ICESat-2 shallow sea sounding signal extraction method and system

Abstract

The invention belongs to the technical field of ocean remote sensing, and discloses an ICESat-2 shallow sea sounding signal extraction method and system, and the method comprises the steps: carrying out the data preprocessing based on the obtained ICESat-2 sounding data; constructing an elliptical neighborhood direction adaptive model based on submarine topographic features, and determining the orientation of an elliptical neighborhood; constructing an elliptical neighborhood size adaptive model considering the depth change, and determining the size of an elliptical neighborhood; refraction correction and tide correction are carried out, and water depth correction of the seabed photon points is completed. Through the density clustering algorithm, accurate detection of the underwater topography of the shallow sea water area is realized, and a new analysis thought is provided for shallow sea water depth detection. The submarine photon point extraction problem caused by non-uniform point cloud density and weak connectivity of the photon counting laser radar ICESat-2 is solved, the sounding precision is remarkably improved, and the long-term requirements of the marine surveying and mapping field on the high-precision and low-cost sounding technology are met.

Description

Technical Field

[0001] The present invention belongs to the field of marine remote sensing technology, and in particular relates to an ICESat-2 shallow sea bathymetric signal extraction method and system. Background Art

[0002] Photon counting (also known as "single-photon") lidar is a new type of lidar technology that uses a high-repetition-rate, low-energy laser emitter and an extremely sensitive single-photon detector to acquire three-dimensional information about the target. Compared to linear lidar, photon counting lidar has a detection sensitivity 2 to 3 orders of magnitude higher, enabling higher resolution and longer detection range. ICESat-2 (Ice, Cloud and land Elevation Satellite-2), a representative single-photon lidar system, was successfully launched by NASA (National Aeronautics and Space Administration) in 2018. It is equipped with the photon counting lidar ATLAS (Advanced Topographic Laser Altimeter System), which can obtain photon point cloud data with higher density and smaller spot size to achieve precise extraction of three-dimensional spatial information of the detected target. ICESat-2 was originally designed to monitor polar ice sheets and global land elevation changes. Fortunately, the ATLAS system's laser pulse wavelength is in the "ocean window" (532nm), allowing some photons to penetrate the water and reach the bottom, enabling it to demonstrate significant performance in ocean remote sensing applications such as shallow water depth measurement and water quality inversion. When using ICESat-2 for seafloor topography measurements, the effective target signals are sea surface signals and seafloor signals. However, due to the high sensitivity of the photon counting lidar system, it is also affected by device factors and environmental factors (such as the atmosphere and water bodies), resulting in a large number of noise photons, which increases the complexity of data processing.

[0003] Existing methods mainly include the density-based spatial clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise), which divides the point cloud into different clusters by spatial density to achieve noise photon removal, but the effect of extracting seabed signals with uneven density and weak connectivity is limited, especially in deep water areas. The directional rotating ellipse model uses a fixed angle as the rotation step to search for the best ellipse neighborhood to extract seabed photon points. This method can extract photon signals well in specific seabed terrain, but the connectivity is weak in deep water areas and the accuracy is reduced. The adaptive elliptical filtering method is used to accurately detect signal photons on the water surface and bottom, and dynamically adjusts the filter shape to adapt to changes in photon density, but it still faces challenges in areas with drastic density changes.

[0004] Through the above analysis, the problems and defects of the existing technology are as follows:

[0005] (1) Due to the uneven density and weak connectivity of ICESat-2 photon counting LiDAR data, it is difficult to accurately extract sea surface and seabed photon signals based solely on proximity distance or density characteristics.

[0006] (2) Existing density-based spatial clustering algorithms have limited effectiveness in extracting seabed signals with uneven density and weak connectivity, especially in deep water areas.

[0007] (3) The existing directional rotating ellipse model has weak connectivity in deep water areas and its accuracy has decreased, while the adaptive ellipse filtering method still faces challenges in areas with drastic density changes. Summary of the Invention

[0008] To overcome the problems of low extraction accuracy of seabed photon signals in complex terrain areas and weak connectivity of effective seabed photon points in related technologies, the present invention discloses an ICESat-2 shallow-water bathymetric signal extraction method and system, and more particularly, an ICESat-2 shallow-water bathymetric signal extraction method and system based on a size- and orientation-adaptive elliptical neighborhood. The technical solution is as follows:

[0009] The present invention is implemented as follows: a method for extracting ICESat-2 shallow water sounding signals, comprising the following steps:

[0010] S1, data preprocessing based on the acquired ICESat-2 bathymetric data;

[0011] S2, constructing an elliptical neighborhood direction adaptive model based on the seabed topography characteristics to determine the orientation of the elliptical neighborhood;

[0012] S3, constructing an ellipse neighborhood size adaptive model that takes into account depth changes and determining the size of the ellipse neighborhood;

[0013] S4, performs refraction correction and tide correction to complete the water depth correction of the seabed photon point.

