CN116630558A - Three-dimensional terrain reconstruction method of pebble bed based on underwater images - Google Patents

Abstract

The pebble bed surface three-dimensional terrain reconstruction method based on the underwater image comprises the following steps: acquiring an underwater image sequence of a pebble bed surface, wherein the image sequence consists of a plurality of underwater images of the pebble bed surface obtained by shooting the water surface of a target area cross section by cross section, a plurality of control points can be shot in each underwater image, the control points are arranged on two sides of the water surface of the target area, the upper surfaces of the control points are painted with regular geometric patterns and have definite center point positions, and the center point positions are defined as coordinates of the control points; performing space calculation on the image sequence based on a motion restoration structure method to obtain an initial encrypted three-dimensional point cloud and a digital orthographic image; and establishing a water surface model, calculating the initial water depth of the gridding point cloud, carrying out refraction correction on the initial water depth of each point of the gridding point cloud, calculating the height of the corrected bed surface, and carrying out secondary correction to obtain the underwater three-dimensional topography of the target pebble bed surface. The method has high spatial resolution and precision, and can not interfere an observed object.

Description

Three-dimensional terrain reconstruction method of pebble bed based on underwater images

technical field

The disclosure belongs to the fields of environment and water conservancy engineering, river geomorphology, and mountain disaster measurement, and in particular relates to a three-dimensional terrain reconstruction method for pebble beds based on underwater images.

Background technique

Pebble rivers are widely distributed in nature, and the riverbed materials have a wide gradation, in which the coarse particles are pebbles and gravels. The wide-graded riverbed materials are easy to form a variety of riverbed structures under the action of water flow, and the three-dimensional characteristics of the bed topography are obvious. The slope of the pebble river is between the bedrock river with a steep slope and the sandy river with a gentle slope. It generally occurs in the transition zone between high and steep mountains and plains. Some lakes originating from the ancient channel of pebble rivers also have pebble beds. . There are a large number of pebble rivers and lakes on the pebble bed in the hinterland and edge of the Qinghai-Tibet Plateau in my country. The continuous topographical changes of the pebble bed are characterized by infrastructure engineering construction, urbanization development, geological disaster prevention, living environment improvement, and water pollution in these areas. Important basic information for governance, biodiversity restoration and protection. However, whether it is field observation or laboratory simulation test, it is very difficult to directly obtain the underwater three-dimensional topography of the pebble bed. In the laboratory environment, it is generally necessary to interrupt the water flow process before measuring the topography of the bed surface because of the condition of measuring the topography after the water is cut off.

With the rapid development of machine vision technology, photogrammetry methods based on visible light have been widely used in the field of topographic measurement, such as the structure from motion (SfM) technology based on multi-eye vision, which can acquire target areas based on overlapping image sequences The high spatial resolution and high-precision 3D topographic information have great potential in the acquisition of pebble bed topography. When the clarity of the water body is high, the camera captures an image sequence above the water surface on the pebble bed. Based on this image sequence, the SfM method can be used to directly reconstruct the 3D underwater topography of the bed surface. However, visible light will be reflected and refracted on the water surface, especially the refraction effect makes the underwater terrain imaging higher than the actual elevation. than low. Moreover, when using the above method to obtain the underwater three-dimensional topography of the pebble bed surface, errors will be introduced due to the influence of the material of the bed surface. Therefore, when the SfM method is applied to directly obtain the underwater three-dimensional topography of the pebble bed surface, it is necessary to remove the influence of water surface refraction and bed surface material.

To sum up, in view of the above-mentioned continuous monitoring requirements of the underwater 3D topography of the pebble bed and the problems existing in the existing technology, it is necessary to develop a special reconstruction method for the 3D topography of the pebble bed based on underwater images.

Contents of the invention

The present disclosure aims to solve at least one of the technical problems existing in the prior art.

