A method for ocean surveying and mapping based on multi-beam bathymetry correction

Abstract

The present invention discloses a method for ocean surveying and mapping based on multi-beam bathymetric correction, belonging to the technical field of ocean surveying and mapping. This method uses a multi-beam sonar array, an acoustic tomography transceiver array, and a laser Raman probe to collect bathymetric echo signals, acoustic wave propagation delay data, and temperature data. This method then fuses these data to construct a dynamic refractive index field model, uses an adaptive variable step-size algorithm to correct for acoustic ray refraction distortion, and performs weighted distribution and data fusion on spatiotemporally adjacent beams to obtain high-confidence water depths and generate bathymetric topographic maps. This method solves the refraction distortion problem through multi-source data fusion, dynamically optimizes the acoustic ray tracking step-size, and uses weighted fusion to suppress noise interference, significantly improving the accuracy and reliability of topographic mapping in complex environments such as deep-sea hydrothermal areas.

Description

Technical Field

[0001] The present invention belongs to the technical field of ocean surveying and mapping, and in particular relates to an ocean surveying and mapping method based on multi-beam bathymetric correction. Background Art

[0002] Marine mapping technology plays a vital role in deep-sea resource exploration, seabed topography mapping, and marine environmental monitoring. Multi-beam bathymetry, a core method in this field, uses a multi-beam sonar array to transmit and receive high-frequency acoustic signals, acquire real-time bathymetric echo data from the seabed, and reconstruct seabed topography models. This method provides wide-scale, high-resolution water depth information and generates contour maps, providing key support for deep-sea engineering operations, geological surveys, and ecosystem research. It overcomes the limitations of traditional single-beam bathymetry, improves mapping efficiency and coverage, and enables detailed exploration of complex seabed landforms, supporting the practical needs of marine resource development, disaster warning, and scientific research.

[0003] However, existing multi-beam bathymetry applications still face a series of technical issues, limiting their accuracy and reliability in complex ocean conditions. First, in highly inhomogeneous media environments, such as deep-sea hydrothermal turbulence zones, the acoustic wave propagation path undergoes refraction distortion due to drastic changes in factors such as temperature and salinity, causing the bathymetric data to deviate from the true seabed topography. This refraction effect prevents the acoustic ray trajectories from accurately reflecting the actual water depth, resulting in significant errors in topographic reconstruction. Second, existing ray tracing methods often use fixed-step algorithms. This makes it difficult to dynamically adjust the calculation step size in areas with sudden changes in sound velocity gradients, such as hydrothermal vent boundaries. This leads to increased cumulative errors in the acoustic ray paths, reduced data accuracy, and distorted mapping results at key topographic features such as steep slopes or rift valleys. Third, single-source data acquisition strategies cannot fully quantify the anomaly distribution of sound velocity caused by hydrothermal turbulence. Traditional methods rely on a single sensor to obtain water depth information and lack the integration of acoustic tomography and temperature data. This makes the refractive index model unable to accurately capture medium inhomogeneities, further exacerbating the deviations in bathymetric results. Fourth, random noise interference caused by turbulent fluctuations leads to high discreteness in water depth data. Existing methods cannot effectively suppress noise through simple filtering or averaging techniques, making the surveying and mapping data unstable in hydrothermal-affected areas, reducing consistency and reliability. In addition, the synchronization of multi-sensor data acquisition also poses a challenge. If acoustic tomography, multi-beam sonar, and temperature measurement equipment are not strictly aligned in time, time drift errors will be introduced, affecting the collaborative analysis of multi-source data and the accuracy of the final terrain reconstruction.

[0004] These technical issues collectively contribute to the limitations of existing multi-beam bathymetry methods in complex deep-sea environments. This not only restricts the application of oceanographic mapping equipment in extreme scenarios such as hydrothermal vents, but also hinders the safety and efficiency of deep-sea engineering operations. Therefore, designing an oceanographic mapping method that can effectively address refraction distortion, dynamically optimize ray tracing, fuse multi-source data, and suppress noise is an urgent need to improve the reliability and engineering applicability of deep-sea topographic mapping. Summary of the Invention

[0005] To solve the problems in the background technology, the present invention provides an ocean surveying and mapping method based on multi-beam bathymetric correction, comprising the following steps:

[0006] S1, uses a multi-beam sonar array to acquire bathymetric echo signals from deep-sea hydrothermal areas, while simultaneously collecting acoustic wave propagation delay data across the hydrothermal area using an acoustic tomography transceiver array, and measuring the hydrothermal plume temperature using a laser Raman probe;

[0007] S2, based on the data collected by S1, fuses the acoustic tomography data and laser Raman data in the refractive index field inversion module to construct a dynamic refractive index field model;

[0008] S3, in the sound ray tracking module, an adaptive variable step size algorithm is used to correct the sound ray refraction distortion and generate a preliminary water depth;

[0009] S4, through the water depth fusion module, weights are assigned and data are fused for adjacent beams in time and space, and high-confidence water depth is output;

[0010] S5, transmit the high-confidence water depth to the surface display and control terminal via the data output interface to generate a contour topographic map.

[0011] Furthermore, in S2, the specific steps of constructing the dynamic refractive index field include:

[0012] S21, divide the target water area into two parts with side length 3D grid;

[0013] S22, construct the fusion objective function in the refractive index field inversion module:

[0014] ;

[0015] in is the objective function value; is the acoustic tomography projection matrix, which is constructed by the acoustic wave path and grid mapping relationship of the acoustic tomography transceiver array; is the refractive index vector of the grid node, the variable to be solved; is the measured acoustic delay vector, acquired through acoustic tomography transceiver array acquisition; 、 is the regularization coefficient, which is set according to the inversion stability requirements; The Raman point selection matrix is ​​used to extract the grid nodes corresponding to the laser Raman probe; is the Raman measured refractive index vector, which is obtained by measuring the temperature with a laser Raman probe and converting it; is the Laplace smoothing operator matrix, used to enforce spatial smoothness of the refractive index field; is the norm square operator, which calculates the sum of the squares of the vector elements;

[0016] S23, iteratively solve the objective function through the conjugate gradient method and output the dynamic refractive index field .

[0017] Furthermore, the specific steps of S3 include:

[0018] S31, real-time calculation of the refractive index gradient at the sound ray path point in the sound ray tracing module and the second-order derivative ;

[0019] S32, dynamically adjust the integration step size according to the gradient mutation intensity:

[0020] ;

[0021] in is the next integration step length; It is a preset maximum step size, set according to the system computing performance constraints; To preset the minimum step size, it is determined based on the wavelength of the sound waves of the multi-beam sonar array; is the normalization coefficient, an empirical parameter used to adjust the step size; is the magnitude of the refractive index gradient, which is calculated in real time at the acoustic ray path point and obtained by performing the first-order spatial differentiation of the refractive index field model output by S2; is the magnitude of the second-order derivative of the refractive index gradient, which is calculated in real time at the acoustic ray path point and obtained by performing the second-order spatial differentiation of the refractive index field model output by S2; is the weight coefficient, which is used to balance the influence of gradient and second-order derivative and is optimized through numerical experiments; is the minimum value operator; is the maximum value operator;

[0022] S33, using the variable step size Runge-Kutta method to solve the acoustic ray trajectory differential equation:

[0023] ;

[0024] in is the sound ray deflection angle vector; is the acoustic path length; η is the refractive index, which is provided by the refractive index field model output by S2; is the partial derivative of the refractive index in the normal direction of the sound line, calculated in real time at the path point; is the derivative of the sound ray deflection angle with respect to the path length.