[0014] In step S1, data preprocessing is performed based on the acquired ICESat-2 bathymetric data, including:

[0015] Set the distance threshold ε and the point number threshold MinPts, and use the DBSCAN algorithm to divide the original photon points into water surface photon points and secondary photon points; use the obtained water surface photon points to determine the average sea surface elevation H surface , the secondary photon points are divided into above-water photon points and underwater photon points;

[0016] Design a strategy for separating deep and shallow water areas based on grid point density, and calculate the total grid density Q and the grid point density per row Q i ; When Q>Q i When , the photon points in the grid row will be defined as deep water area photon points, and the DBSCAN algorithm will be used for denoising; otherwise, they will be defined as shallow water area photon points, and the subsequent water body outlier filtering strategy based on grid point density will be carried out;

[0017] Design a water body outlier filtering strategy based on grid point density, and calculate the point density Q corresponding to each column of grids and each grid in the shallow water area. j and Q ij ; When Q j >Q ij When , the photon points in the corresponding grid are removed as outliers; otherwise, they are retained.

[0018] Further, calculate the total grid density Q and the grid point density per row Q i The calculation formula is:

[0019]

[0020] Where C is the total number of grid photon points, x min ,x max are the distances along the track from the start to the end of the analysis area, y min ,y max are the minimum and maximum elevation data of the underwater point, C i is the total number of photon points in each row of the grid, i is the row number of each grid, and L is the designed grid size.

[0021] Furthermore, the density of photon points per grid in shallow water area is Q j And the point density Q corresponding to each grid ij The calculation formula is:

[0022]

[0023] Where j is the column number of each grid, C j is the total number of photon points in each column of the grid, y min ,y max are the minimum and maximum elevation data of the underwater point, C ij is the number of photon points in each grid, and L is the designed grid size.

[0024] In step S2, an elliptical neighborhood direction adaptive model based on seabed topography characteristics is constructed to determine the orientation of the elliptical neighborhood, including:

[0025] According to the window interval Δx, the area where the underwater point is located is divided into N yellow windows with equal distances along the x direction; from shallow water to deep water, the photon point cloud within each window interval is used to fit and solve the linear regression equation of each window in turn;

[0026] Calculate the correlation coefficient R between the photon point in the window and the fitting equation 2 , evaluate the degree of agreement between the two; set R threshold =0.6, when R 2 >R threshold When , the elliptical neighborhood window direction θ will be set to arctan(k); otherwise, the window interval is set to Δx / 2, and the regression equation and its correlation coefficient R in the new window are recalculated. 2 , repeat this process until the correlation coefficient meets the threshold requirement.

[0027] Furthermore, the window division and the linear regression equation of each window are expressed as:

[0028]

[0029] y=kx+y0

[0030] Where N is the number of windows, Dis is the distance along the track, Δx is the window interval, k is the slope of the linear regression equation, x is the lateral distance, y is the elevation, and y0 is the intercept of the line on the y-axis.

[0031] In step S3, an ellipse neighborhood size adaptive model that takes into account depth variation is constructed to determine the size of the ellipse neighborhood, including:

[0032] Based on the preprocessed data, the underwater points are divided into n depth intervals according to their elevation;

[0033] Use consistent ellipse neighborhood size parameters within each depth interval.

[0034] Furthermore, the expression of the ellipse neighborhood size parameter is:

[0035]

[0036] Where m is the depth interval number as the depth increases; r0, r n are the ellipse neighborhood parameters for the depth intervals when m=1 and m=n, respectively; u is 1 / 2 of the vertical size of the depth interval, and v is 1 / 4 of the vertical size of the depth interval, which are used to determine the ellipse neighborhood size parameters in the shallowest and deepest depth intervals, respectively; r m is the elliptical neighborhood parameter of the depth interval when m=2…n-1; n is determined according to the vertical span of the photon point cloud data, and n=5 is set; h is a custom coefficient with a value range of (1,+∞) and is determined according to the survey area data; a m is the size of the major axis of the ellipse neighborhood, b m is the size of the minor axis of the ellipse neighborhood, y min ,y max They are the minimum and maximum elevation data of underwater points respectively.

[0037] In step S4, refraction correction and tide correction are performed to complete the water depth correction of the seabed photon point, including:

[0038] When laser propagates in air and water, due to the different refractive indices of the two media, the refractive index of air is 1.00029 and the refractive index of water is 1.33, the laser will refract when passing through the air-water interface; according to Snell's law, that is:

[0039] n1sinθ1=n2sinθ2

[0040] Where n1, n2 are the refractive indices of air and water respectively, θ1 is the laser incident angle, and θ2 is the refraction angle;

[0041] Calculate the deflection angle of the laser's propagation direction after entering the water. Because refraction changes the laser's propagation path, the laser's propagation path must be corrected to accurately determine the true depth of the seafloor. Laser propagation speed in water is slower than in air, so the time of flight must be corrected based on the difference in propagation speed between the media. The refraction angle is combined to calculate the laser's propagation distance in water, and the depth information of the seafloor point is corrected based on the geometry of the optical path.

