A method for underwater acoustic and magnetic multimodal fusion processing

Abstract

This invention discloses an underwater acoustic-magnetic multimodal fusion processing method, comprising: preprocessing and spatiotemporally aligning acquired sonar and magnetic data to extract morphological and magnetic feature vectors of pipelines; calculating the optimal matching path between the sonar profile and the magnetic anomaly zone based on the pipeline routing direction through dynamic time warping, establishing an acoustic-magnetic mapping and fusing features to generate a joint vector; and finally determining the pipeline type and status based on the joint features. This method aims to solve the problem that in nearshore port and industrial waters, due to interference from construction waste and pipeline covering layers, a single sonar is insufficient for accurately detecting sewage pipelines and cables. This invention utilizes the complementarity of acoustic and magnetic features to improve the accuracy of pipeline detection in complex scenarios, providing reliable data support for pipeline operation and maintenance in nearshore waters.

Description

Technical Field

[0001] This invention relates to the field of similar devices employing other waves, particularly to acoustic waves and magnetism, specifically to an underwater acoustic-magnetic multimodal fusion processing method. Background Technology

[0002] Currently, with the rapid development of marine resource development, marine engineering construction, and marine monitoring, the demand for underwater target detection and monitoring is increasing. In numerous underwater detection scenarios, such as subsea pipeline inspection, underwater facility status assessment, and underwater obstacle detection, accurate acquisition of target information is crucial. Acoustic detection technology, especially various types of sonar, has become a core means of underwater target detection due to its advantage of underwater propagation. For example, side-scan sonar and multibeam echo sounders can efficiently survey seabed topography and the condition of exposed pipelines; synthetic aperture sonar can provide high-resolution images to identify pipeline surface details; shallow seismic profilers can detect relevant information about shallow buried pipelines; and acoustic positioning systems provide location services for detection platforms and targets.

[0003] Existing technologies, relying solely on acoustic detection, particularly single-type sonar solutions, have significant limitations. In buried pipeline detection, acoustic waves are limited by the type and frequency of the substrate, resulting in insufficient penetration depth for deeply buried pipelines and a sharp decrease in resolution with depth. Strong clutter and multiple reflections generated by geological structures such as rocks and gravels in complex substrates can mask pipeline signals or create false targets. When pipelines have special cladding layers that make their acoustic characteristics similar to the surrounding sediments, an acoustic "stealth" effect occurs, resulting in weak acoustic reflection / scattering signals that are difficult to identify. Furthermore, acoustic waves cannot penetrate metal pipe walls, cannot detect the internal condition of the pipeline, and cannot distinguish pipeline materials or assess the ferromagnetism of metal pipes.

[0004] Therefore, it is necessary to improve the existing underwater acoustomagnetic multimodal fusion processing methods to solve the above problems. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides an underwater acoustic-magnetic multimodal fusion processing method. It aims to solve the problems in the prior art where strong acoustic clutter caused by construction waste in near-shore port and industrial waters makes it difficult for single sonar detection to identify sewage pipelines that are acoustically camouflaged due to the coating layer and cables whose material and corrosion status cannot be determined by sound waves.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: an underwater acoustic-magnetic multimodal fusion processing method, comprising:

[0007] S1. Preprocess the sonar and magnetic data of underwater pipelines;

[0008] S2. Extract the morphological feature vector of the pipeline from the sonar data and extract the magnetic feature vector of the pipeline from the magnetic data.

[0009] S3. Based on the pipeline routing, calculate the matching degree between the sonar profile routing and the magnetic anomaly zone routing, calculate the optimal path through dynamic time warping, and establish an acoustomagnetic mapping; and fuse the morphological feature vector and the magnetic feature vector to generate a joint feature vector.

[0010] The step of calculating the matching degree in step S3 includes:

[0011] The pipeline orientation angle is extracted from sonar data as the sonar profile orientation, and the magnetic anomaly zone orientation angle is extracted from magnetic data.

[0012] Calculate the angular difference between the two orientation angles. ,in, The sonar profile orientation angle. For the orientation angle of the magnetic anomaly zone, when and When they are completely identical, M=1; when they are perpendicular, M=0; when the matching degree corresponding to the difference value is ≥0.8, they are judged to be the same pipeline feature.

[0013] S4. Determine the pipeline type and pipeline status based on the joint feature vector; where the pipeline type is sewage pipeline and cable.

[0014] In a preferred embodiment of the present invention, the preprocessing of sonar data in step S1 includes:

[0015] A nonlocal mean filtering algorithm is adopted to suppress strong reflection clutter generated by construction waste by calculating the similarity weight of neighboring pixels, while preserving the continuous contour signal of the pipeline.

[0016] In a preferred embodiment of the present invention, the preprocessing of the magnetic data in step S1 includes:

[0017] The theoretical geomagnetic field strength is calculated using a geomagnetic model, and this value is subtracted from the original magnetic data to obtain the remaining magnetic field data.

[0018] A sliding window filter is applied to the remaining magnetic field data to remove high-frequency magnetic field fluctuations caused by ship anchoring.

[0019] In a preferred embodiment of the present invention, the sonar data includes: side-scan sonar echo intensity data and shallow seismic profiler reflection coefficient data; the magnetic data includes: total magnetic field intensity data and magnetic field gradient data;

[0020] Preprocessing also includes spatiotemporal alignment, the specific steps of which are: to achieve time synchronization between the sonar device and the magnetic device by triggering pulses through hardware, to convert the distance-angle coordinates of the side-scan sonar into the WGS84 geographic coordinate system by combining GNSS positioning data and INS attitude data, and to unify the coordinates of the magnetic sensor to the same geographic coordinate system by calibrating the tow cable position.

[0021] In a preferred embodiment of the present invention, in step S2, the morphological feature vector includes pipeline length, pipeline width, pipeline orientation angle, and contour continuity; the magnetic feature vector includes magnetic anomaly peak value, magnetic anomaly range, maximum magnetic field gradient value, and magnetic anomaly symmetry.