[0025] Furthermore, the specific steps of data fusion in S4 include:

[0026] S41, in the water depth fusion module, the spatiotemporal adjacent beams are defined as acoustic ray clusters;

[0027] S42, calculate the intra-cluster beam weights:

[0028] ;

[0029] in For the beam weights; Beam and The pointing angle difference is calculated by the beam pointing angle data of the multi-beam sonar array; Beam and The emission time difference is calculated by the timestamp data of the synchronization control unit; is the spatially related scale parameter of turbulence, which is set based on the statistical model of turbulence in the hydrothermal region; is the turbulence time-dependent scale parameter, which is set based on the turbulence statistical model in the hydrothermal region; is the natural exponential operator;

[0030] S43, weighted output fusion water depth:

[0031] ;

[0032] in For the Beam-corrected water depth; For the The original water depth of the beam is output by the ray tracing module; For all beams in the cluster Summation operator.

[0033] Furthermore, the specific steps of S5 are as follows:

[0034] S51: Geographic coordinate mapping, converting the fused water depth data from the sensor coordinate system to the geographic coordinate system, and mapping the sonar measurement values ​​to three-dimensional geodetic coordinates through the coordinate transformation matrix;

[0035] S52: Generate isobath terrain. Use Kriging spatial interpolation algorithm to generate isobath terrain map in the display and control terminal.

[0036] ;

[0037] in: Points to be interpolated Estimated water depth; is the summation operator; is the weight coefficient, which quantifies the spatial correlation through the semivariogram; For known points Observed water depth;

[0038] S53: Accuracy verification, quantitatively evaluating accuracy through the standard deviation formula:

[0039] ;

[0040] in: To correct for the standard deviation of water depth; is the total number of beams; For the Beam-corrected water depth; For the The true water depth of the beam is obtained by measuring the seabed benchmark.

[0041] The present invention provides an ocean surveying and mapping system based on multi-beam bathymetric correction, comprising:

[0042] The pressure-resistant towed body has a navigation sonar installed on its head and a cross-shaped titanium alloy bracket fixed at its bottom. The four extension arms of the bracket form a 45° angle with the central axis of the towed body. Each arm end of the bracket is equipped with an acoustically transparent sealed cabin to protect the internal sensors.

[0043] The multi-source sensor array includes a multi-beam sonar array, an acoustic tomography transceiver array, and a laser Raman probe. The multi-beam sonar array is installed at the center of the towed body's bottom, transmitting and receiving 200-400kHz sound waves and outputting raw sounding data. The acoustic tomography transceiver array contains eight piezoelectric ceramic array elements, symmetrically arranged in acoustically sealed chambers at the four ends of the cross-bracket, and is used to collect acoustic wave propagation delay data across the hydrothermal zone. The laser Raman probe, which passes vertically through the center of the cross-bracket and has a built-in sapphire optical window, is used to measure the hydrothermal plume temperature and convert it into refractive index.

[0044] The synchronization control unit includes a high-precision clock and a trigger controller. The high-precision clock, based on a rubidium atomic clock, connects the multi-beam array's main control board, the tomography array's signal processor, and the laser Raman probe's ADC acquisition module in a star topology via the IEEE 1588 fiber-optic synchronization bus. The trigger controller sends synchronization excitation commands to the multi-beam array and acoustic tomography array to ensure consistent data acquisition timing.

[0045] The data processing unit includes a refractive index field inversion module, an acoustic ray tracing module, and a water depth fusion module. The refractive index field inversion module has a built-in board for fusing acoustic time delay and Raman temperature data to construct a dynamic refractive index field. The acoustic ray tracing module has a built-in FPGA chip, which is interconnected with the refractive index field inversion module to correct acoustic ray refraction distortion. The water depth fusion module has a built-in DSP processor that receives the output data of the acoustic ray tracing module and performs spatiotemporal beam weighted fusion.

[0046] The data output system includes an armored optical cable interface and a display and control terminal. The armored optical cable interface runs through the sealed cabin on the top of the towed body and is directly connected to the data processing unit for transmitting the corrected water depth data to the water surface; the display and control terminal receives data from the armored optical cable interface through optical fiber and generates a contour topographic map in real time.

[0047] In a preferred embodiment, the four extension arms of the cross-shaped titanium alloy bracket form an angle of 45° with the central axis of the tow body, and the piezoelectric ceramic array elements of the acoustic tomography transceiver array are fixed in the acoustically transparent sealed cabin provided at the end of each arm.

[0048] In a preferred solution, a coprocessor is added to the data processing unit and is vertically plugged into the top of the board through an MXM interface to accelerate the refractive index field inversion calculation.

[0049] In the preferred solution, the IEEE 1588 fiber optic synchronization bus connects the main control board of the multi-beam sonar array, the signal processor of the acoustic tomography transceiver array, and the ADC acquisition module of the laser Raman probe in a star topology to achieve microsecond time synchronization.

[0050] The beneficial effects achieved by the present invention are:

[0051] First, the present invention designs a multi-source data fusion method, which significantly improves the accuracy and reliability of deep-sea hydrothermal area topography mapping. By combining acoustic tomography data and laser Raman temperature data, a dynamic refractive index field model is constructed to accurately quantify the abnormal distribution of sound velocity caused by hydrothermal turbulence, thereby effectively solving the refraction distortion problem of traditional single-source sounding in complex environments, ensuring efficient synchronous acquisition and real-time analysis of multi-modal data, and reducing measurement errors caused by medium inhomogeneity in deep-sea mapping.

[0052] Second, the present invention constructs a dynamic refractive index field model. By fusing acoustic time delay and hydrothermal temperature data, and using objective function optimization and conjugate gradient iterative inversion to invert the three-dimensional sound velocity distribution, this scheme achieves accurate modeling of the acoustic characteristics of hydrothermal plumes, significantly enhances the ability to perceive medium heterogeneity, and ensures the smoothness and stability of the model in the gradient mutation area.

[0053] Third, the present invention designs an adaptive sound ray tracking correction technology, which calculates the refractive index gradient and its second-order derivative in real time, dynamically adjusts the integration step size, and combines the variable step size Runge-Kutta method to solve the sound ray trajectory. This mechanism enables the algorithm to automatically reduce the step size in strong turbulence areas to capture details, and expand the step size in flat areas to improve efficiency, thereby effectively overcoming the cumulative error problem of the traditional fixed step size method and ensuring the millimeter-level accuracy and robustness of the sound ray path calculation.

[0054] Fourth, the present invention designs a sound ray cluster fusion correction method. By weightedly fusing adjacent beams in time and space and calculating the weights based on the turbulence-related model, this scheme effectively suppresses random noise interference and significantly improves the stability and consistency of water depth data. The output results show highly smooth characteristics in the hydrothermal influence area, solving the data discreteness problem caused by turbulent fluctuations in traditional methods.

