Bridge underwater disease detection system and method based on array imaging

Abstract

The present application relates to the technical field of bridge disease detection, and in particular to a bridge underwater disease detection system and method based on array imaging, wherein in the disease detection registration module of the system, a "sonar-laser-polarization" three-mode independent solution and double-reference constraint are combined with the layered coaxial integrated design of the marker point "acoustic reflection unit + optical diffuse reflection unit", so as to realize high-precision initial registration of sonar coordinate solution error, laser ranging correction error and polarization three-dimensional conversion error, solve the multi-modal data scale drift problem, and through the scene-based adaptation of "spiral-radial swing trajectory" and "variable pitch spiral trajectory", the bridge pier path scanning module realizes complete coverage of the bridge pier surface disease detection, and effectively solves the coverage gap and attitude mismatch problems of the traditional fixed path.

Description

Technical Field

[0001] This invention relates to the field of bridge defect detection technology, and more specifically, to a bridge underwater defect detection system and method based on array imaging. Background Technology

[0002] Underwater bridge defect detection is crucial for ensuring bridge structural safety. Existing technologies often employ multimodal sensors such as sonar, lidar, and polarization cameras for joint detection. However, traditional multimodal detection systems suffer from significant deficiencies in the initial registration stage: the coordinate system calculations for sonar, lidar, and polarization modes lack effective reference constraints, multimodal data is prone to scale drift, and the marker point design fails to achieve a coordinated reference between acoustic and optical modes. This results in high errors in sonar coordinate calculations and lidar ranging corrections, failing to meet the requirements for sub-millimeter-level bridge defect detection. Furthermore, the spatiotemporal registration accuracy of multimodal data in existing technologies is insufficient, making it difficult to establish a unified spatial reference, which severely affects the accuracy of subsequent defect detection.

[0003] Traditional methods for defect feature extraction and path planning suffer from significant technical bottlenecks: single-modal feature extraction fails to consider complex environmental interferences such as underwater turbulence and sand content, exhibits low sensitivity to subtle defects, and lacks the ability to monitor the dynamic development of defects; path planning typically employs fixed trajectories, which cannot adapt to dynamic changes in the curvature of bridge pier surfaces. The difference between the fixed curvature of the detection loop and the curvature of the bridge pier surface leads to coverage gaps, and sonar / laser field-of-view mismatch causes repeated or missed scans, resulting in low coverage integrity and failing to meet the full-scene detection requirements for underwater bridge defects. These issues result in existing technologies exhibiting low accuracy, incomplete coverage, and poor dynamic adaptability in underwater bridge defect detection, necessitating innovative solutions. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a bridge underwater defect detection system and method based on array imaging.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An underwater bridge defect detection system based on array imaging, including

[0007] The defect detection and registration module configures the detection loop and performs initial spatial registration of multimodal data before detecting defects in the underwater piers of the bridge.

[0008] The modal feature enhancement module is used to enhance the defect sensitivity of multimodal features, extract inter-frame change rate and multi-frame sliding window processing, and fuse 8-dimensional features;

[0009] The pier path scanning module is used for planning defect detection paths for bridge pier curved surfaces.

[0010] Furthermore, the configured detection ring is divided into a lower layer, a middle layer, and an upper layer. The sensor type of the lower layer is sonar, the sensor type of the middle layer is lidar, and the sensor type of the upper layer is polarization camera.

[0011] Furthermore, the initial spatial registration of multimodal data includes spatiotemporal synchronization and multimodal registration of the detection loop, independent three-modal solution, and dual-benchmark constraints.

[0012] Furthermore, the detection ring is subjected to spatiotemporal synchronization and multimodal registration: n marker points are set on the detection ring, and the n marker points are evenly distributed in a ring direction. Each marker point is a cylinder, and the internal structure adopts a layered coaxial integration of "acoustic reflection unit + optical diffuse reflection unit".