[0022] In a preferred embodiment of the present invention, the specific steps of dynamic time warping in step S3 include:

[0023] Extract the acoustic centerline coordinate sequence and the magnetic anomaly trajectory coordinate sequence, and construct the spatial distance matrix between the two sequences;

[0024] The optimal path with the minimum cumulative distance is solved by dynamic programming, and a point-to-point mapping relationship is established.

[0025] When the cumulative distance of the optimal path is ≤0.5m, the morphological feature vector and the magnetic feature vector are bound together.

[0026] In a preferred embodiment of the present invention, the specific steps of feature fusion in step S3 include:

[0027] The morphological feature vector and magnetic feature vector are normalized, and each feature parameter is mapped to the interval [0,1].

[0028] Based on the acoustic-magnetic mapping relationship, three types of interactive features are calculated: material correlation factor, spatial consistency factor, and coupling coefficient.

[0029] A 6-dimensional joint feature vector containing anti-interference factor, scale-invariant features, and scale ratio is constructed by combining dynamic compression.

[0030] In a preferred embodiment of the present invention, the material correlation factor is constructed by fusing the peak value of the magnetic anomaly, the maximum value of the magnetic field gradient, and the pipeline width; the spatial consistency factor is calculated based on the spatial deviation of the dynamic time warp mapping point; and the coupling coefficient reflects the coupling relationship between the magnetic anomaly range and the acousto-magnetic orientation deviation.

[0031] In a preferred embodiment of the present invention, the specific steps for determining the pipeline type in step S4 include:

[0032] When the peak value of the magnetic anomaly is greater than 500 nT and the symmetry of the magnetic anomaly is greater than 0.7, it is determined to be a cable;

[0033] When the peak value of the magnetic anomaly is <200nT, the profile continuity is >0.8, and the coupling coefficient is <0.3, it is determined to be a sewage pipeline with a cladding layer;

[0034] All other cases are classified as standard sewage pipelines.

[0035] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0036] (1) This invention proposes an underwater acoustic-magnetic multimodal fusion processing method to solve the problem that single sonar is difficult to accurately detect sewage pipelines and cables in near-shore port and industrial waters due to interference from construction waste and pipeline covering layers. First, the sonar and magnetic data of underwater pipelines are preprocessed to remove construction waste clutter and ship anchoring interference. Then, the morphological feature vector and magnetic feature vector of the pipeline are extracted respectively. Next, based on the pipeline routing, an acoustic-magnetic mapping is established by calculating the matching degree and dynamic time warping, and a joint vector is generated by fusion. Finally, the type and status of sewage pipelines and cables are determined based on the joint features, and the detection accuracy in complex scenarios is improved by utilizing the complementarity of acoustic-magnetic features.

[0037] (2) This invention calculates the matching degree between the sonar profile and the magnetic anomaly zone, and establishes an acoustomagnetic mapping by combining the dynamic time warping (DTW) algorithm to achieve feature fusion. The matching degree ensures from a macroscopic perspective that the sonar and magnetic features belong to the same pipeline, while DTW adapts to local bending or offset of the pipeline through flexible alignment, and accurately binds the morphological feature vector and the magnetic feature vector. It can effectively avoid feature misalignment caused by irregular pipeline shape or environmental interference. The feature fusion of the prior art does not consider the accuracy of spatial correlation, and is prone to mis-fusion of features of different targets, resulting in distorted results. The accurate mapping of this invention enables the joint feature vector to truly reflect the comprehensive characteristics of the pipeline.

[0038] (3) This invention generates a joint vector containing interactive features through feature fusion. After normalizing morphological and magnetic features, interactive features such as material correlation factors and spatial consistency factors are calculated to construct a multi-dimensional joint vector. The interactive features uncover the intrinsic relationship between morphological and magnetic features. For example, the material correlation factor combines magnetic anomalies with pipeline width to reflect material characteristics, adapting to the differences in pipeline materials in nearshore scenarios, and the influence of the ferromagnetism of cables and the non-metallicity and coating of sewage pipes. The joint vector can reflect both the external morphology of the pipeline and its material properties and spatial consistency. Compared with the fusion method in the prior art that simply splices features without considering the correlation between features, the joint vector information of this method is more comprehensive. A further effect is that it significantly improves the ability to distinguish sewage pipelines with coatings from cables, as well as the accuracy of pipeline status judgment, providing a reliable basis for the operation and maintenance of pipelines in nearshore waters.

[0039] (4) In response to the strong reflection clutter generated by construction waste in nearshore waters, the present invention uses nonlocal mean filtering to preserve the continuous contour of pipelines in sonar data; and in response to the high-frequency magnetic field fluctuations of anchored ships, the magnetic data is calculated using a geomagnetic model and filtered by a sliding window to extract the remaining magnetic field. The main interference sources in the scene are accurately located, and by suppressing clutter and removing fluctuations, the effective signals of pipelines can be effectively preserved, making the pipeline contours in the sonar data clearer and the magnetic anomaly signals in the magnetic data more prominent. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is an overall flowchart of a preferred embodiment of the present invention;

[0042] Figure 2 This is a preprocessing flowchart of a preferred embodiment of the present invention;

[0043] Figure 3 This is a flowchart of the feature extraction process according to a preferred embodiment of the present invention. Detailed Implementation

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0046] Application Overview:

[0047] This application focuses on the waters of near-shore ports and industrial areas, where there is significant interference from ship anchoring and the seabed contains a large amount of construction waste, such as concrete blocks and steel bars. This application targets the detection of submarine pipelines, especially sewage pipelines and cables in the aforementioned scenarios. Conventional techniques have long relied on acoustic detection, which has become the mainstream method for underwater detection due to its long underwater propagation distance and ability to directly obtain target shape information. This path dependence leads to the neglect of its inherent defects: in this scenario, the presence of a large amount of construction waste in the seabed will generate strong clutter and multiple reflections, interfering with the identification of pipeline signals; at the same time, the cladding layer formed by long-term use of sewage pipelines makes their acoustic characteristics similar to the surrounding environment, resulting in weak acoustic reflection signals, and the metal material of cables cannot be accurately identified and evaluated by acoustic waves.

