, and Z-axis coordinates of a grid point located in the underwater region. a Let d be the initial water depth at a given grid point. ij This is the corrected water depth for a specific grid point calculated based on the camera position ij. The average corrected water depth is calculated for the camera positions corresponding to all images of a given grid point within the image range. p and q represent the number of camera array positions and camera numbers that stopped shooting, respectively, for all camera positions visible at that grid point. Z w Z represents the elevation of the water surface model corresponding to the horizontal position of a certain grid point. c is the corrected bed surface elevation for a given grid point; n1 and n2 are the refractive indices of water and air, respectively.

8. The measurement method according to claim 1, characterized in that, In steps (5-2) and (5-3), When rasterizing the initial 3D point cloud C1 g or C1' g If, after deleting grid points with a water depth less than 0, the resulting grid point cloud located below the water surface model WS1 or WS1' still contains some grid points actually located above the water surface model WS1 or WS1', then it is necessary to generate a spatial fitting plane FWS1 or FWS1' for the water surface boundary points and calculate the initial rasterized 3D point cloud C1. g or C1' g The distance to the spatial fitting plane FWS1 or FWS1' is used to perform secondary screening of the three-dimensional points on the water surface. Different screening thresholds are tested until all grid points located above the water surface model WS1 or WS1' are removed, thus finally obtaining the grid point cloud located below the water surface model WS1 or WS1'.

9. The measurement method according to any one of claims 1 to 8, characterized in that, The measurement method further includes obtaining the corresponding underwater three-dimensional topography of the target sandy bed surface by taking the first image sequence P1 obtained in step (4) at different times according to steps (5-1) to (5-3), thereby obtaining the sequence of the three-dimensional topography of the target sandy bed surface changing over time.

10. The measurement method according to any one of claims 1 to 8, characterized in that, The measurement method also includes identifying sand transport zones in the target sandy bed surface, with the following specific steps: S1, Obtain the second image sequence P2 In step (4), let the distance be closer to Y. min The camera position is set as the first camera, close to Y. max The position setting is for the m-th column of cameras, Y max and Y min Let X and Y represent the maximum and minimum coordinates of the target water surface area along the Y-axis, respectively. After all cameras in the camera array simultaneously capture the first image at position i, the position of the camera array remains unchanged, and the cameras in the remaining columns (excluding the first column) capture the second image sequentially at equal time intervals ΔT. After all cameras in the camera array have captured the second image at position i, the camera array moves one vertical spatial step ΔL to position i+1, and the above shooting method is repeated until the camera array moves to X. max The camera array stops shooting at position i, and the second image obtained by the j-th camera at that position is denoted as (P2). ij , composed of all images (P2) ij This constitutes the second image sequence P2; S2. Based on the second image sequence P2 and the target area sandy bed surface obtained in step (5), underwater three-dimensional topography identification of the sediment transport area. Following step (5-1), underwater terrain modeling is performed on the second image sequence P2 to obtain a 3D point cloud. Spatial data registration is performed using surface control points to obtain the initial encrypted 3D point cloud C2 and its corresponding digital orthophoto DOM2. Following step (5-2), refraction correction is performed on the initial encrypted 3D point cloud C2 to obtain the refraction-corrected raster point cloud, which is used as the final underwater 3D raster point cloud C2 obtained based on the second image sequence P2. f ;C1 f With C2 f Subtracting the values ​​yields the raster point cloud ΔC, where each raster point is assigned the value C1. f With C2 f The elevation difference of the grid points on the bed surface is used to identify grid points in the grid point cloud ΔC whose grid point values ​​exceed the error threshold as sand transport areas in the target sandy bed surface.

Images