Underwater topographic survey method

Abstract

The invention discloses an underwater topographic measurement method. In order to solve the problems that in an existing physical model experiment, contact type measurement destroys the terrain, optical measurement is interfered by refraction, and acoustic equipment is high in cost, a measurement scheme integrating flight time correction and refraction geometric compensation is provided. According to the method, a depth camera is arranged to collect scoured terrain point cloud data in real time, Gaussian filtering denoising and outlier elimination are carried out in sequence, a double-medium light path correction model is constructed based on the Snell law, flight time errors caused by light velocity differences and axial and radial offsets caused by air-water interface refraction are calculated respectively, and the accuracy and the reliability of the two-medium light path correction model are improved. And finally, high-precision three-dimensional terrain coordinates are generated. According to the method, low-cost hardware of the depth camera is combined with algorithm compensation, errors caused by refraction and light velocity differences in underwater topographic survey are reduced, experimental results show that the average relative error is smaller than 3%, non-contact and high-precision real-time monitoring of the physical model scoured topography is achieved, and reliable technical support is provided for hydraulic structure safety assessment.

Description

Technical Field

[0001] The present invention belongs to the field of smart water conservancy, and specifically relates to an underwater terrain measurement method, which is suitable for three-dimensional reconstruction and real-time monitoring of scour terrain in hydraulic physical model experiments. Background Art

[0002] Physical model experiments are hampered by bottlenecks in topographic surveying technology for localized scour studies of water-related projects. Existing measurement methods are categorized as static and dynamic. Static measurement requires draining the physical model of water and uses contact measuring tools such as steel rulers and probes. While this method offers single-point accuracy, it destroys the original topography and fails to capture transient processes. Dynamic monitoring relies on non-contact technologies such as ultrasound and laser scanning. While this method allows for real-time measurement, it faces technical obstacles such as high equipment costs and operational complexity. For small-scale physical models, existing technologies struggle to strike a balance between cost-effectiveness, operational efficiency, and dynamic monitoring capabilities.

[0003] The aquatic environment places higher technical demands on scour topography measurements: contact measurements require drainage, resulting in data distortion; optical measurements are significantly affected by water refraction, leading to significant errors; and while acoustic equipment has underwater detection capabilities, its resolution is insufficient and its cost is prohibitive. Especially during dynamic scour processes, complex conditions such as turbid water and transient topographic changes place stringent demands on the measurement system's environmental adaptability, data acquisition frequency, and processing speed. Existing technologies struggle to meet the demands for high-precision, real-time monitoring.

[0004] Three-dimensional reconstruction technology provides a new direction for breaking through traditional limitations: although optical photogrammetry based on computer vision can achieve non-contact high-precision reconstruction, it relies on ideal lighting conditions and data processing is time-consuming; depth cameras have unique advantages due to their real-time depth perception characteristics. Their equipment cost is low and their operation is convenient (a single camera can complete data collection), but their adaptability to underwater environments, dynamic measurement accuracy and robustness in complex scouring scenarios still need in-depth research and verification.

[0005] In summary, current technologies face numerous challenges: high-precision equipment (such as laser scanners) conflicts with experimental economics; complex operational processes (such as the time-consuming calibration-acquisition-reconstruction process of photogrammetry) conflict with the timeliness of dynamic monitoring; and idealized measurement conditions conflict with the real-world hydraulic environment. There is an urgent need to develop new measurement methods that combine high precision, dynamics, and strong environmental adaptability to meet the dynamic monitoring needs of physical model scour terrain research. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing topographic surveying methods. By constructing a model that integrates time-of-flight correction with geometric compensation for refraction, it aims to accurately calculate the axial and radial offsets caused by light refraction at the air-water interface. This method provides real-time dynamic monitoring capabilities, significantly reducing equipment cost and operational complexity while ensuring high-precision measurements. It is particularly well-suited for real-time monitoring of scour terrain in small-scale physical model experiments, providing reliable technical support for water-related project safety assessments and disaster warnings.

