A method for denoising a shallow water single-photon point cloud

Abstract

The present application belongs to the field of single photon water depth detection, and particularly relates to a shallow water single photon point cloud denoising method, comprising: identifying water surface single photon point cloud from the shallow water single photon point cloud to be denoised according to the elevation and performing fine denoising thereon, and obtaining water bottom single photon point cloud to be denoised from the shallow water single photon point cloud; performing signal point number Gaussian function fitting on the water bottom single photon point cloud in the elevation dimension, determining the elevation difference corresponding to a certain confidence interval as the semi-minor axis of the elliptical search domain in the OPTICS algorithm; determining the horizontal feature distance sorting list according to the distribution characteristics of the water bottom single photon point cloud to be denoised; taking each horizontal feature distance in the list as the candidate semi-major axis of the elliptical search domain, determining the value of the semi-major axis based on the MinPts threshold parameter calculation formula, in combination with the semi-minor axis and the preset MinPts threshold parameter value; and inputting the value of the semi-minor axis and the semi-major axis into the OPTICS algorithm to realize denoising of the water bottom single photon point cloud. The present application improves the denoising effect of the shallow water single photon point cloud.

Description

Technical Field

[0001] This invention belongs to the field of single-photon depth detection, and more specifically, relates to a method for denoising single-photon point clouds in shallow water. Background Technology

[0002] Shallow water depth measurement is fundamental to shallow water resource exploration. High-resolution depth sounding data not only contributes to research on the ecological environment and climate change, but also plays a crucial role in environmental monitoring and protection, as well as the rational utilization of resources. Traditional depth sounding methods are conducted via shipboard or airborne methods, but these are inefficient and costly to acquire data. Compared to traditional methods, spaceborne single-photon lidar offers advantages such as wide coverage, high repetition rate, and less susceptibility to external factors, demonstrating enormous potential for shallow water depth sounding.

[0003] In recent years, the detection technology of spaceborne single-photon lidar systems has gradually matured, but the processing of its point cloud data is still in the research stage. Due to environmental factors such as solar background and atmospheric scattering, as well as the influence of the system itself, the received point cloud data contains a large amount of noise signals, which makes it difficult to extract elevation information. In recent years, researchers have carried out research and innovation on the basis of traditional denoising filtering algorithms, forming a number of photon point cloud denoising algorithms to better extract elevation information.

[0004] In 2012, Magruder et al. proposed an improved Canny edge detection denoising algorithm. This algorithm converts the profile photon point cloud into a two-dimensional raster image and then uses image processing techniques to remove noise points. Such algorithms can remove photon noise to a certain extent, but some useful information is lost during the conversion of the photon point cloud into a raster image, reducing the accuracy of the denoising algorithm. In 2017, Xie Feng et al. proposed a denoising algorithm based on an elliptic denoising kernel, which calculates the point density of the photon point cloud and sets parameter thresholds to separate signal from noise. This algorithm is simple in principle, fast in calculation, and performs well in low-noise conditions, but it is ineffective when dealing with high-density clustered noise points. In 2019, Ma et al. proposed a density clustering denoising algorithm based on DBSCAN, which classifies the photon point cloud into signal photons and noise photons using density clustering. This algorithm has good overall denoising performance, but it is ineffective in complex terrain areas. In 2022, Xi Xiaohuan et al. proposed a single-photon point cloud denoising algorithm based on OPTICS, which uses OPTICS clustering to divide photon point cloud data into signal photons and noise photons based on the density and distance distribution characteristics of photon point cloud data. This method has high denoising accuracy and performs well in both strong background noise and complex regions, thus it has been widely applied to the processing of single-photon point cloud data in shallow water. However, this algorithm exhibits poor denoising performance in areas near the seabed. Summary of the Invention

[0005] To address the shortcomings and improvement needs of existing technologies, this invention provides a denoising method for single-photon point clouds in shallow water areas, with the aim of improving the overall denoising effect of single-photon point clouds in shallow water areas.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for denoising single-photon point clouds in shallow water is provided, comprising:

[0007] S1. Based on the fact that the elevation of the water surface is higher than the elevation of the bottom, the single-photon point cloud of the water surface is coarsely identified from the single-photon point cloud of the shallow water area to be denoised and then finely denoised to obtain the denoised single-photon point cloud of the water surface.