[0042] Query the current day's tide information for the study area and normalize the tide level to mean sea level. Using tidal forecasts or measured tide data, obtain the tide height corresponding to each laser measurement time. Then, perform a tidal correction on the seafloor photon point's height. This correction process normalizes all bathymetric data to mean sea level by subtracting or adding the corresponding tide height. This process effectively eliminates the effects of tides on bathymetric results and improves the accuracy of seafloor topography data.

[0043] Another object of the present invention is to provide an ICESat-2 shallow water sounding signal extraction system, which is used to control the ICESat-2 shallow water sounding signal extraction method. The system includes:

[0044] Data preprocessing module, used for data preprocessing based on acquired ICESat-2 bathymetric data;

[0045] The ellipse neighborhood orientation determination module is used to construct an ellipse neighborhood orientation adaptive model based on the seabed topography characteristics and determine the orientation of the ellipse neighborhood;

[0046] The ellipse neighborhood size calculation module is used to build an ellipse neighborhood size adaptive model that takes into account depth changes and determine the size of the ellipse neighborhood;

[0047] The seabed photon point water depth correction module is used to perform refraction correction and tide correction to complete the water depth correction of the seabed photon point.

[0048] In combination with all the above technical solutions, the beneficial effects of the present invention are as follows:

[0049] This paper improves the DBSCAN algorithm by replacing its original circular neighborhood with an elliptical search neighborhood that adapts to photon point cloud density and depth. Furthermore, it constructs an elliptical neighborhood orientation-adaptive model based on seafloor topography characteristics and an elliptical neighborhood size-adaptive model that accounts for depth variations. This method for extracting ICESat-2 shallow-water bathymetric signals, based on size- and orientation-adaptive elliptical neighborhoods, addresses the challenge of extracting seafloor photon points caused by the uneven density and weak connectivity of ICESat-2 photon-counting lidar point clouds, overcoming current challenges in seafloor topography measurements in shallow waters near offshore islands and reefs.

[0050] This method enables accurate detection of shallow-water underwater topography using only a density clustering algorithm, providing a new analytical approach for shallow-water depth detection. It also addresses the challenge of extracting seafloor photon points caused by the uneven density and weak connectivity of ICESat-2 photon-counting lidar point clouds, providing a solid data foundation for the application of multispectral satellite imagery in shallow-water depth inversion for offshore islands and reefs.

[0051] The present invention proposes an efficient and accurate method for extracting shallow-water bathymetric signals from ICESat-2, which can be widely used in fields such as ocean mapping, waterway surveying, marine ecological monitoring, and island and reef measurement. Through the improved adaptive elliptical neighborhood algorithm and DBSCAN classification method, the accuracy and adaptability of bathymetric signal extraction are improved, errors are reduced, and the degree of automation of data processing is improved. By constructing an elliptical neighborhood direction adaptive model based on seabed topography characteristics and an elliptical neighborhood size adaptive model that takes into account depth changes, and combining the DBSCAN algorithm with a grid point density-based strategy, the present invention achieves accurate classification and extraction of photon points, thereby improving the extraction accuracy of shallow-water bathymetric signals. The present invention solves the problems of low accuracy, large errors, and poor adaptability in shallow-water bathymetric signal extraction based on ICESat-2 data. Traditional shallow-water bathymetric methods are limited by water depth changes, refraction effects, tidal changes, and noise interference, resulting in an unstable signal extraction process. The present invention innovatively introduces an elliptical neighborhood adaptive method, so that the extraction of bathymetric signals no longer relies on fixed parameters, but can be dynamically adjusted with changes in seabed topography and water depth, thereby improving adaptability and reliability. Combining DBSCAN and grid point density analysis, breakthroughs have also been made in noise filtering and identification of water anomalies, significantly improving bathymetric accuracy and meeting the long-standing demand for high-precision, low-cost bathymetric technology in the field of marine surveying and mapping. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure;

[0053] Figure 1 This is a flow chart of a method for extracting ICESat-2 shallow-water bathymetric signals provided by an embodiment of the present invention;

[0054] Figure 2 This is a flowchart of the adaptive elliptical neighborhood window model denoising provided by an embodiment of the present invention;

[0055] Figure 3 is a flow chart of a data preprocessing method provided by an embodiment of the present invention; wherein (a) represents the distribution of ICESat-2 photon point clouds, and (b) represents a schematic diagram of a water body outlier filtering strategy based on grid point density;

[0056] Figure 4 Schematic diagram of an elliptical neighborhood direction adaptation model provided by an embodiment of the present invention; wherein (a) is a schematic diagram of elliptical neighborhood direction adaptation and window size, and (b) is a comparison diagram of the elliptical neighborhood adaptive direction and the initial horizontal direction;