[0007] A method for underwater topography measurement, characterized by comprising the following steps:

[0008] Step S1: Arrange the physical model experimental scene, install the depth camera and water level meter;

[0009] Step S2, using a depth camera to collect terrain data before and during the experiment;

[0010] Step S3, terrain data preprocessing, terrain data preprocessing is performed through Gaussian filtering and point cloud discreteness analysis;

[0011] Step S4, correcting the time of flight measured underwater, correcting the error caused by the time of flight of light based on the difference in the speed of light between air and underwater environments;

[0012] Step S5, calculating the axial offset of the underwater refraction measurement, and calculating the axial depth deviation caused by refraction based on Snell's law and the geometric model;

[0013] Step S6, calculating the radial offset correction value of the underwater measurement, correcting the horizontal position deviation caused by refraction, and adjusting the point cloud coordinates in combination with the direction vector;

[0014] Step S7, underwater terrain reconstruction: apply the above correction formula, traverse all point cloud data, and generate the corrected terrain three-dimensional coordinates.

[0015] Furthermore, step S1 is specifically as follows: A depth camera is mounted above the physical model test area, with the camera axis aligned perpendicular to the surface to be measured. For underwater measurement scenarios, a physical model containing the target scour terrain is constructed in a water tank, with controlled water depth and turbidity. A water level meter is also deployed to collect water level parameters.

[0016] Furthermore, step S2 is specifically as follows: Before the experiment began, topographic data was recorded at the initial moment. During the experiment, point cloud data of the underwater scour surface was continuously collected at set intervals. Simultaneously, water level parameters were recorded using synchronized water level meters to ensure that the collected data accurately reflected the dynamic changes in the scour process. The acquired depth images were used for subsequent 3D reconstruction.

[0017] Furthermore, step S3 is specifically as follows: Import the raw point cloud data collected in step S2 into MATLAB and store it as a three-dimensional coordinate matrix, with each point containing the coordinates (X, Y, Z). A Gaussian smoothing filter algorithm is used to suppress random noise. Based on the statistical characteristics of the point cloud density, the average distance between each point and its nearest neighbor is calculated. A neighborhood search algorithm is then used to remove sparse outliers in the point cloud, thereby removing noise points.

[0018] Furthermore, step S4 is specifically a flight time correction calculation, as shown in the following formula (1): Correct the measurement error caused by the difference in the speed of light in air and water, where d pixel The original measured depth of each pixel recorded by the depth camera, c air and c water are the speed of light in air and water, d air is the depth of the air layer, d f is the depth value corrected according to the flight time.

[0019] Furthermore, step S5 is specifically a calculation of axial offset correction, as shown in the following formula (2): The incident angle on the water surface is θ a , the refraction angle is θ w The Z-axis coordinate of the point is updated by formula (2) to correct the axial distortion caused by the refraction of the optical axis.

[0020] Furthermore, step S6 is specifically a radial offset correction calculation, as shown in the following formula (3): d r =(tanθ a -tanθ w )d water (3) The radial offset component is calculated by formula (3) to adjust the X and Y coordinates of the point.

[0021] Furthermore, step S7 is specifically as follows: Coordinate update and integration: Traverse all point cloud data, apply the above correction formula in sequence, and generate the corrected three-dimensional coordinates.

[0022] The beneficial effect of the present invention is that by integrating time-of-flight correction and optical path refraction compensation technology, the error caused by light speed differences and medium refraction in the underwater topography measurement of physical models is effectively reduced. At the same time, the use of a depth camera, with low equipment cost, can achieve dynamic data acquisition and real-time monitoring of key parameters such as the depth and length of scour pits, providing a high-precision, low-cost three-dimensional dynamic topography monitoring solution for hydraulic model experiments such as seawalls and bridge piers. The measurement data of the scour experiment of the physical model of the seawall's back slope show that the average relative error of the underwater topography measurement value of this method is less than 3%, which meets the high-precision dynamic monitoring requirements of the hydraulic model experiment. The error data of the experimental measurement are shown in Tables 1, 2 and 3. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of underwater topography measurement method.

[0024] Figure 2 This is the layout diagram of the underwater topography measurement experiment of the physical model of the back slope of the seawall.

[0025] Figure 3 This is a real-life picture of the underwater topography measurement experiment on the back slope of the seawall.

[0026] Figure 4 This is a schematic diagram of the refraction axial offset of the depth camera underwater measurement.

[0027] Figure 5 This is a schematic diagram of the underwater measurement optical path of the depth camera.

[0028] Figure 6 This is a schematic diagram of the radial offset of refraction measured by a depth camera underwater.

[0029] Figure 7 This is a scene diagram of underwater topography measurement during the scouring experiment.