To this end, the embodiment of the present disclosure provides a three-dimensional terrain reconstruction method based on underwater images of pebble beds with high spatial resolution and precision, easy operation and low cost, including:

(1) It is stipulated that the X-axis is along the length direction of the water surface of the target area, the Y-axis is along the width direction of the water surface of the target area, and the Z-axis is perpendicular to the XY plane; the camera is used to obtain the underwater image sequence of the pebble bed surface, and the fixed The focal length of the camera and the space step along the X-axis and the Y-axis, the lens of the camera is directed perpendicular to the water surface, and the underwater image sequence is a plurality of pebble beds obtained by shooting cross-sections of the water surface in the target area Each underwater image can capture at least 2 control points and two adjacent underwater images meet the overlap requirements. The control points are arranged on both sides of the water surface in the target area, including the upper surface A slab drawn with a regular geometric pattern with a horizontal slope, the geometric pattern has a definite center point position, and the center point position is defined as the coordinates of the control point;

(2) Perform underwater terrain modeling on the underwater image sequence of the pebble bed based on the motion restoration structure method to obtain a three-dimensional point cloud, and use the control points to perform spatial data registration on the three-dimensional point cloud, based on the registered 3D point cloud Obtain the initial encrypted 3D point cloud and generate the corresponding digital orthophoto;

(3) Generate a water surface model containing the target area based on the digital orthophoto, filter out the part below the water surface model in the initial encrypted 3D point cloud, and obtain a 3D point cloud located in the underwater area. The 3D point cloud located in the underwater area is refraction corrected to obtain the corrected underwater terrain and initial water depth;

(4) Using the initial water depth to correct the corrected underwater topography again, so as to obtain the underwater three-dimensional topography measurement result of the pebble bed.

In some embodiments, step (2) specifically includes the following steps:

Perform spatial calculation on the underwater image sequence based on the motion recovery structure method, reconstruct the spatial position of the camera and the orientation of the lens corresponding to each image, and obtain a sparse three-dimensional point cloud of the target area through feature point matching; Identifying and marking the center position of each control point appearing in the underwater image sequence, obtaining the image coordinates of the control point, using the corresponding relationship between the real coordinates of the control point and the image coordinates to perform spatial data registration on the sparse 3D point cloud, Based on the registered 3D point cloud, the encryption method is used to obtain the initial encrypted 3D point cloud, and the corresponding digital orthophoto is generated.

In some embodiments, step (3) specifically includes the following steps:

In the digital orthophoto, a single point is used to calibrate the boundary points of the water surface, and projected to the initial encrypted 3D point cloud to obtain a series of spatial coordinates of the boundary points of the water surface, and a Delaunay triangular mesh is generated based on the boundary points of the water surface as Including the water surface model of the target area; performing rasterization processing on the initial encrypted 3D point cloud, the bed surface elevation value of the raster is the maximum elevation value of all point cloud points contained in the raster range, and the rasterization initial Three-dimensional point cloud; calculate the distance from the gridded initial three-dimensional point cloud to the water surface model as the initial water depth; delete the three-dimensional points in the gridded initial three-dimensional point cloud whose initial water depth is less than 0, and obtain the underwater The grid points of the area; for the camera position corresponding to each image, the actual spatial range corresponding to the image captured by the camera position will be calculated according to the camera orientation angle calculated by the motion recovery structure method, the focal length of the camera, and the size of the photosensitive element of the camera. Grid points that fall within this actual spatial extent are considered visible to this camera position;

For each screened grid point located in the underwater area, refraction correction is performed according to the following formula to obtain the corrected underwater terrain and initial water depth:

d _a =Z _w -Z _a

Among them, j represents a certain grid point in the underwater area visible to the jth camera position, r _j is the angle between the camera optical axis and the water surface at the jth camera position, X _cj and Y _cj are respectively The X-axis and Y-axis coordinates of the j-th camera position, H _wj is the vertical distance between the j-th camera position and the water surface model, X _a , Y _a , Z _a are the X of the certain grid point axis direction, Y axis direction and Z axis direction coordinates, d _a is the initial water depth of the certain grid point, d _j is the corrected water depth of the certain grid point calculated based on the jth camera position, is the average corrected water depth calculated for the camera position corresponding to all the images that include the certain grid point in the image range, Z _w is the elevation of the water surface model corresponding to the horizontal position of the certain grid point, and Z _p is the certain grid point The corrected bed elevation, n ₁ and n ₂ are the refractive indices of water and air, respectively.