[0057] Figure 5: This is a comparison chart of the denoising results of the Molokai Island data along the track and different algorithms provided by an embodiment of the present invention; among them, (a) is the distribution of the original single-photon point cloud along the track, (b) is the official confidence result of ATL03, (c) is the denoising result of the DBSCAN algorithm, (d) is the denoising result of the directional rotating elliptical neighborhood algorithm, and (e) is the denoising result of the adaptive rotating elliptical neighborhood algorithm of the present invention;

[0058] Figure 6 : These are cross-sectional comparison diagrams of the Molokai Island data and the ALB data provided by an embodiment of the present invention; wherein (a) is a cross-sectional comparison diagram using the DBSCAN algorithm, (b) is a cross-sectional comparison diagram using the directional rotated ellipse neighborhood algorithm, and (c) is a cross-sectional comparison diagram using the adaptive rotated ellipse neighborhood algorithm of the present invention;

[0059] Figure 7 3. The figures are comparative analysis diagrams of the results obtained when the elliptical neighborhood orientation provided by the embodiment of the present invention is the optimal orientation; (a) is the denoising result of the adaptively rotated elliptical neighborhood algorithm of the present invention, (b) is the denoising result of a clockwise rotation of π / 8, and (c) is the denoising result of a counterclockwise rotation of π / 8.

[0060] Figure 8 This is an error frequency curve distribution diagram when the elliptical neighborhood orientation provided by an embodiment of the present invention is the optimal orientation. DETAILED DESCRIPTION

[0061] To make the above-mentioned objects, features, and advantages of the present invention more readily apparent, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following description sets forth numerous specific details to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art may make similar modifications without departing from the scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0062] The innovation of this invention lies in: Based on the traditional DBSCAN algorithm, this method innovatively optimizes the fixed circular neighborhood into an elliptical neighborhood, enabling it to adaptively adjust the search range based on the density of the photon point cloud and changes in water depth, thereby improving the accuracy of bathymetric signal extraction. At the same time, an elliptical neighborhood direction adaptive model based on seabed topography is constructed, allowing the neighborhood direction to automatically adjust with changes in terrain, enhancing adaptability to different terrain structures. Furthermore, considering the impact of water depth on the distribution of bathymetric signal, this invention proposes an elliptical neighborhood size adaptive model that takes into account depth changes, allowing the neighborhood size to dynamically change with water depth, thereby optimizing bathymetric results in shallow and deep water areas.

[0063] This method also combines grid point density-based deep-shallow water area separation and water outlier filtering to optimize the DBSCAN clustering process, effectively removing water noise photon points and improving the reliability of the bathymetric data. Furthermore, the integration of refraction correction and tidal correction makes the water depth inversion results more accurate and consistent with the actual seabed topography.

[0064] Compared with the limitations of traditional bathymetric methods that rely on fixed parameters and empirical thresholds, the present invention can achieve high-precision seabed photon point extraction by relying solely on density clustering and adaptive neighborhood adjustment, providing a more intelligent and efficient solution for shallow water depth detection. It has important application value in the fields of ocean mapping, island and reef surveys, and waterway surveys.

[0065] Example 1, as Figure 1 As shown, the ICESat-2 shallow water sounding signal extraction method provided by the embodiment of the present invention specifically includes the following steps:

[0066] S1, data preprocessing based on the acquired ICESat-2 bathymetric data;

[0067] To effectively identify underwater signals and calculate water depth, data preprocessing is required on the raw photon point cloud. This aims to accurately distinguish above-water points, surface points, and underwater points, determine water surface elevation information, and perform coarse denoising of seafloor photon points. Given the uneven photon density of ICESat-2, this method divides the raw photon points into surface photon points, above-water photon points, and underwater photon points. Based on the distribution characteristics of water noise photons, the ICESat-2 underwater photon points are further subdivided into outlier points and local outlier points (noise points that are significantly deviated from and relatively close to seafloor photon points, respectively). The local outliers are then removed to complete the data preprocessing.

[0068] The data preprocessing diagram of the present invention is as follows Figure 3 As shown, Figure 3 (a) shows the distribution of ICESat-2 photon point cloud, Figure 3 (b) Schematic diagram of the strategy for filtering outlier points in water bodies based on grid point density.

[0069] The present invention performs data preprocessing based on the acquired ICESat-2 bathymetric data, specifically including:

[0070] Step 1.1: Use the DBSCAN algorithm (a reasonable distance threshold (ε) and point number threshold (MinPts) need to be set) to divide the original photon points into water surface photon points and secondary photon points, and use the obtained water surface photon points to determine the average sea surface elevation H surface , subdividing the sub-photon points into above-water photon points and underwater photon points.

[0071] Step 1.2: Design a strategy for separating deep and shallow water areas based on grid point density, and calculate the total grid density and the grid point density per row, Q and Q i , as shown in the innovative formula (1) proposed by the present invention. i When , the photon points in the grid row will be defined as deep water area photon points and denoised using the DBSCAN algorithm; otherwise, they will be defined as shallow water area photon points and a subsequent water body outlier filtering strategy based on grid point density will be implemented.

