Submarine topography and land elevation matching method based on multi-source data fusion

Abstract

The invention provides a submarine topography and land elevation matching method based on multi-source data fusion, and belongs to the technical field of geographic information technology and ocean engineering cross fusion. Seamless splicing of sea and land topographies is realized through multi-source data preprocessing, topographic feature correlation extraction, dynamic compensation matching and adaptive weight fusion; according to the method, the problems of data heterogeneity, dynamic environment adaptability and insufficient feature association can be solved, the matching precision can reach + / -3cm, the processing efficiency is also improved, and the method is suitable for coastal zone modeling, ocean engineering and other scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of cross-integration of geographic information technology and marine engineering, and in particular to a method for matching seabed topography and land elevation based on multi-source data fusion. Background Art

[0002] Elevation matching of seabed topography and land can provide millimeter-level precision terrain data support for coastal digital twins, marine engineering design, ecological and environmental monitoring, and other fields, thereby promoting the leapfrog development of integrated land-sea geographic information processing technology. Currently, traditional land-sea topography matching methods have the following drawbacks:

[0003] 1. Data heterogeneity: When the Kriging interpolation method has a resolution difference greater than 5:1, the data fault rate at the land-sea boundary reaches 23% (for example, when 1m seabed data is merged with 0.2m land data);

[0004] 2. Dynamic environmental impact: Factors such as tides and ocean currents cause real-time changes in seabed topography. Existing algorithms do not consider the dynamic submergence characteristics of the intertidal zone. The matching accuracy can deviate by up to ±20 cm at high and low tide levels (e.g., measured data from a port project).

[0005] 3. Insufficient feature correlation: Traditional algorithms are based only on coordinate position matching and ignore the correlation of terrain features (such as the direction of the coastline, the extension relationship between submarine ridges and land bedrock), resulting in poor continuity of the terrain surface after matching.

[0006] Therefore, a seabed topography and land elevation matching method based on multi-source data fusion is proposed to solve the above problems. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method for matching seabed topography and land elevation based on multi-source data fusion, so as to at least solve the above problems.

[0008] The technical solution adopted in the present invention is as follows:

[0009] A method for matching seabed topography and land elevation based on multi-source data fusion, the method comprising the following steps:

[0010] Multi-source heterogeneous data preprocessing:

[0011] The Bursa-Wolf seven-parameter transformation model is used to unify the coordinate system of the seabed sonar discrete point cloud data and the land LiDAR raster data, and the transformation error is controlled within ±5cm.

[0012] Land LiDAR data were resampled to 1m resolution using edge-preserving bilinear interpolation to preserve the details of steep cliffs with slopes > 45°.

[0013] Terrain feature association extraction:

[0014] The Douglas-Peucker algorithm is used to extract inflection points where the curvature of the seabed contour lines changes by more than 15°, and the flow accumulation method is used to identify the apex of the seabed ridge.

[0015] Canny edge detection and chain code tracking are used to extract the concave and convex points of the land coastline with a curvature radius of less than 50m, and the land watershed intersection points are generated using the D8 algorithm;

[0016] Construct a distance and direction weighted feature correlation matrix and eliminate mismatched pairs using a density clustering algorithm;

[0017] Dynamic compensation matching:

[0018] The dynamic elevation correction value of the intertidal zone is calculated based on the harmonic analysis model calibrated with tide gauge data;

[0019] Taking effective feature points as control vertices, thin plate spline transformation is used to achieve elastic deformation matching;

[0020] Adaptive weight fusion and iterative optimization:

[0021] Assign data weights based on the distance between the target point and the coastline, and fuse to generate seamless terrain data;

[0022] The parameters are adjusted by RMSE feedback until the matching accuracy reaches within ±5cm. Furthermore, the formula of the Bursa-Wolf seven-parameter conversion model is:

[0023] ΔX=ΔX0+ε x X+ε y Y+ε z Z+mX

[0024] Among them, X, Y, Z are the coordinates of the source coordinate system, ΔX is the coordinate offset after conversion, ΔX0 is the translation parameter, ε x , ε y , ε z is the rotation parameter and m is the scale parameter.

[0025] Furthermore, the dynamic elevation correction function of the harmonic analysis model is:

[0026]

[0027] Among them, H i is the tidal amplitude, σ i is the angular frequency, θ i is the initial phase, δ i is the local tidal lag angle, the tidal waves include the dominant tidal waves M2, S2, and K1 with an amplitude ratio greater than 80%, and the model parameters are calibrated by historical data of tide gauge stations covering seasonal variations.