[0008] S2. Set the signal point corresponding to the lowest elevation in the denoised water surface single-photon point cloud as the threshold, and take the single-photon signal points in the shallow water single-photon point cloud that are lower than the threshold as the single-photon signal points of the bottom water, so as to obtain the bottom single-photon point cloud to be denoised.

[0009] S3. Perform Gaussian function fitting on the elevation dimension of the underwater single-photon point cloud to be denoised to determine the elevation difference corresponding to the preset confidence interval, which is used as the semi-minor axis of the elliptical search domain in the OPTICS algorithm; determine the horizontal feature distance sorting list based on the distribution characteristics of the underwater single-photon point cloud to be denoised, which represents the horizontal distance sorting list between any signal point in the underwater single-photon point cloud and other signal points; use each horizontal feature distance in the sorting list as a candidate semi-major axis of the elliptical search domain in the OPTICS algorithm, and finally determine the value of the semi-major axis based on the MinPts threshold parameter calculation formula, combined with the semi-minor axis and the preset MinPts threshold parameter value;

[0010] S4. Input the values ​​of the semi-minor axis and semi-major axis into the OPTICS algorithm, and use the OPTICS algorithm to denoise the underwater single-photon point cloud to be denoised, thus completing the denoising of the shallow water single-photon point cloud.

[0011] Furthermore, before executing S1, the method also includes:

[0012] A coarse overall denoising is performed on the shallow water single-photon point cloud to be denoised using a confidence parameter, wherein the confidence parameter is the confidence parameter provided by the acquisition system when acquiring the shallow water single-photon point cloud.

[0013] Furthermore, S1 is implemented as follows:

[0014] Based on the elevation information of the signal points, an elevation distribution histogram of the shallow water single-photon point cloud to be denoised is established. Since the elevation of the water surface is higher than that of the bottom, the elevation interval corresponding to the water surface in the elevation distribution histogram is initially determined. The distribution histogram of this elevation interval is fitted with a Gaussian function, and the single-photon point cloud in the first preset confidence interval is retained as the denoised water surface single-photon point cloud.

[0015] Furthermore, the first preset signal interval is [μ-3σ,μ+3σ], where μ is the mean elevation of all signal points within the elevation interval corresponding to the water surface, and σ is the variance of the elevation of all signal points within the elevation interval corresponding to the water surface.

[0016] Furthermore, S3 includes:

[0017] Through scene adaptation, the value of the semi-minor axis b is determined. Specifically, the underwater single-photon point cloud is divided into multiple segments along the track direction. For each segment, a Gaussian function is fitted to the number of signal points in the elevation dimension. The elevation difference in the second preset signal interval is selected as the vertical distance h of the underwater point cloud contour of that segment. The median of the h values ​​obtained from all segments is taken as the elevation difference h between the upper and lower boundaries of the underwater terrain contour, and the value of h / 2 is taken as the value of the semi-minor axis b.

[0018] Through scene adaptation, the value of the major semi-axis 'a' is determined. Specifically, the K-means nearest neighbor method is used to determine list D. eps Each signal point in the underwater single-photon point cloud is numbered, and the horizontal distance between any two signal points is calculated to construct the horizontal distance distribution matrix D of the underwater single-photon point cloud. nn D nn It is an n x n real symmetric matrix, where n is the total number of signal points in the underwater single-photon point cloud, and the elements D in the matrix are... nn (i, j) is the horizontal distance between the i-th signal point and the j-th signal point; arranging each row of the horizontal distance distribution matrix Dnn in ascending order yields... right Calculate the average of each column to obtain the sorted list D of horizontal feature distances. eps This list is sorted in ascending order of the horizontal distances from any point p in the underwater single-photon point cloud to other points; list D... eps Each value in the equation is used as a candidate semi-major axis a. The threshold parameter value of MinPts is calculated using each candidate semi-major axis a and semi-minor axis b. The candidate semi-major axis a corresponding to the threshold parameter value of MinPts is used as the semi-major axis a of the elliptical search domain of the OPTICS algorithm.

[0019] Furthermore, the second preset confidence interval is a 95% confidence interval.

[0020] Furthermore, the preset value of the threshold parameter of MinPts is 4.

