A method for processing seabed topography images

Abstract

This invention relates to a method for processing seabed topographic images, belonging to the field of computer processing technology. The method includes the following steps: S01, acquiring sonar point cloud data and original optical images during collaborative detection using multi-beam sonar and optical equipment; wherein the sonar point cloud data consists of spatial coordinate components, reflection intensity components, and beam incidence angle components; S02, dynamically dividing the sonar point cloud data into grids to generate an initial grid model matching the resolution of the optical image; S03, embedding a water attenuation compensation factor into the initial grid model to correct the scattering noise of laser scanning and optical images in turbid waters, thereby obtaining a corrected initial grid model. This invention achieves high-fidelity fusion of sonar-laser-optical multi-source data across resolutions, simultaneously solving the problems of scattering noise interference and loss of micro-topographic features in turbid waters, and outputting geometrically accurate and textured seabed topographic images.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a method for processing seabed topographic images. Background Technology

[0002] Seafloor topographic image processing technology is a crucial support for marine resource exploration, seafloor engineering construction, and marine scientific research. With the rapid development of deep-sea exploration technology, heterogeneous data sources such as multibeam sonar, laser scanning, and optical imaging are gradually becoming core methods for seafloor topographic mapping. Especially under the advancement of national marine strategies, the demand for high-precision 3D seafloor modeling continues to grow in fields such as oil and gas resource exploration, subsea pipeline laying, and deep-sea archaeology. Traditional methods mainly rely on single-sensor data (such as multibeam sonar point clouds), improving resolution through beam cross-block segmentation and depth value fusion, a typical example being the transmit / receive beam cross-block segmentation algorithm. However, as multi-source collaborative detection becomes mainstream, alignment distortion and resolution discontinuities in heterogeneous data are becoming increasingly prominent.

[0003] For example, multibeam sonar systems use the transmit and receive beam crossover as the smallest detection unit. The actual area of ​​the acquisition zone is affected by both the transmit beam width and the seabed depth, leading to a sharp decrease in resolution in deep water. Meanwhile, new optical devices such as the LRC-20 over-the-horizon underwater camera have achieved micrometer-level imaging accuracy, capturing extremely subtle topographic features at close range. When these two types of data are fused, the resolution difference can reach orders of magnitude, causing texture distortion or loss of detail using traditional interpolation or downsampling methods. Furthermore, a Chinese patent (CN119399406A, published on 2025-02-07) discloses a rapid 3D modeling method, medium, and system based on seabed point cloud data. While this improves modeling efficiency through point cloud downsampling and hybrid decomposition, it fails to consider the dynamic interference of suspended matter in the water on the optical data. Especially in turbid waters, scattering noise from laser scanning and optical images prevents accurate registration between the initial mesh model constructed by the sonar and the optical details.

[0004] Therefore, it is hoped that a good solution can be found to address the issues of resolution inconsistency and feature distortion in the fusion of multi-source heterogeneous seabed topographic data and to achieve high-fidelity, real-time 3D seabed modeling. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides a method for processing seabed topographic images, the method comprising the following steps:

[0006] S01. When multi-beam sonar and optical equipment are used for collaborative detection, acquire sonar point cloud data and original optical images; wherein, the sonar point cloud data consists of spatial coordinate components, reflection intensity components and beam incidence angle components.

[0007] S02. Dynamically divide the sonar point cloud data into grids to generate an initial grid model that matches the resolution of the optical image.

[0008] S03. Embed a water body attenuation compensation factor into the initial mesh model to correct the scattering noise of laser scanning and optical images in turbid waters, so as to obtain the corrected initial mesh model.

[0009] S04. Based on the corrected initial grid model, extract the structural contour features of the sonar data;

[0010] S05. Decompose the original optical image into a basic brightness layer, a detail texture layer, and a color information layer, and then fuse the structural contour features with the detail texture layer to generate a high-precision texture mapping map;

[0011] S06. Extract the illumination distribution information from the basic brightness layer and the geometric feature information from the high-precision texture map, and fuse the illumination distribution information and the geometric feature information to obtain a fused image of the seabed topography.

[0012] Preferably, the dynamic grid division of the sonar point cloud data in step S02 includes:

[0013] S21. Use the octree spatial partitioning algorithm to divide the sonar point cloud data into multiple subspaces;

[0014] S22. Within each subspace, the mesh size is adaptively adjusted according to the point cloud density;

[0015] S23. Smooth the grid boundaries using a local weighted regression algorithm to generate a continuous initial grid model, or input the sonar point cloud data into a pre-trained dynamic grid generation model to obtain an initial grid model that matches the resolution of the optical image.

[0016] S24. The dynamic mesh generation model is trained using a sonar point cloud dataset and corresponding optical image resolution labels.

