Underwater three-dimensional shape real-time reconstruction method and system

Abstract

The invention discloses an underwater three-dimensional shape real-time reconstruction method and system, relates to the technical field of computer graphics, and solves the problem that the execution speed is limited by memory and algorithm complexity in an incremental point cloud three-dimensional reconstruction process. According to the technical scheme, the method is characterized in that reading and rasterization processing of preprocessed underwater three-dimensional sensing data are realized through a point cloud rasterization module, and meanwhile, a rapid access method of the data is constructed; the grid point cloud splicing and smoothing module queries a grid point cloud sequence connected with the current grid point cloud, and performs smoothing and splicing processing through a smoothing kernel along a boundary; and the incremental three-dimensional morphology reconstruction module realizes read-in analysis, surface reconstruction and modification of raster data. In one acquisition interval of underwater three-dimensional sensing data, related modules complete all processing of incremental updating data, and then real-time performance is guaranteed. By means of the method, smooth splicing and fusion of the incremental three-dimensional surface and the historical three-dimensional surface in the underwater three-dimensional shape real-time reconstruction process are achieved, and the underwater three-dimensional detection efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of computer graphics, and particularly to a method and system for real-time reconstruction of underwater three-dimensional topography. Background Art

[0002] With the development of the marine economy, related researches such as marine construction and seabed exploration have been continuously promoted. Sensing devices such as multibeam sonar and sidescan sonar are widely used in fields such as underwater topographic mapping, marine resource development, and underwater facility detection. The real-time perception and enhanced display of underwater three-dimensional topography can provide immediate feedback for on-site operations and improve the safety and efficiency of the operation and maintenance process. With the current development of underwater three-dimensional mapping equipment towards unmanned solutions, higher requirements are put forward for the real-time performance of the system. Most traditional underwater three-dimensional reconstruction technologies are based on offline processing methods, which require obtaining complete point cloud data first and then performing three-dimensional topography reconstruction. Such methods cannot meet the real-time requirements when dealing with incremental data. For the processing of incremental sensing data, there is currently a lack of mature solutions for real-time generation, stitching, and smoothing of a large amount of terrain grid data.

[0003] In view of the above problems, there is an urgent need to design a modular method to achieve real-time surface reconstruction of incremental underwater three-dimensional topography point cloud data. Summary of the Invention

[0004] The present application provides a method and system for real-time reconstruction of underwater three-dimensional topography, and its technical objective is to achieve real-time surface reconstruction of incremental underwater three-dimensional topography point cloud data.

[0005] The above technical objective of the present application is achieved through the following technical solutions:

[0006] A method for real-time reconstruction of underwater three-dimensional topography includes:

[0007] Step S1: Real-time monitor whether there is input of underwater three-dimensional incremental point cloud data. If so, execute steps S2 - S4; if not, re-execute step S1;

[0008] Step S2: Perform rasterization processing on the input incremental point cloud data and calibrate the data sequence to form rasterized incremental point cloud data that can be queried by index;

[0009] Step S3: Query the grid point cloud sequence in the global point cloud data that is connected to the rasterized incremental point cloud data, and splice and smooth the rasterized incremental point cloud data with the global point cloud data according to the connected grid point cloud sequence;

[0010] Step S4: Perform 3D reconstruction on the rasterized incremental point cloud data after splicing and smoothing, make corresponding modifications to the reconstruction grid of the global point cloud data, then mark the currently processed rasterized incremental point cloud data as the global point cloud data, and go to step S1.

[0011] Further, the step S2 includes:

[0012] Preliminarily rasterize the incremental point cloud data at a given density to generate the smallest rectangular grid point cloud on the xoy plane that can enclose the projection of the original incremental point cloud data;

[0013] Expand the number of rows and columns on the periphery of the smallest rectangular grid point cloud according to the given width parameter to obtain the expanded grid point cloud;

[0014] Assign elevation values to the grid points with valid positions by constructing the mapping relationship from the original incremental point cloud data to the smallest rectangular grid point cloud;

[0015] Retrieve and mark the grid occupancy points outside the expanded grid point cloud that have not been assigned elevation values, and extract the boundary of the expanded grid point cloud;

[0016] Interpolate the grid points with valid positions and unassigned elevation values within the boundary to supplement the missing values inside the expanded grid point cloud to obtain the rasterized incremental point cloud data.