[0021] Furthermore, the formula for calculating the MinPts threshold parameter is as follows:

[0022]

[0023] In the formula, S1 represents the total number of photons in the elliptical search domain; N1 represents the total number of signal points in the underwater single-photon point cloud to be denoised; h and l represent the maximum difference in vertical distance and the maximum difference in distance along the track in the single-photon point cloud, respectively; S2 represents the number of noise photons in the elliptical search domain; ceil represents the rounding function; N2 is the number of signal points within the range h2 above the lowest water depth; h2 is 5m.

[0024] The present invention also provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when executed by a processor, controls the device containing the storage medium to perform a denoising method for shallow water single-photon point clouds as described above.

[0025] In summary, the above-described technical solutions conceived in this invention can achieve the following beneficial effects:

[0026] (1) To address the problem of poor denoising performance of existing OPTICS algorithms in the vicinity of the seabed, this invention proposes a parametrically adaptive OPTICS single-photon point cloud denoising algorithm. The algorithm obtains more suitable input parameters through scene adaptation, namely the semi-minor axis and semi-major axis of the elliptical search domain in the OPTICS algorithm, to ensure the measurement accuracy of photon distribution parameters in the vicinity of the seabed, thereby improving the denoising effect in the vicinity of the seabed. Overall, this invention can improve the denoising effect of single-photon point clouds in shallow water.

[0027] (2) The present invention also proposes to first use a confidence parameter to perform overall coarse denoising on the shallow water single photon point cloud to be denoised. The confidence parameter is the confidence parameter provided by the acquisition system when acquiring the shallow water single photon point cloud. After coarse denoising, the water surface is identified and denoised, which can improve the overall denoising effect.

[0028] (3) When setting the semi-minor axis for scene adaptation, this invention considers that some underwater contours have large undulations. For such unstable underwater contours, directly fitting the signal point number Gaussian function of the point cloud in the elevation dimension will result in a mixed Gaussian distribution, which leads to uncertainty in the elevation difference within the preset information interval and cannot well represent the elevation difference between the upper and lower interfaces of the underwater contour. This invention proposes the idea of ​​segmentation, dividing the underwater single-photon point cloud into multiple segments along the track direction. For each segment of the point cloud, Gaussian function fitting of the signal point number is performed in the elevation dimension. The elevation difference in the second preset information interval is selected as the vertical distance h of the underwater point cloud contour of that segment. The median of the h values ​​obtained from all segments is taken as the elevation difference h between the upper and lower boundaries of the underwater terrain contour, and the value of h / 2 is taken as the value of the semi-minor axis b. This method can more accurately represent the elevation difference between the upper and lower interfaces of the underwater contour.

[0029] (4) When setting the semi-major axis for scene adaptation, this invention considers that a value that is too small will lead to incomplete denoising, while a value that is too large will result in the loss of some underwater contour information. This invention combines the K-mean nearest neighbor method and the calculation formula of MinPts, and selects the minimum value of a corresponding to MinPts of 4 as the value of the semi-major axis, thereby achieving a better denoising effect without losing underwater contour information. Attached Figure Description

[0030] Figure 1 A flowchart illustrating a method for denoising single-photon point clouds in shallow water, provided in an embodiment of the present invention;

[0031] Figure 2 A schematic flowchart of a method for denoising single-photon point clouds in shallow water provided by an embodiment of the present invention;

[0032] Figure 3 These are single-photon point cloud denoising effect images for different shallow water scenes provided in the embodiments of the present invention, wherein (a), (b), (c), and (d) correspond to different datasets. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0034] Example 1

[0035] A method for denoising single-photon point clouds in shallow water, such as... Figure 1 As shown, it includes:

[0036] S1. Based on the fact that the elevation of the water surface is higher than the elevation of the bottom, the single-photon point cloud of the water surface is coarsely identified from the single-photon point cloud of the shallow water area to be denoised and then finely denoised to obtain the denoised single-photon point cloud of the water surface.

[0037] S2. Set the signal point corresponding to the lowest elevation in the denoised water surface single-photon point cloud as the threshold, and take the single-photon signal points in the shallow water single-photon point cloud below the threshold as the single-photon signal points of the bottom water, so as to obtain the bottom single-photon point cloud to be denoised.