[0017] S25. The sonar point cloud dataset is obtained by a multibeam sonar system under different seabed topographic conditions; the optical image resolution label is obtained by an over-the-horizon underwater optical imaging device at the same detection location and is generated by manual annotation.

[0018] Preferably, embedding a water attenuation compensation factor in the initial grid model in step S03 includes:

[0019] S31. Calculate the attenuation coefficient at different wavelengths based on real-time monitoring data of suspended solids concentration in water.

[0020] S32. At each node of the initial mesh model, a dynamic weighting function related to the attenuation coefficient is introduced;

[0021] S33. An anti-interference mesh model is generated by correcting the scattering noise of laser scanning and optical images through weighted averaging.

[0022] Preferably, embedding a water attenuation compensation factor in the initial grid model in step S03 further includes:

[0023] S34. Establish a mathematical model of the concentration and attenuation coefficient of suspended solids in water using the Mie scattering theory;

[0024] S35. Fit the measured data using the least squares method to optimize the mathematical model;

[0025] S36. Embed the optimized mathematical model into each node of the initial mesh model to generate a dynamic attenuation compensation field, or input the initial mesh model into the attenuation compensation model to obtain an anti-interference mesh model.

[0026] The attenuation compensation model was trained using the initial grid model dataset and the corresponding anti-interference grid model labels.

[0027] The initial grid model dataset was obtained through collaborative detection by a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0028] The anti-interference mesh model label is obtained by manually labeling the distribution of scattering noise in turbid water and generating a noise-free mesh model.

[0029] Preferably, the structural contour features extracted from the sonar data in step S04 include:

[0030] S41. Use an edge detection algorithm to extract structural contour lines from the corrected initial mesh model;

[0031] S42. Optimize the continuity of contour lines through morphological operations;

[0032] S43. Convert the optimized contour lines into a binary image to generate a structural contour feature map, or input the corrected initial mesh model into the structural contour extraction model to obtain a structural contour feature map.

[0033] The structural contour extraction model was trained using the modified initial mesh model dataset and the corresponding structural contour labels.

[0034] The corrected initial grid model dataset was obtained through collaborative detection by a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0035] The structural outline labels are generated by manually annotating the structural outlines of the seabed topography.

[0036] Preferably, step S05 involves decomposing the original optical image into a base brightness layer, a detail texture layer, and a color information layer, including:

[0037] S51. Using Retinex theory, the original optical image is decomposed into a base brightness layer and a reflection component layer.

[0038] S52. Extract the detail texture layer from the reflection component layer using a bilateral filtering algorithm;

[0039] S53. Subtract the detail texture layer from the original image to generate the color information layer, or input the original optical image into the image decomposition model to obtain the basic brightness layer, detail texture layer and color information layer.

[0040] The image decomposition model was trained using the original optical image dataset and corresponding hierarchical labels.

[0041] The raw optical image dataset was acquired using an over-the-horizon underwater optical imaging device under different seabed topographical conditions (such as coral reefs, deep-sea plains, and hydrothermal vents).

[0042] The layered labels are generated by manually annotating the basic brightness layer, the detail texture layer, and the color information layer.

[0043] Preferably, the generation of the high-precision texture mapping map in step S05 includes:

[0044] S54. Align the structural contour feature map with the detail texture layer at the pixel level;

[0045] S55. Merge structural contours and detail textures using a weighted fusion algorithm;

[0046] S56. Optimize the edge smoothness of the fusion result through guided filtering algorithm to generate a high-precision texture map, or input the structural contour feature map and detail texture layer into the texture mapping model to obtain a high-precision texture map;

[0047] The texture mapping model is trained using a structural contour feature map dataset, a detail texture layer dataset, and corresponding high-precision texture map labels.

[0048] The structural contour feature map dataset was obtained through collaborative detection by a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0049] The detailed texture layer dataset was acquired using an underwater optical imaging device beyond visual range; the high-precision texture map labels were generated through manual annotation.

[0050] Preferably, the step S06 of obtaining the seabed topographic fusion image includes:

[0051] S61. Map the base brightness layer and the high-precision texture map to independent color spaces to obtain the first mapped image corresponding to the base brightness layer and the second mapped image corresponding to the high-precision texture map, respectively.

[0052] S62. Extract the brightness channel containing illumination distribution information from the first mapped image;

[0053] S63. Extract the texture channel and color channel containing geometric feature information from the second mapped image;

[0054] S64. Merge the luminance channel, texture channel, and color channel to generate a merged image of the seabed topography.

