A method for denoising underwater terrain data based on triangulation

Abstract

A method for denoising underwater terrain data based on triangulation includes the following steps: step 1, performing tidal level correction, sound velocity correction and attitude correction on the collected underwater terrain point cloud data to complete data preprocessing; step 2, gridding the point cloud data and dividing the area according to the terrain characteristics; step 3, in each grid, selecting the median elevation as the sample point, triangulating based on the grid sample points, and calculating the error distance from the point to the triangular surface; step 4, calculating the adaptive threshold according to multi-level error range statistics and dynamic weight optimization; step 5, judging and eliminating noise points according to the threshold, and outputting the purified point cloud data. The present invention combines the dynamic adjustment of the threshold by region, the multi-level error range statistics and the dynamic weight optimization to achieve fast and efficient noise elimination.

Description

Technical Field

[0001] The invention relates to the technical field of underwater terrain data processing, and in particular to an underwater terrain data denoising method based on triangulation. Background Art

[0002] Underwater topographic survey is widely used in the preliminary engineering volume calculation and later engineering acceptance of river dredging and land reclamation projects. Currently, single-beam and multi-beam echo sounders are mostly used, which use the transmitting array of the transducer array to emit fan-shaped sound waves to the bottom of the water, and receive the signal reflected from the bottom of the water through the receiving array for calculation to obtain the water depth data of each point in the measurement area. The collected data is affected by many factors such as turbidity of the water area, the noise of the instrument itself, and the interference caused by the carrying ship, which will generate a lot of noise data. Therefore, noise reduction processing is required after data registration.

[0003] Manufacturers of multi-beam echo sounders generally provide supporting data processing software. In addition, professional software such as QIMERA and CARIS provide data processing functions with richer functions and compatibility with multi-vendor equipment. These software also provide basic data noise reduction filtering algorithms, but in actual applications, there are many types of filters, the filter parameter settings are too complicated, and improper parameters will cause the data to be too smooth or retain too much noise; the effect of some filters depends on the preprocessing of the data; boundary effects may occur, and there is a lack of sufficient neighborhood points at the edge of the data for calculation, resulting in information distortion at the boundary or data discontinuity at the boundary, which will lead to the inability to effectively remove noise, which is particularly obvious in the processing of terrain data. These problems have not been adopted by surveying and mapping engineers in actual work.

[0004] Data denoising has been widely studied as a universal engineering data processing problem. Commonly used algorithms include adaptive filtering, surface fitting, normal correction, etc. The multi-beam point cloud filtering algorithm based on threshold adaptive determination proposed by Shen Wei removes significant outlier noise points by using virtual grid secondary hierarchical statistics combined with k-Medoids slope clustering, but it cannot well preserve the terrain information at the edge of steep slopes and slopes; Wang Jun et al. proposed to use the moving least squares method (MLS) to denoise the point cloud, and fit the smooth surface by finding the least squares solution in the local area. This algorithm requires dense sampling and complex local processing, and is not suitable for situations where large amounts of data need to be processed quickly; Lu Yuren et al. proposed a bilateral filtering point cloud denoising process based on normal correction, which constructs the best connected domain through spatial cell division, and then combines normal correction and bilateral filtering technology to accurately remove noise, but the denoising effect depends largely on the accurate estimation of the initial normal. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an underwater terrain data denoising method based on triangulation, which combines regional dynamic adjustment of thresholds, multi-level error range statistics and dynamic weight optimization to achieve fast and efficient noise removal.

[0006] In order to solve the above technical problems, the present invention is solved by the following technical solutions

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for denoising underwater terrain data based on triangulation, comprising the following steps:

[0009] Step 1: Perform tide level correction, sound velocity correction and attitude correction on the collected underwater terrain point cloud data to complete data preprocessing;

[0010] Step 2: Grid the point cloud data and divide the area according to the terrain features;

[0011] Step 3: In each grid, select feature points as sample points, perform triangulation based on the grid sample points, and calculate the error distance from the point to the triangular surface;

[0012] Step 4: Calculate the adaptive threshold based on multi-level error range statistics and dynamic weight optimization;

[0013] Step 5: Determine and remove noise points according to the threshold value, and output denoised point cloud data.