[0072]

[0073] Where C is the total number of grid photon points, x min ,x max are the distances along the track from the start to the end of the analysis area, y min ,y max are the minimum and maximum elevation data of the underwater point, C i is the total number of photon points in each row of the grid, i is the row number of each grid, and L is the designed grid size.

[0074] Step 1.3: A water body outlier filtering strategy based on grid point density is designed. The point density Q corresponding to each grid column and each grid is calculated in the shallow water area. i and Q ij , as shown in the innovative formula (2) proposed by the present invention. j >Q ij When , the photon points in the corresponding grid are removed as outliers; otherwise, they are retained.

[0075]

[0076] Where j is the column number of each grid, C j is the total number of photon points in each column of the grid, y min ,y max are the minimum and maximum elevation data of the underwater point, C ij is the number of photon points in each grid, and L is the designed grid size.

[0077] S2, constructing an elliptical neighborhood direction adaptive model based on the seabed topography characteristics to determine the orientation of the elliptical neighborhood;

[0078] In order to determine the orientation of the elliptical neighborhood, the photon point cloud is divided into windows with equal intervals along the x direction. By fitting the linear regression equation of the photon points in each window, the slope k is obtained and the window direction θ is determined based on it. The schematic diagram of the elliptical neighborhood direction adaptive model is shown in Figure 4 As shown, Figure 4 (a) is a schematic diagram of elliptical neighborhood direction adaptation and window size. Figure 4(b) is a comparison diagram of the adaptive direction of the ellipse neighborhood and the initial horizontal direction. The dotted ellipse is the horizontal ellipse neighborhood, and the solid ellipse is the ellipse neighborhood determined according to the slope of the straight line.

[0079] The present invention constructs an elliptical neighborhood direction adaptive model based on the seabed topography characteristics, and determines the orientation of the elliptical neighborhood specifically including:

[0080] Step 2.1: Divide the area where the underwater point is located into N yellow windows with equal distances along the x direction according to a certain interval Δx. Then, from shallow water to deep water, use the photon point cloud within each window interval to fit and solve the linear regression equation of each window in turn. The expression is:

[0081]

[0082] y=kx+y0 (4)

[0083] Where N is the number of windows, Dis is the distance along the track, Δx is the window interval, k is the slope of the linear regression equation, x is the lateral distance, y is the elevation, and y0 is the intercept of the line on the y-axis.

[0084] Step 2.2: Calculate the correlation coefficient (R) between the photon points in the window and the fitting equation 2 ) to evaluate the degree of agreement between the two. 2 >R threshold (R threshold =0.6), the elliptical neighborhood window direction θ will be set to arctan(k); otherwise, the window interval needs to be set to Δx / 2, and the regression equation and its correlation coefficient R in the new window need to be recalculated. 2 , repeat this process until the correlation coefficient meets the threshold requirement.

[0085] S3, construct an ellipse neighborhood size adaptive model that takes into account depth changes and determines the size of the ellipse neighborhood.

[0086] To determine the elliptical neighborhood size parameters, this paper proposes an adaptive elliptical neighborhood size model that accounts for depth variations. As the depth of the sea increases, underwater points become increasingly sparse, making it difficult to accurately extract seafloor photon points using fixed elliptical neighborhood window size parameters (a and b). Therefore, this paper divides the depth intervals into distinct elliptical neighborhood size parameters to determine their respective values.

[0087] The present invention constructs an elliptical neighborhood size adaptive model that takes into account depth changes, and determining the size of the elliptical neighborhood specifically includes:

[0088] Step 3.1: Based on data preprocessing, the underwater points are divided into n depth intervals according to elevation (n needs to be determined based on the vertical span of the photon point cloud data, and n=5 is set in the present invention).

[0089] Step 3.2: Use a consistent ellipse neighborhood size parameter in each depth interval, which is expressed as:

[0090]

[0091] Where m is the depth interval number as the depth increases; r0, r n are the ellipse neighborhood parameters for the depth intervals when m=1 and m=n, respectively; u is 1 / 2 of the vertical size of the depth interval, and v is 1 / 4 of the vertical size of the depth interval, which are used to determine the ellipse neighborhood size parameters in the shallowest and deepest depth intervals, respectively; r m is the elliptical neighborhood parameter of the depth interval when m=2…n-1; n is determined according to the vertical span of the photon point cloud data, and n=5 is set; h is a custom coefficient with a value range of (1,+∞) and is determined according to the survey area data; a m is the size of the major axis of the ellipse neighborhood, b m is the size of the minor axis of the ellipse neighborhood, y min ,y max They are the minimum and maximum elevation data of underwater points respectively.

[0092] S4, performs refraction correction and tide correction to complete the water depth correction of the seabed photon point.