[0013] Furthermore, the three-mode independent solution includes independent solution of the sonar coordinate system, independent solution of the laser coordinate system, and independent solution of the polarization coordinate system;

[0014] Sonar coordinate system:

[0015] Algorithm formula: c represents the speed of sound underwater. For the sonar echo phase difference, Where is the sonar frequency, and v is the turbulent relative velocity measured by the inertial measurement unit. To compensate for the time difference in data acquisition; the compensation logic is to use the turbulent velocity v measured in real time by the inertial measurement unit to correct for sound speed fluctuations and relative motion interference.

[0016] Laser coordinate system:

[0017] Optical path correction formula: ; The speed of light in a vacuum. The speed of light underwater is calculated dynamically by monitoring water temperature and salinity in real time. The raw distance is directly measured by the lidar. This is the corrected laser ranging value;

[0018] Polarization coordinate system:

[0019] Conversion process: Extract polarization images of marker points in the three bands of 450nm, 550nm, and 650nm. Calculate the DoP and AoP of each pixel in the 450nm polarization image. Filter the pixels of marker points that satisfy DoP>0.8 and AoP fluctuation<±5°. Mark the intersection area of ​​the filtered pixels as the position of the marker point in the image and output its pixel coordinates (u, v). Convert it into three-dimensional coordinates through the pre-calibrated camera intrinsic parameter matrix and the three-dimensional mapping model.

[0020] Furthermore, the defect sensitivity of multimodal features is enhanced, including single-mode feature transformation, where single modes include sonar mode, laser mode, and polarization mode.

[0021] Furthermore, for the path planning of defect detection for bridge pier curved surfaces, it includes: curvature difference scenario-based compensation trajectory and multi-sensor collaborative step size optimization with field of view constraints.

[0022] Furthermore, for the curvature difference scenario-based compensation trajectory, R is defined as the radius of curvature of the bridge pier. To detect the radius of curvature of the torus, when Trajectory type: Spiral-radial oscillating trajectory; Algorithm logic:

[0023] Axial step size L: based on the effective coverage width of the sonar ;Pick ;

[0024] Circumferential rotation angle : To ensure continuous circumferential coverage;

[0025] Radial oscillation: After each complete rotation of the spiral, it oscillates slightly radially along the pier to compensate for the inner gap;

[0026] when Trajectory type: Variable pitch spiral trajectory; Algorithm logic:

[0027] Axial step size L: increases with R, as shown in the formula. ;

[0028] Circumferential rotation angle It decreases as R increases, ensuring stable circumferential cover density;

[0029] Attitude adjustment: Detects micro-rotations around its own axis.

[0030] Furthermore, considering the differences in the field of view of sonar, laser, and polarization cameras (sonar ±45°, laser ±15°, polarization ±30°), a unified step-size calculation model is designed to ensure multi-modal coverage coordination.

[0031] Effective coverage width calculation: , , ;

[0032] Standardized step size ;

[0033] Dynamic adjustment logic: Every 10 seconds, the real-time curvature R of the bridge pier is obtained through SLAM mapping, and L and rotation angle are recalculated. To adapt to local curvature changes on the surface of the bridge piers;

[0034] Using a sonar-laser fusion SLAM-generated point cloud map of bridge piers, blind spots are detected in real time, and rescanning paths are dynamically inserted.

[0035] Missed area identification: Octree point cloud density analysis is used to mark areas with insufficient density. .

[0036] Local replanning algorithm: In Generate the shortest sub-path for the nearest neighbor, with the following constraints: q: The pose node of the current detection loop, representing the current position in path planning; Missed areas The feature pose nodes represent the target positions that need to be scanned again. : Detection ring with fixed curvature With the current curvature of the bridge pier The absolute difference, with a coefficient of 0.3: the weighting factor for the curvature difference;

[0037] Path insertion and optimization: Insert the supplementary scan sub-path into the main path, and achieve a smooth transition through B-spline interpolation to ensure stable movement of the detection loop;

[0038] Define a multimodal coverage quality function to monitor and detect integrity in real time, triggering rescanning or path replanning: ; This represents the percentage of the area covered by sonar. This represents the percentage of the area covered by the laser. This represents the percentage of the coverage area of ​​the polarization camera.