[0048] Conventional approaches are limited to optimizing acoustic detection technology, attempting to reduce clutter by enhancing signals and improving algorithms, but they have always been unable to overcome the inherent limitations of acoustic waves in complex substrates and special coatings.

[0049] This application breaks with conventional methods. Based on in-depth analysis of a specific scenario, it discovers that the ferromagnetism of metal pipelines such as cables is unaffected by construction waste and cladding layers, serving as a stable detection marker; magnetic field propagation in this scenario is unaffected by construction waste, enabling stable capture of target signals; and there is a correlation between the external morphology information acquired by acoustic waves and the material information acquired by magnetic force regarding the pipeline's routing. Sonar data excels at capturing the external morphology of pipelines but cannot reflect their material composition and internal state; magnetic data excels at reflecting the material properties of pipelines but cannot provide morphological details. Therefore, this application integrates acoustic and magnetic detection, forming a complementary rather than simple superposition approach. By establishing a feature correlation model between the two, accurate verification of sewage pipelines and cables in this scenario can be achieved.

[0050] Exemplary method:

[0051] like Figure 1 As shown, an underwater acousto-magnetic multimodal fusion processing method includes the following steps:

[0052] S1. Preprocess the sonar and magnetic data of underwater pipelines;

[0053] S2. Extract the morphological feature vector of the pipeline from the sonar data, and extract the magnetic feature vector of the pipeline from the magnetic data.

[0054] S3. Based on the pipeline routing, calculate the matching degree between the sonar profile routing and the magnetic anomaly zone routing, calculate the optimal path through dynamic time warping, and establish an acoustomagnetic mapping; fuse the morphological feature vector and the magnetic feature vector to generate a joint feature vector.

[0055] S4. Determine the pipeline type and pipeline status based on the joint feature vector, where the pipeline type is sewage pipeline and cable.

[0056] Sonar data refers to data generated by transmitting sound waves underwater through sonar equipment, receiving reflected or scattered echo signals, and processing them. It can reflect the geometric shape and location information of underwater targets and has high spatial resolution, but is easily affected by environmental interference.

[0057] Specifically, side-scan sonar and shallow seismic profiler are used to acquire the data. The side-scan sonar emits sound waves to both sides and receives the echoes reflected by seabed targets to generate two-dimensional image data. The shallow seismic profiler emits low-frequency sound waves underwater, penetrates the seabed surface sediments, and receives the reflected signals from different strata interfaces to form profile data.

[0058] In the context of nearshore port and industrial area waters covered by this application, the sonar data includes: side-scan sonar echo intensity data and shallow seismic profiler reflection coefficient data. The side-scan sonar echo intensity data reveals the external outline and surface features of sewage pipelines and cables, as well as the distribution of surrounding construction waste, identifying the approximate shape and location of the targets; the shallow seismic profiler reflection coefficient data reflects the burial depth information of the pipelines and the interface characteristics with the surrounding sediments, providing a basis for determining the burial status of the pipelines.

[0059] Side-scan sonar equipment is fixed to the sides of the survey vessel or to an underwater robot to ensure that the direction of sound wave emission covers the area around the port and wharf, the extension area of ​​industrial sewage outlets, and the waters where known pipelines may pass; high-frequency sound waves of 200-500kHz are selected to improve resolution and avoid excessive scattering of low-frequency sound waves in shallow water; the emission beam angle is controlled at 10-20° to reduce sidelobe interference.

[0060] When sound waves come into contact with underwater targets, including sewage pipes, cables, and construction debris, they are reflected back. The echo signal is captured by the receiving transducer, preprocessed, and converted into an electrical signal. The echo intensity data is presented as two-dimensional image pixel values.

[0061] The shallow seismic profiler is installed at the center of the bottom of the survey vessel, with the transmitting transducer facing directly downwards. It targets nearshore seabed sediments, mainly silt and fine sand, with a thickness of 1-5 meters. Low-frequency sound waves of 500-2000Hz are selected to ensure the penetration depth.

[0062] The equipment travels along a survey line perpendicular to the shoreline, emitting low-frequency sound waves that penetrate the seabed surface. These sound waves are reflected at various media interfaces, including the seabed surface, pipeline walls, and sediment layer interfaces. The receiving transducer captures the reflected signals, processes the signals, and calculates the reflection coefficient; the reflection coefficient data is presented in profile form.

[0063] Among them, magnetic data refers to data obtained by measuring the strength and changes of underwater magnetic fields through magnetic detection equipment. It reflects the ferromagnetic characteristics of the target, is unaffected by water and sediment, and has strong penetrating power. In this application, a fluxgate magnetometer and a fiber optic magnetic sensor are used to acquire the data. The fluxgate magnetometer determines the magnetic field strength by measuring the induced electromotive force generated on the sensitive element by the magnetic field; the fiber optic magnetic sensor utilizes the magneto-optic effect to measure the magnetic field by detecting changes in the propagation characteristics of light in the optical fiber.

[0064] In the context of this application, the magnetic data includes total magnetic field strength data collected by a fluxgate magnetometer and magnetic field gradient data collected by a fiber optic magnetic sensor. The total magnetic field strength data can be used to identify ferromagnetic cables because the metal armor of the cable causes abnormal changes in the magnetic field; the magnetic field gradient data can more accurately locate the pipeline, distinguish the pipeline from the magnetic anomalies of surrounding construction waste, and reduce interference.

[0065] The fluxgate magnetometer is suspended 50-100 meters behind the survey vessel by a tow cable, at a depth of 3-5 meters underwater, with a sampling frequency of 1-10Hz. It measures the total magnetic field strength at the location in real time. The total magnetic field strength data is a continuous numerical sequence, including: background magnetic field data, magnetic anomaly data caused by the cable, and magnetic field data of the sewage pipeline area.