[0028] Furthermore, the deformation energy function of the thin plate spline transformation is:

[0029]

[0030] Where φ(r) is the radial basis function, and r = || xx k || is the Euclidean distance from the point to the control vertex, γ is the regularization parameter, and γ = 0, 05, w k To control the vertex weight, x is the two-dimensional coordinate of the point to be deformed, x k is the two-dimensional coordinate of the k-th control vertex, is the second-order Laplacian operator of the deformed surface z′(x,y), and m controls the total number of vertices.

[0031] Furthermore, the density clustering algorithm is DBSCAN, and its parameters include:

[0032] The neighborhood radius is dynamically adjusted based on the data resolution, with a default value of 50m;

[0033] The minimum number of neighborhood points is set to 5.

[0034] Furthermore, the iterative optimization includes the following steps:

[0035] When RMSE>10cm, the direction consistency threshold is gradually reduced to 20° to increase the number of valid feature matching pairs;

[0036] Increase the thin plate spline transformation to control vertex density by 20% in error concentration areas;

[0037] Update the weight coefficient of the distance decay function, and the optimization range is 0.3-0.7.

[0038] Furthermore, when applied to sea ice-covered areas, the dynamic compensation model further includes sea ice thickness and ice surface roughness correction terms:

[0039] Δh ice =k1·d ice +k2·r ice

[0040] Among them, d ice is the ice thickness, r ice is the roughness, k1∈[0.1,0.3], k2∈[0.05,0.15] are the empirical coefficients.

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

[0042] 1. Significantly improved accuracy:

[0043] The elevation matching accuracy of the land-sea interface has been improved from ±15cm of traditional methods to ±3cm, thus meeting the millimeter-level modeling requirements of marine engineering.

[0044] The dynamic compensation model effectively corrects the influence of tides, and the matching error is reduced at high and low tide levels.

[0045] 2. Enhanced robustness:

[0046] Supports multi-source heterogeneous data input (sonar, LiDAR, satellite remote sensing), with a data resolution difference tolerance of 10:1 (e.g., fusion of 1m seabed data with 0.1m land data);

[0047] The feature bidirectional matching and elastic transformation algorithm significantly improves the matching success rate of complex terrain (such as multiple islands and fjords), which is better than traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 It is a schematic diagram of the overall process of a method for matching seabed topography and land elevation based on multi-source data fusion proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The principles and features of the present invention are described below with reference to the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0051] Reference Figure 1 The present invention provides a method for matching seabed topography and land elevation based on multi-source data fusion, the method comprising the following steps:

[0052] Multi-source heterogeneous data preprocessing:

[0053] The Bursa-Wolf seven-parameter transformation model is used to unify the coordinate system of the seabed sonar discrete point cloud data and the land LiDAR raster data, and the transformation error is controlled within ±5cm.

[0054] Land LiDAR data were resampled to 1m resolution using edge-preserving bilinear interpolation to preserve the details of steep cliffs with slopes > 45°.

[0055] For example, the discrete point cloud data of seabed sonar belongs to the WGS84 coordinate system data, and the land LiDAR raster data belongs to the National 2000 coordinate system. The two coordinate systems have problems such as inconsistent coordinate resolution (10:1 tolerance) and incompatible data formats. The traditional method leads to 23% data faults in the boundary zone. By constructing a preprocessing system of seven-parameter transformation + bilinear interpolation, the coordinate transformation error can be controlled within ±5cm, and more than 95% of the terrain details (such as steep cliff slope information loss of <3%) can be retained; standardized feature descriptors (128-dimensional SIFT features) can also be designed to eliminate semantic faults caused by data format differences and achieve a 95% recognition rate of the extension relationship between seabed ridges and land bedrock. By preprocessing multi-source heterogeneous data, seamless cross-modal data splicing can be achieved.

[0056] Terrain feature association extraction:

[0057] The Douglas-Peucker algorithm is used to extract inflection points where the curvature of the seabed contour lines changes by more than 15°, and the flow accumulation method is used to identify the apex of the seabed ridge.