[0055] The present invention has at least the following beneficial effects:

[0056] 1. By using dynamic mesh generation technology, an initial mesh model matching the resolution of optical images is generated from sonar point cloud data. The sonar structural contour features and optical detail texture layers are then integrated, effectively solving the resolution discontinuity problem of sonar, laser, and optical equipment. This avoids texture blurring or loss of details caused by traditional interpolation or downsampling, and significantly improves the geometric accuracy and texture fidelity of seabed topographic images.

[0057] 2. A water body attenuation compensation factor is introduced, and a dynamic mathematical model of suspended matter concentration and attenuation coefficient is established based on Mie scattering theory. The scattering noise of laser scanning and optical images is corrected by weighted averaging, and the forward / backward scattering compensation mechanism is distinguished, which significantly improves the imaging robustness in turbid waters and reduces feature distortion caused by suspended matter interference.

[0058] 3. Image layer decomposition technology is used to separate the basic brightness layer, detail texture layer and color layer of the optical image. Combined with the morphological optimization of the sonar structure contour, a high-precision texture mapping map is generated through pixel-level alignment and weighted fusion to achieve fine restoration of the seabed micro-topography.

[0059] 4. The illumination distribution information of the base brightness layer and the geometric features of the texture map are separated and processed in an independent color space. The consistency of illumination and the authenticity of geometric structure are preserved through channel-level fusion, which solves the problem of uneven illumination caused by water absorption or equipment differences, and outputs a seabed landscape fusion image with natural colors and clear details. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of a method for processing seabed topography images provided in Embodiment 1 of the present invention;

[0062] Figure 2 This is a flowchart of S02 provided in Embodiment 1 of the present invention;

[0063] Figure 3 A flowchart of S03 provided in Embodiment 1 of the present invention;

[0064] Figure 4 A flowchart of S04 provided in Embodiment 1 of the present invention;

[0065] Figure 5 This is a flowchart of S05 provided in Embodiment 1 of the present invention;

[0066] Figure 6 The flowchart for S06 provided in Embodiment 1 of the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0069] Example 1

[0070] This embodiment provides a method for processing seabed topographic images, including the following steps, combined with Figure 1 As shown:

[0071] S01. When multi-beam sonar and optical equipment are used for collaborative detection, acquire sonar point cloud data and original optical images; wherein, the sonar point cloud data consists of spatial coordinate components, reflection intensity components and beam incidence angle components.

[0072] Specifically, this step includes a multi-beam sonar system and optical equipment (laser scanning device and beyond-line-of-sight underwater optical imaging equipment), wherein:

[0073] Multibeam sonar systems use the transmit and receive beam crossover block as the smallest detection unit, and the area of ​​its acquisition region is constrained by both the transmit beam width and the seabed depth.

[0074] Laser scanning devices can acquire high-precision point cloud data, but they are easily affected by suspended matter in turbid waters, resulting in scattering noise.

[0075] Over-the-horizon underwater optical imaging equipment (such as the LRC-20 camera) achieves micron-level imaging accuracy (approximately 5.86 microns) and can capture seabed micro-topographic features, but its effective detection range is limited by the water attenuation coefficient.

[0076] However, during the fusion of multi-source data, there are resolution gaps (orders of magnitude difference) and dynamic interference distortion problems. Traditional interpolation or downsampling methods can lead to texture blurring or loss of detail.

[0077] Furthermore, the transmit beam width of the multibeam sonar system includes an adjustable range of 3°-5°, and the corresponding seabed coverage width increases linearly with depth.

[0078] The operating wavelength of the over-the-horizon underwater optical imaging device includes the blue-green light window of 450nm-550nm, which has the lowest attenuation coefficient in water.

[0079] The spatial coordinate components of sonar point cloud data include three-dimensional coordinates of X, Y, and Z; the reflection intensity component is related to the acoustic impedance of the seabed medium.

[0080] The beam incidence angle component affects the spatial distribution density of point cloud data.

[0081] Using sonar point clouds as vertices, a triangular mesh topology is generated. Mesh density is constrained inversely based on the optical image resolution (e.g., 0.05 mm / pixel). For example, in shallow water (where sonar resolution is high), small meshes are generated to match optical details; in deep water (where sonar resolution is low), the mesh is densified through interpolation. The larger the incident angle, the sparser the spatial distribution of the point cloud. The algorithm automatically identifies low-density areas and uses local weighted regression to smooth boundaries, avoiding mesh breakage. It should be noted that the vertex spacing of the sonar mesh (e.g., 1 m) differs from the optical pixel scale (5.86 μm) by more than five orders of magnitude. Matching is achieved through a layered mapping strategy: 1. The sonar mesh defines the macroscopic geometry of the terrain (e.g., ridge orientation). 2. Optical image pixels are mapped to triangular facets of the sonar mesh, with each facet carrying tens of thousands of optical pixels, achieving a coupling of "coarse framework + fine texture".