Further, in step (3), when the 3D points in the rasterized initial 3D point cloud whose initial water depth is less than 0 are deleted, the grid points located in the underwater area still contain some points that are actually located above the water surface model. Grid points, it is also necessary to generate a space fitting plane of the boundary points of the water surface, calculate the distance from the gridded initial 3D point cloud to the fitting plane, and use this distance to perform secondary screening on the 3D points on the water, and test different Filter the threshold until all the grid points above the water surface model are eliminated, and finally the grid points located in the underwater area are obtained.

Further, in step (3), when the focal plane of the camera is parallel to the water surface, the following approximate correction method is used to calculate the initial water depth of each grid point in the selected underwater area:

_Zp = _Zw -d.

Further, step (4) specifically includes:

The bed elevation Z _p of each point cloud grid in the underwater area after refraction correction is further corrected by the following formula to obtain the final corrected bed elevation Z _p ':

ΔZ=-0.0049d-0.0019

z _p '=z _p -Δz

The underwater three-dimensional topographic measurement results of the pebble bed are composed of the X-axis coordinate X _a , the Y-axis coordinate Y _a and the final corrected bed surface elevation Z _p ' of all grid points.

The beneficial effects of the disclosure are:

1. The underwater three-dimensional topography of the pebble bed surface can be obtained with high spatial resolution and precision; among them, based on the visible light and images used, the reconstructed three-dimensional underwater topography of the pebble bed surface has high spatial resolution, up to submillimeter level ; The overall error is controlled based on the control points, especially when the coordinates of the control points with higher precision can be obtained, the error can be further reduced (such as using a total station and a laser scanner to measure the coordinates of the control points, the error is millimeter level ), in addition, the present disclosure performs secondary correction after the refraction correction, and removes the elevation error introduced by the refraction correction and the material of the bed surface. Therefore, the above two reasons make the reconstructed underwater three-dimensional topography of the pebble bed surface highly accurate , up to millimeter level;

2. It is a non-invasive measurement, which does not interfere with the water body and the bed surface;

3. It is easy to operate and does not need to calibrate the shooting camera.

Description of drawings

Fig. 1 is a flow chart of a three-dimensional terrain reconstruction method for a pebble bed provided by an embodiment of the present disclosure.

(a)-(d) of FIG. 2 are schematic structural diagrams of control points laid out in the reconstruction method provided by the embodiment of the present disclosure.

Fig. 3 is a digital elevation model diagram of the three-dimensional underwater terrain of the pebble bed surface obtained by the reconstruction method provided by an embodiment of the present disclosure, wherein (a) is the underwater digital elevation model of the pebble bed surface before refraction correction, and (b) is the refraction Corrected digital elevation model.

In the figure, 1 is a control point, 1-1 is a frame, 1-2 is a flat plate, 1-3 is a ground nail, 1-4 is a rotating shaft, and 1-5 is a support rod.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, not to limit the present application.

On the contrary, this application covers any alternatives, modifications, equivalent methods and schemes within the spirit and scope of this application as defined by the claims. Further, in order to make the public have a better understanding of the application, some specific details are described in detail in the detailed description of the application below. The present application can be fully understood by those skilled in the art without the description of these detailed parts.

In describing the present disclosure, it is to be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", " Orientation indicated by rear, left, right, vertical, horizontal, top, bottom, inside, outside, clockwise, counterclockwise, etc. The positional relationship is based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present disclosure and simplifying the description, and does not indicate or imply that the referred base or element must have a specific orientation, be constructed and operated in a specific orientation, Therefore, it should not be construed as limiting the present disclosure. In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, a feature defined as "first" and "second" may explicitly or implicitly include one or more of these features. In the description of the present disclosure, "plurality" means two or more, unless otherwise specifically defined.

In the description of the present disclosure, it should be noted that unless otherwise specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection. Connection, or integral connection; can be mechanical connection or electrical connection; can be direct connection or indirect connection through an intermediary, and can be the internal communication of two elements or the interaction relationship between two elements. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present disclosure according to specific situations.