[0058] Canny edge detection and chain code tracking are used to extract the concave and convex points of the land coastline with a curvature radius of less than 50m, and the land watershed intersection points are generated using the D8 algorithm;

[0059] Construct a distance and direction weighted feature correlation matrix and eliminate mismatched pairs using a density clustering algorithm;

[0060] For example, traditional coordinate-level matching ignores the correlation of terrain features (such as the consistency of the direction of the concave and convex points of the coastline and the inflection points of the seabed isobaths), resulting in an 18% mismatching rate in multi-island and fjord areas, and a surface continuity index of only 0.62. A "distance-direction" two-factor correlation matrix (distance weight 60% + direction consistency 40%) can be proposed to reduce the mismatching rate from 15% to below 3% through two-way matching verification. A feature-driven elastic deformation model can also be constructed to make the curvature difference of the sea-land boundary zone <0.1%, and the success rate of complex terrain matching is increased from 57% to 97%. The specific two-way matching verification process is as follows: construct a feature correlation matrix M n×m , where n is the number of seabed feature points, m is the number of land feature points, and the matrix element M ij Represents the Euclidean distance and direction consistency weight of the feature point pair (Formula: M ij =ad ij +(1-α)θ ij α=0.6 is the distance weight coefficient).

[0061] Feature matching optimization is specifically as follows:

[0062] Bidirectional matching verification density filtering is used: DBSCAN clustering is performed on the matching point pairs, and isolated matching pairs are eliminated (neighborhood radius 50m, minimum number of points 5). The final effective matching pair retention rate reaches 83%.

[0063] For detecting inflection points on seabed topography, the Douglas-Peucker algorithm can be used to simplify isobaths and extract inflection points with a curvature change greater than 15° (for example, the extraction rate for inflection points on the 100m isobath reaches 92%). For ridge vertex identification, the flow accumulation method can be used to determine the seabed ridgeline and extract local elevation maxima (with an elevation accuracy of ±10cm). For coastline feature points in land terrain, Canny edge detection and chain code tracking can be combined to extract concave and convex points with a coastline curvature radius less than 50m (such as capes and bay vertices). For ridgeline intersections, the D8 algorithm can be used to extract land watersheds, with sub-pixel positioning accuracy of less than 0.5m.

[0064] Dynamic compensation matching:

[0065] The dynamic elevation correction value of the intertidal zone is calculated based on the harmonic analysis model calibrated with tide gauge data;

[0066] Taking effective feature points as control vertices, thin plate spline transformation is used to achieve elastic deformation matching;

[0067] For example, dynamic factors such as tides (±20cm error) and ocean currents (±30cm error) lead to terrain matching deviations, making traditional rigid transformations unable to adapt to the periodic inundation of intertidal zones and the real-time disturbance of seafloor topography. Harmonic analysis combined with thin plate spline (TPS) elastic transformations can be introduced to achieve ±3cm accuracy correction at high and low tides, reducing tidal matching errors by 80%. A multi-factor dynamic compensation framework (compatible with tide, sea ice, and sediment parameters) can also be established, reducing the RMSE from ±8cm to ±4.5cm in complex scenarios such as polar ice zones and estuarine siltation areas.

[0068] Adaptive weight fusion and iterative optimization:

[0069] Assign data weights based on the distance between the target point and the coastline, and fuse to generate seamless terrain data;

[0070] The parameters are adjusted by the root mean square error (RMSE) feedback until the matching accuracy reaches within ±5cm.

[0071] For example, the distance weighted fusion is specifically as follows:

[0072] Introducing the logarithmic distance decay model

[0073] Where d(x,y) is the distance from the target point to the coastline, and λ = 0.5 is the attenuation coefficient, achieving:

[0074] 0-100m from the coastline: land data weight 0.9-1 (LiDAR high-precision data is preferred);

[0075] 100-500m: weight linear transition (0.7-0.9 on land, 0.1-0.3 on seabed);

[0076] ≥500m: Seabed data weight 0.8-1 (depends on the wide coverage of sonar data).

[0077] The specific accuracy verification and iterative optimization are as follows:

[0078] The root mean square error (RMSE) between the fused terrain and the measured data is calculated. If the RMSE is greater than 5 cm, the feature matching threshold and elastic transformation parameters are automatically adjusted until the accuracy requirements are met.

[0079] Adjust the feature matching threshold (e.g., reduce the threshold from 30° to 20°);

[0080] Optimize TPS control vertex density (increase 20% feature points in areas with error > 10cm);

[0081] When the number of iterations is ≤5, the RMSE converges to within 2.8cm.

[0082] The existing method relies on manual parameter adjustment (adaptation period of 2-3 weeks), which cannot meet the requirements of millimeter-level accuracy (±5cm) for marine engineering and real-time performance (30 minutes / 100km for emergency scenarios). 2 )need.