[0017] Further, during the process of rasterizing the underwater three-dimensional incremental point cloud data, key information of the grid data is stored through different variables, including:

[0018] Coordinate array, an array used to store the coordinate information of the grid point cloud;

[0019] Status array, an array that maps the grid position to the index of the coordinate array;

[0020] Transformation vector, the transformation vector from the grid position to the global coordinate system.

[0021] Further, the step S3 includes:

[0022] Extract the range of the current grid point cloud and construct a range record list;

[0023] Query the grid point cloud sequence in the global point cloud data that is connected to the rasterized incremental point cloud data through the range record list;

[0024] Construct a smoothing kernel, move the center of the smoothing kernel along the boundary of the current grid point cloud, traverse the boundary of the current grid point cloud, and then determine whether the traversal of the boundary of the current grid point cloud is completed. If not, perform a step process on the smoothing kernel; then traverse and query the sequence of adjacent grid point clouds, and perform smoothing kernel processing on them in turn. After each processing, determine whether the adjacent grid point cloud has been traversed. If so, the smoothing kernel continues to perform a step process along the boundary of the current grid point cloud until the traversal of the boundary of the current grid point cloud is completed, and finally complete the smoothing and stitching of the rasterized incremental point cloud data and the global point cloud data;

[0025] Among them, the current grid point cloud represents the latest input incremental point cloud data after rasterization in the incremental reconstruction process; the adjacent grid point cloud represents the point cloud adjacent to the current grid point cloud in the global point cloud data; the global point cloud data represents all rasterized point cloud data except the current grid point cloud.

[0026] Further, the smoothing kernel is generated with a certain grid point on the boundary as the center, its side length is an even multiple of the grid density, and its side length does not exceed 2 times the outer extension length of the smallest rectangular grid point cloud.

[0027] Further, the smoothing and stitching process includes:

[0028] When there are both the current grid point cloud and the adjacent grid point cloud at the same grid position, the smoothing kernel sets the elevation value of this grid position to the average of the elevation values of the current grid point cloud and the adjacent grid point cloud;

[0029] When the number of adjacent grid point clouds within the current smoothing kernel exceeds a given threshold, assign elevation values to the grid occupancy points inside the smoothing kernel through an interpolation algorithm to form new valid grid points.

[0030] Further, the step S4 includes:

[0031] Combine the coordinate array, grid density, and coordinate system conversion vector to analyze the original spatial coordinates of the rasterized incremental point cloud data after smoothing and stitching;

[0032] According to the relative position relationship of the grid points implicitly included in the status array, combine with the original spatial coordinates of the current grid point cloud to perform three-dimensional topography reconstruction of the incremental point cloud data;

[0033] According to the elevation data information that needs to be modified in the adjacent grid point cloud after smoothing kernel processing, correspondingly modify the node coordinates of the three-dimensional grid of the adjacent grid point cloud.

[0034] Further, traverse the data in the status array. Except for the elements in the first row and the last column of the status array, query the values of the status array corresponding to a certain point and its adjacent points directly above, directly to the right, and diagonally above. When there are 3 valid points in the positions, connect them to form 1 triangular patch. When there are 4 valid points in the positions, connect them to form 2 triangular patches. Repeat this step until the incremental point cloud data reconstruction is completed.

[0035] Further, the method for calculating the elevation value includes the nearest neighbor algorithm, the average value algorithm, and the inverse distance weighted algorithm.