[0039] set up ,when : Initiate local replanning, prioritizing areas with poor multimodal coordination; when Continue executing the main path to reduce computational overhead.

[0040] Furthermore, the method for detecting underwater bridge defects based on array imaging comprises the following steps:

[0041] Step 1: Detection ring configuration and initial multimodal registration;

[0042] Step 2: Multimodal data processing and feature enhancement;

[0043] Step 3: Scan path planning for bridge pier surface adaptation.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The defect detection and registration module of this invention features independent three-modal calculation and dual-reference constraints for "sonar-laser-polarization," combined with a layered coaxial integrated design of "acoustic reflection unit + optical diffuse reflection unit" for marker points. This achieves high-precision initial registration, mitigating sonar coordinate calculation errors, laser ranging correction errors, and polarization 3D conversion errors, thus solving the multimodal data scale drift problem. The modal feature enhancement module extracts single-modal defect features such as phase difference anomalies, depth residuals, and DoP anomalies, and combines them with dynamic feature mining using inter-frame change rate and a 20-frame sliding window. By integrating an 8-dimensional feature space, the sensitivity of underwater bridge pier defect identification is improved. The pier path scanning module addresses the contradiction between the fixed detection ring curvature and the dynamic changes in pier curvature. Through scenario-based adaptation of "spiral-radial swing trajectory" and "variable pitch spiral trajectory", combined with sonar-laser field of view collaborative step size optimization, SLAM-driven missing area completion and dynamic feedback of multimodal coverage quality function Q≥0.95, it achieves complete coverage of pier surface defect detection and effectively solves the coverage gap and attitude mismatch problems of traditional fixed path.

[0046] The method of this invention achieves accurate monitoring of underwater bridge defects in environments with high sediment content and high turbulent flow velocity through a technical closed loop of "registration-planning-feature enhancement". Attached Figure Description

[0047] Figure 1 This is a schematic diagram of a bridge underwater defect detection system based on array imaging.

[0048] Figure 2 The principle block diagram for independent solution of the three modes;

[0049] Figure 3 A flowchart for modeling curvature differences. Detailed Implementation

[0050] Example 1: Refer to Figures 1 to 3 The underwater bridge defect detection system based on array imaging includes a defect detection registration module, a modal feature enhancement module, and a pier path scanning module.

[0051] A detection ring is configured for the underwater piers of the bridge. The diameter of the detection ring is adjusted according to the diameter of the underwater pier. The detection ring is divided into a lower layer (acoustic zone), a middle layer (optical zone), and an upper layer (polarization zone). The sensor type of the lower layer (acoustic zone) is a miniature multibeam sonar, with 8 groups, a parameter of 500kHz, and a resolution of 0.1mm when the water depth is ≤10m. The sensor type of the middle layer (optical zone) is a 16-channel lidar, with a parameter of 905nm, a scanning frequency of 200Hz, and an angular resolution of 0.1°. The sensor type of the upper layer (polarization zone) is an 8-channel multi-band polarization camera, with parameters of 450 / 550 / 650nm three bands and an 8-channel beam splitter prism design.

[0052] Spatiotemporal synchronization and multimodal registration of the detection ring: n marker points are set on the detection ring, with n being at least 3. The n marker points are evenly distributed in a circumferential direction, such as the 3 marker points being evenly distributed in a 120° circumferential direction. Each marker point is a cylinder, and the internal structure adopts a layered coaxial integration of "acoustic reflection unit + optical diffuse reflection unit" to achieve "one device, dual-modal reference".