[0093] Because the propagation speed of bathymetric lasers in air and water is inconsistent, they refract as they pass through the water. This creates a discrepancy between the seafloor topography displayed by the raw laser data and the actual seafloor position. Therefore, to obtain accurate seafloor depth information for photon points, water depth correction is required for the denoised photon points.

[0094] The present invention performs refraction correction and tide correction to complete the water depth correction of the seabed photon point, specifically including:

[0095] Step 4.1: Using the different refractive indices of laser light in air and water, and Snell's law of refraction, calculate the corrected depth information of the seabed point and perform refraction correction.

[0096] Step 4.2: Query the local tide information for the day and perform tide correction to obtain accurate seabed topography information.

[0097] In embodiment 2, the ICESat-2 shallow water sounding signal extraction system provided by the embodiment of the present invention includes:

[0098] The data preprocessing module is used to preprocess the acquired ICESat-2 bathymetric data using a grid point density-based strategy for separating deep and shallow water areas and a strategy for filtering outliers in the water body.

[0099] The ellipse neighborhood orientation determination module is used to construct an ellipse neighborhood orientation adaptive model based on the seabed topography characteristics and determine the orientation of the ellipse neighborhood;

[0100] The ellipse neighborhood size calculation module is used to build an ellipse neighborhood size adaptive model that takes into account depth changes and determine the size of the ellipse neighborhood;

[0101] The seabed photon point water depth correction module is used to use the different refractive indices of laser in air and water to perform refraction correction and tidal correction through Snell's refraction law to complete the water depth correction of the seabed photon point.

[0102] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0103] In order to verify the performance of the ICESat-2 shallow water bathymetric signal extraction method based on size and orientation adaptive elliptical neighborhood proposed in the present invention, the ICESat-2ATL03 (20230528gt1l) data set in the waters of Molokai Island, Hawaii was used for verification. The ATLAS system carried by ICESat-2 operates at an orbital altitude of about 500km, covering the area of the earth from 88°N north latitude to 88°S south latitude. The ATLAS system emits a green laser with a wavelength of 532nm, which has a strong ability to penetrate water and can detect a maximum depth of about 40m in extremely clear water. ICESat-2 emits 3 groups of 6 parallel laser beams in the form of each group containing one strong and one weak laser to obtain photon point cloud information. Among them, strong laser signals have stronger anti-noise interference capabilities and higher point density than weak laser signals, so the present invention mainly uses strong laser data for analysis. The ATL03 dataset is a Level-2 global positioning photon product provided by ICESat-2. It records the longitude, latitude, and ellipsoid height of a single photon, providing source data for the water depth detection of the present invention.

[0104] To demonstrate the effectiveness of the proposed method, the present invention uses the ATL03 official confidence level, the DBSCAN algorithm, and the oriented rotated elliptical neighborhood algorithm to compare the experimental results with the experimental results of the proposed algorithm. The accuracy and continuity of seabed topography extraction are analyzed from both qualitative and quantitative perspectives. The comparison data used is the ALB (Airborne LiDAR Bathymetry) data near Molokai Island. Figure 5 This is a comparison chart of the Molokai Island data along the track and the denoising results between different algorithms. Figure 5 (a) is the original single-photon point cloud distribution along the trajectory, Figure 5 (b) is the official confidence result of ATL03, Figure 5(c) is the denoising result of the DBSCAN algorithm. Figure 5 (d) is the denoising result of the directional rotating ellipse neighborhood algorithm. Figure 5 (e) is the denoising result of the algorithm of the present invention; Figure 6 This is a cross-section comparison of Molokai Island data and ALB data. Figure 6 (a) is the cross-section comparison diagram of the DBSCAN algorithm. Figure 6 (b) is a cross-section comparison diagram of the directional rotation ellipse neighborhood algorithm. Figure 6 (c) is a cross-sectional comparison diagram of the algorithm of the present invention.

[0105] The results were analyzed from a qualitative perspective. Figure 5 (b) is the official confidence result of ATL03, but it cannot accurately extract the effective photon signal from the seabed. Figure 5 (c) is the extraction result using the DBSCAN algorithm. Since the conventional circular neighborhood does not consider the direction and density changes of its seabed effective signals, the maximum extraction depth is only about 10m, resulting in some seabed signals in deep water areas (such as Figure 5 (c) The box part) is misjudged as a noise point, and important information about the seabed topography in the deep water area is lost. Figure 5 In (d), after using the directional rotating ellipse neighborhood algorithm, the extraction effect of single photons from the seabed in deep water areas is improved, but the continuity of the extraction results is poor (e.g. Figure 5 (boxed part in (d)), it is difficult to obtain a relatively complete single-photon point cloud of the seabed. Figure 5 (e) shows that the algorithm proposed in this invention can effectively remove most of the noise points when extracting dense signals in shallow water and sparse signals in deep water, and successfully retain the seabed single photons in the deep water area.