In the present disclosure, unless otherwise clearly stated and limited, a first feature being "on" or "under" a second feature may include direct contact between the first and second features, and may also include the first and second features Not in direct contact but through another characteristic contact between them. Moreover, "above", "above" and "above" the first feature on the second feature include that the first feature is directly above and obliquely above the second feature, or simply means that the first feature is horizontally higher than the second feature. "Below", "beneath" and "under" the first feature to the second feature include that the first feature is directly below and obliquely below the second feature, or simply means that the first feature has a lower level than the second feature.

Referring to FIG. 1 , the method for reconstructing the three-dimensional terrain of the pebble bed based on the underwater image provided by the embodiment of the present disclosure includes:

(1) It is stipulated that the X-axis is along the length direction of the water surface of the target area, the Y-axis is along the width direction of the water surface of the target area, and the Z-axis is perpendicular to the XY plane; the camera is used to obtain the underwater image sequence of the pebble bed surface, and the fixed The focal length of the camera and the space step along the X-axis and Y-axis, the lens of the camera is basically perpendicular to the water surface, and the acquired underwater image sequence is obtained by taking cross-sections of the water surface in the target area. Underwater image composition, at least two control points 1 can be photographed in each underwater image, and two adjacent underwater images meet the overlap requirements (generally, the overlapping area of two adjacent underwater images in the X-axis and Y-axis directions is required It should not be lower than the set threshold. It should be noted that the overlapping area requirements of two adjacent underwater images in the X-axis and Y-axis directions can be the same or different. In this embodiment, the two requirements are the same, and the set thresholds are both taken as 80%), referring to Fig. 2, the control point 1 is arranged on both sides of the water surface in the target area, including the frame 1-1 fixed on the water surface of the target area on both sides of the water surface and the flat plate 1-2 connected with the frame 1-1 in rotation , the slope of the flat plate 1-2 remains level, and the upper surface of the flat plate 1-2 is painted with a regular geometric pattern composed of black and white two colors, and has a clear center point position, which is defined as the coordinate of the control point 1; control The side length of point 1 should ensure that the center point of control point 1 can be clearly identified in the camera image. In Figure 2, (a) is the shape of the control point when the placement position is horizontal, (b) is the shape of the control point when the placement position is a slope, (c) is the bottom view of the control point, and (d) is the ground nail of the control point , fixed on the outer frame of the control point through the screw segment.

Furthermore, the control point 1 is a solid rectangular (mostly square) plastic flat plate with a thickness of 5-10 mm, and the side length is determined according to the height of the camera. The principle of value selection is that the center of the control point 1 can be clearly identified in the camera image point location. The plate 1-2 of the control point is connected to one side of the frame 1-1 through the rotating shaft 1-4, and the other side of the plate 1-2 and the frame 1-1 is connected through the support rod 1-5, and the frame 1-1 has a screw hole Used to install ground nails 1-3. The control point 1 is generally installed on the shore, and the ground nails 1-3 are embedded into the bed surface to fix the control point 1, and then the slope of the flat plate 1-2 is adjusted to the basic level through the rotating shaft 1-4, so that the camera can take pictures clearly from the top.

It should be noted that during field measurements, drones equipped with cameras can also be used instead of cameras for shooting. During the shooting process, the camera lens should be kept perpendicular to the water surface to reduce reflections on the water surface.

Furthermore, when the control point 1 is arranged on the water-free ground around the water surface in the target area, the control point 1 should be as close to the water surface as possible, and the distance between the two adjacent control points 1 on the same bank is relatively uniform, and the distance is related to the height and focal length of the camera. The layout principle is that at least 2 control points 1 can be captured in each camera image, and preferably more than 4 control points 1 can be captured at the same time. The distance between two adjacent control points 1 on the same bank is preliminarily determined by the following formula. Regardless of whether the direction of water flow can be determined in the target area, the distance between control points can also be preliminarily selected according to the following criteria. It should be noted that after the preliminary arrangement of control points 1, it is necessary to test whether the above-mentioned standard of capturing at least two control points 1 in each camera image can be met after the camera is installed. If not, it is necessary to adjust the number of control points 1 and spacing until the requirements are met.

L _G ≤0.5W

Among them, L _G is the distance between two adjacent control points on the same bank, and W is the actual spatial size corresponding to the width direction of a single camera image (the smaller value of the two side lengths of the image).