[0038] S3. Through scene adaptation, determine the input parameters of the OPTICS algorithm, namely the semi-minor axis b and semi-major axis a of the elliptic search domain:

[0039] To denoise the underwater single-photon point cloud, a Gaussian function is fitted to the number of signal points along the elevation dimension to determine the elevation difference corresponding to a pre-set confidence interval, which serves as the semi-minor axis of the elliptical search domain in the OPTICS algorithm. Based on the distribution characteristics of the underwater single-photon point cloud to be denoised, a horizontal feature distance ranking list is determined, representing the ranking list of horizontal distances between any signal point in the underwater single-photon point cloud and other signal points. Each horizontal feature distance in the ranking list is used as a candidate semi-major axis of the elliptical search domain in the OPTICS algorithm. Based on the MinPts threshold parameter calculation formula, combined with the semi-minor axis and the pre-set MinPts threshold parameter value, the value of the semi-major axis is finally determined.

[0040] S4. Input the values ​​of the semi-minor axis and semi-major axis into the OPTICS algorithm, and use the OPTICS algorithm to denoise the underwater single-photon point cloud to be denoised, thus completing the denoising of the shallow water single-photon point cloud.

[0041] The OPTICS algorithm is explained below:

[0042] (1) Concepts of elliptic search domain and algorithm input parameters

[0043] The OPTICS algorithm uses the input point as the center and determines the distribution parameters of the input point through an ellipse of a certain size and a threshold parameter. The region covered by the ellipse is called the elliptical search domain. The semi-minor axis *b*, semi-major axis *a*, and the threshold parameter *MinPts* are the three input parameters of the OPTICS algorithm. Among the three input parameters, the threshold parameter *MinPts* can be obtained from *b* and *a*, as shown in the formula:

[0044]

[0045] In the formula: S1 is the total number of photons in the elliptical search domain; N1 is the total number of photons in the given sample; h and l are the maximum differences between the vertical distance and the distance along the track in the sample photon point cloud, respectively; S2 is the number of noise photons in the elliptical search domain; ceil represents the floor function. N2 is the number of signal points within a range of h2 upwards from the lowest water depth; h2 is 5m.

[0046] Therefore, the OPTICS algorithm can be considered to have only two input parameters: the semi-minor axis b and the semi-major axis a of the elliptic search domain.

[0047] (2) Denoising process of OPTICS algorithm

[0048] Substituting the obtained input parameters into the OPTICS algorithm (in this embodiment, the input parameter values ​​are obtained through scene adaptation), the core distance list CD and reachable distance list RD of the underwater photon point cloud can be obtained. Then, the reachable distance threshold RD of the remaining point cloud data is obtained using the Otsu's method (OTSU). th If RD is less than RD th Photons are considered signal points, while other points are considered noise points, thus achieving the separation of noise and signal.

[0049] Among them, the concepts of core distance CD and reachability distance RD.

[0050] If the number of photon points within the elliptical search domain centered at point p is greater than or equal to the density threshold MinPts, then point p is considered a core point. If point p is a core point, the minimum distance that makes point p a core point is called the core distance of point p, denoted by CD(p). RD represents the reachability distance. If p is a core object, then RD(q, p) is the larger of the distance dist(p, q) between sample point q and core point p and the core distance CD(p); when p is not a core point, RD(q, p) is meaningless. CD and RD are shown in the formula:

[0051]

[0052]

[0053] In the formula, N ab (p) represents the number of points in the neighborhood centered at p, with a and b as the major and minor axes. The definition of dist(p, q) is as follows:

[0054]

[0055] Where x represents the distance along the track, y represents the elevation value, and a and b represent the semi-major axis and semi-minor axis of the ellipse, respectively.

[0056] For photon point cloud data, each input point corresponds to a reachable distance RD. A larger RD value indicates a greater distance from other points, while a smaller RD value indicates a closer distance. Based on the characteristics of the RD list corresponding to photon point cloud data, the Otsu's method (OTSU) is used to obtain the reachable distance threshold RD. th If the reachable distance RD corresponding to sample point p is less than RD th If the signal is positive, then sample point p is a signal point; otherwise, sample point p is a noise point. This can be achieved by passing through the reachable distance threshold RD. th Ultimately, this achieves the separation of signal points and noise points.