[0082] S02. Dynamically divide the sonar point cloud data into grids to generate an initial grid model that matches the resolution of the optical image.

[0083] Specifically, such as Figure 2 As shown, the dynamic meshing of sonar point cloud data in the above steps includes:

[0084] S21. Use the octree spatial partitioning algorithm to divide the sonar point cloud data into multiple subspaces;

[0085] S22. Within each subspace, the mesh size is adaptively adjusted according to the point cloud density;

[0086] S23. Smooth the grid boundaries using a local weighted regression algorithm to generate a continuous initial grid model, or input the sonar point cloud data into a pre-trained dynamic grid generation model to obtain an initial grid model that matches the resolution of the optical image.

[0087] S24. The dynamic mesh generation model is trained using a sonar point cloud dataset and corresponding optical image resolution labels.

[0088] S25. The sonar point cloud dataset was obtained by a multibeam sonar system under different seabed topography; the optical image resolution labels were obtained by an over-the-horizon underwater optical imaging device at the same detection location and generated by manual annotation.

[0089] In the above technique, the overall space of the sonar point cloud is first recursively divided into an octree structure, with each subspace representing a cubic region of different scales. Within each subspace, the point cloud density (e.g., the number of points per unit volume) is calculated in real time.

[0090] If the density is higher than the threshold (such as in a shallow water steep slope area), the subspace is further divided into smaller levels to generate a fine-grained mesh to match the optical high resolution;

[0091] If the density is below a threshold (such as in a flat area of ​​deep water), the segmentation is stopped, and the coarse-grained mesh is retained to avoid redundant calculations.

[0092] After segmentation, the mesh boundaries of adjacent subspaces are smoothly interpolated using a local weighted regression algorithm to eliminate mesh gaps caused by abrupt density changes and generate a continuous and seamless initial mesh model.

[0093] When dealing with complex terrains (such as seamounts and hydrothermal vents), the octree algorithm may lose geomorphic features due to regular geometric segmentation. In such cases, a pre-trained dynamic mesh generation model is employed. The model input consists of point cloud data collected by multibeam sonar in different typical terrains (flat seabed, slopes, seamounts), labeled with the resolution annotation data of the corresponding optical camera (e.g., LRC-20). The model learns the mapping relationship from the spatial distribution features of the point cloud (such as curvature variations and elevation variance) to the target mesh size. Inputting real-time sonar point clouds, the model automatically identifies the terrain category and outputs a non-uniform mesh topology. For example, it outputs a high-density triangular mesh at the edges of steep slopes (to conform to optical details) and a sparse quadrilateral mesh in flat areas (to improve computational efficiency).

[0094] S03. Embed a water body attenuation compensation factor into the initial mesh model to correct the scattering noise of laser scanning and optical images in turbid waters, so as to obtain the corrected initial mesh model.

[0095] Specifically, such as Figure 3 As shown, the water attenuation compensation factor embedded in the initial grid model includes:

[0096] S31. Calculate the attenuation coefficient at different wavelengths based on real-time monitoring data of suspended solids concentration in water.

[0097] S32. At each node of the initial mesh model, a dynamic weight function related to the attenuation coefficient is introduced;

[0098] S33. An anti-interference mesh model is generated by correcting the scattering noise of laser scanning and optical images through weighted averaging.

[0099] Furthermore, it also includes:

[0100] S34. Establish a mathematical model of the concentration and attenuation coefficient of suspended solids in water using the Mie scattering theory;

[0101] S35. Fit the measured data using the least squares method to optimize the mathematical model;

[0102] S36. Embed the optimized mathematical model into each node of the initial mesh model to generate a dynamic attenuation compensation field, or input the initial mesh model into the attenuation compensation model to obtain an anti-interference mesh model.

[0103] The attenuation compensation model was trained using the initial grid model dataset and the corresponding anti-interference grid model labels;

[0104] The initial grid model dataset was obtained through collaborative detection using a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0105] The anti-interference mesh model label is obtained by manually labeling the distribution of scattered noise in turbid water and generating a noise-free mesh model.

[0106] The above techniques establish a quantitative relationship between the concentration of suspended solids in water and the attenuation of light / sound waves. Higher suspended solids concentrations result in stronger scattering effects at specific wavelengths (e.g., 532nm for laser light, 450–550nm for optical blue-green light). A dynamic weighting function is embedded in each node of the initial mesh model (e.g., a vertex of a Delaunay triangle). ,in: Here, is the wavelength-dependent attenuation coefficient, d is the real-time suspended matter concentration, and C is the propagation path length. Forward scattering compensation terms (for laser scanning) and backscattering compensation terms (for optical imaging) are used to correct signal distortion caused by different scattering mechanisms. Forward scattering causes laser point cloud diffusion, while backscattering causes optical image fogging; these two require differentiated processing.