[0030] Figure 8 This is the initial topographic map of the scour experiment on the back slope of the seawall.

[0031] Figure 9 It is the equilibrium topographic map of the scour experiment on the back slope of the seawall. DETAILED DESCRIPTION

[0032] The following describes the implementation of the present invention in detail, with reference to the accompanying figures and examples. This example examines the scouring of the backslope of a seawall. Based on similarity theory, a 1:100 scale physical model was constructed to simulate the scouring process on the backslope under overflow conditions. A controllable flume system was used to regulate the flow velocity and water level. A Kinect v2 depth camera (resolution 512×424, frame rate 30Hz) was used to collect real-time point cloud data of the scouring surface and simultaneously record water level parameters.

[0033] Step S1: Arrange the scour experiment scene of the physical model of the back slope of the seawall and install a depth camera. Based on similarity theory, a scaled physical model of the back slope of the seawall was constructed, overflow conditions were simulated through controllable water flume experiments, and non-contact dynamic monitoring and three-dimensional terrain reconstruction of the scouring process were achieved by combining water level meters and depth cameras. Figure 2 Arrange the scouring experiment of the physical model of the back slope of the seawall. Set up the depth camera above the sand surface of the local scouring physical model of the back slope of the seawall and shoot vertically downward. The depth camera can capture the depth image of the scouring surface of the back slope of the seawall. The real scene of the experiment process captured by the depth camera is as follows: Figure 3 shown.

[0034] Step S2: Use a depth camera to photograph the scour surface, record the initial terrain of the experiment and the terrain of the underwater scour surface during the experiment, and record the terrain data in the form of point cloud.

[0035] Step S3: Point cloud Gaussian filtering denoising and outlier removal: Step S31: Point cloud Gaussian filtering The structural noise and environmental noise in the depth image appear as randomly distributed isolated noise points. This embodiment uses a two-dimensional Gaussian filter to smooth them. The Gaussian kernel function is defined as: Where σ is the standard deviation of the Gaussian kernel, which controls the smoothing strength of the filter. This is implemented in MATLAB using the imgaussfilt function with the following parameters: kernel size 5*5, σ = 1.2. This parameter effectively suppresses small-scale noise while preserving edge sharpness. Step S32: Remove outliers from point cloud After converting the depth map into a 3D point cloud, the statistical properties of the point cloud density are used to remove sparse outliers. The specific steps are as follows: Domain density calculation: For each point in the point cloud, the average distance between it and its n nearest neighbors is calculated (n = 50 in this study); Threshold determination: The global average distance and standard deviation are calculated, and a threshold is set. If the average distance between a point and its n nearest neighbors is greater than the threshold, the point is considered an outlier. MATLAB implementation: Call the pcdenoise function and control the filtering strength through the NumNeighbors and Threshold parameters, or use a custom neighborhood search based on the pointcloud class (findNearestNeighbors function) to remove outliers from the point cloud.

[0036] Step S4: Calculate the time-of-flight offset of the underwater point cloud. The depth camera is placed in the air, and the infrared beam must pass through two different media: the air between the depth camera lens and the water surface, and the water between the water surface and the surface point of the object being measured. The infrared beam emitted from the depth camera will therefore undergo a medium conversion. The distance (d pixel) will be the sum of the two distances, namely the distance between the camera lens and the water surface (d air ) and the distance from the water surface to the actual object or surface (d water Similarly, the time it takes for the infrared beam to reach the actual object surface through the depth camera (t pixel ) is also divided into two different parts, namely the flight time of light in air (t air ) and the flight time of light in water (t water ). The measured distance and time are divided into two different medium calculation parts, as shown in formulas (5) and (6): d pixel =d air +d water (5) t pixel =t air +t water (6) The calculation formula of the depth camera flight time before correction can be expressed by formula (7). Formula (5), (6) and formula (7) can be combined to obtain formula (8), which is to calculate the new underwater distance (d corrected ). Corrected measured distance (d newpixel ) can be expressed by formula (9) to update the new depth value of each pixel. d corrected =t water *c water (8) d newpixel =d air +d corrected (9) In the above formula, c water represents the speed of light in water, c air Represents the speed of light in air. air The data can be obtained from the measurement data of the water level meter during the experiment. Arranging formula (5) to formula (9) can obtain formula (1). The new depth value calculated based on the principle that light travels more slowly in water is smaller in absolute value than the value directly measured by the depth camera and is closer to the actual distance between the surface point of the underwater object and the plane of the depth camera lens. The same calculation is performed for each depth pixel to obtain the time-of-flight corrected depth at each pixel.