[0057] To address the issue of poor denoising performance of existing OPTICS algorithms in the vicinity of the underwater surface, this embodiment provides a parametrically adaptive OPTICS single-photon point cloud denoising algorithm, such as... Figure 2 As shown, the overall approach is as follows: First, the point clouds of the water surface and the bottom are processed separately, separating the photons of the water surface and the bottom, and extracting the signal photons of the water surface. The differences in the water surface and bottom environments result in different noise levels near their contours, thus necessitating separate processing of the photons from the water surface and the bottom. For the photons of the water surface, a general denoising method is used, which is not limited to any specific method. For the photons of the bottom, the input parameters of the OPTICS algorithm, namely the major and minor semi-axes of the elliptical search domain, are first obtained through scene adaptation. Then, the adaptively obtained input parameters are substituted into the OPTICS algorithm for fine denoising.

[0058] This embodiment addresses the issue of poor denoising performance of existing OPTICS algorithms in areas near the seabed by proposing a parametrically adaptive OPTICS single-photon point cloud denoising algorithm. The algorithm adapts to the scene to obtain more suitable input parameters, specifically the semi-minor and semi-major axes of the elliptical search domain in the OPTICS algorithm, ensuring the accuracy of photon distribution parameter measurements in areas near the seabed and thus improving denoising performance in these areas. Overall, this invention can improve the denoising effect of single-photon point clouds in shallow water.

[0059] As a preferred embodiment, before performing S1, the method further includes:

[0060] A confidence parameter is used to perform overall coarse denoising on the shallow water single-photon point cloud to be denoised. The confidence parameter is the one provided by the acquisition system when acquiring the shallow water single-photon point cloud. For example, the confidence parameter provided with the ATL03 dataset is used to perform overall coarse denoising on the photon point cloud in the ATL03 dataset.

[0061] As a preferred implementation, the above-mentioned S1 is implemented as follows:

[0062] Based on the elevation information of the signal points, an elevation distribution histogram of the shallow water single-photon point cloud to be denoised is established. Since the elevation of the water surface is higher than that of the bottom, the elevation interval corresponding to the water surface in the elevation distribution histogram is initially determined. The distribution histogram of this elevation interval is fitted with a Gaussian function, and the single-photon point cloud in the first preset confidence interval is retained as the finely denoised water surface single-photon point cloud.

[0063] As a preferred embodiment, the first preset signal interval is [μ-3σ,μ+3σ], where μ is the mean elevation of all signal points within the elevation interval corresponding to the water surface, and σ is the variance of the elevation of all signal points within the elevation interval corresponding to the water surface.

[0064] In other words, to execute the method of this embodiment, the elevation distribution histogram of the overall photon point cloud is first required. From the elevation information corresponding to the histogram, it can be seen that higher elevation peaks correspond to surface photons, and lower peaks correspond to bottom photons. The mean μ and variance σ of the photon point cloud concentrated on the surface are obtained by fitting the elevation distribution histogram with a Gaussian function. Then, photon point clouds with elevations in the range of [μ-3σ, μ+3σ] are retained as signal points for surface photons. Finally, the elevation corresponding to the lowest point among the surface photon signal points is selected as the boundary point between the surface and the bottom. The photon point cloud below the boundary point is considered bottom photon.

[0065] As a preferred embodiment, S3 above includes:

[0066] Through scene adaptation, the value of the semi-minor axis b is determined. Specifically, the underwater single-photon point cloud is divided into multiple segments along the track direction. For each segment, a Gaussian function is fitted to the number of signal points in the elevation dimension. The elevation difference of the second preset signal interval is selected as the vertical distance h of the underwater point cloud contour of that segment. The median of the h values ​​obtained from all segments is taken as the elevation difference h between the upper and lower boundaries of the underwater topographic contour, and the value of h / 2 is taken as the value of the semi-minor axis b.

[0067] In other words, the semi-minor axis b is obtained by dividing the point cloud data into multiple segments and estimating the vertical distance h of the underwater contour piecewise. Each segment can be fitted with a Gaussian function, and the width of the 95% confidence interval can be selected as the value of h for that segment. To eliminate outliers that may be obtained from some segments, the median of the results obtained from multiple segments is taken as the width h of the underwater terrain contour, and the value of h / 2 is taken as the value of the semi-major axis b.