[0036] An underwater three-dimensional topography real-time reconstruction system, which is used for the underwater three-dimensional topography real-time reconstruction method. The reconstruction system includes:

[0037] A monitoring module that monitors in real time whether there is underwater three-dimensional incremental point cloud data input. If so, transfer to the point cloud rasterization module; if not, continue monitoring.

[0038] A point cloud rasterization module that rasterizes the input incremental point cloud data and calibrates the data sequence to form rasterized incremental point cloud data that can be queried by index.

[0039] A raster point cloud stitching and smoothing module that queries the raster point cloud sequence in the global point cloud data that is connected to the rasterized incremental point cloud data, and stitches and smooths the rasterized incremental point cloud data and the global point cloud data according to the connected raster point cloud sequence.

[0040] An incremental three-dimensional topography reconstruction module that performs three-dimensional reconstruction on the rasterized incremental point cloud data after stitching and smoothing processing, makes corresponding modifications to the reconstruction grid of the global point cloud data, and then marks the currently processed rasterized incremental point cloud data as the global point cloud data and transfers to the monitoring module.

[0041] The beneficial effects of this application are as follows:

[0042] (1) Ensure the real-time performance of data processing through an efficient data structure. By establishing a mutual mapping between the raster point cloud position and the raster point cloud coordinates, realize the smooth stitching processing of the raster point cloud at the boundary and the fast access to the data during surface reconstruction, with low memory overhead and implementation complexity.

[0043] (2) Ensure reasonable connection of the incremental point cloud data through a smoothing kernel. When the surface of the incremental point cloud data is reconstructed into a three-dimensional grid representation, there may be discontinuous gaps or overlaps at the connection. By using a smoothing kernel to perform raster boundary search and perform efficient smoothing and stitching processing, ensure the correct representation of the three-dimensional topography. Description of the Drawings

[0044] Figure 1It is a schematic diagram of the composition structure and process of the underwater three-dimensional topography real-time reconstruction system in the embodiment of the present application;

[0045] Figure 2 It is a schematic diagram of the storage and representation of grid point cloud data in the embodiment of the present invention;

[0046] Figure 3 It is a schematic diagram of the execution principle of the smoothing kernel in the embodiment of the present invention;

[0047] Figure 4 It is a schematic diagram of the existence of voids and overlaps in the smoothing kernel in the embodiment of the present invention;

[0048] Figure 5 It is a schematic diagram of the existence of voids in the smoothing kernel in the embodiment of the present invention;

[0049] Figure 6 It is a schematic diagram of the grid point cloud reconstruction principle in the embodiment of the present invention. Specific embodiments

[0050] The technical solution of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments are only used to explain the present invention and do not limit the protection scope of the present invention.

[0051] Figure 1 It is a schematic diagram of the composition structure and process of the underwater three-dimensional topography real-time reconstruction system in the embodiment of the present application, as Figure 1 shown. The point cloud rasterization module realizes the reading and rasterization processing of the underwater three-dimensional incremental point cloud data after preprocessing, and outputs raster point cloud information; the raster point cloud splicing and smoothing module receives the current raster point cloud input, queries the raster point cloud sequence connected to it, and realizes the smoothing and splicing processing by marching along the boundary through the smoothing kernel; the incremental three-dimensional topography reconstruction module realizes the reading, parsing, surface reconstruction and modification of the raster data.

[0052] The underwater three-dimensional topography real-time reconstruction method described in the present application includes:

[0053] Step S1: Real-time monitor whether there is underwater three-dimensional incremental point cloud data input. If so, execute steps S2 - S4; if not, re-execute step S1.

[0054] Step S2: Perform rasterization processing on the input incremental point cloud data and calibrate the data sequence to form rasterized incremental point cloud data queried by index.

[0055] Preferably, the step S2 includes:

[0056] 101: Generation of planar grid point cloud. The underwater three-dimensional incremental point cloud data is preliminarily rasterized at a given raster density to generate the smallest rectangular grid point cloud that can enclose the projection of the original incremental point cloud data on the xoy plane.

[0057] In this application, the underwater three-dimensional incremental point cloud data has 2.5D characteristics.

