Underwater concrete three-dimensional reconstruction method and system based on fusion of linear structured light and stereoscopic vision

Abstract

The invention relates to the technical field of underwater concrete three-dimensional reconstruction, in particular to an underwater concrete three-dimensional reconstruction method and system based on linear structured light and stereoscopic vision fusion. The method comprises the steps of obtaining a depth map and a texture map of underwater concrete based on a line structured light module and a stereoscopic vision module, and obtaining a preliminary depth data sequence and a surface texture sequence; performing environment compensation and image enhancement on the depth data and the texture data to generate data after environment correction; based on the corrected data, performing registration and matching on the depth and texture data by using a multi-modal fusion algorithm to generate a fusion point cloud data model; motion compensation and deformation repair are carried out on the point cloud model, deformation and structure offset caused by water flow are corrected, and a repaired point cloud model is generated; the repaired point cloud is input into a three-dimensional model reconstruction engine, a high-precision underwater concrete three-dimensional model is generated, intelligent analysis is conducted on the surface of the model, and damage and structural integrity of the model are evaluated. The underwater concrete three-dimensional reconstruction method effectively improves the precision of underwater concrete three-dimensional reconstruction, solves the problem of model distortion in complex environments such as water flow and turbid water quality, and has efficient and accurate structure detection and maintenance functions.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater three-dimensional reconstruction, and in particular to a method and system for underwater concrete three-dimensional reconstruction based on the fusion of line structured light and stereoscopic vision. Background Art

[0002] At present, the inspection and maintenance of underwater concrete structures face many challenges, especially in complex underwater environments. Due to factors such as water flow, turbid water quality, and unstable lighting, traditional underwater structure inspection methods are difficult to obtain high-precision three-dimensional data models. Existing underwater three-dimensional reconstruction technologies mostly rely on a single sensor or imaging method, such as sonar imaging or laser scanning. These technologies often cannot effectively deal with the impact of water flow on point cloud deformation, and there are large errors when processing complex textures on the surface of underwater structures.

[0003] Especially in dynamic water flow environments, the deformation and offset of the structure surface caused by water flow pose great challenges to the 3D reconstruction process. Traditional reconstruction methods are unable to effectively compensate for the deformation caused by water flow in a timely manner, resulting in serious errors in the reconstruction results. Therefore, how to use a variety of technical means to solve the fusion problem of depth data and surface texture data in complex underwater environments and correct the model deformation caused by water flow is still a key issue in current technology.

[0004] In response to this technical challenge, the underwater 3D reconstruction method based on the fusion of line structured light and stereo vision provides a new solution. By combining line structured light sensors and stereo vision sensors, high-precision depth acquisition and surface detail capture of underwater concrete structures can be achieved. At the same time, with the help of data fusion algorithms and adaptive repair technology, the deformation and structural offset problems caused by water flow can be effectively solved, and the accuracy and integrity of the 3D model of underwater concrete structures can be improved. Summary of the invention

[0005] The present invention provides a method and system for underwater concrete three-dimensional reconstruction based on the fusion of line structured light and stereo vision, so as to solve the problem of how to achieve high-precision and real-time three-dimensional reconstruction of underwater concrete structures in a complex underwater environment based on the fusion technology of line structured light and stereo vision, and correct the deformation and structural offset caused by water flow to ensure the accuracy and integrity of the model.

[0006] In order to solve the above technical problems, the present invention provides a method for underwater concrete 3D reconstruction based on the fusion of line structured light and stereo vision, comprising:

[0007] Based on the line structured light module and stereo vision module, the depth map and texture map of underwater concrete are obtained to obtain the preliminary depth data sequence and surface texture sequence;

[0008] Performing environmental compensation and image enhancement processing on the depth data sequence and the surface texture sequence, repairing uneven illumination and noise interference, and generating a depth map and a texture map after environmental correction;

[0009] Based on the depth map and texture map corrected by the environment, a multimodal fusion algorithm is used to register and match the depth and texture data to generate a fused point cloud data model;

[0010] Performing motion compensation and deformation repair on the fused point cloud data model to correct the spatial deformation and structural offset caused by water flow, and obtaining a repaired point cloud model sequence;

[0011] The repaired point cloud model sequence is input into a three-dimensional model reconstruction engine to generate a high-precision three-dimensional model of underwater concrete, and intelligent analysis is performed on the damage and structural integrity of the model surface.

[0012] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of obtaining the depth map and texture map of the underwater concrete specifically includes:

[0013] Use the line structured light module to project known structured light to obtain the depth data of underwater concrete;

[0014] The stereo vision module is used to collect texture images of underwater concrete from multiple perspectives to generate texture data.