[0106] The results were analyzed quantitatively: Based on the ALB data near Molokai, the depth-corrected seafloor signal was used to evaluate the bathymetric effectiveness of the proposed algorithm. The root mean square error (RMSE) and mean absolute error (MAE) were calculated using equations (7) and (8). Table 1 shows the accuracy comparison results.

[0107] Table 1 Evaluation of water depth accuracy of ICESat-2 seabed point extraction

[0108]

[0109]

[0110] As can be seen from Table 1, the results obtained using the DBSCAN algorithm are inferior to the actual water depth values in terms of RMSE and MAE evaluation parameter results. However, when the water depth data obtained using the two elliptical neighborhood algorithms are compared, the above evaluation parameters are improved to varying degrees, and the accuracy of extracting valid seabed points is further enhanced. However, compared with the directional rotation elliptical neighborhood algorithm, the size-direction adaptive elliptical neighborhood algorithm proposed in this invention has further improved in all indicators. Figure 6 As can be seen, the main reason for the improved accuracy is likely due to the limited application of the directional rotating elliptical neighborhood algorithm in complex offshore terrain, which fails to accurately remove noise points. An overview of the entire table shows that the accuracy of seabed points extracted by the adaptive rotating elliptical neighborhood algorithm is 0.60m lower in RMSE and 0.28m lower in MAE than the DBSCAN algorithm. This means that the algorithm proposed in this paper can effectively remove noise point clouds while retaining the seabed signal point cloud data.

[0111] The algorithm proposed in this paper shows significant advantages in processing seabed point cloud data. By verifying that the elliptical neighborhood orientation determined by the proposed algorithm is the optimal orientation, this paper takes the Molokai Island data as an example and further analyzes the comparison with other rotation angles (such as clockwise and counterclockwise rotation of π / 8). It is found that the algorithm of this invention has higher accuracy in extracting seabed point cloud data. The specific results are as follows: Figure 7 As shown, Figure 7 (a) is the denoising result of the algorithm of the present invention, Figure 7 (b) is the denoising result after clockwise rotation of π / 8. Figure 7 (c) is the denoising result of counterclockwise rotation of π / 8. The elliptical neighborhood direction adaptive model proposed in this invention can intelligently adjust the orientation of the neighborhood according to the changes in the seabed terrain features. This feature enables the algorithm to capture specific underwater structures and terrain features more effectively. In comparison, the algorithms of clockwise rotation of π / 8 and counterclockwise rotation of π / 8 often cannot adapt to the changes in terrain when processing different terrain features, resulting in deviations in the extraction results. Figure 7 It can be seen that the continuity of the seabed point cloud extracted with a fixed rotation angle is poor and important terrain information may be missed. The algorithm proposed in this invention can better adapt to these changes by adaptively adjusting the orientation of the elliptical neighborhood, thereby improving the extraction accuracy. Figure 8 The error frequency distribution curves for three different seabed removal directions compared with ALB data are shown, along with an evaluation of bathymetric accuracy. In comparison, the adaptive rotating elliptical neighborhood algorithm yields the most concentrated error frequency distribution, achieving the highest RMSE of 0.53 m. This demonstrates that the directions determined by the proposed algorithm better adapt to variations in seabed topography and achieve higher extraction accuracy.

[0112] The above description is only a preferred specific implementation method of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with this technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for extracting ICESat-2 shallow water bathymetry signals, characterized in that: The method comprises the following steps: S1, data preprocessing based on the acquired ICESat-2 bathymetric data; S2, constructing an elliptical neighborhood direction adaptive model based on the seabed topography characteristics to determine the orientation of the elliptical neighborhood; S3, constructing an ellipse neighborhood size adaptive model that takes into account depth changes and determining the size of the ellipse neighborhood; S4, performs refraction correction and tide correction to complete the water depth correction of the seabed photon point.

2. The ICESat-2 shallow water sounding signal extraction method according to claim 1, characterized in that: In step S1, data preprocessing is performed based on the acquired ICESat-2 bathymetric data, including: Set the distance threshold ε and the point number threshold MinPts, and use the DBSCAN algorithm to divide the original photon points into water surface photon points and secondary photon points; use the obtained water surface photon points to determine the average sea surface elevation H surface , the secondary photon points are divided into above-water photon points and underwater photon points; Design a strategy for separating deep and shallow water areas based on grid point density, and calculate the total grid density Q and the grid point density per row Q i ; When Q>Q i When , the photon points in the grid row will be defined as deep water area photon points, and the DBSCAN algorithm will be used for denoising; otherwise, they will be defined as shallow water area photon points, and the subsequent water body outlier filtering strategy based on grid point density will be carried out; Design a water body outlier filtering strategy based on grid point density, and calculate the point density Q corresponding to each column of grids and each grid in the shallow water area. j and Q ij ; When Q j >Q ij When , the photon points in the corresponding grid are removed as outliers; otherwise, they are retained.