[0107] When the physical model parameters are unknown, a pre-trained attenuation compensation model (deep learning network) is used: The input is a noisy initial mesh model (including sonar coordinates and optical textures). The output is an interference-resistant mesh model. The model training is based on a real-world dataset: sonar-optical data are simultaneously collected in various turbid water bodies (such as the Yellow River estuary and the Yangtze River estuary), and noise-free mesh labels are manually added. The network directly predicts the compensation field by learning the spatial distribution patterns of scattered noise (such as the aggregation characteristics of near-bottom suspended matter).

[0108] S04. Based on the corrected initial grid model, extract the structural contour features of the sonar data;

[0109] Specifically, such as Figure 4 As shown, the above steps include:

[0110] S41. Use an edge detection algorithm to extract structural contour lines from the corrected initial mesh model;

[0111] S42. Optimize the continuity of contour lines through morphological operations;

[0112] S43. Convert the optimized contour lines into a binary image to generate a structural contour feature map, or input the corrected initial mesh model into the structural contour extraction model to obtain a structural contour feature map.

[0113] The structural contour extraction model was trained using the modified initial mesh model dataset and the corresponding structural contour labels.

[0114] The revised initial grid model dataset was obtained through collaborative detection using a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0115] The structural outline labels are generated by manually annotating the structural outlines of the seabed topography.

[0116] In the aforementioned technology, based on real-time monitored water suspended solids concentration data, a mathematical model of the attenuation coefficients of light and sound waves at different wavelengths is established using Mie scattering theory. This model correlates parameters such as suspended solids particle size distribution and refractive index with the attenuation intensity, and optimizes the model parameters by fitting measured data using the least squares method. A dynamic weighting function is embedded at each node of the initial mesh model (i.e., the vertex of the Delaunay triangular mesh). ,in, Let β(λ) be the attenuation coefficient corresponding to wavelength β(λ), and d be the propagation path length from the node to the sensor. This function is applied to laser point clouds (forward scattering compensation) and optical images (backscattering compensation), respectively:

[0117] Laser scanning: compensates for point cloud position drift caused by forward scattering of suspended objects, and restores the real terrain by adjusting the point cloud coordinate weights;

[0118] Optical imaging: Suppresses image haze caused by backscattering and enhances texture contrast.

[0119] Finally, an anti-interference mesh model is generated, whose node coordinates and attribute values ​​are corrected by weighted averaging to eliminate the influence of dynamic noise.

[0120] S05. Decompose the original optical image into a basic brightness layer, a detail texture layer, and a color information layer, and then fuse the structural contour features and the detail texture layer to generate a high-precision texture map.

[0121] Specifically, such as Figure 5 As shown, the above steps decompose the original optical image into a base brightness layer, a detail texture layer, and a color information layer, including:

[0122] S51. Using Retinex theory, the original optical image is decomposed into a base brightness layer and a reflection component layer.

[0123] S52. Extract the detail texture layer from the reflection component layer using a bilateral filtering algorithm;

[0124] S53. Subtract the detail texture layer from the original image to generate the color information layer, or input the original optical image into the image decomposition model to obtain the basic brightness layer, detail texture layer and color information layer.

[0125] The image decomposition model was trained using the original optical image dataset and its corresponding hierarchical labels.

[0126] The raw optical image dataset was acquired using an over-the-horizon underwater optical imaging device under different seabed topographical conditions (such as coral reefs, deep-sea plains, and hydrothermal vents).

[0127] Layered labels are generated by manually annotating the basic brightness layer, detail texture layer, and color information layer.

[0128] After obtaining the anti-interference mesh model, this step first uses an edge detection algorithm (such as the Canny operator) to scan the gradient abrupt change regions of the mesh model. Since the corrected mesh has embedded water attenuation compensation factors (forward / backward scattering compensation), the reflection intensity values ​​of the mesh nodes are closer to the characteristics of the real seabed medium. The Canny operator identifies terrain abrupt change lines (such as trench and ridge boundaries) through double-threshold hysteresis processing (high threshold captures strong edges, low threshold connects weak edges), generating an initial contour. Subsequently, morphological operations (dilation and erosion) repair the broken contour: dilation fills small gaps caused by noise or resolution differences, and erosion removes isolated noise points, ultimately generating a continuous closed structural contour. This contour is converted into a binary feature map (contour line as 1, background as 0), serving as the geometric skeleton of the sonar data.