[0015] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of performing environmental compensation and image enhancement on the depth data sequence and the surface texture sequence comprises:

[0016] Based on the light intensity adaptive compensation algorithm, the influence of uneven lighting is corrected;

[0017] Use spectral filtering image enhancement algorithm for denoising and contrast enhancement;

[0018] Perform artifact removal and edge optimization on the enhanced depth map and texture map.

[0019] Furthermore, the underwater concrete 3D reconstruction method based on line structured light and stereoscopic vision fusion, the step of registering and matching the depth and texture data using a multimodal fusion algorithm based on the depth map and texture map after environmental correction includes:

[0020] Perform pixel-level feature matching on the depth map and texture map to generate a multimodal matching data sequence;

[0021] A three-dimensional point cloud data model is constructed based on the multimodal matching data sequence.

[0022] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of generating a fused point cloud data model comprises:

[0023] Use 3D reconstruction algorithm to fuse depth data with texture data to generate a point cloud data model;

[0024] The point cloud data model is preliminarily optimized.

[0025] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of performing motion compensation and deformation repair on the fused point cloud data model comprises:

[0026] Based on water flow environment parameters and dynamic target detection algorithm, the influence of water flow on the motion of point cloud model is analyzed to generate motion compensation parameters;

[0027] The deformation and offset of the point cloud model caused by the water flow are corrected based on the motion compensation parameters.

[0028] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of inputting the repaired point cloud model sequence into the 3D model reconstruction engine to generate a high-precision 3D model includes:

[0029] Generate a geometric model of underwater concrete based on the repaired point cloud data using a 3D modeling engine;

[0030] The geometric model is optimized in detail and its accuracy is improved.

[0031] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of intelligently analyzing the damage and structural integrity of the model surface comprises:

[0032] Based on the depth change and surface texture change, automatic defect detection algorithm is applied to identify damage;

[0033] A multi-scale analysis is performed on the damaged area to identify structural defects and calculate the damage extent.

[0034] Furthermore, in the underwater concrete 3D reconstruction method based on the fusion of line structured light and stereoscopic vision, the step of generating an intelligent analysis report comprises:

[0035] Assess damage location, damage extent and its impact on structural integrity and generate a complete underwater concrete structure health report;

[0036] Based on the evaluation results, repair suggestions and optimization solutions are provided.

[0037] Furthermore, an underwater concrete 3D reconstruction system based on the fusion of line structured light and stereoscopic vision is provided, the system comprising:

[0038] A data acquisition module that acquires depth and texture data of underwater concrete surfaces from a variety of sensors;

[0039] The data processing module performs preliminary processing on the collected depth data and surface texture data;

[0040] The multimodal data fusion module uses a multimodal fusion algorithm to register and match the depth data and texture data based on the depth map and texture map after environmental correction to generate a fused point cloud data model;

[0041] A deformation repair and motion compensation module performs motion compensation and deformation repair on the fused point cloud data model to correct the spatial deformation and structural offset caused by water flow;

[0042] The 3D model reconstruction and damage detection module generates a high-precision 3D model of underwater concrete based on the repaired point cloud model, and performs intelligent analysis on the surface damage and structural integrity of the model;

[0043] Intelligent analysis and report generation module, which generates detailed intelligent analysis reports based on the repaired 3D model and damage detection results, and provides structural repair suggestions;

[0044] The system monitoring and adaptive optimization module continuously monitors the operating status of the system and adaptively optimizes model parameters and system settings according to environmental changes.

[0045] The key innovative features of the present invention include:

[0046] (1) Line structured light and stereo vision fusion technology: The present invention combines a line structured light sensor with a stereo vision sensor and utilizes the data fusion of the two to achieve high-precision underwater concrete surface reconstruction, solving the problem that a single sensor cannot efficiently capture both depth and texture information at the same time.

[0047] (2) Environmental compensation and image enhancement: By performing environmental compensation and image enhancement on depth data and texture data, uneven lighting and noise interference are effectively repaired, thereby ensuring high data quality and providing accurate input for subsequent 3D reconstruction and analysis.

[0048] (3) Multimodal data fusion algorithm: The multimodal data fusion algorithm is used to align and match the depth map and the texture map to generate a fused point cloud data model, thereby ensuring the accurate fusion of depth data and texture information and enhancing the authenticity and detail expression of the underwater 3D model.