After the layout of the control points is determined, use RTK GPS, total station or laser scanner and other equipment with high measurement accuracy to measure the coordinate position of the center point of the control points. Taking the position of RTK GPS, total station or laser scanner as the origin, the horizontal direction along the water flow direction is the positive direction of the X-axis, the horizontal direction vertical to the left is the positive direction of the Y-axis, and the vertical upward direction is the positive direction of the Z-axis; When the water flow direction is not specified, the width direction of the water surface is the Y-axis direction, and the horizontal direction perpendicular to it is the X-axis direction, and the positive directions of the X-axis and Y-axis are determined according to the image shooting direction.

During the measurement process, the camera shooting parameters (image size, image format, aperture, shutter, ISO, etc.) are controlled to remain unchanged, and the movement speed remains unchanged. The pause time of the camera in the cross-section measurement is determined by whether a clear image can be obtained, and it is generally maintained at 1s to eliminate the impact of the camera shaking on the shooting when the camera stops moving. If the lighting conditions are good when shooting, you can keep the shutter speed above 1/500s under the premise of clear imaging, and you can keep the camera moving at a constant speed without stopping. The setting of the moving speed U is also subject to clear imaging and sharp edges. If a uniform motion is used, the camera shutter needs to be set at a fixed time interval to shoot, and the time interval Δt is calculated and set according to the following formula:

All the images captured by the camera are arranged in chronological order to form the underwater image sequence of the pebble bed, which is recorded as P={P ₁ ,P ₂ ,...,P _l ,...P _q }, P _l is the lth image captured by the camera An underwater image of a pebble bed, l∈[1,q]. q is the total number of images,

(2) Based on the Structure from Motion (SfM) method, the underwater terrain is modeled to obtain a 3D point cloud, and the control points are used for spatial data registration.

Based on the structure from motion (SfM) method, the underwater image sequence P is spatially calculated, and the corresponding camera spatial position and lens orientation of each image are reconstructed, and the sparse 3D point cloud of the target area is obtained through feature point matching. The center position of each control point appearing in the underwater image sequence P is identified and marked by manual or automatic methods, and the image coordinates of the control points are obtained. The manual method is to visually identify the position of the center of the control point in each image The automatic method recognizes the control point mark area in the image through image processing (such as edge detection) or training convolutional neural network (CNN), and then uses the geometric shape of the set rule (such as circle, ellipse, rectangle, etc., depending on After approximating the actual shape of the control point pattern), the centroid is calculated as the image coordinate of the control point center. Use the corresponding relationship between the real coordinates of the control points and the image coordinates to perform spatial data registration on the sparse 3D point cloud, and use encryption methods (such as multi-view stereo, MVS) to obtain the initial encrypted 3D point cloud based on the registered 3D point cloud , and generate the corresponding digital orthophoto.

(3) Refraction correction is performed on the initial encrypted 3D point cloud to obtain the corrected underwater topography and water depth.

In the digital orthophoto image containing the target area, a single point is used to manually calibrate the water surface boundary points to obtain a series of plane space coordinates of the water surface boundary points, and then projected to the initial encrypted 3D point cloud to obtain the vertical coordinates of the water surface boundary points. A Delaunay triangular mesh is generated based on the boundary points of the water surface as a water surface model. The initial encrypted 3D point cloud is rasterized, and the grid size is determined by the spatial resolution requirements of the data analysis, which should be smaller than the spatial resolution required by the data analysis. The bed surface elevation of the grid is the maximum value of the elevation of all point cloud points contained in the grid range, and the initial gridded 3D point cloud is obtained. Calculate the distance from the rasterized initial 3D point cloud to the water surface model as the initial water depth. The 3D points whose initial water depth is less than 0 in the rasterized initial 3D point cloud can be deleted, that is, the grid points located above the water surface model can be deleted. If the grid points selected based on the above screening criteria still contain some grid points that are actually located above the water surface model, it is also necessary to generate the space fitting plane of the water surface boundary points, and calculate the distance between the rasterized initial 3D point cloud and the fitting plane. Distance, use this distance to perform secondary screening on the 3D points on the water, test different screening thresholds until all the grid points above the water surface model are eliminated, so as to obtain the grid points located in the underwater area. For the camera position corresponding to each image, the camera orientation angle (pitch angle, roll angle, and yaw angle) calculated by the SfM method, the focal length of the camera, and the size of the photosensitive element of the camera are used to calculate the corresponding The actual spatial range, the grid points falling within the actual spatial range are considered to be visible to the camera position, and each grid point is generally visible to multiple camera positions.