[0014] Preferably, the area division is based on terrain relief and point density, and is divided into different areas according to the following rules:

[0015] Steep area: the undulation is greater than the set threshold;

[0016] Flat area: the fluctuation is lower than the set threshold and the point density is moderate;

[0017] Sparse area: low point density;

[0018] Among them, the terrain relief and point density are calculated as follows:

[0019] ; density

[0020] Where Z is the elevation of the measuring point, Zmax is digits, Zmin is digits, N is the total number of points in the grid, and A is the area of ​​the grid.

[0021] Preferably, the feature point is the point with the minimum distance to other points, and the feature point P i The distance to other points and Si The calculation formula is:

[0022] .

[0023] Preferably, the grid size is dynamically adjusted according to the noise density, point density and terrain undulation, and the calculation formula is:

[0024]

[0025] in: , It is an adjustment coefficient used to balance the effects of density and fluctuation.

[0026] Preferably, the error distance is calculated as follows: traverse all points in the grid and calculate the error distance from the point to the triangle surface:

[0027]

[0028] Among them, (a, b, c, d) are the plane equation coefficients of the triangle patch, ( , , ) are the coordinates of the current point.

[0029] Preferably, the multi-level error range statistics includes the following steps:

[0030] Calculate the mean value μ of the error distance:

[0031]

[0032] Then define multiple error ranges. Specifically, define the point ratio of the first error range [0, μ] :

[0033]

[0034] Define the second error range [μ,μ+ ] :

[0035]

[0036] in, is the standard deviation, is the total number of points, , are the number of points in each range.

[0037] Preferably, the adaptive threshold is calculated by the following formula:

[0038]

[0039] in, , ; , is the initial weight, , is the weight adjustment coefficient.

[0040] Preferably, the noise removal comprises the following steps: if the error distance of a point If the threshold is exceeded, it is marked as noise and removed.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The present invention provides a method for denoising underwater terrain data based on triangulation, which combines dynamic adjustment of thresholds by region, multi-level error range statistics and dynamic weight optimization to achieve fast and efficient noise removal.

[0043] The present invention dynamically adjusts the size of the grid by analyzing the density, undulation and distribution characteristics of point cloud data to optimize subsequent calculation efficiency and accuracy.

[0044] The present invention can dynamically adjust the threshold according to regional characteristics, adapt to complex terrain, accurately identify noise points through multi-level error range statistics combined with dynamic weight optimization; and accelerate large-scale data calculations by combining balanced grid partitioning and multi-threading technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0047] Figure 2 Schematic diagram of spectral image before denoising underwater terrain data Figure 1 ;

[0048] Figure 3 Schematic diagram of spectral image of underwater terrain data denoised by this method Figure 1 ;

[0049] Figure 4 Schematic diagram of spectral image before denoising underwater terrain data Figure 2 ;

[0050] Figure 5 Schematic diagram of spectral image of underwater terrain data denoised by this method Figure 2. DETAILED DESCRIPTION

[0051] The present invention is further described in detail below with reference to the accompanying drawings.

[0052] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be used for other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0053] See also Figure 1-Figure 5 , a method for denoising underwater terrain data based on triangulation, comprising the following steps:

[0054] Step 1: Perform tide level correction, sound velocity correction and attitude correction on the collected underwater terrain point cloud data to complete data preprocessing. The purpose is to correct the tide level, sound velocity and attitude errors in the collected point cloud data and provide standardized input for subsequent algorithms. Data preprocessing includes the following steps:

[0055] (1) Tide level correction:

[0056]

[0057] in, is the elevation of the measuring point, is the tide level correction value.

[0058] (2) Sound speed correction:

[0059] Correct the sound wave propagation error according to the sound velocity profile:

[0060]

[0061] in, and are the actual sound speed and the reference sound speed respectively.

[0062] Posture Correction:

[0063] Correction of measurement deviation caused by the tilt of the ship.

[0064] Step 2: Grid the point cloud data and divide the area according to the terrain characteristics.