[0049] The following are its main beneficial effects:

[0050] The present invention realizes high-precision three-dimensional reconstruction of underwater concrete structures by combining line structured light and stereoscopic vision fusion technology, and can effectively cope with the challenges brought by factors such as uneven lighting, turbid water and dynamic water flow in complex underwater environments. Compared with traditional single sensor systems, the present invention uses multimodal data fusion technology to accurately obtain the depth information and texture details of the underwater concrete surface, thereby greatly improving the accuracy and completeness of the three-dimensional model. In addition, the use of motion compensation and deformation repair algorithms can timely correct the point cloud deformation and structural offset caused by water flow, avoiding the influence of water flow interference on model accuracy. Overall, the present invention not only improves the accuracy and efficiency of underwater concrete three-dimensional reconstruction, but also effectively reduces manual intervention and optimizes the detection and maintenance process of underwater structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of a process for underwater concrete 3D reconstruction based on the fusion of line structured light and stereo vision provided in an embodiment of the present application;

[0052] Figure 2 This is a structural block diagram of an underwater concrete 3D reconstruction system based on the fusion of line structured light and stereo vision provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] Example 1: Reference Figure 1 , is a flow chart of a method for underwater concrete 3D reconstruction based on the fusion of line structured light and stereoscopic vision provided by an embodiment of the present invention. The process may at least include steps S100-S500:

[0056] S100, based on the line structured light module and the stereo vision module, a depth map and a texture map of underwater concrete are obtained to obtain a preliminary depth data sequence and a surface texture sequence.

[0057] S200: Performing environmental compensation and image enhancement processing on the depth data sequence and the surface texture sequence, repairing uneven illumination and noise interference, and generating a depth map and a texture map after environmental correction.

[0058] S300, based on the depth map and texture map after environment correction, use a multimodal fusion algorithm to align and match the depth and texture data to generate a fused point cloud data model.

[0059] S400, performing motion compensation and deformation repair on the fused point cloud data model, correcting the spatial deformation and structural offset caused by water flow, and obtaining a repaired point cloud model sequence.

[0060] S500, input the repaired point cloud model sequence into the 3D model reconstruction engine to generate a high-precision 3D model of underwater concrete, and perform intelligent analysis on the surface damage and structural integrity of the model.

[0061] Step S100 at least includes steps S110-S130:

[0062] S110 , acquiring structured light reflection data of the target underwater concrete surface based on the line structured light module, and generating a preliminary depth map sequence.

[0063] Laser light source projection: Based on the line structured light module, the projected light source is projected at a known angle θ i It is projected onto the target underwater concrete surface to generate line structured light stripes. During this process, the underwater ambient light intensity I e and laser power P l are recorded synchronously as environmental parameters for depth measurement.

[0064] Reflected light acquisition: The image sensor collects light at a preset frequency f c The reflected light signal is collected synchronously to generate a grayscale image G(x i ,y i ), where (x i ,y i ) represents pixel coordinates.

[0065] Depth calculation: through the projection point of line structured light (u i ,v i ) and the center point of the light streak captured by the camera (x i ,y i ), using the formula:

[0066]

[0067] Among them, D i is the depth value corresponding to pixel i, b is the baseline distance between the light source and the camera, and f is the focal length of the camera.

[0068] Depth map generation: depth value D of all pixels i Combine to form a preliminary depth map D = D i |i∈(1,n), the depth map is used for subsequent data fusion.

[0069] S120, collecting multi-view images of the target underwater concrete based on the stereo vision module to generate a preliminary surface texture sequence.

[0070] Binocular image acquisition: Use the left and right cameras of the binocular stereo vision module to collect two images with different viewing angles. L (x,y) and I R (x,y). The image resolution is set to (W,H), where W is the width and H is the height.

[0071] Disparity calculation: Use the disparity matching algorithm to match the feature points of the left and right images. The disparity formula is:

[0072] d(x,y)=|I L (x,y)-I R (x,y)|

[0073] Among them, d(x,y) is the disparity value of each pixel, I L (x,y) and I R (x, y) is the pixel grayscale value of the left and right views. Texture map generation: The disparity map obtained by disparity matching is combined with the left camera image I L (x, y), generate the surface texture map T(x, y).

[0074] S130 , performing time synchronization and spatial registration on the structured light depth map and the stereoscopic vision texture sequence to obtain a preliminary depth data sequence and a surface texture sequence.

[0075] Time synchronization: Use the global timestamp synchronization system to synchronize the depth map D output by the structured light module. i It is time synchronized with the texture map T(x,y) output by the stereo vision module to ensure that the two data sequences in the same time frame are accurately matched.

[0076] Spatial registration: In the depth map D i Define a unified three-dimensional coordinate system O(x,y,z) on the texture map T(x,y), and use the internal and external parameters to calibrate the matrix K int (internal parameter matrix) and K ext (External parameter matrix) for spatial registration, the formula is as follows:

[0077]

[0078] Among them, P i represents the coordinates of the 3D point cloud after registration, K int and K ext is the camera's intrinsic and extrinsic calibration matrix.