[0055] Fifth, the present invention designs an ocean mapping system based on multi-beam bathymetry correction, with an integrated sensor layout and hardware collaborative design, including a cross-shaped titanium alloy bracket to optimize the coverage of multi-source sensors, a sound-transmitting sealed cabin at each arm end to protect the piezoelectric ceramic array elements of the acoustic tomography array, and a vertical through-the-core layout of the laser Raman probe to ensure direct contact with the core of the hydrothermal plume; the synchronization control unit adopts a high-precision rubidium atomic clock and an IEEE 1588 optical fiber synchronization bus in a star topology to achieve microsecond time synchronization, ensuring strict timing alignment of multi-sensor data acquisition; the data processing unit combines GPU, FPGA and DSP hardware acceleration modules, and connects to the coprocessor through the MXM interface to execute refractive index field inversion, sound ray tracking and water depth fusion algorithms in real time, thereby improving the accuracy and reliability of deep-sea hydrothermal area topography mapping, overcoming the data distortion problem of traditional systems in turbulent environments, and enhancing the stability and consistency of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a comparison chart of the original water depth and the corrected water depth in Example 1;

[0057] Figure 2 3D spatial distribution diagram of the refractive index field in the hydrothermal vent area in Example 1;

[0058] Figure 3 Graph showing the gradient and step size changes of the adaptive ray tracing in Example 1;

[0059] Figure 4 A comparison diagram of the original bathymetric topography and the corrected topographic contour lines in Example 1;

[0060] Figure 5 is the distribution histogram of the original water depth error and the corrected error in Example 1;

[0061] Figure 6 This is a flow chart of an ocean surveying and mapping method based on multi-beam bathymetric correction according to the present invention;

[0062] Figure 7 This is a structural framework diagram of an ocean surveying and mapping system based on multi-beam bathymetric correction according to the present invention. DETAILED DESCRIPTION

[0063] The technical solutions of the present invention will be described clearly and completely below in conjunction with the drawings in the present invention. In addition, the forms of the various structures described in the following embodiments are merely examples. The present invention is not limited to the various structures described in the following embodiments. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0064] The present invention designs a marine surveying and mapping system based on multi-beam bathymetric correction, referring to Figure 7 , including the following structure:

[0065] The pressure-resistant towed body serves as the system's platform, equipped with a navigation sonar mounted on its head for real-time positioning and mapping, ensuring spatial accuracy in data collection. A cross-shaped titanium alloy bracket is affixed to the bottom of the towed body. Its four extending arms form a precise 45° angle with the towed body's central axis. This layout maximizes the coverage of the acoustic tomography array while enhancing structural stability to withstand the high-pressure deep-sea environment. Acoustically transparent sealed chambers are located at the ends of each arm to protect the internal sensors from seawater corrosion while allowing sound waves to penetrate unimpeded, providing an ideal operating environment for the acoustic tomography transceiver array.

[0066] A multi-source sensor array directly serves the data acquisition needs of Method S1. A multi-beam sonar array is installed at the center of the towed vessel's bottom. It transmits and receives high-frequency acoustic waves in the 200–400 kHz range to acquire seafloor echo signals and output raw bathymetric data. Its central location ensures a 150° acoustic coverage angle, enabling the capture of fine topographic features. However, it is susceptible to interference from refraction by hydrothermal turbulence.

[0067] The acoustic tomography transceiver array consists of eight piezoelectric ceramic elements, symmetrically arranged within an acoustically sealed chamber at the four ends of a cross-bracket. This layout maximizes element spacing. By alternately transmitting 10 kHz broadband pulses, the array collects acoustic wave propagation delay data across the hydrothermal region, providing information on the distribution of sound velocity anomalies for refractive index field inversion.

[0068] A laser Raman probe passes vertically through the center of the cross-bracket, with a built-in sapphire optical window directly contacting the core of the hydrothermal plume. Raman spectroscopy measures temperature in real time and converts it into refractive index. This vertical penetration ensures the probe is located at the intersection of acoustic tomography paths, improving the accuracy of acoustic-thermal data fusion.

[0069] The synchronization control unit ensures spatial and temporal alignment of multiple sensors to meet parallel acquisition requirements. A high-precision clock generates a reference timing signal based on a rubidium atomic clock. This clock connects the multibeam array main control board, the tomography array signal processor, and the laser Raman probe ADC acquisition module in a star topology via an IEEE 1588 fiber-optic synchronization bus. This topology achieves microsecond-level time synchronization, eliminating inter-sensor timing drift.

[0070] After receiving the clock signal, the trigger controller sends synchronous excitation instructions to the multi-beam array and acoustic tomography array to ensure that the timing of sound wave emission is strictly consistent and avoid cross interference.

[0071] The data processing unit implements the correction algorithm. The refractive index field inversion module is built-in with the NVIDIA Jetson AGX Xavier board, which receives the acoustic delay and Raman refractive index data, performs the inversion calculation of the fusion objective function, and outputs the dynamic refractive index field. Its GPU acceleration capability supports fast convergence within 50 iterations.

[0072] The ray tracing module, equipped with a Xilinx UltraScale FPGA chip, is directly connected to the refractive index field inversion module via a high-speed data bus, reading refractive index field data in real time. The FPGA parallelly calculates the gradient and second-order derivatives of the ray path points, executes the dynamic step size algorithm of S32 and the Runge-Kutta method of S33, and outputs a preliminary corrected water depth.

[0073] The depth fusion module, equipped with a TI TMS320C6678 DSP processor, receives preliminary depth data from the ray tracing module and executes the ray cluster fusion algorithm. The DSP calculates beam weights and performs weighted averaging in real time to suppress turbulent noise.

[0074] The coprocessor is vertically plugged into the top of the NVIDIA Jetson AGX Xavier board through the MXM interface, accelerating the conjugate gradient method solution and improving the real-time performance of the inversion module.

[0075] The NVIDIA Jetson AGX Xavier board is a high-performance edge computing module with the following CPUs:

[0076] 8-core ARM Cortex-A57 / Carmel v82, main frequency up to 226GHz, supports multi-threaded task scheduling.

[0077] The GPU is a 512-core Volta architecture GPU + 64 Tensor Cores, providing 32 TOPS of INT8 computing power and supporting parallel deep learning inference.

[0078] The storage is 32GB LPDDR4x memory + 32GB eMMC storage, which meets the caching requirements of multiple models.

[0079] Xilinx UltraScale FPGAs are programmable logic devices that enable low-latency, high-throughput hardware acceleration.

[0080] The logic density is 16nm FinFET+ process, and a single chip integrates up to 2.85 million logic units, supporting complex algorithm hardening.

[0081] The computing unit consists of 9,000+ DSP slices, providing 27x18-bit MAC computing power and a peak computing power of 12M ops / s.

[0082] The high-speed interface is an integrated 3275Gb / s transceiver that natively supports protocols such as PCIe Gen4 and 100G Ethernet.

[0083] TI TMS320C6678 DSP is a multi-core digital signal processor designed for high-precision real-time signal processing.

[0084] The multi-core design is an 8-core C66x DSP, supporting fixed / floating-point operations and a peak performance of 320 GMAC / 160 GFLOPs.

[0085] The storage system consists of 512KB L2 cache per core + shared 4MB L2, with a 64-bit DDR3 interface.

[0086] High-speed interconnection integrates HyperLink and SRIO 21 to achieve low-latency collaboration among multiple chips.

[0087] The data output system corresponds to terrain generation. An armored fiber optic cable interface runs through the sealed cabin atop the towed body and directly connects to the output of the data processing unit. The fused, corrected water depth is transmitted to the water surface via optical fiber in real time, with a transmission delay of less than 10 ms.

[0088] After receiving the optical cable data, the display and control terminal performs Kriging interpolation to generate a depth contour map and calculates the accuracy index to intuitively display the correction effect.

[0089] A navigation sonar is mounted on the nose of the pressure-resistant towed vehicle, continuously outputting the towed vehicle's three-dimensional position coordinates, providing a spatial reference for the entire surveying and mapping system. A cross-shaped titanium alloy bracket is rigidly attached to the towed vehicle's bottom. The bracket's four extension arms extend at a precise 45° angle to the towed vehicle's central axis, maximizing the acoustic sensor's physical baseline. Acoustically transparent capsules are sealed at each arm end. These hollow capsules are filled with a pressure-equalizing medium to prevent external seawater pressure from being transmitted to the internal sensors.