[0053] Acoustic reflector unit (adapted to sonar)

[0054] Structure: A 3mm diameter titanium alloy reflective sphere (density 4.5g / cm³, acoustic impedance ≈ 3 times that of seawater, achieving efficient sound reflection) is embedded at the bottom, and the surface is processed into a 30° cone-shaped scattering surface (instead of a traditional spherical surface).

[0055] Function: To enable sonar (500kHz, beam angle ±45°) to receive stable echoes within a ±45° viewing angle (echo intensity standard deviation <5%, traditional spherical echo attenuation >30% at the edge viewing angle).

[0056] Optical diffuse reflection unit (for laser / polarization cameras)

[0057] Core structure: The top integrates a 1mm diameter alumina ceramic target (resistant to seawater corrosion, hardness HRA>90), and the surface adopts a "dual-band functional film" design.

[0058] 550nm high reflectivity film: vapor-deposited silicon dioxide-titanium oxide multilayer film, with a reflectivity of >95% for lidar in the 905nm band;

[0059] 450nm fluorescent coating: Phosphate material doped with europium ions, which absorbs 450nm polarized light and emits 590nm fluorescence, enhancing the uniqueness of polarization camera identification (DoP>0.8 in the 450nm band, forming a strong difference from DoP<0.3 in the background scattering).

[0060] The disease detection and registration module completes the initial registration through independent solution of three modes: sonar, laser, and polarization, plus dual reference constraints.

[0061] Sonar coordinate system (phase difference-Doppler joint algorithm):

[0062] Algorithm formula: c represents the speed of sound underwater. For the sonar echo phase difference, =500kHz is the sonar frequency, and v is the turbulent relative velocity measured by the IMU (the IMU is configured on the detection loop). To compensate for the time difference in data acquisition; the compensation logic is as follows: using the turbulent velocity v measured in real time by the inertial measurement unit (IMU), the sound velocity fluctuation and relative motion interference are corrected to make the sonar coordinate solution error <0.2mm;

[0063] Laser coordinate system (turbulent optical path model correction):

[0064] Optical path correction formula: ; The speed of light in a vacuum. The speed of light underwater is calculated dynamically by monitoring water temperature and salinity in real time. The raw distance is directly measured by the lidar. This is the corrected laser ranging value;

[0065] Polarization coordinate system (multi-band polarization feature transformation):

[0066] Conversion process:

[0067] Polarization images of the marker points were extracted in three bands: 450nm (fluorescence), 550nm (high reflectivity), and 650nm (background). 450nm (fluorescence): The europium ion fluorescent coating on the marker point surface absorbs 450nm polarized light and emits 590nm fluorescence (Stokes shift), making the marker point's polarization signal unique in the 450nm band (DoP>0.8). 550nm (high reflectivity): The silica-titanium oxide multilayer film of the marker point is highly reflective to the lidar at 905nm (near-infrared) and also highly reflective in the 550nm (visible light) band (aiding localization and enhancing robustness). 650nm (background): Background noise was collected for differential processing (removing interference from water scattering and biological attachment).

[0068] Hardware support: A customized polarization camera with an 8-channel beam splitter prism, which simultaneously acquires polarization images in three bands (each band contains 0° / 45° / 90° / 135° linear polarization + RCP / LCP circular polarization, for a total of 8 channels).

[0069] Pixel coordinates are located using the characteristics of 450nm fluorescence: DoP>0.8 and AoP fluctuation<±5°. The location logic is as follows: only pixels that simultaneously satisfy DoP>0.8 and AoP fluctuation<±5° are retained as marker areas, and their pixel coordinates (u, v) are output.

[0070] The system converts the pre-calibrated camera intrinsic parameter matrix into 3D coordinates using a 3D mapping model, with an error of <0.3mm.