[0068] Through scene adaptation, the value of the major semi-axis 'a' is determined. Specifically, the K-means nearest neighbor method is used to determine list D. eps Each signal point in the underwater single-photon point cloud is numbered, and the horizontal distance between any two signal points (i.e., the distance along the track direction) is calculated to construct the horizontal distance distribution matrix D of the underwater single-photon point cloud.nn D nn It is an n x n real symmetric matrix, where n is the total number of signal points in the underwater single-photon point cloud, and the elements D in the matrix are... nn (i, j) is the horizontal distance between the i-th signal point and the j-th signal point; the horizontal distance distribution matrix D nn Sort each row in ascending order to get right Calculate the average of each column to obtain the sorted list D of horizontal feature distances. eps This list is sorted in ascending order of the horizontal distances from any point p in the underwater single-photon point cloud to other points; list D... eps Each value in the equation is used as a candidate semi-major axis a. The threshold parameter value of MinPts is calculated using each candidate semi-major axis a and semi-minor axis b. The candidate semi-major axis a corresponding to the threshold parameter value of MinPts is used as the semi-major axis a of the elliptical search domain of the OPTICS algorithm.

[0069] In other words, obtaining the semi-major axis 'a': First, obtain list D using the K-means nearest neighbor method. eps This list is equivalent to an ascending sort of the horizontal distances from a point p to other points. This list is used as the parameter list D for the semi-major axis a. a D a Substituting each value in the table and the known semi-minor axis b into the threshold parameter formula of MinPts, we can obtain the threshold parameter list D of MinPts. m According to D m and D a It can obtain the minimum value of 'a' corresponding to different MinPts. It can find the minimum value of 'a' corresponding to MinPts equal to 4 (empirical value) and use it as the value of the semi-major axis of the elliptic search domain.

[0070] The formula for calculating the MinPts threshold parameter is as follows:

[0071]

[0072] In the formula, S1 represents the total number of photons in the elliptical search domain; N1 represents the total number of signal points in the underwater single-photon point cloud to be denoised; h and l represent the maximum difference in vertical distance and the maximum difference in distance along the track in the single-photon point cloud, respectively; S2 represents the number of noise photons in the elliptical search domain; ceil represents the rounding function; N2 is the number of signal points within the range h2 above the lowest water depth; h2 is 5m.

[0073] This embodiment uses photon point cloud data from the Yongle Atoll area in the South China Sea for experimental verification, providing high-precision point cloud images for subsequent water depth measurement and underwater topographic mapping.

[0074] Figure 3This paper demonstrates the denoising effect of this embodiment on photon point clouds under different scenarios. As shown in the figures, this embodiment can achieve denoising of photon point clouds in various scenarios with good results. The average F-value of this embodiment in different scenarios reaches 97.52%, which is higher than the average F-value of 97.26% of the existing OPTICS algorithm, indicating an improvement in overall denoising performance. Furthermore, the average cohesion of this embodiment in the underwater region is 0.0322, which is lower than the average cohesion of 0.0373 of the existing OPTICS algorithm, resulting in fewer noise points and better denoising performance in the underwater region.

[0075] Example 2

[0076] A computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed by a processor, it controls the device where the storage medium is located to perform a denoising method for shallow water single-photon point clouds as described above.

[0077] The relevant technical solutions are the same as in Embodiment 1, and will not be repeated here.