[0090] The multibeam sonar array is rigidly bolted to the center of the underside of the towed body, with its acoustic emitting surface flush with the lower surface of the towed body, ensuring that the sound waves cover a 150-degree fan-shaped area vertically downward. The eight piezoelectric ceramic elements of the acoustic tomography transceiver array are embedded in acoustically sealed chambers at the four ends of the cross-bracket. Each element is connected to the signal processor via waterproof cables passing through the bracket's internal channels. This layout allows for a spacing of 2.5 meters between adjacent elements, forming an acoustic interference network across the hydrothermal zone. The sapphire window of the laser Raman probe extends vertically through the center node of the cross-bracket, and its fiber optic sensor cable extends upward along the bracket's central axis to the data processing unit, enabling lossless transmission of hydrothermal core temperature data.

[0091] The high-precision clock's rubidium atomic clock core generates a master clock signal using the IEEE 1588 protocol. This signal is distributed via a fiber-optic synchronization bus in a star topology to three terminals: the first connects to the clock receiver on the multibeam sonar array's main control board; the second connects to the clock interface of the acoustic tomography transceiver array's signal processor; and the third connects to the sampling trigger of the laser Raman probe's ADC acquisition module. The trigger controller's command output port is hardwired to the multibeam array's transmit circuit and the tomography array's pulse generator, ensuring that both initiate acoustic wave transmission within 1 microsecond of receiving the synchronization excitation command.

[0092] The NVIDIA Jetson board in the refractive index field inversion module directly receives the time-delayed data stream from the acoustic tomography array and the refractive index vector from the laser Raman probe via a PCIe data channel. Its parallel computing core performs acoustic-thermal fusion modeling, and the output is pushed to the ray tracing module in real time via a 40Gbps fiber channel. The ray tracing module's Xilinx FPGA chip acquires the refractive index field data stream via a high-speed serial interface. Its internal logic unit synchronously calculates the gradient field and ray trajectories. The corrected water depth matrix is ​​written directly into the DSP buffer of the water depth fusion module via a DMA channel. The TI DSP processor in the water depth fusion module reads the raw water depth data by ray clusters. After completing spatiotemporal weighted fusion, it packages and transmits high-confidence water depth values ​​to the transmit buffer of the armored fiber optic cable interface.

[0093] Electromagnetic shielding ducts are embedded within the titanium alloy arms of the cross support to house all sensor cables and protect them from interference from deep-sea eddies. The three modules of the data processing unit are installed in a stacked format: the Jetson board for the refractive index field inversion module is located on the bottom layer, with the coprocessor plugged vertically into the MXM interface on top; the FPGA board for the ray tracing module is supported in the middle layer by copper pillars; and the DSP board for the water depth fusion module is fixed to the top layer via a backplane connection, forming a compact three-level processing stack. The metal-sealed connector of the armored fiber optic cable interface runs through the pressure chamber at the top of the towed body, and its internal optoelectronic conversion module connects to the output port of the data processing unit via a board-to-board connector.

[0094] The raw water depth data collected by the multi-beam sonar array undergoes refraction distortion correction in the ray tracking module. This process uses the gradient field model constructed by the refractive index field inversion module in real time, forming a closed loop from data acquisition to ray correction. The corrected water depth output by the water depth fusion module is uploaded to the display and control terminal via the armored optical cable interface. The terminal's built-in Kriging interpolation engine converts the discrete point cloud into a contour topographic map, achieving end-to-end output from sensor data to visualization products. The high-precision clock of the synchronization control unit uniformly sets the sampling time of all sensors, ensuring that the acoustic delay, hydrothermal temperature, and beam pointing angle are strictly aligned on the time axis.

[0095] Reference Figure 6 This paper designs an oceanographic mapping method based on multi-beam bathymetry correction, which specifically includes the following steps: S1: Multi-source data acquisition; S1 is the process of acquiring multimodal data of deep-sea hydrothermal areas through an integrated sensor system, aiming to provide comprehensive input for subsequent correction. The specific implementation includes three parallel data acquisition stages:

[0096] Multibeam bathymetric data acquisition: A multibeam sonar array transmits 256 beams at a frequency of 200–400 kHz, with a coverage angle of 150°. It receives seafloor echo signals and records raw depth measurements (draw). These beams are refracted and distorted by hydrothermal turbulence, causing the measured water depth to fluctuate between 1,832 and 2,145 meters. Multibeam sonar arrays are seafloor topography mapping devices that calculate water depth by measuring the round-trip time of sound waves. High-frequency sound waves can capture fine topographic features, but are susceptible to interference from medium inhomogeneities.

[0097] Acoustic tomography data acquisition: The eight piezoelectric ceramic array elements of the acoustic tomography transceiver array are symmetrically arranged at the four ends of the cross bracket, alternately emitting 10 kHz broadband pulses to measure the acoustic wave propagation delay data across the hydrothermal zone. The array records the changes in the propagation time of sound waves passing through turbulent areas and captures the abnormal distribution of sound speed. Acoustic tomography is a technology that inverts the characteristics of the medium through the propagation delay of sound waves. Its core is to establish a mapping relationship between the sound wave path and the medium parameters. Examples of measured data are as follows: ;in: Represents the measured sound wave propagation delay vector (unit: second); square brackets [] represent vectors.

[0098] To collect hydrothermal temperature data, a laser Raman probe is inserted vertically into the core of the hydrothermal plume. Laser light is emitted through a sapphire optical window, and the Raman scattering spectrum is analyzed, measuring the temperature change ΔT at a 10 Hz sampling rate. This technique is based on the principle of Raman scattering, whereby laser excitation of molecules in a medium produces a characteristic frequency-shifted optical signal whose intensity is positively correlated with temperature. For example, a peak temperature ΔT of 348°C is measured, and the refractive index is calculated using the physical conversion formula: ;in: represents the refractive index measured by laser Raman spectroscopy (dimensionless); 1.33 represents the refractive index of background seawater (dimensionless); represents the temperature-refractive index conversion coefficient (unit: °C−1); ΔT represents the temperature change (unit: °C).

[0099] The final output refractive index vector: ;in: Represents the refractive index vector (dimensionless); square brackets [] denote a vector.

[0100] S2: Construct a dynamic refractive index field model. This step is performed in the refractive index field inversion module. By fusing acoustic tomography data and laser Raman temperature data, a three-dimensional dynamic refractive index field model of the hydrothermal area is constructed. It specifically includes the following sub-steps:

[0101] S21: Water body grid division, divide the target water area into (unit: meter) three-dimensional grid, where This is a preset grid spacing used to discretize the computational space. The total number of grid nodes is determined by the target water area size. Meshing is based on the principle of spatial discretization, ensuring that the model resolution can capture the sudden gradient changes at the hydrothermal plume boundary.

[0102] S22: Fusion objective function construction, defining the optimization objective function in the refractive index field inversion module:

[0103] ;in:

[0104] is the objective function value (dimensionless) that needs to be minimized;

[0105] is the acoustic tomography projection matrix (dimensionless), which is constructed by the acoustic wave path and grid mapping relationship of the acoustic tomography transceiver array, and its dimension is ( is the number of grid nodes);

[0106] is the refractive index vector of the grid node (dimensionless), which is the variable to be solved;

[0107] is the measured acoustic delay vector (unit: seconds), acquired through acoustic tomography transceiver array acquisition;

[0108] and is the regularization coefficient (dimensionless), which is set according to the inversion stability requirements (e.g. , );

[0109] The Raman point selection matrix (dimensionless) is used to extract the grid nodes corresponding to the laser Raman probe;

[0110] is the Raman measured refractive index vector (dimensionless), and the temperature is measured by the laser Raman probe according to the formula Convert to get is the temperature change);

[0111] is the Laplace smoothing operator matrix, used to enforce spatial smoothness of the refractive index field;

[0112] is the norm square operator, which calculates the sum of the squares of the vector elements.