For each filtered grid point located in the underwater area, the initial water depth is corrected according to the following formula:

d _a =z _w -Z _a

Among them, j represents a certain grid point in the underwater area visible to the jth camera position, r _j is the angle between the camera optical axis and the water surface at the jth camera position, X _cj and Y _cj are respectively The X-axis and Y-axis coordinates of the j-th camera position, H _wj is the vertical distance between the j-th camera position and the water surface model, X _a , Y _a , Z _a are the X of the certain grid point axis direction, Y axis direction and Z axis direction coordinates, d _a is the initial water depth of the certain grid point, d _j is the corrected water depth of the certain grid point calculated based on the jth camera position, is the average corrected water depth calculated for the camera position corresponding to all the images that include the certain grid point in the image range, Z _w is the elevation of the water surface model corresponding to the horizontal position of the certain grid point, and Z _p is the certain grid point The corrected bed elevation, n ₁ and n ₂ are the refractive indices of water and air, respectively.

When the measurement accuracy is low, or the focal plane of the camera is strictly parallel to the water surface, the following approximate correction method can also be used to calculate the initial water depth of each grid point in the selected underwater area:

z _p =z _w -d

In the embodiment of the present disclosure, through step (3), the influence of the refraction of visible light passing through the water surface on the SfM-based three-dimensional reconstruction result is eliminated.

(4) Use the initial water depth to correct the corrected underwater terrain again, so as to obtain the underwater three-dimensional topographic measurement results of the pebble bed.

The bed surface elevation Z _p of each point cloud grid in the underwater area after refraction correction needs to be further corrected by the following formula (the unit of length is m) to obtain the final corrected bed surface elevation Z _p ':

ΔZ=-0.0049d-0.0019

Z _p '=Z _p -ΔZ

The underwater three-dimensional topographic measurement results of the pebble bed are composed of the X-axis coordinate X _a , the Y-axis coordinate Y _a and the final corrected bed surface elevation Z _p ' of all grid points.

The embodiment of the present disclosure further eliminates the error in the elevation of the bed surface after the refraction correction through step (4). The error is related to the refraction correction itself, and is also related to the material of the bed surface. This error has a linear relationship with water depth, so it can be corrected by the linear formula provided.

In one embodiment of the method for reconstructing the three-dimensional topography of the pebble bed surface of the present disclosure, the three-dimensional topography of the pebble bed surface is reconstructed based on the underwater image of the indoor water tank. The particle size range of the pebbles making up the bed surface is 9mm-50mm, with a median particle size of about 15mm. The width of the bed is 2.2m, and the length along the measurement is 4m. The test uses a single-camera solution (Canon 80D SLR), and the height of the camera from the bed is about 3.4m. The focal length of the camera is fixed at 18mm, the photosensitive element size is 22.3×14.9mm ² , and the photo resolution is 6000×4000pix ² . Arrange 4 control points in the measurement area, each bank has 2 control points, the control points are made of plastic plates with a side length of 10cm, located on the shore close to the water surface, and the 2 control points on the same bank are along the direction of the water flow (X-axis direction ) spacing is 2m. The real coordinates of the control points are measured with a total station with an accuracy of 3 mm. This embodiment is used to illustrate the processing flow of the underwater image of the pebble bed surface.

The specific process is as follows:

1. Turn on the tank circulation system and set a constant inflow rate of 40L/s. After the flow is stable, the measurement is started. The spatial interval of the horizontal and vertical movement of the camera is 0.5m, and it is a stepping motion, and the stop is 1s between two adjacent steps. During the shooting process, the aperture was set to 4.0, the shutter speed was higher than 1/80s, ISO=1600-2000, and a total of 101 images were collected for the target area.