[0079] Joint sequence generation: Time synchronization and spatial registration of point clouds P i Recombine to form a joint data sequence M i =(x i ,y i ,D i ,T i )|i∈(1,n), which is used for subsequent data fusion and reconstruction steps.

[0080] Step S200 at least includes steps S210-S230:

[0081] S210 , performing illumination balance adjustment on the preliminary depth data sequence and the surface texture sequence based on a light intensity adaptive compensation algorithm.

[0082] Input data identification: the preliminary depth data sequence D obtained from step S130 i =D i,j |i∈(1,M),j∈(1,N) and texture image T i =T i,j |i∈(1,M),j∈(1,N) are input into the current module, and M×N is the total number of pixels.

[0083] Light intensity estimation: using the ambient light intensity I output by the light sensor e The mean brightness of the image Estimate the illumination distribution bias Ibias, the formula is:

[0084]

[0085] in, is the average brightness value of the image.

[0086] Adaptive compensation adjustment: for texture map T i and the depth map D i , perform illumination compensation operation, the correction formula is as follows:

[0087] and

[0088] Among them, T' i,j is the texture image pixel after illumination equalization, D' i,jis the corrected depth map pixel, and γ is the illumination compensation coefficient.

[0089] S220: Apply an image enhancement algorithm based on spectral filtering to perform image denoising and contrast enhancement on the depth data sequence and the surface texture sequence.

[0090] High-frequency noise filtering: The texture map T' is filtered by a spectral filter. i,j and depth map D' i,j Apply Fast Fourier Transform (FFT) to filter out high-frequency noise components. The formula is as follows:

[0091] F(u,v)=F(T' i,j ) and G(u,v)=F(D' i,j )

[0092] Among them, F(u,v) and G(u,v) represent the depth and texture maps in the frequency domain.

[0093] Spectral enhancement filter design: Design a Gaussian low-pass filter H(u,v) to suppress high-frequency noise:

[0094]

[0095] Filtering results:

[0096] T” i,j =F -1 (F(u,v)·H(u,v)) and D” i,j =F -1 (G(u,v)·H(u,v))

[0097] Among them, T i,j and D' i,j is the pixel value after denoising and enhancement, and σ is the standard deviation parameter of the filter.

[0098] Contrast enhancement: Apply the histogram equalization algorithm to adjust the brightness histogram distribution of the texture image and the depth image to a uniform range. The formula is:

[0099] T'' i,j =HE(T" i,j ) and D'' i,j =HE(D" i,j )

[0100] Here, HE(·) represents the histogram equalization operation.

[0101] S230 , performing artifact elimination and edge optimization on the depth map and texture map after illumination compensation and enhancement, to generate a depth map and texture map after environment correction.

[0102] Artifact detection and removal: In the depth map D''i,j With texture map T'' i,j In the above method, the edge of the object is extracted by an edge detection operator (such as the Canny operator) to detect the boundary of the artifact.

[0103] Artifact removal formula:

[0104]

[0105] in, and represents the depth and texture image pixels after artifact removal, τ d With τ t is the threshold parameter for edge detection.

[0106] Edge optimization processing: Use edge smoothing algorithm to enhance the border based on Laplace smoothing equation:

[0107]

[0108] in, represents the Laplace operator, and λ is the edge enhancement weight parameter.

[0109] Environmental correction result generation: The processed depth map and texture map data are combined and The output is used as a depth data sequence and a texture map sequence after environment correction for use by the next module S300.

[0110] Step S300 at least includes steps S310-S330:

[0111] S310, performing pixel-level feature matching on the depth map and texture map after environment correction based on a multi-view geometric matching algorithm to generate a multimodal matching data sequence.

[0112] Input data loading: Depth map after environment correction output from S230 and texture map Is input to the current module. Each pixel (i, j) contains depth value and texture feature value:

[0113]

[0114] Feature point extraction and descriptor construction: Use the scale-invariant feature transform (SIFT) or fast feature point detection (FAST) algorithm to extract the feature points from the Extract feature points (x k ,y k ), corresponding to the descriptor Φ k , the formula is as follows:

[0115] where k∈(1,K)

[0116] Among them, Φ k is the descriptor of the kth feature point, and K is the total number of feature points extracted.

[0117] Feature matching and multimodal association: Use the Euclidean distance matching algorithm to match feature points between multi-view images:

[0118]

[0119] Store matching point pairs as a multimodal matching data sequence:

[0120]

[0121] S320, constructing a three-dimensional point cloud for the multimodal matching data sequence to generate a preliminary point cloud model.

[0122] Three-dimensional reconstruction formula: For each pixel in the matching data sequence, the three-dimensional space point coordinates (X, Y, Z) are calculated according to the parallax principle:

[0123]

[0124] Where f is the focal length of the camera. b is the baseline distance of the binocular camera. (c x ,c y ) is the optical center coordinate of the camera.