[0129] The image is treated as the product of the illumination and reflection components. First, a multi-scale Gaussian convolution is performed on the original optical image to simulate the attenuation effect of light in seawater, separating the basic brightness layer (containing the light and dark distribution caused by non-uniform illumination and light absorption by the water). Next, the reflection component layer is processed by bilateral filtering: its spatial domain Gaussian kernel preserves large-scale smooth regions, while its value domain Gaussian kernel protects areas of abrupt gradient changes (such as sand ripples and reef edges), thereby extracting the detail texture layer (containing only high-frequency information of micro-topography). Finally, the difference between the original image and the detail texture layer is calculated to extract the color information layer (containing low-frequency color features such as sediment color and biological attachment).

[0130] Furthermore, in the above embodiments, generating a high-precision texture map includes:

[0131] S54. Align the structural contour feature map with the detail texture layer at the pixel level;

[0132] S55. Merge structural contours and detail textures using a weighted fusion algorithm;

[0133] S56. Optimize the edge smoothness of the fusion result through guided filtering algorithm to generate a high-precision texture map, or input the structural contour feature map and detail texture layer into the texture mapping model to obtain a high-precision texture map;

[0134] The texture mapping model is trained using a structural contour feature map dataset, a detail texture layer dataset, and corresponding high-precision texture map labels.

[0135] The structural contour feature map dataset was obtained through collaborative detection using a multibeam sonar system and an over-the-horizon underwater optical imaging device.

[0136] The detailed texture layer dataset was acquired using an underwater optical imaging device beyond visual range; the high-precision texture map labels were generated through manual annotation.

[0137] In the aforementioned technique, the binarized structural contour feature map extracted by sonar (such as the edge of a trench) is first coordinate-registered with the detailed texture layer of the optical image (such as sediment texture). Since the sonar mesh model has been dynamically partitioned and matched to the optical resolution, the two have a basis for alignment in spatial scale. An affine transformation matrix is ​​used to correct for micrometer-level offsets caused by platform sway or differences in detection angle, ensuring strict overlap between the contour lines and texture pixels. An α-mixing algorithm is employed to dynamically balance the contribution weights of the contour and texture.

[0138] Give higher weight (α>0.7) to the sonar structure near the outline (such as the mid-ocean ridge boundary) to enhance the rigidity of the geometric framework;

[0139] In flat areas (such as deep-sea plains), the focus is on optical texture weight (α<0.3) to preserve the fine layers of sediments.

[0140] After fusion, edge optimization is performed using guided filtering: the sonar contour is used as a guide map to constrain the filtering direction of the texture layer, ensuring a natural transition of the fused boundary along the terrain and eliminating jagged edges. When a pre-trained texture mapping model is used, the model learns cross-modal feature associations through an encoder-decoder architecture.

[0141] Encoder side: A two-branch CNN is used to extract the topological features (such as edge curvature) of the contour map and the local statistical features (such as gradient distribution) of the texture layer, respectively.

[0142] Fusion layer: Automatically focuses on key areas (such as the boundary between steep and gentle slopes) through an attention mechanism to suppress false textures in optical images caused by water refraction;

[0143] Decoder side: During upsampling reconstruction, an adversarial loss function (GAN) is introduced to ensure that the output texture map has both structural accuracy and visual realism.

[0144] S06. Extract the illumination distribution information from the basic brightness layer and the geometric feature information from the high-precision texture mapping map respectively, and fuse the illumination distribution information and geometric feature information to obtain the seabed landform fusion image.

[0145] Specifically, such as Figure 6 As shown, the above steps include:

[0146] S61. Map the base brightness layer and the high-precision texture map to independent color spaces to obtain the first mapped image corresponding to the base brightness layer and the second mapped image corresponding to the high-precision texture map, respectively.

[0147] S62. Extract the brightness channel containing illumination distribution information from the first mapped image;

[0148] S63. Extract the texture channel and color channel containing geometric feature information from the second mapped image;

[0149] S64. Merge the luminance channel, texture channel, and color channel to generate a merged image of the seabed topography.

[0150] In the aforementioned technology, the base brightness layer mapping converts the base brightness layer containing global illumination information (such as sunlight transmission and equipment supplemental lighting) to the Lab color space, extracting the L channel as a carrier of illumination distribution. This channel is stripped of color interference and purified into a grayscale brightness gradient of 0~100%, accurately reflecting the absorption / scattering effect of water on light (especially the attenuation characteristics of blue-green light bands in the 450-550nm range).

[0151] Texture mapping involves simultaneously converting a high-precision texture map (containing geometric structure and micro-topographic texture) to Lab space, separating the a channel (red and green hues) and the b channel (blue and yellow hues). Both channels together convey the reflective properties of the seabed medium (such as the red of coral and the yellow and brown of sediments), while texture details are expressed through the spatial frequency variations of the a / b channels.