2. Use the SfM method to perform three-dimensional reconstruction of the obtained image, obtain a sparse three-dimensional point cloud, manually mark the center position of the control point in each image as the image coordinate of the control point, and use the corresponding relationship between the real coordinates of the control point and the image coordinates to Sparse 3D point cloud is used for spatial registration, and based on the registered 3D point cloud, the MVS method is used to encrypt and calculate the initial encrypted 3D point cloud, see (a) in Figure 3, and the corresponding digital orthophoto image is generated with a spatial resolution of 0.65mm/pix.

3. In the digital orthophoto map, use a single point to manually calibrate the water surface boundary points, obtain the spatial coordinates of 195 water surface boundary points and export them. Use CloudCompare software to generate a Delaunay triangular mesh based on the water surface boundary points as the water surface model. The initial encrypted 3D point cloud is processed into a 0.005m grid, and the bed elevation of the grid is the maximum value of the elevation of all point cloud points contained in the grid range to obtain the gridded initial 3D point cloud, and calculate its The distance to the water surface is taken as the initial water depth. Delete the grid points above the water surface model according to the initial water depth less than 0. Since there is a problem in calculating the distance of the triangular mesh at the boundary of the water surface model, the space fitting plane of the water surface boundary points is generated, and different distance thresholds are tested for secondary screening until Delete all the grid points above the water surface, and finally get the grid points in the underwater area. At this time, the grid point data includes grid point coordinates X _a , Y _a , Z _a , water surface elevation Z _w and initial water depth d _a .

4. Export the grid points into csv format, export the camera position (three-dimensional coordinates, and three angles representing the orientation) data from the SfM method, and prepare the photosensitive element data of the camera itself, based on the above three aspects of data for underwater The grid points are corrected for refraction, and the corrected water depth of each grid point is obtained and the corrected bed elevation Z _p . Finally, according to the empirical formula, the bed surface elevation Z _p is further corrected to obtain the final corrected bed surface elevation Z _p ', and the underwater three-dimensional terrain is generated, as shown in (b) in Fig. 3 .

In the description of this specification, references to the terms "one embodiment," "some embodiments," "exemplary embodiments," "example," "specific examples," or "some examples" are intended to mean that the implementation A specific feature, structure, material, or characteristic described by an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations 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 any one or more embodiments or examples.

Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present application. The scope of the application is defined by the claims and their equivalents.

Claims (6)