[0083] By developing an adaptive weight function driven by error feedback (logarithmic distance decay model, λ=0.5), the weight of land data in the area 0-100m from the coastline can be set to 0.9-1, and the detail retention rate can reach 93%; a three-level iterative optimization mechanism can also be established (feature matching threshold adjustment + TPS vertex density optimization + weight function correction), the first processing compliance rate is >90%, and the full process automated processing time is shortened by 87.5% (from 4 hours to 30 minutes).

[0084] The present invention achieves seamless splicing of sea and land terrain through multi-source data preprocessing, terrain feature correlation extraction, dynamic compensation matching and adaptive weight fusion. It can solve the problems of data heterogeneity, dynamic environmental adaptability and insufficient feature correlation, and the processing efficiency is also improved. It is suitable for scenarios such as coastal modeling and marine engineering.

[0085] The formula of the Bursa-Wolf seven-parameter transformation model is:

[0086] ΔX=ΔX0+ε x X+ε y Y+εz Z+mX

[0087] Among them, X, Y, Z are the coordinates of the source coordinate system, ΔX is the coordinate offset after conversion, ΔX0 is the translation parameter, ε x , ε y , ε z is the rotation parameter and m is the scale parameter.

[0088] For example, the seabed WGS84 coordinates are converted to the National 2000 coordinate system through three translation parameters, three rotation parameters, and one scale parameter, and the error is controlled within ±5 cm through the iterative least squares method; land DEM resampling: for 0.1m high-resolution LiDAR data, bilinear interpolation + edge-preserving filtering is used to retain ≥95% of the terrain details at a 1m grid resolution (such as the slope information loss of the cliff edge is <3%).

[0089] You can also perform noise filtering and hole repair on the preprocessed data:

[0090] For seafloor data processing, a TIN model can be constructed based on the Delaunay triangulation network, the local standard deviation of each water depth point can be calculated, and outliers exceeding ±3σ can be eliminated (for example, the elimination rate of ship bottom noise points reaches 98%). For land DEM restoration, a terrain roughness weight is introduced through an improved Kriging algorithm. When repairing holes, adjacent 10×10 grid data with consistent slope and aspect are preferentially referenced. The local terrain relief error of the restored DEM is less than 2cm.

[0091] The dynamic elevation correction function of the harmonic analysis model is:

[0092]

[0093] Among them, H i is the tidal amplitude, σ i is the angular frequency, θ i is the initial phase, δ i is the local tidal lag angle, the tidal waves include the dominant tidal waves M2, S2, and K1 with an amplitude ratio greater than 80%, and the model parameters are calibrated by historical data of tide gauge stations covering seasonal variations.

[0094] The deformation energy function of the thin plate spline transformation is:

[0095]

[0096] Where φ(r) is the radial basis function, and r = || xx k || is the Euclidean distance from the point to the control vertex, γ is the regularization parameter, and γ = 0, 05, w k To control the vertex weight, x is the two-dimensional coordinate of the point to be deformed, xk is the two-dimensional coordinate of the k-th control vertex, is the second-order Laplacian operator of the deformed surface z′(x,y), and m controls the total number of vertices.

[0097] The density clustering algorithm is DBSCAN, and its parameters include:

[0098] The neighborhood radius is dynamically adjusted based on the data resolution, with a default value of 50m;

[0099] The minimum number of neighborhood points is set to 5.

[0100] The iterative optimization comprises the following steps:

[0101] When RMSE>10cm, the direction consistency threshold is gradually reduced to 20° to increase the number of valid feature matching pairs;

[0102] Increase the thin plate spline transformation to control vertex density by 20% in error concentration areas;

[0103] Update the weight coefficient of the distance decay function, and the optimization range is 0.3-0.7.

[0104] When applied to sea ice-covered areas, the dynamic compensation model further includes sea ice thickness and ice surface roughness correction terms:

[0105] Δh ice =k1·d ice +k2·r ice

[0106] Among them, d ice is the ice thickness, r ice is the roughness, k1∈[0.1,0.3], k2∈[0.05,0.15] are the empirical coefficients.

[0107] Specific application cases of the above methods:

[0108] Modeling the terrain of the land-sea boundary zone of a coastal city (Xiamen Bay area)

[0109] The first step is to implement the environment and data input, taking Table 1 as an example

[0110] Table 1

[0111]

[0112]

[0113] Second step implementation

[0114] Multi-source data preprocessing (12 minutes)

[0115] Coordinate unification:

[0116] Seabed data conversion:

[0117] The Bursa-Wolf seven-parameter method was used, and three translation parameters (ΔX = -82.5m, ΔY

[0118] =-137.6m, ΔZ = -72.3m), 3 rotation parameters (ε x =0.0012°,ε y =0.0008°,ε z =0.0015°) and one scale parameter (m=0.999987). After conversion and verification by GPS measurement, the coordinate errors are all less than ±5cm.