[0058] Preferably, during the rasterization process of the incremental three-dimensional data, the processing should be performed with a consistent raster density.

[0059] 102: Raster peripheral expansion. As Figure 6 shown, to facilitate adding data outside the rectangular raster during the subsequent boundary smoothing process of the incremental point cloud data, the number of rows and columns on the periphery of the minimum rectangular raster point cloud is expanded according to the given width parameter to obtain the expanded raster point cloud.

[0060] 103: Raster point elevation calculation. By constructing the mapping relationship from the original incremental point cloud data to the minimum rectangular raster point cloud, elevation values are assigned to the raster points with valid positions.

[0061] Preferably, the methods for elevation calculation include the nearest neighbor algorithm, the average value algorithm, and the inverse distance weighted algorithm.

[0062] 104: Raster boundary extraction. The raster occupancy points outside the expanded raster point cloud that have not been assigned elevation values are retrieved and labeled to extract the boundary of the expanded raster point cloud.

[0063] Preferably, the external raster occupancy points are retrieved and labeled through the breadth-first algorithm.

[0064] 105: Interpolation of raster points inside the boundary. Due to the disordered distribution characteristics of the original incremental point cloud data, there are raster points with valid positions but not assigned elevation values inside the raster point cloud boundary. Interpolation is performed on the raster points with valid positions and not assigned elevation values inside the boundary to supplement the missing values inside the expanded raster point cloud, obtaining the rasterized incremental point cloud data.

[0065] Preferably, the optional interpolation algorithms include the bilinear interpolation method, the bicubic interpolation method, and the Kriging interpolation method.

[0066] Preferably, during the rasterization process of the underwater three-dimensional incremental point cloud data, the key information of the raster data is stored through different variables, such as Figure 2 shown, including:

[0067] 201: Coordinate array, that is, the array Points used to store the coordinate information of the raster point cloud. The x, y, and z coordinate values of the raster point cloud are recorded through vectors, and the coordinate vectors can be accessed through indexes. Among them, the x and y coordinates are represented by the row and column numbers where the raster points are located, and their values multiplied by the raster density are the actual coordinate values.

[0068] 202: Status array, which is an array Status that maps grid positions to coordinate array indices. The status array realizes the mapping from the projection position of the grid point cloud on the xoy plane to its three-dimensional coordinates in a two-dimensional form. The two-dimensional indices of the array represent the relative row and column positions of the corresponding grid points, and the non-negative values of the array elements represent the indices of the valid points at the corresponding positions in the coordinate array; for the grid occupancy points along the outer boundary of the grid point cloud that are not assigned elevation values, including the outer expansion of the grid, their values are assigned -1.

[0069] 203: Transformation vector, that is, the transformation vector from the grid position to the global coordinate system.

[0070] Specifically, since the two-dimensional indices of the array can only record the relative positions of the grid points, the transformation vector from it to the global coordinate system needs to be calculated while generating the status array.

[0071] The above information enables fast access to different types of data and ensures the real-time nature of the processing process. Further, the extraction of the grid boundary in 104 is specifically implemented through the status array of 202. By traversing the outermost rectangular boundary of the grid except for the outer expansion and performing a search inward along the vertical or horizontal direction, the grid positions where the status array values are non-negative and are first searched are marked as the grid boundary.

[0072] Step S3: Query the grid point cloud sequence in the global point cloud data that is adjacent to the rasterized incremental point cloud data, and splice and smooth the rasterized incremental point cloud data and the global point cloud data according to the adjacent grid point cloud sequence.

[0073] Specifically, the step S3 includes:

[0074] 301: Extract the range of the current grid point cloud and construct a range record list. When processing the incremental point cloud data each time, the maximum and minimum values of the x coordinate and y coordinate of the corresponding grid point cloud are respectively recorded through the range record list, that is, x min 、x max 、y min and y max , and record the coverage range of the current grid point cloud.