[0113] S23: Refractive index field solution, using the conjugate gradient method to iteratively solve the objective function and output the dynamic refractive index field The conjugate gradient method is an efficient optimization algorithm that only requires first-order derivative information and converges after 50 iterations on an NVIDIA Jetson board. The inversion results quantify the acoustic distortion effects of hydrothermal turbulence, such as the center of the vent. , background value .

[0114] S3: Adaptive ray tracing correction. This step is implemented in the ray tracing module. It accurately solves the ray trajectory by dynamically adjusting the integration step size and corrects the refraction distortion. It specifically includes the following sub-steps:

[0115] S31: Gradient and second-order derivative calculation, real-time calculation of the refractive index gradient at the sound path point and its second-order derivative , both of which are obtained by spatial differentiation of the refractive index field model output by S2. For example, near the hydrothermal vent, , , reflecting the medium heterogeneity and turbulence mutation characteristics.

[0116] S32: Dynamic step size adjustment, adaptively adjust the integration step size according to the gradient mutation intensity:

[0117] ;

[0118] in:

[0119] is the next integration step length (unit: meter);

[0120] The preset maximum step length (unit: meter) is set according to the system calculation performance;

[0121] The preset minimum step size (unit: meter) is determined based on the wavelength of the multibeam sonar array sound waves (e.g. 0.01 meter corresponds to a wavelength of 5 mm at 300 kHz);

[0122] is the normalization coefficient (unit: ), empirical parameters, used to adjust the step size;

[0123] is the refractive index gradient, calculated in real time at the sound ray path point;

[0124] is the second-order derivative of the refractive index gradient, calculated in real time at the sound ray path point;

[0125] is a weight coefficient (dimensionless), used to balance the influence of the gradient and the second-order derivative, optimized through numerical experiments (usually set to 1.0);

[0126] is the minimum value operator;

[0127] is the maximum value operator.

[0128] S33: Solve the sound ray trajectory, using the variable step size Runge-Kutta method to solve the sound ray trajectory differential equation:

[0129] ;

[0130] in:

[0131] is the sound ray deflection angle vector (unit: radian);

[0132] is the length of the sound path (unit: meter);

[0133] η is the refractive index (dimensionless), provided by the refractive index field model;

[0134] is the partial derivative of the refractive index in the direction normal to the sound ray (unit: ), calculated in real time at the waypoints;

[0135] is the derivative of the sound ray deflection angle with respect to the path length (unit: rad / m).

[0136] The Runge-Kutta method is a fourth-order numerical integration algorithm suitable for solving complex differential equations. In ocean acoustics, the trajectory of the sound line is affected by the change of the refractive index of the medium and its path is determined by the differential equation describe, is the sound ray deflection angle, is the path length, η is the refractive index, is the normal gradient. The present invention uses the variable step-size Runge-Kutta method to solve the equation, and the process is as follows:

[0137] Dynamic step size adjustment: according to the refractive index gradient and its second-order derivative , according to the formula:

[0138] ;

[0139] Adaptive adjustment of integration step size Based on the wavelength of sound waves, In areas with sudden gradient changes, the step size is automatically reduced to millimeters to avoid cumulative errors; in areas with gentle gradients, the step size is increased to 0.8 m to improve computational efficiency.

[0140] Fourth-order Runge-Kutta iteration: The state vector of the current sound ray position (position , deflection angle ) is the initial value, and four slope evaluations are performed at each step:

[0141] ;

[0142] in Provided by the real-time refractive index model. Local truncation error Ensure accuracy in strong gradient areas.

[0143] This method overcomes the failure of traditional fixed-step-size methods in hydrothermal turbulent regions by employing a gradient-driven variable-step-size strategy. In regions with abrupt gradient changes, the step-size is reduced to capture sudden changes in ray curvature and avoid ray path distortion. In regions with gentle gradients, the step-size is expanded to accelerate computation, maintaining a real-time performance of 2.2 seconds per frame. Ultimately, high-precision ray tracking is achieved, reducing the depth error in hydrothermal regions from an initial 142 meters to 14 meters.

[0144] This algorithm introduces adaptive numerical methods of fluid mechanics into ocean acoustics, accurately depicts the propagation of sound rays in inhomogeneous media, solves the problem of refraction distortion caused by hydrothermal turbulence, and significantly improves the accuracy of deep-sea topography reconstruction.

[0145] S4: Ray cluster fusion correction. This step is performed in the water depth fusion module and suppresses turbulence noise by weighted fusion of adjacent spatiotemporal beams. It specifically includes the following sub-steps:

[0146] S41: Sound ray cluster definition: A collection of spatiotemporally adjacent beams is defined as a sound ray cluster, such as a 3×3 beam array and beams within three consecutive transmission cycles. This utilizes the spatiotemporal correlation of sound waves propagating in turbulent media to improve data reliability.

[0147] S42: Weight calculation, calculate the intra-cluster beam weight based on the turbulence spatiotemporal correlation model:

[0148] ;

[0149] in:

[0150] For the beam weight (dimensionless);

[0151] Beam and The pointing angle difference (unit: radian) is calculated from the beam pointing angle data of the multi-beam sonar array;

[0152] Beam and The emission time difference (unit: seconds) is calculated by the timestamp data of the synchronization control unit;

[0153] is the turbulence spatial correlation scale parameter (unit: radian), which is set based on the turbulence statistical model in the hydrothermal region;

[0154] is the turbulence time-dependent scale parameter (unit: seconds), which is set based on the turbulence statistical model in the hydrothermal region;

[0155] is the natural exponentiation operator.

[0156] S43: Water depth fusion, weighted output fusion water depth:

[0157] ;

[0158] in:

[0159] For the Beam-corrected water depth (unit: meters);

[0160] For the The original water depth of the beam (unit: meter), output by the ray tracing module;

[0161] For all beams in the cluster sum operator;

[0162] S5: Terrain generation and verification. This step uses the data output system to visualize and verify the corrected water depth:

[0163] S51: Geographic coordinate mapping, converting the fused water depth data from the sensor coordinate system to the geographic coordinate system, and mapping the sonar measurement values ​​to three-dimensional geodetic coordinates through the coordinate transformation matrix.

[0164] S52: Generate isobath terrain. Use Kriging spatial interpolation algorithm to generate isobath terrain map in the display and control terminal.

[0165] ;

[0166] in:

[0167] Points to be interpolated Estimated water depth (in meters);

[0168] For the summation operator ( is the number of known points);

[0169] is the weight coefficient (dimensionless), which quantifies the spatial correlation through the semivariogram;

[0170] For known points The water depth observation value (unit: meter).

[0171] The Kriging algorithm preserves the characteristics of terrain mutations. Kriging is an optimal interpolation method based on spatial statistics. Its core is to quantify the spatial correlation of geographic data through the variogram function, using the spatial distribution structure and weights of known sampling points ( ), an unbiased estimate of the unknown position ( ), and also provides the interpolation standard deviation. Its mathematical form is:

[0172] ;

[0173] The weight Determined by solving the following system of equations:

[0174] ;

[0175] is the variation function, is the Lagrange multiplier; the display and control terminal receives the corrected water depth data transmitted through the armored optical cable interface , whose geographic coordinates have been mapped to geodetic coordinates through the coordinate transformation matrix.