[0066] The fiber optic magnetic sensor employs an array design, fixed to the robotic arm of the underwater robot. A 632.8nm helium-neon laser is used as the laser source, and the photodetector has a sampling frequency of 10-50Hz to capture subtle changes in the magnetic field. The sensor array simultaneously measures the magnetic field strength at different spatial points. The magnetic field gradient is obtained by calculating the ratio of the difference in magnetic field strength between adjacent sensor units to their spacing. The magnetic field gradient data is a spatial gradient value sequence, including: the gradient peak at the cable location, gradient data from the construction waste area, and gradient data from the background area.

[0067] like Figure 2 As shown, preprocessing includes denoising the sonar and magnetic data.

[0068] For sonar data, a nonlocal mean filtering algorithm is used for noise reduction, with a window size of 0.7×0.7m. Since construction waste in the scene is mostly 0.5-1m in size, this window can effectively distinguish between isolated clutter and continuous pipeline contours. For strong reflection clutter generated by construction waste in near-shore port and industrial area waters, the similarity weight of neighboring pixels is calculated to suppress isolated high-value noise points and preserve the continuous contour signal of pipelines.

[0069] Its core calculation formula is as follows: ,in, The coordinates in the side-scan sonar echo intensity data are... The original pixel value, which reflects the intensity of acoustic wave reflection at the corresponding location in the waters of near-shore ports and industrial areas; The coordinates after filtering are: The pixel values, i.e., the pixel values ​​of the sonar data after noise reduction; Indicates A neighborhood window centered on the center, containing multiple pixels. Used to select and Pixels with similar characteristics; Represents pixels For pixels The weights are determined by calculating the similarity of the local regions where two pixels are located. The higher the similarity, the greater the weight, thereby suppressing isolated high-value noise points generated by construction waste and preserving the continuous contour signal of the pipeline.

[0070] Because the strong reflection of construction waste can obscure pipeline signals and affect the accuracy of subsequent feature extraction, processing can effectively remove clutter interference, making the pipeline outline clearer and improving the signal-to-noise ratio.

[0071] For magnetic data, the theoretical geomagnetic field strength of the detection area is first calculated using the IGRF geomagnetic model. The theoretical geomagnetic field strength is then subtracted from the original magnetic data to obtain the residual magnetic field data ΔT, thus eliminating the influence of the Earth's main magnetic field. Then, based on the ship's navigation trajectory and electromagnetic interference model, a sliding window filter is applied to the ΔT data to remove high-frequency magnetic field fluctuations caused by anchored ships and retain low-frequency pipeline magnetic anomaly signals.

[0072] The formula for calculating the residual magnetic field data in magnetic data preprocessing is as follows: ,in, This indicates the waters near near-shore ports and industrial areas, with coordinates as follows: The residual magnetic field data at time t can highlight local magnetic anomalies caused by sewage pipes and cables, especially metal-armored cables, and eliminate the masking effect of the Earth's main magnetic field. Indicates the fluxgate magnetometer in coordinates The raw total magnetic field strength data collected at time t includes multiple components such as the Earth's main magnetic field, pipeline magnetic anomalies, and ship anchoring interference. This indicates that the coordinates were calculated by calling the IGRF geomagnetic model. The theoretical geomagnetic field strength at time t reflects the baseline strength of the Earth's main magnetic field in the near-shore detection area.

[0073] Since the Earth's main magnetic field and ship interference can mask the magnetic anomalies of pipelines, the above processing can highlight the magnetic characteristics of pipelines, laying the foundation for subsequent magnetic anomaly extraction and improving the reliability of the data.

[0074] Spatiotemporal alignment is achieved by hardware-triggered time synchronization of sonar and magnetic data, ensuring timestamp error ≤1ms; by combining GNSS and INS positioning data, the relative distance coordinates of sonar and the sensor coordinates of magnetics are unified into the same geographic coordinate system, so that the two are consistent in space.

[0075] To ensure strict time-series matching between sonar and magnetic data, a multi-device hardware synchronization triggering scheme is adopted: a high-precision clock module is integrated into the main control system, and the acquisition and control units of the side-scan sonar, shallow seismic profiler, fluxgate magnetometer, and fiber optic magnetic sensor are connected through synchronization trigger lines.

[0076] When the system starts, the main control module preprocesses the underwater pipeline sonar and magnetic data every 1ms and sends a synchronization pulse signal. Each device immediately starts a data acquisition after receiving the pulse and embeds the timestamp of the pulse into the header information of the acquired data. For array acquisition of fiber optic magnetic sensors, the synchronization signal is additionally transmitted through optical fiber to avoid delay in electrical signal transmission.

[0077] The core of spatial alignment is to unify the transformation of the relative coordinate system of sonar data and the sensor coordinate system of magnetic data to the WGS84 geographic coordinate system;

[0078] Raw data from side-scan sonar and shallow seismic profiling are expressed in range-angle relative coordinates. Combining positioning data from shipborne GNSS and attitude data from INS, the relative distance of the sonar is converted to geographic coordinates using a coordinate transformation formula:

[0079] Lateral distance is converted into east-west offset using the heading angle;

[0080] The longitudinal distance is converted into a north-south offset by the cumulative distance of the GNSS track;

[0081] The depth data from the shallow seismic profiler, combined with the ship's draft and tide data, is converted into an absolute depth relative to sea level.

[0082] Ultimately, each pixel of the side-scan sonar and each profile sampling point of the shallow seismic profiler are assigned unique coordinates, including longitude, latitude, and depth.

[0083] The position of the fluxgate magnetometer's drag cable is calibrated in the following way:

[0084] Based on the tow cable length, ship heading angle, and underwater depth, the offset coordinates of the sensor relative to the ship's GNSS antenna are calculated. By superimposing the real-time geographic coordinates of the ship, the absolute coordinates of the magnetometer are obtained.

[0085] The coordinates of the fiber optic magnetic sensor are directly linked to the underwater robot's positioning system: the GNSS or ultra-short baseline positioning system on the robot provides its geographic coordinates, which, combined with the joint angle sensor data of the robotic arm, calculate the spatial coordinates of each sensor unit.