[0037] Step S5: Calculate the refraction axial offset of the underwater point cloud. The point cloud acquired by the depth camera underwater shows obvious axial distortion (such as Figure 4(As shown in Figure 2). This distortion is more pronounced in the depth image of the plane, where the reconstructed 3D plane appears concave, convex toward the camera along the corners. The virtual image of the measured point travels along the normal to the actual measurement point, and the error increases as the incident angle of the infrared beam increases as the actual measurement point moves away from the center of the optical axis. To simplify the problem, it is assumed that there is no horizontal offset between the infrared emitter and the depth camera sensor, and the main axis of the depth camera is aligned with the optical axis of the image sensor. According to Snell's law, it can be shown that the axial offset of the depth (d a ) depends on the angle of incidence and refraction as well as the measured depth value. The incident angle in air of the infrared beam from the depth camera is defined as θ a , the angle of refraction in water is defined as θ w The known parameters here are the uncorrected coordinates of the measured point P, the air layer distance (d air ) and water depth d water Using the sine law, the displacement of depth d a It can be calculated by formula (2). Among them, θ a ,θ w The coordinates of the measured point P and the water depth d water The parameters are calculated and the calculation process is as follows: (1) Coordinate system and variable definition like Figure 5 As shown, the camera position is set to the coordinate origin (0,0,0). Shoot vertically downward, with the Z axis vertically downward. Assume that the measured coordinates of the measured point P are (X,Y,Z), and the light incident on the water surface is point Q (x,y,z), with an incident angle of θ a , the refraction angle is θ w , the air layer distance is d air , water depth d water . (2) Horizontal displacement proportional relationship Since the light received by the camera is vertical in the air, the light in the water needs to adjust the horizontal displacement to satisfy the law of refraction. The coordinates of the intersection point on the water surface are Q(x,y,z), then: Horizontal displacement in water: Vertical displacement in water: Zz; Horizontal displacement in air: Vertical displacement in air: z; According to Snell's law, the incident angle θ a (in air) and the refraction angle θ w (In water) satisfaction: n a sinθ a=n w sinθ w The refractive index of air is n a =1, water refractive index n w =1.33. (3) Geometric relationship and proportion derivation The ratio of horizontal displacement to vertical displacement is inversely proportional to the refractive index: Right now: The coordinates of the water surface intersection point Q are proportional to the coordinates of point P: x = kX y = kY. Where k is the proportional factor. Substituting into the geometric relationship equation, we get: Solve for the proportional factor k: (4) Q point coordinates Substitute k into x=kX, y=kY to obtain the coordinates of the intersection point with the water surface: Finally, the coordinates of point Q are: (5) From the geometric relationship, we can get θ a ,θ w expression:

[0038] Step S6: Calculate the refraction radial offset of the underwater point cloud. The point cloud acquired by the depth camera underwater also shows radial distortion (such as Figure 6 As shown in Figure 2, the measured point will be deflected away from the normal. When the actual measurement point is far from the center of the optical axis, the error increases with the increase of the incident angle of the infrared beam. This phenomenon manifests itself as the outline of an object in water appearing larger than it actually is when viewed from air. According to the geometric relationship, the radial offset can be obtained as: d r =(tanθ a -tanθ w )d water (3) Radial offset (d r ) represents the horizontal displacement of a point in the X, Y plane due to refraction. The direction must be determined based on the original coordinates (X, Y) of the point. The unit vector for: The radial offset (dr ) are assigned to the X and Y axes by direction:

[0039] Step S7: During the experiment, the underwater topography measurement scene is as follows: Figure 7 As shown. According to the calculated flight time offset, refraction axial offset, and refraction radial offset, the underwater point cloud coordinate data is updated to obtain underwater three-dimensional terrain data. Comparing the initial terrain and the scour equilibrium terrain of the scour experiment of the physical model of the back slope of the seawall, the initial terrain is as follows Figure 8 As shown, the scour balance topography is Figure 9 shown. The measured 3D topographic data was compared with the steel tape measurements. The steel tape measurements served as experimental reference values, while the depth camera measurements served as measured values. By comparing the reference and measured values, the performance of the underwater topographic measurement method was evaluated and the reliability of the measurement process was determined. The steel tape reference values ​​for the scour length at the dike foot, maximum scour depth, and maximum scour depth length, as well as the depth camera measurements and their associated errors, were obtained from the scour experiment on the back slope of the seawall. Tables 1, 2, and 3 show these values. AE represents absolute error, and RE represents relative error. Table 1 Measurement values ​​and errors of embankment foot scour length Scour length at the embankment foot (L b ) In terms of measurement, the mean absolute error (MAE) of the Kinect v2 depth camera is about 0.2cm and the mean relative error (MAPE) is about 2.3%. Table 2 Maximum scour depth measurement values ​​and errors At the maximum scour depth (h s ) In terms of measurement, the mean absolute error (MAE) of the Kinect v2 depth camera is about 0.2cm and the mean relative error (MAPE) is about 1.6%. Table 3 Maximum scour length measurement value and error At the maximum scour length (L s ) measurement, the mean absolute error (MAE) of depth camera measurement is about 2.3cm and the mean relative error (MAPE) is about 2.6%. In summary, the present invention demonstrates high practical value in the reconstruction of complex scour terrain in physical models. Its non-contact measurement characteristics can effectively make up for the shortcomings of traditional terrain measurement methods, and the overall average error is less than 3%.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for underwater topography measurement, characterized in that: The following steps are involved: Step S1: Arrange the physical model experimental scene and install the depth camera and water level meter; Step S2, using the depth camera to collect terrain point cloud data before and during the experiment, and synchronously record water level change parameters; Step S3, preprocessing the collected original point cloud data, including suppressing random noise through Gaussian filtering and removing outliers based on the statistical characteristics of point cloud density; Step S4, correcting the time-of-flight error based on the difference in light propagation speed in air and water to generate a depth value after time-of-flight correction; Step S5, based on Snell's law and the geometric model, calculate the axial depth offset caused by refraction at the air-water interface and update the Z-axis coordinate of the point; Step S6, calculating the radial horizontal offset caused by refraction and adjusting the X and Y axis coordinates of the point; In step S7, all point cloud data are traversed, and the correction formulas of steps S4 to S6 are applied to generate corrected three-dimensional terrain coordinates to achieve dynamic reconstruction and monitoring of underwater terrain.

2. The underwater topography measurement method according to claim 1, characterized in that: In step S1, the depth camera is fixed above the physical model experimental area, with its axis perpendicular to the surface to be measured, and is synchronously arranged with the water level meter to collect water level parameters.

3. The underwater topography measurement method according to claim 1, characterized in that: In step S3, the Gaussian filter uses a two-dimensional Gaussian kernel function for smoothing, and the outlier removal is performed by calculating the average distance between each point in the point cloud and the nearest neighboring point, combined with a global statistical threshold to determine and remove sparse outliers.

4. The underwater topography measurement method according to claim 1, characterized in that: In step S4, the flight time error correction formula is: where d pixel is the original measured depth recorded by the depth camera, c air and c water are the speed of light in air and water, d air is the distance of the air layer.

5. The underwater topography measurement method according to claim 1, characterized in that: In step S5, the axial depth offset calculation formula is: Among them, θ a is the incident angle in air, θ w is the angle of refraction in water, d water For water depth.

6. The underwater topography measurement method according to claim 1, characterized in that: In step S6, the radial offset calculation formula is: <h2 style=";text-align:left;direction:ltr">d<h2 style=";text-align:left;direction:ltr"> r <h2 style=";text-align:left;direction:ltr"> (tanθ)<h2 style=";text-align:left;direction:ltr"> a <h2 style=";text-align:left;direction:ltr"> -tanθ<h2 style=";text-align:left;direction:ltr"> w <h2 style=";text-align:left;direction:ltr"> )d<h2 style=";text-align:left;direction:ltr"> water And based on the direction vector, the offset is assigned to the X and Y axes and the horizontal coordinate is updated.

7. The underwater topography measurement method according to claim 1, characterized in that: The depth camera is an infrared depth camera, which is suitable for dynamically collecting point cloud data of the underwater scouring surface and synchronously associating it with the water level meter data.

8. The underwater topography measurement method according to claim 1, characterized in that: The method is suitable for real-time monitoring of scour terrain in small-scale hydraulic physical model experiments, including three-dimensional dynamic reconstruction of local scour such as seawalls, riverbeds, and bridge piers.

Images