[0125] Point cloud data generation: Combine each 3D point coordinate with its texture value to generate a preliminary point cloud model P i :

[0126]

[0127] The point cloud model is further passed to the error correction and coordinate registration module.

[0128] S330, based on the adaptive registration algorithm, coordinate calibration and error correction are performed on the preliminary point cloud model to generate a fused point cloud data model.

[0129] Coordinate system conversion: based on camera external parameter calibration matrix K ext And the internal calibration matrix K int , for the preliminary point cloud model P i Perform coordinate transformation and convert to the global coordinate system:

[0130]

[0131] Error correction: Use the nearest neighbor iterative closest point (ICP) algorithm to correct the errors in the point cloud data model. The correction formula is as follows:

[0132]

[0133] Among them, ΔP i is the point cloud error correction vector, which is generated by ICP optimization iteration.

[0134] Fusion point cloud generation: The corrected point cloud points Recombine and generate a fused point cloud data model:

[0135] P final = {P i *|i∈(1,N)}

[0136] Step S400 at least includes steps S410-S430:

[0137] S410, based on the fluid environment parameters and the dynamic target detection algorithm, analyzing the motion trajectory of the dynamic object in the water flow environment, and generating a motion compensation parameter sequence.

[0138] Input data loading: fused point cloud data model output from S330 Input to the current step, and at the same time input the water flow velocity V from the environmental sensor module w , fluid pressure P w and the water flow direction vector

[0139] Water flow model construction: Based on the fluid mechanics formula, a water flow fluid model is constructed to calculate the P of each point cloud point. i Movement displacement affected by water flow

[0140]

[0141] in, V is the water flow direction vector. w is the water flow velocity. i Point cloud point P i exposure time.

[0142] Dynamic target detection and compensation parameter calculation: Using Kalman filter to estimate the actual motion trajectory of point cloud points

[0143]

[0144] Where K is the Kalman gain, which represents the weighted proportion of the estimation error.

[0145] Motion compensation parameter generation: Motion compensation parameter sequence is generated for subsequent deformation repair steps.

[0146] S420, inputting the motion compensation parameter sequence into a deformation repair algorithm to correct the point cloud model distortion and structure offset caused by water flow.

[0147] Warp vector field construction: Based on the motion compensation parameter sequence Ω i , construct the distortion vector field of the point cloud model

[0148]

[0149] in, Point cloud point P i The distortion correction vector of .

[0150] Application of deformation repair algorithm: Use the nearest neighbor (NN) based deformation repair algorithm to repair the affected point cloud points P i Perform position correction:

[0151]

[0152] in, is the coordinate of the repaired point cloud point. λ is the deformation repair coefficient (dynamically adjusted according to the confidence of the point cloud point).

[0153] Structural constraint optimization: Use the structural constraint optimization model to correct the overall deformation of the point cloud model through the Laplace constraint optimization formula:

[0154]

[0155] in, It is the coordinates of the globally optimized point cloud points. is the geometric center point of the neighborhood points.

[0156] S430, performing three-dimensional spatial reconstruction and precision optimization on the repaired point cloud model to generate a repaired point cloud model sequence.

[0157] 3D spatial reconstruction of point cloud model: The repaired point cloud model Input to the 3D modeling engine and generate a 3D surface model through the triangular mesh reconstruction algorithm:

[0158]

[0159] Among them, M 3D The reconstructed 3D mesh model.

[0160] Model accuracy optimization: Use point cloud accuracy optimization algorithms to repair edge noise and surface gaps in the reconstructed model:

[0161]

[0162] Among them, Mopt A 3D model optimized for accuracy.

[0163] Repaired point cloud model output: The final generated repaired point cloud model sequence It is output to provide input data for subsequent 3D model analysis and defect detection modules.

[0164] Step S500 at least includes steps S510-S530:

[0165] S510, based on the repaired point cloud model sequence, use a 3D model reconstruction engine to construct a geometric model to generate a complete 3D model.

[0166] Input data loading: The repaired point cloud model sequence output from step S430 is input into the 3D model reconstruction engine.

[0167] 3D mesh generation: Use the Delaunay triangulation algorithm to construct a point cloud 3D mesh model M mesh , the formula is as follows:

[0168]

[0169] Among them, M mesh The generated triangular mesh model.

[0170] Surface fitting: Use the surface fitting algorithm based on normal vectors to smooth the point cloud mesh model and generate a reconstructed surface model M. surf , the formula is as follows:

[0171]

[0172] Among them, w i It is the surface weight, which is dynamically adjusted according to the normal vector direction and the distance between the neighboring points.