[0071] Improved ICP algorithm (turbulence robust version):

[0072] To address sensor jitter caused by turbulence, the traditional ICP algorithm is improved with a combination of "scale constraint + motion compensation":

[0073] Scale constraints (introduced by geometric priors):

[0074] Define the sonar-laser scale mapping relationship: sonar wavelength λ = 0.3 mm (500 kHz), laser wavelength λ = 905 nm, and establish the scale transformation matrix: ;

[0075] The registration process forces the adherence to this scale relationship to avoid scale drift in multimodal data (traditional ICP is prone to 10%-20% scale bias).

[0076] Motion compensation (turbulence vector pre-correction):

[0077] Extracting turbulent motion vectors from IMU composite timestamps Pre-compensation is performed on the "nearest point search range" of the ICP: (Traditional search radius is fixed at 10mm, which easily misses the actual corresponding points);

[0078] The modal feature enhancement module enhances the defect sensitivity of multimodal features;

[0079] Modality: Sonar; Original feature: 3D point cloud Defect-sensitive feature transformation: phase difference is common : , >10°, the physical meaning is sound propagation phase distortion (cracks / corrosion causing changes in acoustic impedance); The sonar echo phase value at the current detection point. The "health baseline phase" (such as the historical frame phase of the disease-free area of ​​the bridge pier, or the phase fitted by the disease-free points in the neighborhood) in the same area is used as the "disease-free reference value".

[0080] Modality: LiDAR, Original feature: Depth map (z), Defect-sensitive feature transformation: Depth residual : Crack >0.2mm, physically meaning abrupt changes in surface depth (geometric features of cracks / scour); The current depth value (unit: mm) measured by the lidar reflects the distance from the pier surface to the detection ring. : "Local healthy surface" fitted by RANSAC algorithm (using neighborhood disease-free points to fit a continuous plane / surface to simulate the geometric shape of disease-free area).

[0081] Modal: Polarization camera; Original features: DoP / AoP / R / B; Defect-sensitive feature transformation: DoP anomaly : Disease The physical meaning is a sudden change in polarization (material anisotropy in biological adhesion / corrosion); : The degree of polarization at the current point (0≤DoP≤1, reflecting the degree of polarization of light; the higher the DoP, the more significant the polarization characteristics). Background polarization degree in the same area (such as the mean DoP value of disease-free areas, or the background value extracted by Gaussian filtering);

[0082] Based on the characteristic dynamic changes caused by underwater flow disturbance (velocity ≤ 0.5 m / s), the following is extracted:

[0083] Inter-frame change rate: such as the time derivative of sonar phase difference ( ), the time derivative of the laser depth residual ( ), highlighting dynamically developing diseases (such as crack expansion and intensified erosion);

[0084] Multi-frame sliding window: Spatiotemporally stitching 20 consecutive frames of data to form a spatiotemporal cube (T=20, X=100, Y=100), preserving the dynamic evolution information of the disease.

[0085] It integrates 8-dimensional features including geometry (x, y, z), acoustic phase (Δφ), optical depth (Δz), and polarization (ΔDoP, AoP, R / B);

[0086] Pier path scanning module, pier surface adaptation path planning;

[0087] When the curvature of the detection ring is fixed (set to be...), (e.g., design value 3m), while the pier curvature When things change dynamically, the following challenges are faced:

[0088] Coverage gap: There is a curvature difference between the detection ring and the pier surface ( This results in the sensor's field of view not being able to fully fit, creating a scanning blind spot;

[0089] Attitude mismatch: Under a fixed curvature, the projection area of ​​the sensor field of view (such as sonar ±45°, laser ±15°) of the detection ring on the curved surface of the pier changes with R. Traditional equidistant paths are prone to repeated scanning or missed scanning.

[0090] The lack of dynamic adjustment means that the curvature difference cannot be compensated by hardware deformation, and it is necessary to rely entirely on the intelligent adaptation of the path planning algorithm.

[0091] Curvature difference modeling and trajectory generation strategy:

[0092] against The resulting cover gap (R is the radius of curvature of the pier, To detect the radius of curvature of the toroidal ring, a "spiral offset" or "variable pitch spiral" trajectory is designed, and the curvature difference is compensated by adjusting the translation / rotation attitude.