[0119] Land data resampling

[0120] The 0.5m resolution LiDAR point cloud was used to generate a 1m×1m DEM through bilinear interpolation. After edge-preserving filtering, the elevation gradient retention rate of the beach steep slope (slope > 45°) reached 98%.

[0121] Noise filtering and hole repair:

[0122] Seabed data:

[0123] A Delaunay triangulation was constructed, and the local standard deviation (σ = 1.2 m) of each water depth point was calculated. 1200 outliers (accounting for 0.6%) with water depth values ​​greater than the mean + 3σ (i.e., greater than 5.6 m, with the actual maximum water depth being 4.8 m) were removed.

[0124] Land DEM:

[0125] 23 cavities were detected (the largest area is 200m 2 ), the improved Kriging algorithm was used to repair the DEM using the adjacent 10×10 grid data (average altitude 15m, standard deviation 0.8m). The local roughness error of the repaired DEM was less than 1.5cm.

[0126] Terrain feature correlation extraction (takes 15 minutes)

[0127] Feature point detection:

[0128] Inflection point of the isobath:

[0129] The 10 m isobath was simplified using Douglas-Peucker simplification (with a tolerance of 0.5 m), and 2800 inflection points with curvature changes greater than 15° were extracted, which were distributed at the junction of submarine ridges and depressions.

[0130] Ridge apex:

[0131] Twelve submarine ridgelines were identified using the water accumulation method, and 400 local elevation maximum points were extracted (average elevation -12.5 m, standard deviation 1.0 m).

[0132] Coastline feature points:

[0133] Canny edge detection combined with chain code tracking was used to extract 3,500 concave and convex points such as headlands and bay vertices, including 2,200 key points with a curvature radius of less than 50m.

[0134] Ridgeline intersection:

[0135] The D8 algorithm extracts 8 land watersheds with an intersection positioning accuracy of 0.3m, for a total of 2000 intersections.

[0136] Feature matching optimization:

[0137] Construct a 6300×5500 dimensional feature correlation matrix and calculate the Euclidean distance d ij (Unit: m) Angle θ with the direction ij (normalized to [0,1]), weight α=0.6.

[0138] Bidirectional matching verification: 4200 pairs were obtained through forward matching, 1400 pairs of one-way matches were eliminated through reverse verification, and 2800 pairs of valid matches were retained after DBSCAN density filtering (neighborhood radius 50m, minimum number of points 5), and the false matching rate was reduced from 15% to 2.8%.

[0139] Dynamic compensation matching (took 20 minutes)

[0140] Modeling Tidal Impacts:

[0141] Harmonic analysis: Extract the three tides M2 (amplitude H = 1.2m, angular frequency σ = 1.932rad / h, initial phase θ = 2.1rad), S2 (H = 0.8m, σ = 2.034rad / h, θ = 1.8rad), and K1 (H = 0.5m, σ = 0.0727rad / h, θ = 3.2rad), and construct a dynamic correction function: Δh(t) = 1.2cos(1.932t+2.1)+0.8cos(2.034t+1.8)+0.5cos(0.0727t+3.2)

[0142] TPS elastic transformation: Using 2800 pairs of feature points as control vertices and setting the regularization parameter γ = 0.05, after 10 iterations, the Gaussian curvature difference of the land-sea boundary zone was reduced from 0.3% to 0.08% (the target is < 0.1%).

[0143] Adaptive weight fusion (8 minutes)

[0144] Distance-weighted fusion:

[0145] Weight function The land weight in the area 0-100m from the coastline is 0.9-1, the seabed weight 500m away is 0.8-1, and the middle area has a linear transition.

[0146] Fusion calculation: Using a weighted average method, areas with the highest land LiDAR data weight (such as beaches) retain 0.5-meter-level terrain details, while areas dominated by submarine sonar data (such as the seabed) are smoothly displayed with a 1-meter resolution trend.

[0147] Accuracy verification and iteration:

[0148] The first fusion RMSE = 4.5cm (<5cm threshold), no iteration required; compared with the traditional method RMSE = 12.8cm, the efficiency of meeting the standard is improved by 64%.

[0149] Step 3: Implementation results and comparative analysis

[0150] Table 2 shows the comparison of quantitative indicators.