[0086] This is done because sonar and magnetic data come from different acquisition devices, and the temporal and spatial differences can lead to data mismatch, affecting the fusion effect. Spatiotemporal alignment ensures the correspondence between the two types of data in time and space, providing an accurate data foundation for subsequent feature fusion and correlation analysis, and improving the accuracy of fusion processing.

[0087] After preprocessing and spatiotemporally aligning the sonar and magnetic data in step S1, noise interference in the data is eliminated and the consistency between the two in time and space is ensured, laying a reliable data foundation for subsequent feature extraction. Next, step S2 is performed to extract feature vectors that reflect pipeline characteristics from the processed sonar and magnetic data.

[0088] like Figure 3 As shown, in step S2, the morphological feature vector of the pipeline is extracted from the sonar data, and the magnetic feature vector of the pipeline is extracted from the magnetic data.

[0089] Among them, the morphological feature vector refers to a set of quantitative parameters extracted from sonar data that can reflect the external geometric shape and spatial distribution characteristics of the pipeline. These parameters can effectively distinguish the pipeline from the surrounding environment and reflect the pipeline's own morphological attributes.

[0090] Extracting the morphological feature vector of the pipeline from sonar data, specifically:

[0091] The preprocessed and spatiotemporally aligned side-scan sonar echo intensity data is binarized, and a threshold is set to separate the pipeline contour region from the background region, resulting in a binarized contour image of the pipeline.

[0092] Based on the binarized contour image, the length feature of the pipeline is extracted, which is the total number of pixels of the continuous contour of the pipeline in the image multiplied by the spatial resolution of the sonar image (m / pixel) to obtain the pipeline length in meters. This length can reflect the extension range of the pipeline in the waters of near-shore ports and industrial areas.

[0093] The width feature of the pipeline is extracted by calculating the maximum pixel span perpendicular to the pipeline direction in the contour image and then multiplying it by the spatial resolution to obtain the pipeline width. This can be used to distinguish between sewage pipelines and cables with different diameters.

[0094] The directional angle features of the pipeline are extracted, and the Hough transform is used to detect straight line segments in the contour image. The angle between the straight line segment and the horizontal direction is calculated, which reflects the extension direction of the pipeline within the detection area.

[0095] Extract the pipeline's contour continuity features and statistically analyze the percentage of continuous pixels in the contour image. The ratio of continuous pixels to total pixels is calculated; the closer this value is to 1, the more complete the pipeline contour and the less obstruction it receives from construction waste. In areas with dense construction waste, the ratio is calculated per 10m... 3 Containing 3-5 concrete blocks, the continuity value of the complete pipeline outline is usually >0.6. If C <0.4, it indicates severe shading, which needs to be verified in conjunction with magnetic data.

[0096] The extracted length (L), width (W), orientation angle (θ), and profile continuity (C) constitute the pipeline's morphological feature vector, i.e., morphological feature vector = .

[0097] Among them, the magnetic feature vector refers to a set of quantitative parameters extracted from magnetic data that can reflect the ferromagnetic properties of pipelines and the abnormal distribution of magnetic fields. It can be used to identify the material properties of pipelines and for precise positioning.

[0098] Extracting the magnetic feature vector of the pipeline from the magnetic data, specifically:

[0099] A sliding window search was performed on the preprocessed and spatiotemporally aligned residual magnetic field data. The window size was 2m×2m to identify magnetic anomaly regions with ΔT values ​​> 300nT. These regions correspond to ferromagnetic pipelines, especially cables.

[0100] Extract the peak characteristics of magnetic anomalies, that is, the maximum value of ΔT within the magnetic anomaly region. The larger the value, the stronger the ferromagnetism of the pipeline, and the more likely it is a cable.

[0101] Extracting the magnetic anomaly range features involves calculating the area of ​​the magnetic anomaly region, which is calculated as the number of pixels multiplied by the square of the spatial resolution. This area is related to the length and diameter of the pipeline, and helps determine the pipeline size.

[0102] Extracting the maximum value feature of the magnetic field gradient: From the magnetic field gradient data collected by the fiber optic magnetic sensor, the maximum value of the gradient in the magnetic anomaly region is extracted. This value can reflect the rate of change of the magnetic field at the pipeline edge, which helps to accurately locate the pipeline boundary.

[0103] Extract the symmetry features of magnetic anomalies, calculate the degree of symmetry of ΔT values ​​within the magnetic anomaly region about the central axis of the region, and the sum of ΔT differences of symmetrical pixels / total number of pixels. Magnetic anomalies of cables usually have high symmetry, while magnetic anomalies of sewage pipelines have low symmetry.

[0104] The magnetic eigenvector of the pipeline is constructed from the extracted magnetic anomaly peak value (P), magnetic anomaly range (A), maximum magnetic field gradient (G), and magnetic anomaly symmetry (S), i.e., magnetic eigenvector = .

[0105] In this step, by extracting morphological and magnetic feature vectors, the original sonar and magnetic data are transformed into quantified parameters with clear physical meaning. This achieves data dimensionality reduction while preserving the key characteristics of the pipeline. The morphological feature vector describes the pipeline from a geometric perspective, while the magnetic feature vector reflects the pipeline's characteristics from a ferromagnetic perspective. This provides effective input parameters for subsequent feature fusion and pipeline identification, improving the distinguishability of features and the model's recognition accuracy.

[0106] Step S3: Calculate the matching degree between the sonar profile and the magnetic anomaly zone based on the pipeline orientation, establish an acoustomagnetic mapping, and fuse the morphological feature vector and the magnetic feature vector to generate a joint feature vector.

[0107] Among them, pipeline orientation refers to the direction in which the pipeline extends in the waters of near-shore ports and industrial areas. It is reflected by both the orientation of the sonar profile and the orientation of the magnetic anomaly zone, and is an important link between sonar data and magnetic data.