[0075] 302: Query the grid point cloud sequence in the global point cloud data that is adjacent to the rasterized incremental point cloud data through the range record list;

[0076] 303: Construct a smoothing kernel, move the center of the smoothing kernel along the boundary of the current grid point cloud, traverse the boundary of the current grid point cloud, and then determine whether the traversal of the boundary of the current grid point cloud is completed. If not, perform a step process on the smoothing kernel; then traverse and query the adjacent grid point cloud sequence, and perform smoothing kernel processing on it in turn. After each processing, determine whether the adjacent grid point cloud has been traversed. If so, the smoothing kernel continues to perform a step process along the boundary of the current grid point cloud until the traversal of the boundary of the current grid point cloud is completed, and finally complete the smoothing and splicing of the rasterized incremental point cloud data and the global point cloud data;

[0077] Among them, the current grid point cloud represents the latest input incremental point cloud data after rasterization in the incremental reconstruction process; the adjacent grid point cloud represents the point cloud adjacent to the current grid point cloud in the global point cloud data; the global point cloud data represents all rasterized point cloud data except the current grid point cloud. After the current grid point cloud completes this 3D reconstruction, it is also marked as global point cloud data.

[0078] Preferably, the in-kernel region processing in 303 is as Figure 3 shown, and the consistency and continuity of the current grid point cloud and the adjacent grid point cloud are achieved through local processing. The smoothing kernel is generated with a certain boundary grid point as the center, and is a square kernel with a side length that is an even multiple of the grid density, and its side length does not exceed 2 times the outer extension length of the smallest rectangular grid point cloud. When performing each smoothing step, the grid points within its coverage are processed. Considering the principle of the smoothing algorithm, since the point cloud adopts an incremental processing method, for all adjacent grid point cloud sequences, the points at the same grid position have the same elevation coordinate value.

[0079] For different distribution relationships between the adjacent grid point cloud and the current grid point cloud in the smoothing kernel, different operations need to be performed, specifically including:

[0080] 3031: Smoothing of conflicting grid positions. As Figure 4 shown, in some cases, there are both the current grid point cloud and the adjacent grid point cloud at the same grid position, that is, there are conflicting elevation values. Then the smoothing kernel sets the elevation value of this grid position to the mean value of the current conflict values.

[0081] 3032: Completion of vacant grid positions. As Figure 4 and Figure 5 shown, there are some grid positions in the smoothing kernel where there is no point cloud, that is, grid occupancy points. At this time, if the number of adjacent grid point clouds in the current smoothing kernel exceeds a given threshold, the elevation values are assigned to the grid occupancy points inside through an interpolation algorithm. For the interpolated data, to save the overhead of recalculating the coordinate array and status array for the adjacent grid point cloud, it is added to the current grid point cloud.

[0082] 3033: Specifically, the vacant grid positions that need to be filled may be outside the boundary of the current grid point cloud. The grid periphery expansion in step 102 provides reserved positions outside the grid boundary of the smoothing kernel, ensuring the correct filling of the gaps between two adjacent grid point clouds.

[0083] Step S4: Perform three-dimensional reconstruction on the rasterized incremental point cloud data after stitching and smoothing processing, make corresponding modifications to the reconstruction grid of the global point cloud data, and then mark the currently processed rasterized incremental point cloud data as the global point cloud data, and go to step S1.

[0084] Specifically, the step S4 includes:

[0085] 401: Parse the original spatial coordinates of the rasterized incremental point cloud data after smoothing stitching processing in combination with the coordinate array, grid density, and coordinate system transformation vector.

[0086] 402: Perform three-dimensional shape reconstruction of the incremental point cloud data based on the relative position relationship of grid points implicitly included in the status array and in combination with the original spatial coordinates of the current grid point cloud.