[0176] The spatial interpolation process is as follows: modeling the variogram, calculating the semivariogram based on the spatial distribution of the corrected water depth points ( is the point spacing), fit the model. Weight solution and interpolation, for the target grid point , solve for the weight , weighted fusion of water depths of neighboring known points , generating a continuous water depth field.

[0177] Based on the interpolated grid data, smooth isobaths are generated to accurately depict the seabed topography. Through spatial correlation modeling, details such as steep slopes and rift valleys are accurately restored in the hydrothermal vent area; interpolated standard deviation layers are generated. Identify low-confidence areas such as sparse survey lines to guide re-survey decisions.

[0178] S53: Accuracy verification, quantitatively evaluating accuracy through the standard deviation formula:

[0179] ;

[0180] in:

[0181] is the standard deviation of the corrected water depth (unit: meter);

[0182] is the total number of beams;

[0183] For the Beam-corrected water depth (unit: meters);

[0184] For the The true water depth of the beam (unit: meter), obtained by measuring the seabed benchmark.

[0185] Verification results show that the maximum error in the hydrothermal area is reduced from 150 meters to 15 meters, and the error in the background area is stable at ±1.0 meter.

[0186] In Example 1, this example selected an active hydrothermal vent region in the eastern Pacific as the target area. This region hosts multiple high-temperature vents with core temperatures reaching 365°C, while the background seawater temperature remains stable at 2°C. Hydrothermal plumes form a turbulent influence zone with a radius of approximately 100 meters. The surveying and mapping system was deployed on a pressure-resistant towed platform (Model: DeepTow-7000) and operated at a depth of 2,500 meters. The towed platform sailed along the designated survey line at a speed of 1.5 knots (approximately 0.77 m / s), maintaining a height of 200 meters above the seafloor. All sensors were time-synchronized via an IEEE 1588 fiber-optic synchronization bus. The data processing unit utilized a VIDIA Jetson AGX Xavier board for S2, a Xilinx UltraScale FPGA for S3, and a TI TMS320C6678 DSP for S4, executing algorithms in real time.

[0187] S1: Multi-source data acquisition; a multibeam sonar array transmits 256 beams (covering 150°) at a frequency of 300 kHz to collect raw water depth data. Due to the interference of hydrothermal turbulence, the raw water depth values ​​of a typical beam (such as beam ID 128) fluctuate between 1900 and 2050 meters (example raw water depth sequence: The eight piezoelectric ceramic elements of the acoustic tomography transceiver array alternately transmit 10 kHz broadband pulses, and the measured sound wave propagation delay data is as follows:

[0188] Normal seawater reference time delay: 1.000 s. The laser Raman probe is inserted vertically into the core of the hydrothermal plume and the temperature change is measured at a sampling rate of 10 Hz. (peak 365°C), and calculate the refractive index through the physical conversion formula:

[0189] ; Output refractive index vector:

[0190] ;

[0191] S2: Build a dynamic refractive index field model. The refractive index field inversion module divides the target water area (150 m × 150 m × 100 m) into a 3D grid with 5-meter spacing (27,000 nodes). Build the fusion objective function:

[0192] ;

[0193] in, (8×27,000 sparse matrix), (Raman point selection matrix), (Laplacian smoothing operator) strictly matches the system description; regularization coefficient , The conjugate gradient method was used to solve the problem (50 iterations, convergence tolerance ), output refractive index field Inversion results: nozzle center , plume edge , background area .

[0194] S3: Adaptive ray tracing correction, the ray tracing module calculates the refractive index gradient of 128 path points in real time and the second-order derivative :Near vent area , Far field . Dynamically adjust the integration step size:

[0195] ;

[0196] Parameter settings: , , , The calculated step size is 0.010 m near the nozzle and 0.80 m in the far field. The variable step size Runge-Kutta method is used to solve the acoustic ray trajectory differential equation:

[0197] ; Corrected beam 128 water depth: (original value 1925 m), local error < 0.5%.

[0198] S4: Sound ray cluster fusion correction, the water depth fusion module defines the temporal and spatial adjacent sound ray clusters. Calculation weight:

[0199] ;

[0200] Turbulence parameters , Example weights for beam 128: beam 127 ( , ) , beam 128 , beam 129 ( , ) . Weighted fusion output:

[0201] ;

[0202] Beam 128 Final water depth: , the standard deviation dropped from 38.5 m to 3.8 m.

[0203] S5: Topography generation and verification: The data output system transmits the corrected water depth to the display and control terminal via the armored optical cable interface. After geographic coordinate mapping, Kriging interpolation is used to generate the contour topographic map:

[0204] ;

[0205] Accuracy verification formula:

[0206] ;

[0207] Based on the seabed benchmark measurement, the results show that the maximum error in the hydrothermal area is reduced from 142 m to 14 m, and the error in the background area is reduced from ±25 m to ±0.8 m. The technical indicators of the surveying accuracy after correction in this embodiment are the same as those before correction. Figure 5 and as shown in Table 1.

[0208] Figure 5 This is an error distribution histogram obtained according to Example 1, showing the distribution comparison of the original water depth error and the error after correction according to the present invention in a deep-sea hydrothermal area environment.

[0209] Figure 5 In the example, the vertical axis is "frequency", which indicates the number of samples that appear within a specific error interval. It has no specific unit and is a statistical count value, indicating the frequency of occurrence of different error values. Figure 5 It can be seen that the original error distribution range is wide and dispersed, indicating that the uncorrected water depth data has significant uncertainty and large deviations under the influence of hydrothermal turbulence, which reflects the problem of insufficient accuracy of traditional bathymetric methods in complex environments. The distribution of the corrected error is highly concentrated and narrow, indicating that the method of the present invention effectively compresses the error fluctuations and significantly improves the stability and reliability of the measurement results. Specifically, the distribution of the original error shows a wide range of fluctuations, while the corrected error is almost concentrated near zero error, which directly proves the effectiveness of the adaptive ray tracking and ray cluster fusion technology of the present invention, which can overcome the distortion problem caused by the sudden change of gradient in the hydrothermal area. The significant convergence of the error distribution further verifies the feasibility of the multi-source data fusion and dynamic correction algorithm of the present invention in a real deep-sea environment, greatly reduces the bathymetric error, and improves the mapping accuracy by 90%, providing high-precision and reliable terrain data support for deep-sea resource exploration.

[0210] Table 1 Comparison of the surveying and mapping accuracy technical indicators after correction and before correction in Example 1

[0211] Performance indicators Raw data Corrected data Improvement rate Bathymetric standard deviation 38.5 m 3.8 m 90.1% Maximum error of hydrothermal center 142 m 14 m 90.1% Error range of background area ±25 m ±0.8 m 96.8% Data processing delay — 2.2 s / frame —

[0212] As can be seen from Table 1, the method of the present invention significantly improves surveying and mapping accuracy in highly turbulent environments. Multi-source data fusion accurately quantifies sound velocity anomalies, adaptive ray tracing overcomes the cumulative error of fixed-step-size algorithms, and ray cluster fusion effectively suppresses turbulent noise. Specifically, the method of the present invention significantly improves surveying and mapping accuracy in highly turbulent environments while meeting the real-time requirements of deep-sea engineering for terrain reconstruction. The refractive index field inversion module accurately quantifies the distribution of sound velocity anomalies, adaptive ray tracing effectively overcomes the cumulative error of traditional fixed-step-size algorithms in gradient mutation areas, and ray cluster fusion technology significantly suppresses random fluctuations caused by turbulence. The resulting topographic map clearly reproduces the geological characteristics of the hydrothermal vent area, providing reliable technical support for deep-sea resource exploration.