[0108] The sonar profile orientation is the pipeline extension direction extracted from the pipeline binarized profile image of the sonar data, while the magnetic anomaly zone orientation is the pipeline extension direction extracted from the magnetic anomaly region of the magnetic data.

[0109] Matching degree is a quantitative indicator used to measure the consistency between the orientation of the sonar profile and the orientation of the magnetic anomaly zone. The higher the value, the closer the correlation between the two.

[0110] Acoustomagnetic mapping refers to establishing the correspondence between pipeline features in sonar data and pipeline features in magnetic data, providing a basis for feature fusion.

[0111] The joint feature vector is a comprehensive feature vector obtained by fusing the morphological feature vector and the magnetic feature vector. It combines the characteristics of both and can more comprehensively reflect the properties of the pipeline.

[0112] The matching degree between the sonar profile and the magnetic anomaly zone is calculated based on the pipeline orientation. Specifically:

[0113] Obtain the pipeline's orientation angle from the morphological feature vector. , as a quantization parameter for the orientation of the sonar profile; It is the angle between the straight line segment and the horizontal direction extracted from the binarized contour image of the pipeline through Hough transform;

[0114] Perform a Hough transform on the magnetic anomaly regions in the magnetic data, detect the straight segments of the magnetic anomaly zone, and calculate their angles with the horizontal direction. , as a quantitative parameter for the orientation of the magnetic anomaly zone;

[0115] Calculate the degree of matching ,in, The sonar profile orientation angle. For the orientation angle of the magnetic anomaly zone, when and When they are perfectly aligned, M=1; when they are perpendicular, M=0; the above is rigid alignment.

[0116] In near-shore port and industrial area waters, if M≥0.8, it can be preliminarily determined that the sonar profile and the magnetic anomaly zone correspond to the same pipeline. This is because the pipeline route is relatively stable in this scenario, allowing for a certain measurement error.

[0117] Furthermore, the optimal path mapping is solved by dynamically warping the acoustic centerline and the magnetic anomaly trajectory;

[0118] Extracting the coordinate sequence of the acoustic centerline from sonar data ,in, For the first The planar coordinates of each sampling point reflect the center position of the pipeline in the sonar field of view;

[0119] Extracting coordinate sequences of magnetic anomaly trajectories from magnetic data ,in, For the The planar coordinates of the peak magnetic anomalies correspond to the locations where the pipeline's ferromagnetism is strongest.

[0120] Calculate the Euclidean distance between each pair of points on the acoustic centerline and the magnetic anomaly trajectory, and construct the distance matrix. Among them, elements ,express and Distance in space.

[0121] The distance matrix quantifies the degree of spatial misalignment of all pairs of points on two trajectories. Based on the idea of ​​dynamic programming, we find the optimal path from the upper left corner to the lower right corner of the matrix to minimize the cumulative distance of the path.

[0122] In nearshore port and industrial waters, pipelines may be subject to local bending due to construction deviations or sedimentation. DTW’s flexible alignment allows for non-uniform trajectory shifts in time and space, making it more adaptable to complex environments compared to rigid alignment.

[0123] Establish a one-to-one correspondence between the acoustic centerline sampling points and the magnetic anomaly trajectory sampling points based on the optimal path. This mapping can correct the spatial misalignment of the two trajectories, enabling precise spatial correlation between sonar morphological features and magnetic features.

[0124] The acoustomagnetic mapping is established based on the matching degree (M≥0.8) and the optimal path;

[0125] Based on the calculation of a path matching degree M≥0.8, and combined with the DTW optimal path mapping, further screening of truly correlated sonar and magnetic features is performed:

[0126] If the cumulative distance of the optimal path is ≤0.5m, then the acoustic centerline and the magnetic anomaly trajectory are determined to correspond to the same pipeline, and the morphological feature vector is... With magnetic eigenvectors Map and bind according to the optimal path;

[0127] If the cumulative distance is greater than 0.5m, it is necessary to re-examine whether it is a false magnetic anomaly caused by construction waste or a pipeline outline misidentified by sonar, and eliminate interference.

[0128] The morphological feature vector and the magnetic feature vector are fused to generate a joint feature vector, as follows:

[0129] The morphological and magnetic feature vectors are normalized to map each feature parameter to the [0,1] interval, thus eliminating dimensional differences.

[0130] Traditional feature fusion methods, such as simple weighting or splicing, cannot fully explore the inherent physical relationships between acoustic and magnetic data and are easily affected by environmental noise. This method leverages the complementary nature of sonar data in accurately depicting pipeline geometry and magnetic data in effectively reflecting pipeline material properties. It employs interactive feature modeling and feature-level deep association strategies to deeply integrate morphological geometry information with material physical properties, achieving more accurate and robust feature representation.

[0131] The joint feature vector is formed by dynamically compressing interactive features and derived features, including: material correlation factor, spatial consistency factor, coupling coefficient, anti-interference factor, scale-invariant features and scale ratio, and six-dimensional features.

[0132] In terms of feature interaction, a variety of factors and coefficients are constructed for different situations to achieve accurate pipeline identification; specifically, these include: material association factor, spatial consistency factor, and coupling coefficient.

[0133] Material correlation factors are combined with magnetic anomaly intensity, edge sharpness and pipeline width to form a composite index, revealing the correlation between ferromagnetic materials and tubular structures, and reflecting the difference between good cable profile continuity and poor sewage pipeline profile continuity.

[0134] The spatial consistency factor quantifies the matching degree of acoustic and magnetic trajectories by mapping the spatial deviation of the DTW points, and measures the consistency of the spatial distribution of acoustic and magnetic features. The closer the value is to 1, the higher the probability that the sonar morphology and magnetic anomaly come from the same pipeline.

[0135] To address the contradiction in the acoustic and magnetic characteristics of the covered sewage pipe, namely, the clear sonar profile but weak magnetic signal, a coupling coefficient is constructed to reveal the coupling relationship between the magnetic anomaly range and the acoustic deviation, thereby suppressing interference from broken or obstructed pipelines.