[0087] As Figure 6 shown, the three-dimensional reconstruction steps of the three-dimensional shape reconstruction and display module are performed based on the coordinate array and the status array, which implicitly includes three-dimensional grid information. Traverse the data in the status array, that is, Figure 6 the elements in the grid position points in except for the first row and the last column, query the values of the status array corresponding to a certain point and its adjacent four points, namely the directly above, directly right, and upper right points. When there are 3 valid position points, they can be connected to form 1 triangular patch. When there are 4 valid position points, they can be connected to form 2 triangular patches. Repeat this step until the construction of the complete three-dimensional grid is completed. For the points existing in the grid point cloud periphery expansion after smoothing kernel processing, they also participate in the generation of the grid.

[0088] 403: Modify the node coordinates of the three-dimensional grid of the adjacent grid point cloud correspondingly according to the elevation data information that needs to be modified in the adjacent grid point cloud after smoothing kernel processing.

[0089] Through the incremental underwater three-dimensional data real-time reconstruction and stitching smoothing method, the scanned three-dimensional data can quickly execute the preprocessing, rasterization, and three-dimensional reconstruction processes. Within one acquisition interval of underwater three-dimensional sensing data, the relevant modules complete all processing of the incremental update data, thereby ensuring real-time performance.

[0090] The above specific embodiments are only for explaining the technical concept and structural features of the present invention, aiming to enable those skilled in the art to implement it accordingly. However, the above content does not limit the protection scope of the present invention. Any equivalent changes or modifications made based on the spirit and essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A real-time underwater three-dimensional topography reconstruction method, characterized in that Including: Step S1: Monitor in real time whether there is underwater three-dimensional incremental point cloud data input. If so, execute steps S2 - S4; if not, re-execute step S1. Step S2: Perform rasterization processing on the input incremental point cloud data, and calibrate the data sequence to form rasterized incremental point cloud data that can be queried by index. Step S3: Query the raster point cloud sequence in the global point cloud data that is adjacent to the rasterized incremental point cloud data, and splice and smooth the rasterized incremental point cloud data with the global point cloud data according to the adjacent raster point cloud sequence. Step S4: Perform three-dimensional reconstruction on the rasterized incremental point cloud data after splicing and smoothing processing, modify the reconstruction grid of the global point cloud data accordingly, then mark the currently processed rasterized incremental point cloud data as the global point cloud data, and go to step S1.

2. The real-time underwater three-dimensional topography reconstruction method according to claim 1, wherein The said step S2 includes: Preliminarily rasterize the incremental point cloud data at a given density to generate the smallest rectangular raster point cloud on the xoy plane that can enclose the projection of the original incremental point cloud data. Expand the number of rows and columns on the periphery of the smallest rectangular raster point cloud according to the given width parameter to obtain the expanded raster point cloud. By constructing the mapping relationship from the original incremental point cloud data to the smallest rectangular raster point cloud, assign elevation values to the raster points with valid positions. Retrieve and mark the raster occupancy points outside the expanded raster point cloud that have not been assigned elevation values, and extract the boundary of the expanded raster point cloud. Interpolate the raster points with valid positions and unassigned elevation values inside the boundary to supplement the missing values inside the expanded raster point cloud to obtain the rasterized incremental point cloud data.

3. The real-time underwater three-dimensional topography reconstruction method according to claim 2, wherein, During the process of rasterizing the underwater three-dimensional incremental point cloud data, store the key information of the raster data through different variables, including: Coordinate array, an array used to store the coordinate information of the raster point cloud. Status array, an array that maps the raster position to the index of the coordinate array. Transformation vector, the transformation vector from the raster position to the global coordinate system.