[0213] Figure 1 This is a comparison chart of the original water depth and the corrected water depth generated based on the actual test data of Example 1, showing the significant optimization effect of the correction algorithm of the present invention on the measurement results of 256 beams from violent fluctuations to stable convergence in the deep-sea hydrothermal turbulence environment. Figure 1 It can be seen that the original water depth data shows violent fluctuations in the area affected by hydrothermal turbulence, with its values ​​randomly distributed between 1900 meters and 2050 meters. In particular, in the core area of ​​the hydrothermal vents with beam IDs 100 to 150, the data discreteness reaches its maximum value. After being processed by the correction algorithm of the present invention, the water depth data converges to around 1978 meters, forming a smooth and stable distribution curve, and its fluctuation amplitude is compressed to within the range of ±0.5 meters. The measurement point of beam ID 128 specifically shown in the figure shows that the original water depth value of 1925 meters is increased to 1978.2 meters after correction. This value deviates from the true water depth value of 1978 meters by only 0.2 meters. In the entire surveying area, the standard deviation of the original water depth data is as high as 38.5 meters, while the standard deviation of the corrected data is sharply reduced to 3.8 meters.

[0214] The multi-source data fusion technology proposed in the present invention successfully quantified the abnormal distribution of sound velocity caused by hydrothermal turbulence, and effectively solved the problem of sound wave refraction distortion. The adaptive ray tracking algorithm achieves millimeter-level accuracy in the gradient mutation area by dynamically adjusting the integral step size, significantly reducing the cumulative error of the traditional fixed step size method in the strong turbulence area. The spatiotemporal weighted fusion technology of ray clusters effectively suppresses random noise interference and compresses the maximum measurement error in the hydrothermal vent area from the order of 150 meters to the order of 15 meters. The entire set of processing methods, while maintaining a real-time processing speed of 2.2 seconds per frame, has increased the overall surveying and mapping accuracy by more than 90%, fully verifying the engineering applicability and technical advantages of the present invention in complex deep-sea environments.

[0215] Figure 2This is a 3D spatial distribution diagram of the refractive index field model obtained according to Example 1 at the z=50m horizontal slice, which intuitively shows the refractive index change in the hydrothermal vent area.

[0216] from Figure 2 As can be seen in the figure, the refractive index near the origin of the coordinates in the central region of the vent reaches its highest value of approximately 1.340, representing a distinct red hotspot. As distance increases, the refractive index gradually decreases to a background value of approximately 1.330, transitioning to blue. This indicates that the impact of hydrothermal turbulence on sound velocity is concentrated in the core region of the vent. This distribution validates the effectiveness of the multi-source data fusion of acoustic tomography and laser Raman in step S2 of the present invention. Because the smooth gradient of the refractive index field is directly derived from the coordinated inversion of acoustic time delay data and temperature data, this method demonstrates that it can accurately quantify the acoustic distortion effects of hydrothermal plumes.

[0217] Figure 2 The continuity and smoothness of the medium refractive index field without sudden changes or discontinuous areas confirm the stability of the inversion algorithm, which is due to the regularization constraints and conjugate gradient iterations, and reflects the feasibility of the invention, because in actual deep-sea environments, the model can reliably capture medium inhomogeneities. The significant increase in refractive index within a radius of about 50 meters within the hydrothermal influence range is consistent with the description in Example 1, which supports the effectiveness of the correction accuracy of the present invention. Through this field model, subsequent ray tracking correction can reduce the sounding error from the original 142 meters to 14 meters, greatly improving the reliability of terrain reconstruction. The overall distribution map proves the engineering feasibility of multi-sensor fusion in complex deep-sea environments. The dynamic refractive index field directly solves the failure problem of traditional single-source sounding in hydrothermal areas, ensuring that the surveying and mapping system can still output high-confidence results under extreme conditions.

[0218] Figure 3 It is the adaptive sound ray tracking trajectory and step size change diagram obtained according to Example 1, which intuitively shows the change law of the refractive index gradient, gradient second-order derivative and integral step size dynamically adjusted by the algorithm on the hydrothermal turbulent propagation path of the sound ray.

[0219] from Figure 3 It can be seen that in the initial section of the sound path (0-20 meters), the refractive index gradient ∇η and the second-order derivative of the gradient δ∇η simultaneously reach peak values ​​of 0.018 m⁻¹ and 0.15 m⁻², indicating that the algorithm successfully captures the strong turbulence mutation characteristics in the core area of ​​the hydrothermal vent, verifying the present invention's ability to perceive complex media.

[0220] The integration step size Δs is automatically compressed to a minimum value of 0.01 m in the gradient mutation area of ​​0-20 m, and expanded to more than 0.8 m in the gradient gentle area >50 m, proving that the algorithm can dynamically allocate computing resources according to the medium heterogeneity, solving the problem of cumulative error of the traditional fixed step size method in strong turbulence areas.

[0221] The second-order derivative curve of the gradient shows a sharper attenuation inflection point 20 meters away from the nozzle boundary, indicating that as an indicator of the gradient change rate, it can provide early warning of changes in medium properties, provide forward-looking signals for step size adjustment, and enhance the response sensitivity of the algorithm.

[0222] In the first 0-10 meters of the path with the steepest gradient, the step length is reduced to 0.01 meters, which is the order of the sound wave wavelength, to ensure that the accuracy of solving the differential equation of the sound line deflection angle is controlled within 0.5%.

[0223] All three curves show smooth transition characteristics without mutations or oscillations. Combined with the real-time processing speed of 2.2 seconds per frame in Example 1, it is proved that the variable step size algorithm has strong robustness on FPGA hardware and meets the actual measurement requirements of deep-sea mapping equipment.

[0224] Figure 3 The adaptive ray tracing in step S3 of the present invention was verified, and the failure problem of the traditional fixed step-size method in the hydrothermal gradient mutation zone was overcome through the dynamic step-size mechanism; combined with the dual-parameter control of gradient and second-order derivative, the millimeter-level precision calculation of the ray trajectory was achieved.

[0225] Figure 4 This is a comparison diagram of the original bathymetric topography and the corrected topography obtained according to Example 1. Figure 4 As can be seen in the original bathymetric topography, the central region of the hydrothermal vents exhibits severe distortion, with chaotic distribution of bathymetric contours that fail to accurately reflect the true seafloor topography. This demonstrates the complete failure of traditional surveying methods in the turbulent hydrothermal environment. The corrected bathymetric contours exhibit a natural, continuous gradient, successfully recreating a bowl-shaped topographic structure that extends evenly outward from the vents, validating the effectiveness of the multi-source data fusion technology proposed in this paper. The corrected image clearly shows topographic details in the vent region, without any breaks or mutations in the original data, demonstrating that the adaptive ray tracing algorithm accurately compensates for the effects of acoustic refraction.

[0226] The spacing between depth contours remains uniform across the entire survey area, and the background terrain features are reliably discernible, demonstrating that the ray cluster fusion technique effectively suppresses random interference noise caused by turbulence. The two maps use the same depth scale, but the topographic features differ significantly. This method reduces the maximum depth measurement error in the hydrothermal area from 142 meters to 14 meters, fully meeting the accuracy requirements of deep-sea engineering surveying.