[0093] Scene 1: The bridge piers are more curved ( ,like ):

[0094] Trajectory type: Spiral-radial oscillating trajectory.

[0095] Algorithm logic:

[0096] Axial step size L: based on the effective coverage width of the sonar (d is the distance between the detection ring and the bridge pier) (where the sonar beam angle is taken) Ensure a 30% overlap rate;

[0097] Circumferential rotation angle : To ensure continuous circumferential coverage;

[0098] Radial oscillation: After each complete rotation of the spiral, the bridge pier oscillates slightly radially. ), to compensate for the inner gap (the curvature of the detection ring is more gradual, making it easy to miss the inner side of the bridge pier).

[0099] Scene 2: The bridge piers are more gently sloping ( ,like ):

[0100] Trajectory type: Variable pitch spiral trajectory.

[0101] Algorithm logic:

[0102] Axial step size L: increases with increasing R ( The formula is ;

[0103] Circumferential rotation angle : Decreases as R increases ( ), ensuring stable circumferential coverage density (e.g., maintaining one scan point per degree circumferentially);

[0104] Attitude adjustment: Detects micro-rotation around its own axis ( This allows the fan-shaped scanning surface of the lidar to better fit the curved surface of the bridge pier.

[0105] Multi-sensor collaborative step size optimization with field of view constraints: Taking into account the differences in field of view of sonar, laser, and polarization camera (sonar ±45°, laser ±15°, polarization ±30°), a unified step size calculation model is designed to ensure multi-modal coverage collaboration.

[0106] Effective coverage width calculation (based on sonar, taking the minimum value): , , ; Uniform step size This ensures that even the narrowest field of view sensor can provide continuous coverage.

[0107] Dynamic adjustment logic: Obtain the real-time curvature R of the bridge piers through SLAM mapping every 10 seconds, and recalculate. and rotation angle To adapt to local curvature changes on the surface of the bridge pier (such as abrupt curvature changes caused by scour pits).

[0108] Using sonar-laser fusion SLAM to generate bridge pier point cloud maps, real-time detection of coverage blind spots (point cloud density) is achieved. If a scan is deemed missed, a replacement scan path will be dynamically inserted.

[0109] Missed area identification: Octree point cloud density analysis is used to mark areas with insufficient density. .

[0110] Local Replanning Algorithm (Improved RRT): In Generate the shortest sub-path for the nearest neighbor, with the following constraints: (Prioritize scan trajectories with small curvature differences and short paths to avoid excessive detours). q: The pose node of the current detection loop (containing 3D coordinates x, y, z and attitude angles roll, pitch, yaw), representing the current position in path planning; Missed areas The feature pose nodes (usually the center point or boundary point of the missed area) represent the target positions that need to be scanned again. : Detection ring with fixed curvature With the current curvature of the bridge pier The absolute difference, with a coefficient of 0.3: the weighting factor of the curvature difference, verified through experiments (when the weight > 0.5, the path is excessively detoured; when it < 0.2, the curvature difference compensation is insufficient).

[0111] Path insertion and optimization: Insert the supplementary scan sub-path into the main path and achieve a smooth transition through B-spline interpolation to ensure stable movement of the detection loop (avoid sensor jitter caused by sudden stops and turns).

[0112] Define a multimodal coverage quality function to monitor and detect integrity in real time, triggering rescanning or path replanning: ; This represents the percentage of the area covered by sonar. This represents the percentage of the area covered by the laser. This represents the percentage of the coverage area of ​​the polarization camera.

[0113] set up (Can be set to 0.95), when : Initiate local replanning, prioritizing the rescanning of areas with poor multimodal coordination (such as areas only covered by sonar, or areas not covered by laser / polarization); when Continue executing the main path to reduce computational overhead.