[0173] Model output: Output complete 3D model M final , for use by the defect detection module.

[0174] S520, based on the automatic defect detection algorithm, multi-scale damage identification is performed on the surface texture and depth changes of the 3D model.

[0175] Texture and depth map expansion: for the three-dimensional model M final Expand and generate surface texture map T model (x,y) and the depth map D model (x,y), ensure that the texture and depth map correspond to each other The coordinates of .

[0176] Application of damage detection algorithm: Apply multi-scale surface defect detection algorithm to detect potential damage points through depth changes and texture grayscale differences. def and texture outlier point T def :

[0177]

[0178]

[0179] in, Represents the gradient change. τ d With τ t is the detection threshold.

[0180] Damage feature extraction: Use morphological operators to dilate and detect edges of damage points and extract defect feature regions R def .

[0181] S530, perform feature analysis and integrity assessment on structural abnormality data in the three-dimensional model, and generate an intelligent analysis report of underwater concrete.

[0182] Structural abnormality detection: The detected damaged area R def Perform geometric analysis and calculate the damage area A def and depth change value D var :

[0183]

[0184] D var =max(D def )-min(D def )

[0185] Where ΔA(x,y) represents the area increment of each grid cell.

[0186] Integrity assessment: Use a multi-level structural integrity assessment model to calculate the structural health score S based on the damage area and depth change values. health :

[0187]

[0188] Among them, A model Represents the total surface area of ​​the three-dimensional model. model Represents the average depth value of the model.

[0189] w 1 and w 2 Weight factor for the structural health score.

[0190] Intelligent analysis report generation: Generate a complete structural anomaly detection report, including damage location, damage area, depth variation range and structural health score. This report will be used as the final output for underwater concrete structure monitoring and maintenance decision-making.

[0191] The key innovative features of the present invention include:

[0192] (1) Line structured light and stereo vision fusion technology: The present invention combines a line structured light sensor with a stereo vision sensor and utilizes the data fusion of the two to achieve high-precision underwater concrete surface reconstruction, solving the problem that a single sensor cannot efficiently capture both depth and texture information at the same time.

[0193] (2) Environmental compensation and image enhancement: By performing environmental compensation and image enhancement on depth data and texture data, uneven lighting and noise interference are effectively repaired, thereby ensuring high data quality and providing accurate input for subsequent 3D reconstruction and analysis.

[0194] (3) Multimodal data fusion algorithm: The multimodal data fusion algorithm is used to align and match the depth map and the texture map to generate a fused point cloud data model, thereby ensuring the accurate fusion of depth data and texture information and enhancing the authenticity and detail expression of the underwater 3D model.

[0195] The following are its main beneficial effects:

[0196] The present invention realizes high-precision three-dimensional reconstruction of underwater concrete structures by combining line structured light and stereoscopic vision fusion technology, and can effectively cope with the challenges brought by factors such as uneven lighting, turbid water and dynamic water flow in complex underwater environments. Compared with traditional single sensor systems, the present invention uses multimodal data fusion technology to accurately obtain the depth information and texture details of the underwater concrete surface, thereby greatly improving the accuracy and completeness of the three-dimensional model. In addition, the use of motion compensation and deformation repair algorithms can timely correct the point cloud deformation and structural offset caused by water flow, avoiding the influence of water flow interference on model accuracy. Overall, the present invention not only improves the accuracy and efficiency of underwater concrete three-dimensional reconstruction, but also effectively reduces manual intervention and optimizes the detection and maintenance process of underwater structures.

[0197] Embodiment 2: Figure 2 FIG. 1 is a block diagram of a system for 3D reconstruction of underwater concrete based on the fusion of line structured light and stereo vision according to an embodiment of the present invention. Figure 2 As shown, the system may include:

[0198] Data acquisition module 10: used to obtain depth and texture data of underwater concrete surface from multiple sensors. Specifically, this module includes:

[0199] Line structured light sensor: This sensor projects known structured light onto the underwater concrete surface, captures the reflection information of the light, and thus obtains the depth data of the concrete surface.

[0200] Stereo vision sensor: Use binocular cameras to collect texture images of underwater concrete from multiple perspectives and provide surface texture information.

[0201] Environmental parameter sensor: used to monitor changes in light intensity, temperature, flow rate, etc. in the underwater environment to ensure the effectiveness and accuracy of the collected data.

[0202] Through the above-mentioned multiple sensors, depth data and surface texture images related to underwater concrete structures are collected in real time to ensure the diversity and accuracy of data sources and provide basic data for subsequent processing and modeling.