[0078] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for denoising shallow water single-photon point clouds, the method comprising: include: S1. Based on the fact that the elevation of the water surface is higher than the elevation of the bottom, the single-photon point cloud of the water surface is coarsely identified from the single-photon point cloud of the shallow water area to be denoised and then finely denoised to obtain the denoised single-photon point cloud of the water surface. S2. Set the signal point corresponding to the lowest elevation in the denoised water surface single-photon point cloud as the threshold, and take the single-photon signal points in the shallow water single-photon point cloud that are lower than the threshold as the single-photon signal points of the bottom water, so as to obtain the bottom single-photon point cloud to be denoised. S3. Perform Gaussian function fitting on the elevation dimension of the underwater single-photon point cloud to be denoised to determine the elevation difference corresponding to the preset confidence interval, which is used as the semi-minor axis of the elliptical search domain in the OPTICS algorithm; determine the horizontal feature distance sorting list based on the distribution characteristics of the underwater single-photon point cloud to be denoised, which represents the horizontal distance sorting list between any signal point in the underwater single-photon point cloud and other signal points; use each horizontal feature distance in the sorting list as a candidate semi-major axis of the elliptical search domain in the OPTICS algorithm, and finally determine the value of the semi-major axis based on the MinPts threshold parameter calculation formula, combined with the semi-minor axis and the preset MinPts threshold parameter value; S4. Input the values ​​of the semi-minor axis and semi-major axis into the OPTICS algorithm, and use the OPTICS algorithm to denoise the underwater single-photon point cloud to be denoised, thus completing the denoising of the shallow water single-photon point cloud. Wherein, S3 includes: Determine the shorter semi-axis through scene adaptation. b Specifically, the value of is determined by dividing the underwater single-photon point cloud into multiple segments along the track direction, fitting a Gaussian function to the number of signal points in each segment along the elevation dimension, and selecting the elevation difference of the second preset signal interval as the vertical distance of the underwater point cloud profile for that segment. h Values; obtained for all segments h The median value is taken as the elevation difference between the upper and lower boundaries of the underwater topographic profile. h and take h The value of / 2 is used as the semi-minor axis. b The value of ; Through scene adaptation, the value of the major semi-axis 'a' is determined. Specifically, the K-means nearest neighbor method is used to determine the list. Each signal point in the underwater single-photon point cloud is numbered, the horizontal distance between each pair of signal points is calculated, and a horizontal distance distribution matrix of the underwater single-photon point cloud is constructed. ,in yes n OK n A column of real symmetric matrices. n The total number of signal points in the underwater single-photon point cloud, and the elements in the matrix. It is the first i The signal point to the j The horizontal distance between each signal point; the horizontal distance distribution matrix Sort each row in ascending order to get ;right Calculate the average of each column to obtain a sorted list of horizontal feature distances. Use this list as any point in the underwater single-photon point cloud. p Sort the list in ascending order of horizontal distances to other points; Each value in the matrix is ​​used as a candidate semi-major axis. a Using each candidate semi-major axis a With semi-short axis b Calculate the threshold parameter value of MinPts; select the candidate semi-major axis corresponding to the threshold parameter value of MinPts as a preset value. a The semi-major axis of the elliptic search domain in the OPTICS algorithm a .

2. The method of claim 1, wherein, Before executing S1, the method also includes: A coarse overall denoising is performed on the shallow water single-photon point cloud to be denoised using a confidence parameter, wherein the confidence parameter is the confidence parameter provided by the acquisition system when acquiring the shallow water single-photon point cloud.

3. The method of claim 1, wherein, The implementation of S1 is as follows: Based on the elevation information of the signal points, an elevation distribution histogram of the shallow water single-photon point cloud to be denoised is established. Since the elevation of the water surface is higher than that of the bottom, the elevation interval corresponding to the water surface in the elevation distribution histogram is initially determined. The distribution histogram of this elevation interval is fitted with a Gaussian function, and the single-photon point cloud in the first preset confidence interval is retained as the denoised water surface single-photon point cloud.

4. The method of claim 3, wherein, The first preset signal interval is [0.5, 1.5], wherein, is the mean of the elevations of all signal points in the elevation interval corresponding to the water surface, is the variance of the elevations of all signal points in the elevation interval corresponding to the water surface, is the variance of the elevations of all signal points in the elevation interval corresponding to the water surface.

5. The method of claim 1, wherein, The second preset confidence interval is a 95% confidence interval.

6. The method of claim 1, wherein, The preset value for the threshold parameter of MinPts is 4.

7. The method of claim 1, wherein, The formula for calculating the MinPts threshold parameter is as follows: ； In the formula, represents the total number of photons within the elliptical search domain; represents the total number of signal points in the single-photon point cloud to be denoised; h and l respectively represent the maximum difference of the vertical distance and the maximum difference of the along-track distance in the single-photon point cloud; represents the number of noise photons within the elliptical search domain; ceil represents the rounding function; is the number of signal points in the range from the lowest water depth to the water surface; is the number of signal points in the range from the lowest water depth to the water surface; is 5 m.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is run by a processor, it controls the device on which the storage medium is located to perform a denoising method for shallow water single-photon point clouds as described in any one of claims 1 to 7.

Images