[0136] Specifically, feature interactions include:

[0137] A single sonar signal cannot identify the material of a sewage pipe encased in a protective layer, and a single magnetic field cannot distinguish between steel reinforcement in construction waste and actual cables. Material correlation factors are used to construct a composite index for material identification by fusing magnetic anomaly intensity (P), edge sharpness (G), and pipe width (W).

[0138] Material correlation factor: Among them, molecules Multiplying the two can highlight the magnetic properties and structural integrity of the pipeline; cables have high magnetic peaks and high gradients due to their metal armor, while the magnetic gradients of steel reinforcement fragments are scattered. This is used to reduce the influence of tube diameter differences on magnetic signals, making the features scale invariant. Nonlinear mapping of the contour continuity reveals a strong correlation between ferromagnetic materials and tubular structures. Cables, due to their metallic materials, are typically... The magnetic signal of sewage pipelines is usually weakened by the coating. .

[0139] Construction debris can cause sonar profile breaks, and magnetic data may drift due to mooring interference. The spatial consistency factor quantifies the acoustic-magnetic trajectory matching degree by mapping the spatial deviation of the DTW points.

[0140] Spatial consistency factor: ,in, Constants for scene adaptation; The sum of spatial deviations between the sonar profile and the magnetic anomaly zone is calculated by using DTW mapping points to measure the consistency of the spatial distribution of acoustic and magnetic features. Used for normalization to eliminate the influence of the number of mapping points N. The closer this factor value is to 1, the higher the probability that the sonar morphology and magnetic anomaly originate from the same pipeline, and the stronger the resistance to environmental interference such as construction waste.

[0141] The acoustic-magnetic characteristics of the coated sewage pipe are contradictory: the sonar profile is clear (high C value) but the magnetic signal is weak (low P value); the coupling coefficient reveals the coupling relationship between the magnetic anomaly range (A) and the acoustic orientation deviation (Δθ).

[0142] Coupling coefficient: ,in, This represents the difference between the sonar profile orientation angle and the magnetic anomaly zone orientation angle. To avoid the minimum value where the denominator is 0, This reflects the coupling relationship between the range of magnetic anomalies and the deviation in acoustic-magnetic orientation; the greater the deviation, the higher this value. Considering the effective length of the pipeline, suppress interference from broken or blocked pipelines; the cable has high consistency between magnetic anomalies and its direction due to its uniform metal material.

[0143] The aforementioned interactive features and derived features are dynamically compressed. The dynamically compressed features include: anti-interference factor, scale-invariant features, and scale ratio, constructing a 6-dimensional joint feature vector. ;in, As an anti-interference factor, element-wise product Real pipelines require the coexistence of magnetic anomalies and acoustic profiles. This represents element-wise multiplication; As a scale-invariant characteristic, local deformation of the pipeline will cause changes in the pipeline width (W) and magnetic gradient (G), but the proportional relationship between the two remains stable. The angle between the direction of magnetic field gradient change and the width of the tubular structure is calculated using the arctangent function in the four quadrants. This eliminates the influence of differences in pipe dimensions such as different pipe diameters and focuses on the proportional relationship between the sharpness of the magnetic anomaly edge and the pipe width. To differentiate between short-distance strong magnetic anomalies and long-distance linear targets, the magnetic anomaly intensity per unit length of pipeline is characterized by scale ratio.

[0144] In this step, DTW path alignment overcomes the limitations of simple path matching. By using flexible matching to adapt to pipeline trajectory deformation in complex nearshore environments, it reduces the spatial correlation error of acoustic and magnetic features. Combined with matching degree screening and optimal path mapping, it further eliminates false associations of interfering targets such as construction waste, providing a more reliable foundation for feature fusion. Ultimately, the joint feature vector has both morphological integrity and material specificity, providing more comprehensive feature support for pipeline identification.

[0145] In step S4, the pipeline type and pipeline status are determined based on the joint feature vector, where the pipeline type is sewage pipeline and cable.

[0146] Among them, pipeline type identification refers to distinguishing whether the detected pipeline is a sewage pipeline or an electrical cable. These two types of pipelines differ significantly in terms of material, function, and testing requirements.

[0147] Sewage pipelines are mostly made of non-metallic materials, such as concrete and plastic, and are mainly used to transport sewage. They may have a biological coating on the outside. Cables are mostly made of metallic materials, such as copper core with steel armor sheath, and are used to transmit electricity or signals. They have strong ferromagnetism.

[0148] Pipeline condition refers to the integrity of the pipeline. For sewage pipelines, the main focus is on the integrity of their outer sheath; for cables, the main focus is on the corrosion of their metal sheath and the continuity of their overall structure.

[0149] Pipeline type is determined based on joint feature vectors, specifically:

[0150] Different types of pipelines, such as sewage pipelines and cables, differ in morphological and magnetic characteristics. By combining the features in the joint feature vector, the types of pipelines can be distinguished.

[0151] In nearshore port and industrial waters, cables typically exhibit high magnetic anomaly peaks and symmetrical distributions due to the strong ferromagnetism of their metallic sheaths. In contrast, sewage pipelines, mostly made of non-metallic materials, show lower magnetic anomaly peaks and less symmetry. Based on these characteristics, setting threshold values ​​can differentiate between pipeline types.

[0152] Cable identification: When the following conditions are met and When the condition is met, it is determined to be a cable; or if the condition is satisfied... , and At that time, it was determined to be a cable;

[0153] Sewage pipeline identification:

[0154] Sewage pipes with cladding: when the following conditions are met , and When the condition is not met, it is determined to be a sewage pipe with a cladding layer; standard sewage pipe: when it does not meet the cable identification conditions and does not meet the characteristics of a sewage pipe with a cladding layer, it is determined to be a standard sewage pipe.

[0155] Pipeline status is determined based on joint feature vectors, specifically:

[0156] The integrity of a pipeline is an important indicator for assessing its normal operation. By combining the pipeline's profile continuity and magnetic anomalies, the integrity of the pipeline can be comprehensively judged.