[0203] Data processing module 20: used to perform preliminary processing on the collected depth data and surface texture data. Specifically, this module includes:

[0204] Data denoising and outlier processing: Denoising and filtering are performed on depth data and texture images to remove noise and outliers caused by the underwater environment (such as turbid water, unstable lighting, etc.) to ensure high data quality.

[0205] Data standardization and synchronization: The depth data and texture images from different sensors are synchronized in space and time so that they can be effectively registered in a unified coordinate system.

[0206] Through this module, the collected raw data is effectively preprocessed, laying the foundation for subsequent data fusion and modeling processes.

[0207] Multimodal data fusion module 30: Based on the depth map and texture map after environmental correction, the multimodal fusion algorithm is used to register and match the depth and texture data to generate a fused point cloud data model. Specifically, this module includes:

[0208] Depth and texture matching: Use a geometric matching algorithm to perform pixel-level feature matching on depth data and texture data to ensure the consistency of the information between the two.

[0209] 3D point cloud construction: The matched depth data and texture data are reconstructed in 3D to generate a point cloud data model to accurately capture the geometric shape of the underwater concrete surface.

[0210] The output of this module is a fused point cloud data model containing depth and texture information, which provides a detailed three-dimensional data foundation for subsequent model repair and optimization.

[0211] Deformation repair and motion compensation module 40: performs motion compensation and deformation repair on the fused point cloud data model to correct the spatial deformation and structural offset caused by water flow. Specifically, this module includes:

[0212] Motion compensation: Based on the parameters of the fluid environment and the dynamic target detection algorithm, the motion trajectory of dynamic objects in the water flow environment is analyzed, the motion compensation parameters are generated, and applied to the correction of point cloud data.

[0213] Deformation repair: Use the deformation repair algorithm to correct the distortion and surface changes of the point cloud model caused by water flow according to the motion compensation parameters.

[0214] This module ensures that deformations in the point cloud data model affected by water flow are corrected, improving the accuracy and consistency of the model.

[0215] 3D model reconstruction and damage detection module 50: based on the repaired point cloud model sequence, a high-precision 3D model of underwater concrete is generated, and intelligent analysis of the damage and structural integrity of the model surface is performed.

[0216] Specifically, this module includes:

[0217] Geometric modeling: Use a 3D modeling engine to geometrically reconstruct the repaired point cloud data to generate a complete 3D model.

[0218] Automatic defect detection: A multi-scale damage recognition algorithm based on depth changes and surface textures can automatically detect damaged areas and structural defects on the model surface.

[0219] Structural health assessment: Combining depth data and texture information, through feature analysis and integrity assessment, an intelligent analysis report of underwater concrete structures is generated, providing key information such as damage location and severity.

[0220] This module provides accurate data support for subsequent inspection and repair of underwater concrete structures.

[0221] Intelligent analysis and report generation module 60: Generates a detailed intelligent analysis report based on the repaired 3D model and damage detection results, and provides structural repair suggestions. This module includes:

[0222] Damage analysis and assessment: Identify potential hazardous areas and assess their damage levels by analyzing structural anomalies in the point cloud model.

[0223] Repair suggestions and optimization plans: Based on the results of the structural health assessment, we provide optimized repair plans and suggestions to help engineers perform accurate structural repairs in underwater environments.

[0224] System monitoring and adaptive optimization module 70:

[0225] Continuously monitor the operating status of the system and adaptively optimize model parameters and system settings according to environmental changes to ensure that the system operates stably and efficiently in various complex environments. This module includes:

[0226] Real-time monitoring: Monitor system status in real time to identify and adjust potential system bottlenecks or failures.

[0227] Adaptive optimization: Dynamically adjust model parameters and algorithm settings based on changes in the underwater environment and data feedback to improve the robustness and accuracy of system operation.

[0228] Beneficial effects: The present invention provides a method and system for underwater concrete three-dimensional reconstruction based on the fusion of line structured light and stereoscopic vision, which can efficiently obtain the three-dimensional data model of the concrete structure in a complex underwater environment. Through multimodal data fusion and advanced deformation repair algorithms, the system can accurately repair the point cloud deformation caused by water flow and achieve high-precision three-dimensional reconstruction of the underwater concrete surface. Combined with automatic defect detection and structural health assessment, the system can monitor the integrity of underwater concrete structures in real time and provide accurate repair solutions, greatly improving the inspection and maintenance efficiency of underwater projects, reducing manual intervention, and has significant technical advantages and application value.