3. The ICESat-2 shallow water sounding signal extraction method according to claim 2, characterized in that: Calculate the total grid density Q and the grid point density per row Q i The calculation formula is: Where C is the total number of grid photon points, x min ,x max are the distances along the track from the start to the end of the analysis area, y min ,y max are the minimum and maximum elevation data of the underwater point, C i is the total number of photon points in each row of the grid, i is the row number of each grid, and L is the designed grid size.

4. The ICESat-2 shallow water sounding signal extraction method according to claim 2, characterized in that: Water body outlier filtering strategy based on grid point density, the grid point density per column of photon points in shallow water area Q j The point density Q corresponding to each grid ij The calculation formula is: Where j is the column number of each grid, C j is the total number of photon points in each column of the grid, y min ,y max are the minimum and maximum elevation data of the underwater point, C ij is the number of photon points in each grid, and L is the designed grid size.

5. The ICESat-2 shallow water sounding signal extraction method according to claim 1, characterized in that: In step S2, an elliptical neighborhood direction adaptive model based on seabed topography characteristics is constructed to determine the orientation of the elliptical neighborhood, including: According to the window interval Δx, the area where the underwater point is located is divided into N yellow windows with equal distances along the x direction; from shallow water to deep water, the photon point cloud within each window interval is used to fit and solve the linear regression equation of each window in turn; Calculate the correlation coefficient R between the photon point in the window and the fitting equation 2 , evaluate the degree of agreement between the two; set R threshold =0.6, when R 2 >R threshold When , the elliptical neighborhood window direction θ will be set to arctan(k); otherwise, the window interval is set to Δx / 2, and the regression equation and its correlation coefficient R in the new window are recalculated. 2 , repeat this process until the correlation coefficient meets the threshold requirement.

6. The ICESat-2 shallow water sounding signal extraction method according to claim 5, characterized in that: The window division and the linear regression equation expression of each window are: y=kx+y0 Where N is the number of windows, Dis is the distance along the track, Δx is the window interval, k is the slope of the linear regression equation, x is the lateral distance, y is the elevation, and y0 is the intercept of the line on the y-axis.

7. The ICESat-2 shallow water sounding signal extraction method according to claim 1, characterized in that: In step S3, an ellipse neighborhood size adaptive model that takes into account depth variation is constructed to determine the size of the ellipse neighborhood, including: Based on the preprocessed data, the underwater points are divided into n depth intervals according to their elevation; Use consistent ellipse neighborhood size parameters within each depth interval.

8. The ICESat-2 shallow water sounding signal extraction method according to claim 7, characterized in that: The expression for the ellipse neighborhood size parameter is: Where m is the depth interval number as the depth increases; r0, r n are the ellipse neighborhood parameters for the depth intervals when m=1 and m=n, respectively; u is 1 / 2 of the vertical size of the depth interval, and v is 1 / 4 of the vertical size of the depth interval, which are used to determine the ellipse neighborhood size parameters in the shallowest and deepest depth intervals, respectively; r m is the elliptical neighborhood parameter of the depth interval when m=2…n-1; n is determined according to the vertical span of the photon point cloud data, and n=5 is set; h is a custom coefficient with a value range of (1,+∞) and is determined according to the survey area data; a m is the size of the major axis of the ellipse neighborhood, b m is the size of the minor semi-axis of the ellipse neighborhood, y min ,y max They are the minimum and maximum elevation data of underwater points respectively.

9. The ICESat-2 shallow water sounding signal extraction method according to claim 1, characterized in that: In step S4, refraction correction and tide correction are performed to complete the water depth correction of the seabed photon point, including: When laser light propagates in air and water, due to the different refractive indices of the two media (air has a refractive index of 1.00029 and water has a refractive index of 1.33), the laser light will refract when passing through the air-water interface. According to Snell's law, we get: n1sinθ1=n2sinθ2 Where n1, n2 are the refractive indices of air and water respectively, θ1 is the laser incident angle, and θ2 is the refraction angle; Calculate the deflection angle of the laser's propagation direction after it enters the water body. Since refraction causes the laser propagation path to change, the laser propagation path needs to be corrected to accurately determine the true depth information of the seabed point.

10. An ICESat-2 shallow water bathymetric signal extraction system, characterized in that: The system is used to control the ICESat-2 shallow water sounding signal extraction method according to any one of claims 1 to 9, and the system comprises: Data preprocessing module, used for data preprocessing based on acquired ICESat-2 bathymetric data; The ellipse neighborhood orientation determination module is used to construct an ellipse neighborhood orientation adaptive model based on the seabed topography characteristics and determine the orientation of the ellipse neighborhood; The ellipse neighborhood size calculation module is used to build an ellipse neighborhood size adaptive model that takes into account depth changes and determine the size of the ellipse neighborhood; The seabed photon point water depth correction module is used to perform refraction correction and tide correction to complete the water depth correction of the seabed photon point.

Images