[0152] Furthermore, adaptive histogram equalization is performed on the L channel to compress overexposed areas (bright spots) and brighten dark areas (the bottom of the trench), eliminating illumination gaps caused by water attenuation. Simultaneously, a turbidity compensation coefficient (based on the Mie scattering model) is injected to suppress brightness abrupt changes caused by suspended matter.

[0153] The corrected L channel is recombine with the optimized a / b channels to generate a Lab color space fused image. The L channel weight is dynamically adjusted (deep water area weight +15% to compensate for light attenuation) to ensure consistent brightness across the entire scene. When converting back to RGB color space, a color gamut clipping algorithm is used to project colors outside the display's color gamut towards the displayable boundary (such as the intense orange-red of deep-sea hydrothermal vents), preventing color overflow. Simultaneously, the human eye's visual response curve is overlaid to enhance the visual clarity of details in dark areas.

[0154] This embodiment employs dynamic mesh generation technology to generate an initial mesh model from sonar point cloud data that matches the resolution of the optical image. It then integrates sonar structural contour features with optical detail texture layers, effectively resolving the resolution fragmentation problem between sonar, laser, and optical equipment. This avoids texture blurring or detail loss caused by traditional interpolation or downsampling, significantly improving the geometric accuracy and texture fidelity of seabed topographic images. Secondly, a water attenuation compensation factor is introduced. Based on Mie scattering theory, a dynamic mathematical model of suspended matter concentration and attenuation coefficient is established. Weighted averaging corrects scattering noise from laser scanning and optical images, and distinguishes between forward and backward scattering compensation mechanisms, significantly improving imaging robustness in turbid waters and reducing feature distortion caused by suspended matter interference. Furthermore, image layer decomposition technology separates the basic brightness layer, detail texture layer, and color layer of the optical image. Combined with morphological optimization of the sonar structural contour, pixel-level alignment and weighted fusion generate a high-precision texture map, achieving refined restoration of seabed micro-topography. Furthermore, the illumination distribution information of the base brightness layer and the geometric features of the texture map are processed separately in an independent color space. Channel-level fusion preserves the consistency of illumination and the authenticity of geometric structure, solving the problem of uneven illumination caused by water absorption or equipment differences, and outputting a fused image of the seabed with natural colors and clear details.

[0155] Example 2

[0156] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps:

[0157] When multibeam sonar and optical equipment work together to detect, sonar point cloud data and raw optical images are acquired; the sonar point cloud data consists of spatial coordinate components, reflection intensity components and beam incidence angle components.

[0158] Dynamically mesh the sonar point cloud data to generate an initial mesh model that matches the resolution of the optical image;

[0159] A water attenuation compensation factor is embedded in the initial mesh model to correct the scattering noise of laser scanning and optical images in turbid waters, so as to obtain the corrected initial mesh model.

[0160] Based on the corrected initial grid model, structural contour features of the sonar data are extracted;

[0161] The original optical image is decomposed into a basic brightness layer, a detail texture layer, and a color information layer. Then, the structural contour features and the detail texture layer are fused to generate a high-precision texture map.

[0162] Illumination distribution information from the base brightness layer and geometric feature information from the high-precision texture map are extracted separately, and then the illumination distribution information and geometric feature information are fused to obtain a fused image of the seabed topography.

[0163] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0164] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0165] Example 3

[0166] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps:

[0167] When multibeam sonar and optical equipment work together to detect, sonar point cloud data and raw optical images are acquired; the sonar point cloud data consists of spatial coordinate components, reflection intensity components and beam incidence angle components.

[0168] Dynamically mesh the sonar point cloud data to generate an initial mesh model that matches the resolution of the optical image;

[0169] A water attenuation compensation factor is embedded in the initial mesh model to correct the scattering noise of laser scanning and optical images in turbid waters, so as to obtain the corrected initial mesh model.

[0170] Based on the corrected initial grid model, structural contour features of the sonar data are extracted;

[0171] The original optical image is decomposed into a basic brightness layer, a detail texture layer, and a color information layer. Then, the structural contour features and the detail texture layer are fused to generate a high-precision texture map.

[0172] Illumination distribution information from the base brightness layer and geometric feature information from the high-precision texture map are extracted separately, and then the illumination distribution information and geometric feature information are fused to obtain a fused image of the seabed topography.