[0114] The disease detection and registration module of the above system uses independent three-modal calculation of "sonar-laser-polarization" and dual-reference constraints, combined with the hierarchical coaxial integrated design of "acoustic reflection unit + optical diffuse reflection unit" for marker points. This achieves high-precision initial registration of sonar coordinate calculation errors, laser ranging correction errors, and polarization three-dimensional conversion errors, solving the problem of multimodal data scale drift. The modal feature enhancement module extracts single-modal defect features such as phase difference anomalies, depth residuals, and DoP anomalies, and combines them with dynamic feature mining using inter-frame change rate and a 20-frame sliding window. By integrating an 8-dimensional feature space, the sensitivity of underwater bridge pier defect identification is improved. The pier path scanning module addresses the contradiction between the fixed detection ring curvature and the dynamic changes in pier curvature. Through scenario-based adaptation of "spiral-radial swing trajectory" and "variable pitch spiral trajectory", combined with sonar-laser field of view collaborative step size optimization, SLAM-driven missing area completion and dynamic feedback of multimodal coverage quality function Q≥0.95, it achieves complete coverage of pier surface defect detection and effectively solves the coverage gap and attitude mismatch problems of traditional fixed paths.

[0115] Example 2: A method for detecting underwater bridge defects based on array imaging, the steps of which are as follows:

[0116] Step 1: Detection ring configuration and initial multimodal registration;

[0117] Step 2: Multimodal data processing and feature enhancement;

[0118] Step 3: Scan path planning for bridge pier surface adaptation.

[0119] The above method, through a technical closed loop of "registration-planning-feature enhancement", has achieved accurate monitoring of underwater bridge defects in environments with high sediment content and high turbulent flow velocity.

[0120] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.

[0121] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0122] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0123] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0124] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0125] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0126] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0127] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A bridge underwater defect detection system based on array imaging, characterized in that, include The defect detection and registration module configures the detection loop and performs initial spatial registration of multimodal data before detecting defects in the underwater piers of the bridge. The configured detection ring is divided into a lower layer, a middle layer, and an upper layer. The sensor type of the lower layer is sonar, the sensor type of the middle layer is lidar, and the sensor type of the upper layer is polarization camera. Initial spatial registration of multimodal data includes spatiotemporal synchronization and multimodal registration of the detection loop, independent three-modal solution, and dual-benchmark constraints; Spatiotemporal synchronization and multimodal registration of the detection ring: n marker points are set on the detection ring. The n marker points are evenly distributed in a ring direction. Each marker point is a cylinder. The internal structure adopts a layered coaxial integration of "acoustic reflection unit + optical diffuse reflection unit". The modal feature enhancement module is used to enhance the defect sensitivity of multimodal features, extract inter-frame change rate and multi-frame sliding window processing, and fuse 8-dimensional features; Based on the dynamic changes in features caused by underwater flow disturbance, the following can be extracted: Inter-frame variation rate: the time derivative of sonar phase difference and the time derivative of laser depth residual, highlighting dynamically developing defects; Multi-frame sliding window: Spatiotemporally stitching 20 consecutive frames of data to form a spatiotemporal cube, preserving the dynamic evolution information of the disease; It integrates 8-dimensional features including geometry (x, y, z), acoustic phase (Δφ), optical depth (Δz), and polarization (ΔDoP, AoP, R / B); Enhancement processing is applied to improve the defect sensitivity of multimodal features, including: feature transformation of single modes, which include sonar mode, laser mode, and polarization mode; The pier path scanning module is used for planning detection paths for bridge pier curved surfaces to adapt to defects. For bridge pier surface adaptation defect detection path planning, it includes: curvature difference scenario-based compensation trajectory and multi-sensor collaborative step size optimization with field of view constraints.