[0065] 1. The purpose of gridding point cloud data is to divide massive data into small blocks (grids) to improve processing efficiency.

[0066] Gridding method: evenly divide the point cloud data on the XY plane:

[0067]

[0068] in, , The grid size should be adjusted according to the data characteristics (recommended range: 0.2 to 1000 meters)

[0069] 2. Regional division: Divide into different areas based on terrain undulation and point density.

[0070] Setting of terrain relief threshold in regional division: This application uses elevation data to represent terrain relief, takes the maximum elevation difference as the measure of relief, and sets a relief threshold (i.e. 50 meters). If the maximum elevation difference of an area is greater than 50 meters, it is divided into a steep area; otherwise, it is further judged based on point density;

[0071] Further setting of the point density threshold: point density is measured by calculating the number of points per unit area, and a moderate point density range is set (i.e., 10 to 50 points per square meter). If the point density of an area falls within this range, it is classified as a flat area; if it is lower than the minimum value of this range (i.e., less than 10 points per square meter), the point density is low, and it is classified as a sparse area;

[0072] According to the above rules, it is divided into different areas:

[0073] Steep area: the undulation is greater than the set threshold;

[0074] Flat area: the fluctuation is lower than the set threshold and the point density is moderate;

[0075] Sparse area: low point density;

[0076] The terrain relief is calculated as:

[0077]

[0078] The point density is calculated as:

[0079] Density

[0080] Where Z is the elevation of the measuring point, Zmax is digits, Zmin is The number of digits, N is the total number of points in the grid, and A is the area of ​​the grid;

[0081] The max(Z) and min(Z) are selected to avoid the influence of noise.

[0082] Step 3: In each grid, select the median elevation as the sample point, perform triangulation based on the grid sample points, and calculate the error distance from the point to the triangle surface.

[0083] 1. Sample point selection: In each grid, feature points are selected as sample points to avoid the influence of extreme points.

[0084] The feature point is the minimum distance to other points, and the feature point P i The distance to other points and S i The calculation formula is:

[0085] .

[0086] 2. Triangulation: Use sample points to perform triangulation and generate triangular patches that meet the following conditions:

[0087] Empty circumcircle: There are no other points in the circumcircle of the triangle;

[0088] Maximize Minimum Angle: Avoids generating long and thin triangles.

[0089] 3. Error distance calculation, traverse all points in the grid and calculate the error distance from the point to the triangle surface:

[0090]

[0091] Among them, (a, b, c, d) are the plane equation coefficients of the triangle patch, ( , , ) are the coordinates of the current point.

[0092] Step 4: Calculate the adaptive threshold based on multi-level error range statistics and dynamic weight optimization.

[0093] 1. Multi-level error range statistics

[0094] The multi-level error range statistics include the following steps:

[0095] (1) Calculate the average value μ of the error distance:

[0096]

[0097] (2) Define multiple error ranges, specifically:

[0098] Define the ratio of points in the first error range [0,μ] :

[0099]

[0100] Define the second error range [μ,μ+ ] :

[0101]

[0102] in, is the standard deviation, is the total number of points, , are the number of points in each range.

[0103] 2. Dynamic weight optimization

[0104] according to , , dynamically calculate weights , :

[0105] ,

[0106] in, , is the initial weight, , is the weight adjustment coefficient.

[0107] 3. Adaptive Threshold Formula

[0108] Calculate the threshold based on the weight and error range ratio:

[0109]

[0110] 4. Regional optimization

[0111] Adjust weight parameters in different regions:

[0112] Steep areas:

[0113] ,

[0114] Flat area:

[0115] ,

[0116] It should be noted that the determination of initial weights and initial adjustment coefficients is as follows:

[0117] 1. Initial weight , Determination of:

[0118] Rule of thumb: Set initial values ​​based on data characteristics or refer to similar methods in the literature.

[0119] Experimental tuning: Gradually find the optimal value through experimental adjustment.

[0120] Data statistical analysis: Estimate initial weights based on the mean and standard deviation of data errors.

[0121] 2. Initial weight adjustment coefficient , Determination of:

[0122] Rule of thumb: The weight adjustment coefficient is verified through empirical setting or experimentation.