[0173] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for processing seabed topographic images, characterized in that, The method includes the following steps: S01. When multi-beam sonar and optical equipment are used for collaborative detection, acquire sonar point cloud data and original optical images; wherein, the sonar point cloud data consists of spatial coordinate components, reflection intensity components and beam incidence angle components. S02. Dynamically divide the sonar point cloud data into grids to generate an initial grid model that matches the resolution of the optical image. S03. Embed a water body attenuation compensation factor into the initial mesh model to correct the scattering noise of laser scanning and optical images in turbid waters, so as to obtain the corrected initial mesh model. S04. Based on the corrected initial grid model, extract the structural contour features of the sonar data; S05. Decompose the original optical image into a basic brightness layer, a detail texture layer, and a color information layer, and then fuse the structural contour features with the detail texture layer to generate a high-precision texture mapping map; S06. Extract the illumination distribution information from the basic brightness layer and the geometric feature information from the high-precision texture map, and fuse the illumination distribution information and the geometric feature information to obtain a fused image of the seabed topography.

2. The method for processing seabed topography images according to claim 1, characterized in that, The dynamic grid division of the sonar point cloud data in step S02 includes: S21. Use the octree spatial partitioning algorithm to divide the sonar point cloud data into multiple subspaces; S22. Within each subspace, the mesh size is adaptively adjusted according to the point cloud density; S23. Smooth the grid boundaries using a local weighted regression algorithm to generate a continuous initial grid model, or input the sonar point cloud data into a pre-trained dynamic grid generation model to obtain an initial grid model that matches the resolution of the optical image. S24. The dynamic mesh generation model is trained using a sonar point cloud dataset and corresponding optical image resolution labels. S25. The sonar point cloud dataset is obtained by a multibeam sonar system under different seabed topographic conditions; the optical image resolution label is obtained by an over-the-horizon underwater optical imaging device at the same detection location and is generated by manual annotation.

3. The method for processing seabed topography images according to claim 1, characterized in that, The step S03, which involves embedding a water attenuation compensation factor into the initial grid model, includes: S31. Calculate the attenuation coefficient at different wavelengths based on real-time monitoring data of suspended solids concentration in water. S32. At each node of the initial mesh model, a dynamic weighting function related to the attenuation coefficient is introduced; S33. An anti-interference mesh model is generated by correcting the scattering noise of laser scanning and optical images through weighted averaging.

4. The method for processing seabed topography images according to claim 3, characterized in that, The water attenuation compensation factor embedded in the initial grid model in step S03 further includes: S34. Establish a mathematical model of the concentration and attenuation coefficient of suspended solids in water using the Mie scattering theory; S35. Fit the measured data using the least squares method to optimize the mathematical model; S36. Embed the optimized mathematical model into each node of the initial mesh model to generate a dynamic attenuation compensation field, or input the initial mesh model into the attenuation compensation model to obtain an anti-interference mesh model.

5. The method for processing seabed topography images according to claim 1, characterized in that, The structural contour features extracted from the sonar data in step S04 include: S41. Use an edge detection algorithm to extract structural contour lines from the corrected initial mesh model; S42. Optimize the continuity of contour lines through morphological operations; S43. Convert the optimized contour lines into a binary image to generate a structural contour feature map, or input the corrected initial mesh model into the structural contour extraction model to obtain a structural contour feature map.

6. The method for processing seabed topography images according to claim 1, characterized in that, In step S05, the original optical image is decomposed into a base brightness layer, a detail texture layer, and a color information layer, including: S51. Using Retinex theory, the original optical image is decomposed into a base brightness layer and a reflection component layer. S52. Extract the detail texture layer from the reflection component layer using a bilateral filtering algorithm; S53. Subtract the detail texture layer from the original image to generate the color information layer, or input the original optical image into the image decomposition model to obtain the basic brightness layer, detail texture layer and color information layer.

7. The method for processing seabed topography images according to claim 1, characterized in that, The generation of the high-precision texture mapping map in step S05 includes: S54. Align the structural contour feature map with the detail texture layer at the pixel level; S55. Merge structural contours and detail textures using a weighted fusion algorithm; S56. Optimize the edge smoothness of the fusion result through guided filtering algorithm to generate a high-precision texture map, or input the structural contour feature map and detail texture layer into the texture mapping model to obtain a high-precision texture map.

8. The method for processing seabed topography images according to claim 1, characterized in that, The step S06 of obtaining the seabed topography fusion image includes: S61. Map the base brightness layer and the high-precision texture map to independent color spaces to obtain the first mapped image corresponding to the base brightness layer and the second mapped image corresponding to the high-precision texture map, respectively. S62. Extract the brightness channel containing illumination distribution information from the first mapped image; S63. Extract the texture channel and color channel containing geometric feature information from the second mapped image; S64. Merge the luminance channel, texture channel, and color channel to generate a merged image of the seabed topography.

9. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the seabed topography image processing method as described in any one of claims 1-8.

10. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the seabed topography image processing method as described in any one of claims 1-8.

Images