[0229] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for underwater concrete 3D reconstruction based on the fusion of line structured light and stereo vision, characterized in that: The method comprises: Based on the line structured light module and stereo vision module, the depth map and texture map of underwater concrete are obtained to obtain the preliminary depth data sequence and surface texture sequence; Performing environmental compensation and image enhancement processing on the depth data sequence and the surface texture sequence, repairing uneven illumination and noise interference, and generating a depth map and a texture map after environmental correction; Based on the depth map and texture map corrected by the environment, a multimodal fusion algorithm is used to register and match the depth and texture data to generate a fused point cloud data model; Performing motion compensation and deformation repair on the fused point cloud data model to correct the spatial deformation and structural offset caused by water flow, and obtaining a repaired point cloud model sequence; The repaired point cloud model sequence is input into a three-dimensional model reconstruction engine to generate a high-precision three-dimensional model of underwater concrete, and intelligent analysis is performed on the damage and structural integrity of the model surface.

2. The underwater concrete 3D reconstruction method based on the fusion of line structured light and stereo vision according to claim 1 is characterized in that: The step of obtaining the depth map and texture map of underwater concrete specifically includes: Use the line structured light module to project known structured light to obtain the depth data of underwater concrete; The stereo vision module is used to collect texture images of underwater concrete from multiple perspectives to generate texture data.

3. The underwater concrete 3D reconstruction method based on the fusion of line structured light and stereo vision according to claim 1 is characterized in that: The step of performing environmental compensation and image enhancement on the depth data sequence and the surface texture sequence comprises: Based on the light intensity adaptive compensation algorithm, the influence of uneven lighting is corrected; Use spectral filtering image enhancement algorithm for denoising and contrast enhancement; Perform artifact removal and edge optimization on the enhanced depth map and texture map.

4. The underwater concrete 3D reconstruction method based on the fusion of line structured light and stereo vision according to claim 1 is characterized in that: The step of registering and matching the depth and texture data using a multimodal fusion algorithm based on the depth map and texture map after environment correction includes: Perform pixel-level feature matching on the depth map and texture map to generate a multimodal matching data sequence; A three-dimensional point cloud data model is constructed based on the multimodal matching data sequence.

5. The underwater concrete 3D reconstruction method based on the fusion of line structured light and stereo vision according to claim 1 is characterized in that: The step of generating a fused point cloud data model comprises: Use 3D reconstruction algorithm to fuse depth data with texture data to generate a point cloud data model; The point cloud data model is preliminarily optimized.

6. The underwater concrete 3D reconstruction method based on line structured light and stereoscopic vision fusion according to claim 1 is characterized in that: The step of performing motion compensation and deformation repair on the fused point cloud data model comprises: Based on water flow environment parameters and dynamic target detection algorithm, the influence of water flow on the motion of point cloud model is analyzed to generate motion compensation parameters; The deformation and offset of the point cloud model caused by the water flow are corrected based on the motion compensation parameters.

7. The underwater concrete 3D reconstruction method based on line structured light and stereoscopic vision fusion according to claim 1 is characterized in that: The step of inputting the repaired point cloud model sequence into the 3D model reconstruction engine to generate a high-precision 3D model comprises: Generate a geometric model of underwater concrete based on the repaired point cloud data using a 3D modeling engine; The geometric model is optimized in detail and its accuracy is improved.

8. The underwater concrete 3D reconstruction method based on line structured light and stereoscopic vision fusion according to claim 1 is characterized in that: The step of intelligently analyzing the damage and structural integrity of the model surface comprises: Based on the depth change and surface texture change, automatic defect detection algorithm is applied to identify damage; A multi-scale analysis is performed on the damaged area to identify structural defects and calculate the damage extent.

9. The method for underwater concrete 3D reconstruction based on line structured light and stereoscopic vision fusion according to claim 1, characterized in that: The steps of generating the intelligent analysis report include: Assess damage location, damage extent and its impact on structural integrity and generate a complete underwater concrete structure health report; Based on the evaluation results, repair suggestions and optimization solutions are provided.

10. An underwater concrete 3D reconstruction system based on the fusion of line structured light and stereo vision, characterized in that: The system comprises: A data acquisition module that acquires depth and texture data of underwater concrete surfaces from a variety of sensors; The data processing module performs preliminary processing on the collected depth data and surface texture data; The multimodal data fusion module uses a multimodal fusion algorithm to register and match the depth data and texture data based on the depth map and texture map after environmental correction to generate a fused point cloud data model; A deformation repair and motion compensation module performs motion compensation and deformation repair on the fused point cloud data model to correct the spatial deformation and structural offset caused by water flow; The 3D model reconstruction and damage detection module generates a high-precision 3D model of underwater concrete based on the repaired point cloud model, and performs intelligent analysis on the surface damage and structural integrity of the model; Intelligent analysis and report generation module, which generates detailed intelligent analysis reports based on the repaired 3D model and damage detection results, and provides structural repair suggestions; The system monitoring and adaptive optimization module continuously monitors the operating status of the system and adaptively optimizes model parameters and system settings according to environmental changes.

Images