[0123] Regression Analysis: The coefficients are determined through regression analysis.

[0124] Dynamic optimization: Dynamic weight adjustment coefficients are optimized through iterative optimization.

[0125] Step 5: Determine and remove noise points based on the threshold, and output the purified point cloud data.

[0126] Noise point judgment: For each point, if >Threshold, mark as noise.

[0127] Remove noise: Remove noise points and retain the purified data points.

[0128] Save results: Output the purified point cloud data as a grid for subsequent projects.

[0129] The present invention can dynamically adjust the threshold according to regional characteristics, adapt to complex terrain, accurately identify noise points through multi-level error range statistics combined with dynamic weight optimization, and accelerate large-scale data calculations by combining balanced grid partitioning and multi-threading technology.

[0130] The present invention analyzes the density, fluctuation and distribution characteristics of point cloud data and dynamically adjusts the size of the grid to optimize subsequent calculation efficiency and accuracy, thereby achieving better technical effects.

[0131] Specifically, the program framework is as follows:

[0132] Analyze data characteristics:

[0133] Point density: Calculate the distribution density of data points and identify high-density and low-density areas.

[0134] Terrain Relief: Identifies areas of steep and flat terrain by looking at the range of elevation values.

[0135] Set the grid adjustment rules:

[0136] Smaller meshes are used in high-density areas to ensure that local details are preserved.

[0137] Smaller grids are used in areas with large terrain undulations, and complex areas are depicted in detail.

[0138] Larger grids are used in low-density and flat areas to improve computational efficiency.

[0139] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

Claims

1. A method for denoising underwater terrain data based on triangulation, characterized in that: The following steps are involved: Step 1: Perform tide level correction, sound velocity correction and attitude correction on the collected underwater terrain point cloud data to complete data preprocessing; Step 2: Grid the point cloud data and divide the area according to the terrain features; Step 3: In each grid, select feature points as sample points, perform triangulation based on the grid sample points, and calculate the error distance from the point to the triangular surface; Step 4: Calculate the adaptive threshold based on multi-level error range statistics and dynamic weight optimization; Step 5: Determine and remove noise points according to the threshold value, and output denoised point cloud data; The multi-level error range statistics include the following steps: Calculate the mean value μ of the error distance: , Then define multiple error ranges. Specifically, define the point ratio of the first error range [0, μ] : , Define the second error range [μ,μ+ ] : , in, is the standard deviation, is the total number of points, , are the number of points in each range respectively; The adaptive threshold is calculated by the following formula: , in, , ; , is the initial weight, , is the weight adjustment coefficient.

2. The method for denoising underwater terrain data based on triangulation according to claim 1, characterized in that: The area division is based on terrain relief and point density, and is divided into different areas according to the following rules: Steep area: the undulation is greater than the set threshold; Flat area: the fluctuation is lower than the set threshold and the point density is moderate; Sparse area: low point density; Among them, the terrain relief and point density are calculated as follows: ; density ; Where Z is the elevation of the measuring point, Zmax is digits, Zmin is digits, N is the total number of points in the grid, and A is the area of ​​the grid.

3. The method for denoising underwater terrain data based on triangulation according to claim 2, characterized in that: The feature point is the minimum distance to other points, and the feature point P i The distance to other points and S i The calculation formula is: .

4. The method for denoising underwater terrain data based on triangulation according to claim 1, characterized in that: The grid size is dynamically adjusted according to the noise density, point density and terrain relief, and its calculation formula is: ,in: , It is an adjustment coefficient used to balance the effects of density and fluctuation.

5. The method for denoising underwater terrain data based on triangulation according to claim 1, characterized in that: The error distance is calculated as follows: traverse all points in the grid and calculate the error distance from the point to the triangle surface: , where (a, b, c, d) are the plane equation coefficients of the triangle patch, ( , , ) are the coordinates of the current point.

6. The method for denoising underwater terrain data based on triangulation according to claim 1, characterized in that: The noise point removal comprises the following steps: if the error distance of a point If the threshold is exceeded, it is marked as noise and removed.

Images