1. A three-dimensional terrain reconstruction method based on underwater images, characterized in that it comprises: (1) It is stipulated that the X-axis is along the length direction of the water surface of the target area, the Y-axis is along the width direction of the water surface of the target area, and the Z-axis is perpendicular to the XY plane; the camera is used to obtain the underwater image sequence of the pebble bed surface, and the fixed The focal length of the camera and the space step along the X-axis and the Y-axis, the lens of the camera is directed perpendicular to the water surface, and the underwater image sequence is a plurality of pebble beds obtained by shooting cross-sections of the water surface in the target area Each underwater image can capture at least 2 control points and two adjacent underwater images meet the overlap requirements. The control points are arranged on both sides of the water surface in the target area, including the upper surface A slab drawn with a regular geometric pattern with a horizontal slope, the geometric pattern has a definite center point position, and the center point position is defined as the coordinates of the control point; (2) Perform underwater terrain modeling on the underwater image sequence of the pebble bed based on the motion restoration structure method to obtain a three-dimensional point cloud, and use the control points to perform spatial data registration on the three-dimensional point cloud, based on the registered 3D point cloud Obtain the initial encrypted 3D point cloud and generate the corresponding digital orthophoto; (3) Generate a water surface model containing the target area based on the digital orthophoto, filter out the part below the water surface model in the initial encrypted 3D point cloud, and obtain a 3D point cloud located in the underwater area. The 3D point cloud located in the underwater area is refraction corrected to obtain the corrected underwater terrain and initial water depth; (4) Using the initial water depth to correct the corrected underwater topography again, so as to obtain the underwater three-dimensional topography measurement result of the pebble bed. 2. The three-dimensional terrain reconstruction method on the cobble bed surface according to claim 1, wherein step (2) specifically comprises the following steps: Perform spatial calculation on the underwater image sequence based on the motion recovery structure method, reconstruct the spatial position of the camera and the orientation of the lens corresponding to each image, and obtain a sparse three-dimensional point cloud of the target area through feature point matching; Identifying and marking the center position of each control point appearing in the underwater image sequence, obtaining the image coordinates of the control point, using the corresponding relationship between the real coordinates of the control point and the image coordinates to perform spatial data registration on the sparse 3D point cloud, Based on the registered 3D point cloud, the encryption method is used to obtain the initial encrypted 3D point cloud, and the corresponding digital orthophoto is generated. 3. The three-dimensional terrain reconstruction method on the cobble bed surface according to claim 1, wherein step (3) specifically comprises the following steps: In the digital orthophoto, a single point is used to calibrate the boundary points of the water surface, and projected to the initial encrypted 3D point cloud to obtain a series of spatial coordinates of the boundary points of the water surface, and a Delaunay triangular mesh is generated based on the boundary points of the water surface as Including the water surface model of the target area; performing rasterization processing on the initial encrypted 3D point cloud, the bed surface elevation value of the raster is the maximum elevation value of all point cloud points contained in the raster range, and the rasterization initial Three-dimensional point cloud; calculate the distance from the gridded initial three-dimensional point cloud to the water surface model as the initial water depth; delete the three-dimensional points in the gridded initial three-dimensional point cloud whose initial water depth is less than 0, and obtain the underwater The grid points of the area; for the camera position corresponding to each image, the actual spatial range corresponding to the image captured by the camera position will be calculated according to the camera orientation angle calculated by the motion recovery structure method, the focal length of the camera, and the size of the photosensitive element of the camera. Grid points that fall within this actual spatial extent are considered visible to this camera position; For each screened grid point located in the underwater area, refraction correction is performed according to the following formula to obtain the corrected underwater terrain and initial water depth: d _a =Z _w -Z _a Among them, j represents a certain grid point in the underwater area visible to the jth camera position, r _j is the angle between the camera optical axis and the water surface at the jth camera position, X _cj and Y _cj are respectively The X-axis and Y-axis coordinates of the j-th camera position, H _wj is the vertical distance between the j-th camera position and the water surface model, X _a , Y _a , Z _a are the X of the certain grid point axis direction, Y axis direction and Z axis direction coordinates, d _a is the initial water depth of the certain grid point, d _j is the corrected water depth of the certain grid point calculated based on the jth camera position, is the average corrected water depth calculated for the camera position corresponding to all the images that include the certain grid point in the image range, Z _w is the elevation of the water surface model corresponding to the horizontal position of the certain grid point, and Z _p is the certain grid point The corrected bed elevation, n ₁ and n ₂ are the refractive indices of water and air, respectively. 4. The three-dimensional terrain reconstruction method on the cobble bed surface according to claim 3, characterized in that, in step (3), after deleting the three-dimensional points whose initial water depth is less than 0 in the gridded initial three-dimensional point cloud, the obtained If the grid points located in the underwater area still contain some grid points that are actually located above the water surface model, it is also necessary to generate a space fitting plane of the water surface boundary points, and calculate the gridded initial 3D point cloud to the simulated Using this distance to perform secondary screening on the 3D points above the water, test different screening thresholds until all the grid points above the water surface model are eliminated, and finally the grid points located in the underwater area are obtained. 5. The three-dimensional terrain reconstruction method of pebble bed surface according to claim 3, characterized in that, in step (3), when the focal plane of the camera is parallel to the water surface, the following approximate correction method is used to filter out the underwater area The initial water depth of each grid point is calculated: _Zp = _Zw -d. 6. according to claim 3,4 or 5 described three-dimensional terrain reconstruction method of pebble bed surface, it is characterized in that, step (4) specifically comprises: The bed surface elevation Z _p of each point cloud grid in the underwater area after refraction correction is further corrected by the following formula to obtain the final corrected bed surface elevation Z _p ′: ΔZ=-0.0049d-0.0019 Z _p '=Z _p -ΔZ The underwater three-dimensional topographic measurement results of the pebble bed are composed of the X-axis coordinate X _a , the Y-axis coordinate Y _a and the final corrected bed surface elevation Z _p ′ of all grid points.