[0227] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for ocean surveying and mapping based on multi-beam bathymetric correction, characterized in that: The following steps are involved: S1, uses a multi-beam sonar array to acquire bathymetric echo signals from deep-sea hydrothermal areas, while simultaneously collecting acoustic wave propagation delay data across the hydrothermal area using an acoustic tomography transceiver array, and measuring the hydrothermal plume temperature using a laser Raman probe; S2, based on the data collected by S1, fuses the acoustic tomography data and laser Raman data in the refractive index field inversion module to construct a dynamic refractive index field model; S3, in the sound ray tracking module, an adaptive variable step size algorithm is used to correct the sound ray refraction distortion and generate a preliminary water depth; S4, through the water depth fusion module, weights are assigned and data are fused for adjacent beams in time and space, and high-confidence water depth is output; S5, transmit the high-confidence water depth to the surface display and control terminal via the data output interface to generate a depth contour map; In S2, the specific steps of constructing the dynamic refractive index field include: S21, divide the target water area into two parts with side length 3D grid; S22, construct the fusion objective function in the refractive index field inversion module: ； in is the objective function value; is the acoustic tomography projection matrix, which is constructed by the acoustic wave path and grid mapping relationship of the acoustic tomography transceiver array; is the refractive index vector of the grid node, the variable to be solved; is the measured acoustic delay vector, acquired through acoustic tomography transceiver array acquisition; 、 is the regularization coefficient, which is set according to the inversion stability requirements; The Raman point selection matrix is ​​used to extract the grid nodes corresponding to the laser Raman probe; is the Raman measured refractive index vector, which is obtained by measuring the temperature with a laser Raman probe and converting it; is the Laplace smoothing operator matrix, used to enforce spatial smoothness of the refractive index field; is the norm square operator, which calculates the sum of the squares of the vector elements; S23, iteratively solve the objective function through the conjugate gradient method and output the dynamic refractive index field .

2. The method according to claim 1, characterized in that The specific steps of S3 include: S31, real-time calculation of the refractive index gradient at the sound ray path point in the sound ray tracing module and the second-order derivative ; S32, dynamically adjust the integration step size according to the gradient mutation intensity: ； in is the next integration step length; It is a preset maximum step size, set according to the system computing performance constraints; To preset the minimum step size, it is determined based on the wavelength of the sound waves of the multi-beam sonar array; is the normalization coefficient, an empirical parameter used to adjust the step size; is the magnitude of the refractive index gradient, which is calculated in real time at the acoustic ray path point and obtained by performing the first-order spatial differentiation of the refractive index field model output by S2; The magnitude of the second-order derivative of the refractive index gradient is calculated in real time at the acoustic ray path point and is obtained by performing the second-order spatial differentiation of the refractive index field model output by S2; is the weight coefficient, which is used to balance the influence of gradient and second-order derivative and is optimized through numerical experiments; is the minimum value operator; is the maximum value operator; S33, using the variable step size Runge-Kutta method to solve the acoustic ray trajectory differential equation: ； in is the sound ray deflection angle vector; is the length of the sound path; η is the refractive index, which is provided by the refractive index field model output by S2; is the partial derivative of the refractive index in the normal direction of the sound line, calculated in real time at the path point; is the derivative of the sound ray deflection angle with respect to the path length.

3. The method according to claim 1, characterized in that The specific steps of data fusion in S4 include: S41, in the water depth fusion module, the spatiotemporal adjacent beams are defined as acoustic ray clusters; S42, calculate the intra-cluster beam weights: ； in For the beam weights; Beam and The pointing angle difference is calculated by the beam pointing angle data of the multi-beam sonar array; Beam and The emission time difference is calculated by the timestamp data of the synchronization control unit; is the spatially related scale parameter of turbulence, which is set based on the statistical model of turbulence in the hydrothermal region; is the turbulence time-dependent scale parameter, which is set based on the turbulence statistical model in the hydrothermal region; is the natural exponential operator; S43, weighted output fusion water depth: ； in For the Beam-corrected water depth; For the The original water depth of the beam is output by the ray tracing module; For all beams in the cluster Summation operator.

4. The method according to claim 1, wherein The specific steps of S5 are as follows: S51: Geographic coordinate mapping, converting the fused water depth data from the sensor coordinate system to the geographic coordinate system, and mapping the sonar measurement values ​​to three-dimensional geodetic coordinates through the coordinate transformation matrix; S52: Generate isobath terrain. Use Kriging spatial interpolation algorithm to generate isobath terrain map in the display and control terminal. ； in: Points to be interpolated Estimated water depth; is the summation operator; is the weight coefficient, which quantifies the spatial correlation through the semivariogram; For known points Observed water depth; S53: Accuracy verification, quantitatively evaluating accuracy through the standard deviation formula: ； in: To correct for the standard deviation of water depth; is the total number of beams; For the Beam-corrected water depth; For the The true water depth of the beam is obtained by measuring the seabed benchmark.

5. An oceanographic mapping system based on multi-beam bathymetric correction, used to implement the method according to any one of claims 1 to 4, characterized in that: include: The pressure-resistant towed body has a navigation sonar installed on its head and a cross-shaped titanium alloy bracket fixed at its bottom. The four extension arms of the bracket form a 45° angle with the central axis of the towed body. Each arm end of the bracket is equipped with an acoustically transparent sealed cabin to protect the internal sensors. The multi-source sensor array includes a multi-beam sonar array, an acoustic tomography transceiver array, and a laser Raman probe. The multi-beam sonar array is installed at the center of the towed body's bottom, transmitting and receiving 200-400kHz sound waves and outputting raw sounding data. The acoustic tomography transceiver array contains eight piezoelectric ceramic array elements, symmetrically arranged in acoustically sealed chambers at the four ends of the cross-bracket, and is used to collect acoustic wave propagation delay data across the hydrothermal zone. The laser Raman probe, which passes vertically through the center of the cross-bracket and has a built-in sapphire optical window, is used to measure the hydrothermal plume temperature and convert it into refractive index. The synchronization control unit includes a high-precision clock and trigger controller. The high-precision clock is based on a rubidium atomic clock and is connected to the main control board of the multi-beam array, the signal processor of the tomographic array, and the ADC acquisition module of the laser Raman probe in a star topology through the IEEE1588 optical fiber synchronization bus. The trigger controller sends synchronous excitation instructions to the multi-beam array and the acoustic tomography array to ensure consistent data acquisition timing; The data processing unit includes a refractive index field inversion module, an acoustic ray tracing module, and a water depth fusion module. The refractive index field inversion module has a built-in board for fusing acoustic time delay and Raman temperature data to construct a dynamic refractive index field. The acoustic ray tracing module has a built-in FPGA chip, which is interconnected with the refractive index field inversion module to correct acoustic ray refraction distortion. The water depth fusion module has a built-in DSP processor that receives the output data of the acoustic ray tracing module and performs spatiotemporal beam weighted fusion. The data output system includes an armored optical cable interface and a display and control terminal. The armored optical cable interface runs through the sealed cabin on the top of the towed body and is directly connected to the data processing unit for transmitting the corrected water depth data to the water surface; the display and control terminal receives data from the armored optical cable interface through optical fiber and generates a contour topographic map in real time.

6. The system according to claim 5, characterized in that: The four extension arms of the cross-shaped titanium alloy bracket form an angle of 45 degrees with the central axis of the tow body, and the piezoelectric ceramic array elements of the acoustic tomography transceiver array are fixed in the acoustically transparent sealed cabin arranged at the end of each arm.

7. The system according to claim 5, characterized in that: The data processing unit is additionally provided with a coprocessor which is vertically plugged into the top of the board through an MXM interface and is used to accelerate the refractive index field inversion calculation.

8. The system according to claim 5, characterized in that: The IEEE 1588 fiber optic synchronization bus connects the main control board of the multi-beam sonar array, the signal processor of the acoustic tomography transceiver array, and the ADC acquisition module of the laser Raman probe in a star topology to achieve microsecond-level time synchronization.

Images