[0157] Integrity assessment model ,in, It is the benchmark magnetic anomaly value for similar pipelines;

[0158] Furthermore, for pipelines identified as cables, the focus is on the maximum magnetic field gradient and the continuity of the profile. The maximum magnetic field gradient reflects the edge changes of the cable's metal sheath. If the maximum magnetic field gradient changes abruptly, it indicates that there may be sheath corrosion or damage at that location. A profile continuity of <0.8 indicates that the cable may have a local break or be severely obstructed by construction debris, affecting its normal function.

[0159] For pipelines identified as sewage pipelines, pay close attention to changes in outline continuity and width characteristics: an outline continuity of <0.7 indicates that the outer covering of the sewage pipeline may be damaged or detached; if the width characteristic increases significantly in a certain section, it may be due to local expansion of the covering or deformation of the pipeline.

[0160] Based on the above characteristics, the pipeline status is classified into different levels:

[0161] Status Level 1 (Intact): All characteristics are within the normal range;

[0162] Status Level 2 (Minor Anomaly): There are slight deviations in individual features, but they do not affect the overall function;

[0163] Status Level 3 (Moderate Abnormality): Multiple features deviate, and there may be local damage;

[0164] Status Level 4 (Severe Anomaly): The characteristics deviate significantly, posing a major safety hazard, and immediate repair is required.

[0165] The method of determining pipeline type and status based on joint feature vectors makes full use of the complementarity of morphological and magnetic features: the material type is accurately distinguished by magnetic features, avoiding misjudgment caused by the interference of the coating layer due to single sonar data; the pipeline status is comprehensively evaluated by the integrated analysis of morphological and magnetic features, overcoming the limitation that single magnetic data cannot reflect changes in the external morphology of the pipeline.

[0166] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for underwater acoustic-magnetic multimodal fusion processing, characterized in that, Includes the following steps: S1. Preprocess the sonar and magnetic data of underwater pipelines; S2. Extract the morphological feature vector of the pipeline from the sonar data and extract the magnetic feature vector of the pipeline from the magnetic data. S3. Based on the pipeline routing, calculate the matching degree between the sonar profile and the magnetic anomaly zone, calculate the optimal path through dynamic time warping, and establish an acoustomagnetic mapping. The morphological feature vector and the magnetic feature vector are then fused to generate a joint feature vector. The step of calculating the matching degree in step S3 includes: The pipeline orientation angle is extracted from sonar data as the sonar profile orientation, and the magnetic anomaly zone orientation angle is extracted from magnetic data. Calculate the angular difference between the two orientation angles. ,in, The sonar profile orientation angle. For the orientation angle of the magnetic anomaly zone, when and When they are completely identical, M=1; when they are perpendicular, M=0; when the matching degree corresponding to the difference value is ≥0.8, they are judged to be the same pipeline feature. The specific steps of dynamic time warping in step S3 include: Extract the acoustic centerline coordinate sequence and the magnetic anomaly trajectory coordinate sequence, and construct the spatial distance matrix between the two sequences; The optimal path with the minimum cumulative distance is solved by dynamic programming, and a point-to-point mapping relationship is established. When the cumulative distance of the optimal path is ≤0.5m, bind the morphological feature vector and the magnetic feature vector; The specific steps of feature fusion in step S3 include: The morphological feature vector and magnetic feature vector are normalized, and each feature parameter is mapped to the interval [0,1]. Based on the acoustic-magnetic mapping relationship, three types of interactive features are calculated: material correlation factor, spatial consistency factor, and coupling coefficient. A 6-dimensional joint feature vector containing anti-interference factor, scale-invariant features, and scale ratio is constructed by combining dynamic compression; S4. Determine the pipeline type and pipeline status based on the joint feature vector; where the pipeline type is sewage pipeline and cable.

2. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: The preprocessing of sonar data in step S1 includes: A nonlocal mean filtering algorithm is adopted to suppress strong reflection clutter generated by construction waste by calculating the similarity weight of neighboring pixels, while preserving the continuous contour signal of the pipeline.

3. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: The preprocessing of the magnetic data in step S1 includes: The theoretical geomagnetic field strength is calculated using a geomagnetic model, and this value is subtracted from the original magnetic data to obtain the remaining magnetic field data. A sliding window filter is applied to the remaining magnetic field data to remove high-frequency magnetic field fluctuations caused by ship anchoring.

4. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: The sonar data includes: side-scan sonar echo intensity data and shallow seismic profiler reflection coefficient data; the magnetic data includes: total magnetic field intensity data and magnetic field gradient data. Preprocessing also includes spatiotemporal alignment, the specific steps of which are: to achieve time synchronization between the sonar device and the magnetic device by triggering pulses through hardware, to convert the distance-angle coordinates of the side-scan sonar into the WGS84 geographic coordinate system by combining GNSS positioning data and INS attitude data, and to unify the coordinates of the magnetic sensor to the same geographic coordinate system by calibrating the tow cable position.

5. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: In step S2, the morphological feature vector includes pipeline length, pipeline width, pipeline orientation angle, and contour continuity; the magnetic feature vector includes magnetic anomaly peak value, magnetic anomaly range, maximum magnetic field gradient value, and magnetic anomaly symmetry.

6. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: The material correlation factor is constructed by fusing the peak value of magnetic anomaly, the maximum value of magnetic field gradient, and pipeline width; the spatial consistency factor is calculated based on the spatial deviation of the dynamic time warp mapping point; the coupling coefficient reflects the coupling relationship between the magnetic anomaly range and the acousto-magnetic orientation deviation.

7. The underwater acoustic-magnetic multi-mode fusion processing method according to claim 1, characterized in that: The specific steps for determining the pipeline type in step S4 include: When the peak value of the magnetic anomaly is greater than 500 nT and the symmetry of the magnetic anomaly is greater than 0.7, it is determined to be a cable; When the peak value of the magnetic anomaly is <200nT, the profile continuity is >0.8, and the coupling coefficient is <0.3, it is determined to be a sewage pipeline with a cladding layer; All other cases are classified as standard sewage pipelines.

Images