2. The underwater bridge defect detection system based on array imaging according to claim 1, characterized in that, The three-mode independent solution includes independent solution of the sonar coordinate system, independent solution of the laser coordinate system, and independent solution of the polarization coordinate system. Sonar coordinate system: Algorithm formula: c represents the speed of sound underwater. For the sonar echo phase difference, Where is the sonar frequency, and v is the turbulent relative velocity measured by the inertial measurement unit. To compensate for the time difference in data acquisition; the compensation logic is to use the turbulent velocity v measured in real time by the inertial measurement unit to correct for sound speed fluctuations and relative motion interference. Laser coordinate system: Optical path correction formula: ; The speed of light in a vacuum. The speed of light underwater is calculated dynamically by monitoring water temperature and salinity in real time. The raw distance is directly measured by the lidar. This is the corrected laser ranging value; Polarization coordinate system: Conversion process: Extract polarization images of marker points in the three bands of 450nm, 550nm, and 650nm. Calculate the DoP and AoP of each pixel in the 450nm polarization image. Filter the pixels of marker points that satisfy DoP>0.8 and AoP fluctuation<±5. Mark the intersection area of ​​the selected pixels as the position of the marker point in the image and output its pixel coordinates (u,v). Convert it into three-dimensional coordinates through the pre-calibrated camera intrinsic parameter matrix and the three-dimensional mapping model.

3. The underwater bridge defect detection system based on array imaging according to claim 1, characterized in that, Curvature difference scenario-based compensation trajectory, where R is defined as the radius of curvature of the bridge pier. To detect the radius of curvature of the torus, when Trajectory type: Spiral-radial oscillating trajectory; Algorithm logic: Axial step size L: based on the effective coverage width of the sonar d is the distance between the detection ring and the bridge pier. For the sonar beam angle; take ; Circumferential rotation angle : To ensure continuous circumferential coverage; Radial oscillation: After each complete rotation of the spiral, it oscillates slightly radially along the pier to compensate for the inner gap; when Trajectory type: Variable pitch spiral trajectory; Algorithm logic: Axial step size L: increases with increasing R, as shown in the formula. ; Circumferential rotation angle The density decreases as R increases, ensuring stable circumferential cover density; Attitude adjustment: Detects micro-rotations around its own axis.

4. The underwater bridge defect detection system based on array imaging according to claim 1, characterized in that, Considering the differences in the field of view of sonar, laser, and polarization cameras, a unified step-size calculation model is designed to ensure multimodal coverage coordination: Effective coverage width calculation: , , ; Standardized step size ; Using a sonar-laser fusion SLAM-generated point cloud map of bridge piers, blind spots are detected in real time, and rescanning paths are dynamically inserted. Missed area identification: Octree point cloud density analysis is used to mark areas with insufficient density. ; Local replanning algorithm: In Generate the shortest sub-path for the nearest neighbor, with the following constraints: q: The pose node of the current detection loop, representing the current position in path planning; Missed areas The feature pose nodes represent the target locations that need to be scanned again. : Detection ring with fixed curvature With the current curvature of the bridge pier The absolute difference, with a coefficient of 0.3: the weighting factor for the curvature difference; Path insertion and optimization: Insert the supplementary scan sub-path into the main path, and achieve a smooth transition through B-spline interpolation to ensure stable movement of the detection loop; Define a multimodal coverage quality function to monitor detection integrity in real time and trigger rescanning or path replanning: ; This represents the percentage of the area covered by sonar. This represents the percentage of the area covered by the laser. This represents the percentage of the coverage area of ​​the polarization camera. set up ,when : Initiate local replanning, prioritizing areas with poor multimodal coordination; when Continue executing the main path.

5. A method for detecting underwater bridge defects based on array imaging, applied to the underwater bridge defect detection system based on array imaging as described in any one of claims 1-4, characterized in that, The steps are as follows: Step 1: Detection ring configuration and initial multimodal registration; Step 2: Multimodal data processing and feature enhancement; Step 3: Scan path planning for bridge pier surface adaptation.

Images