[0151] Table 2

[0152]

[0153] This example verifies the feasibility and superiority of the algorithm using real engineering data. Key innovations are implemented in the following aspects:

[0154] 1. Multi-source heterogeneous fusion: Successfully processed 1m seabed and 0.5m land data, retaining 93% of details, far exceeding the 65% of traditional methods;

[0155] 2. Dynamic environmental adaptation: Dynamic compensation based on local tidal parameters reduces the error in high and low tide matching by 80%, meeting the real-time requirements of coastal projects;

[0156] 3. Intelligent iterative optimization: Through RMSE-driven parameter adjustment, the target is achieved on the first pass, avoiding the multiple manual parameter adjustments required by traditional methods and improving efficiency by over 70%.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for matching seabed topography and land elevation based on multi-source data fusion, characterized in that: The method comprises the following steps: Multi-source heterogeneous data preprocessing: The Bursa-Wolf seven-parameter transformation model is used to unify the coordinate system of the seabed sonar discrete point cloud data and the land LiDAR raster data, and the transformation error is controlled within ±5cm. Land LiDAR data were resampled to 1m resolution using edge-preserving bilinear interpolation to preserve the details of steep cliffs with slopes > 45°. Terrain feature association extraction: The Douglas-Peucker algorithm is used to extract inflection points where the curvature of the seabed contour lines changes by more than 15°, and the flow accumulation method is used to identify the apex of the seabed ridge. Canny edge detection and chain code tracking are used to extract the concave and convex points of the land coastline with a curvature radius of less than 50m, and the land watershed intersection points are generated using the D8 algorithm; Construct a distance and direction weighted feature correlation matrix and eliminate mismatched pairs using a density clustering algorithm; Dynamic compensation matching: The dynamic elevation correction value of the intertidal zone is calculated based on the harmonic analysis model calibrated with tide gauge data; Taking effective feature points as control vertices, thin plate spline transformation is used to achieve elastic deformation matching; Adaptive weight fusion and iterative optimization: Assign data weights based on the distance between the target point and the coastline, and fuse to generate seamless terrain data; The parameters are adjusted by the root mean square error (RMSE) feedback until the matching accuracy reaches within ±5cm.

2. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1, characterized in that: The formula of the Bursa-Wolf seven-parameter transformation model is: ΔX=ΔX0+ε x X+e y Y+e z Z+mX Among them, X, Y, Z are the coordinates of the source coordinate system, ΔX is the coordinate offset after conversion, ΔX0 is the translation parameter, ε x , ε y , ε z is the rotation parameter and m is the scale parameter.

3. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1 is characterized in that: The dynamic elevation correction function of the harmonic analysis model is: Among them, H i is the tidal amplitude, σ i is the angular frequency, θ i is the initial phase, δ i is the local tidal lag angle, the tidal waves include the dominant tidal waves M2, S2, and K1 with an amplitude ratio greater than 80%, and the model parameters are calibrated by historical data of tide gauge stations covering seasonal variations.

4. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1, characterized in that: The deformation energy function of the thin plate spline transformation is: Where φ(r) is the radial basis function, and r = || xx k || is the Euclidean distance from the point to the control vertex, γ is the regularization parameter, and γ=0,05, e k To control the vertex weight, x is the two-dimensional coordinate of the point to be deformed, x k is the two-dimensional coordinate of the k-th control vertex, is the second-order Laplacian operator of the deformed surface z′(x,y), and m controls the total number of vertices.

5. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1 is characterized in that: The density clustering algorithm is DBSCAN, and its parameters include: The neighborhood radius is dynamically adjusted based on the data resolution, with a default value of 50m; The minimum number of neighborhood points is set to 5.

6. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1, characterized in that: The iterative optimization comprises the following steps: When RMSE>10cm, the direction consistency threshold is gradually reduced to 20° to increase the number of valid feature matching pairs; Increase the thin plate spline transformation to control vertex density by 20% in error concentration areas; Update the weight coefficient of the distance decay function, and the optimization range is 0.3-0.

7.

7. The method for matching seabed topography and land elevation based on multi-source data fusion according to claim 1, characterized in that: When applied to sea ice-covered areas, the dynamic compensation model further includes sea ice thickness and ice surface roughness correction terms: Δh ice =k1·d ice +k2·r ice Among them, d ice is the ice thickness, r ice is the roughness, k1∈[0.1,0.3], k2∈[0.05,0.15] are the empirical coefficients.

Images