4. The real-time underwater three-dimensional topography reconstruction method according to claim 3, wherein The said step S3 includes: Extract the range of the current raster point cloud and construct a range record list. Query the raster point cloud sequence in the global point cloud data that is adjacent to the rasterized incremental point cloud data through the range record list. Construct a smoothing kernel, make the center of the smoothing kernel travel along the boundary of the current raster point cloud, traverse the boundary of the current raster point cloud, and then judge whether the traversal of the boundary of the current raster point cloud is completed. If not, perform a stepping process on the smoothing kernel; then traverse and query the adjacent raster point cloud sequence and perform smoothing kernel processing on it in turn. After each processing, judge whether the traversal of the adjacent raster point cloud is completed. If so, the smoothing kernel continues to perform a stepping process along the boundary of the current raster point cloud until the traversal of the boundary of the current raster point cloud is completed, and finally complete the smoothing and splicing of the rasterized incremental point cloud data and the global point cloud data. Among them, the current raster point cloud represents the latest input incremental point cloud data after rasterization during the incremental reconstruction process; the adjacent raster point cloud represents the point cloud adjacent to the current raster point cloud in the global point cloud data; the global point cloud data represents all rasterized point cloud data except the current raster point cloud.

5. The real-time underwater three-dimensional topography reconstruction method according to claim 4, characterized in that The smoothing kernel is generated centered on a certain grid point of the boundary. Its side length is an even multiple of the grid density, and its side length does not exceed 2 times the peripheral extension length of the minimum rectangular grid point cloud.

6. The real-time underwater three-dimensional topography reconstruction method according to claim 4 or 5, characterized in that The smoothing and stitching process includes: When there are both the current grid point cloud and the adjacent grid point cloud at the same grid position, the smoothing kernel sets the elevation value of this grid position to the average of the elevation values of the current grid point cloud and the adjacent grid point cloud; When the number of adjacent grid point clouds within the current smoothing kernel exceeds a given threshold, the elevation values are assigned to the grid occupancy points inside the smoothing kernel through an interpolation algorithm to form new valid grid points.

7. The real-time underwater three-dimensional topography reconstruction method according to claim 6, characterized in that The step S4 includes: Combining the coordinate array, grid density, and coordinate system transformation vector to analyze the original spatial coordinates of the rasterized incremental point cloud data after smoothing and stitching processing; According to the relative position relationship of the grid points implicitly included in the status array, and combining with the original spatial coordinates of the current grid point cloud, performing three-dimensional topography reconstruction of the incremental point cloud data; According to the elevation data information that needs to be modified in the adjacent grid point cloud after smoothing kernel processing, correspondingly modify the node coordinates of the three-dimensional grid of the adjacent grid point cloud.

8. The real-time underwater three-dimensional topography reconstruction method according to claim 7, wherein, Traverse the data in the status array. Except for the elements in the first row and the last column of the status array, query the values of the status array corresponding to a certain point and the 4 points directly above, directly to the right, and upper right adjacent to it. When there are 3 valid points in the positions, connect them to form 1 triangular patch. When there are 4 valid points in the positions, connect them to form 2 triangular patches. Repeat this step until the reconstruction of the incremental point cloud data is completed.

9. The real-time underwater three-dimensional topography reconstruction method according to claim 2, wherein The methods for calculating the elevation value include the nearest neighbor algorithm, average value algorithm, and inverse distance weighted algorithm.

10. An underwater three-dimensional topography real-time reconstruction system, characterized in that, This reconstruction system is used for the underwater three-dimensional topography real-time reconstruction method described in any one of claims 1-9. This reconstruction system includes: A monitoring module that monitors in real time whether there is underwater three-dimensional incremental point cloud data input. If so, it transfers to the point cloud rasterization module; if not, it continues to monitor; A point cloud rasterization module that rasterizes the input incremental point cloud data and calibrates the data sequence to form rasterized incremental point cloud data for query by index; A grid point cloud stitching and smoothing module that queries the grid point cloud sequence adjacent to the rasterized incremental point cloud data in the global point cloud data, and stitches and smooths the rasterized incremental point cloud data and the global point cloud data according to the adjacent grid point cloud sequence; An incremental three-dimensional topography reconstruction module that performs three-dimensional reconstruction on the rasterized incremental point cloud data after stitching and smoothing processing, correspondingly modifies the reconstruction grid of the global point cloud data, and then marks the currently processed rasterized incremental point cloud data as the global point cloud data and transfers to the monitoring module.

Images