Sonar image geometric correction method based on image optimization

Abstract

The invention discloses a sonar image geometric correction method based on image optimization, and relates to the technical field of underwater sonar imaging, and the method comprises the steps: collecting a multi-source underwater observation data packet including sonar image original data, platform motion state data and underwater environment data, and carrying out the preprocessing; extracting sonar image original data from the multi-source underwater observation data packet, performing geometric coarse correction to obtain a pre-corrected sonar image, and performing motion compensation on the obtained pre-corrected sonar image to generate a motion corrected sonar image; constructing a joint optimization objective function, carrying out geometric fine tuning on the motion correction sonar image by adopting a sliding time window strategy, and outputting a geometric fine correction sonar image; geometric distortion caused by non-uniform motion of a platform, attitude disturbance and acoustic propagation distortion is effectively eliminated, inter-frame consistency and submarine topography continuity of sonar images are remarkably improved, and the method is suitable for high-precision operation scenes such as underwater surveying and mapping, target recognition and three-dimensional reconstruction.

Description

Sonar Image Geometric Correction Method Based on Image Optimization Technical Field

[0001] This invention relates to the field of underwater sonar imaging technology, specifically a sonar image geometric correction method based on image optimization. Background Technology

[0002] Sonar imaging technology, as a core tool for underwater detection and mapping, is widely used in marine resource exploration, underwater pipeline inspection, and underwater target identification. The quality of sonar images directly affects the efficiency and accuracy of underwater operations, while the geometric accuracy of the images is the foundation for subsequent quantitative analysis, 3D reconstruction, and multi-voyage data fusion. However, due to the characteristics of underwater acoustic propagation and the complexity of platform motion, raw sonar images generally suffer from severe geometric distortion, necessitating effective correction methods. Firstly, existing technologies heavily rely on high-precision external sensors to provide platform position, attitude, and altitude information, lacking adaptive correction capabilities in scenarios where sensors are missing, malfunctioning, or cost-constrained, making it difficult to meet the practical needs of miniaturized, low-cost underwater platforms. Firstly, traditional methods rely excessively on high-precision inertial navigation sensors when compensating for platform motion distortion. When sensor data is affected by noise or is missing, the correction accuracy drops sharply, and there is a lack of ability to fuse and utilize multi-source motion information. Secondly, existing image enhancement and denoising techniques often sacrifice geometric edge sharpness in the process of suppressing inherent speckle noise in sonar images, making it difficult to achieve the dual goals of noise suppression and geometric feature preservation. How to construct a sonar image geometric correction method that does not rely on high-precision external sensors, has environmental adaptive correction capabilities, and can accurately preserve geometric features while suppressing noise is the problem we need to solve. To this end, we now present a sonar image geometric correction method based on image optimization. Summary of the Invention

[0003] The purpose of this invention is to provide a sonar image geometric correction method based on image optimization.

[0004] The objective of this invention can be achieved through the following technical solution: a sonar image geometric correction method based on image optimization, comprising the following steps: Step S1: acquiring multi-source underwater observation data packets, wherein the multi-source underwater observation data packets include raw sonar image data, platform motion state data, and underwater environment data, and preprocessing the obtained multi-source underwater observation data packets; Step S2: performing geometric coarse correction on the raw sonar image data to obtain the corresponding pre-corrected sonar image, and performing motion compensation on the obtained pre-corrected sonar image to obtain a motion-corrected sonar image; Step S3: constructing a joint optimization objective function, performing geometric fine-tuning on the motion-corrected sonar image to obtain a geometrically fine-corrected sonar image.

[0005] Furthermore, the process of acquiring multi-source underwater observation data packets includes: deploying several data acquisition units on the underwater operating platform, each data acquisition unit including a side-scan sonar transducer array deployed at the bottom of the underwater operating platform, an inertial navigation unit deployed at the geometric center inside the underwater operating platform, and a temperature, salinity, and depth (TSD) meter deployed outside the platform; acquiring raw sonar image data through the side-scan sonar transducer array, acquiring platform motion state data through the inertial navigation unit, and acquiring underwater environmental data through the TSD meter; the raw sonar image data is a continuous stream generated by the side-scan sonar. The data consists of multiple frames of raw sonar images and imaging parameters corresponding to each pixel; the imaging parameters corresponding to each pixel include echo arrival time, initial beam transmission angle, range sampling interval, and sonar pulse transmission time; the platform motion status data includes platform position, platform instantaneous velocity, roll angle, pitch angle, and heading angle synchronized with each frame of image; the underwater environment data includes the current water temperature profile, salinity profile, and pressure data; the raw sonar image data, platform motion status data, and underwater environment data are uniformly packaged to obtain a multi-source underwater observation data package.

[0006] Furthermore, the preprocessing of the obtained multi-source underwater observation data packets includes: extracting the timestamp of each frame of the original sonar image; using the timestamp of each frame of the original sonar image as a reference, performing linear interpolation calculations on the platform motion state data and underwater environment data; based on the linearly interpolated underwater environment data, using the empirical formula for sound speed, obtaining the sound speed profile that varies with depth; and then using the cubic spline interpolation method to fit the sound speed profile into a sound speed function.

[0007] Furthermore, the process of extracting raw sonar image data from the multi-source underwater observation data package for geometric coarse correction includes: for any pixel in the raw sonar image, extracting its corresponding echo arrival time and initial beam emission angle; discretizing the sound velocity profile along the depth direction into several thin layers, emitting sound rays from the side-scan sonar transducer array at the initial beam emission angle, and performing path integration; within each propagation step, determining the grazing angle of the sound ray based on the sound velocity function at the current depth and Snell's law; obtaining the true horizontal coordinates and bottom depth corresponding to any pixel according to the iterative solution strategy, performing the iterative solution strategy on all pixels in the raw sonar image to obtain the complete geographic coordinate mapping relationship; resampling the original pixel grayscale values ​​to the horizontal distance domain according to their corrected true horizontal coordinates to generate the geometrically coarsely corrected pre-corrected sonar image and the corresponding initial depth map.

[0008] Furthermore, the motion compensation process for the pre-corrected sonar image includes: obtaining the true track displacement corresponding to each row of pixels in the pre-corrected sonar image based on the platform's instantaneous velocity at the time of each sonar pulse transmission; determining the projection coverage area of ​​adjacent frames of the pre-corrected sonar image on the underwater plane based on the platform's motion state data and sonar geometric parameters, thus obtaining an effective overlapping area; calculating the pixel-level optical flow field within the effective overlapping area to obtain an optical flow observation displacement sequence; constructing a one-dimensional Kalman filter for optimal fusion using the true track displacement and optical flow observation displacement as complementary information, defining the filter state variable as the optimal track displacement; outputting the optimal track displacement sequence after optimal fusion via Kalman filtering; resampling the pre-corrected sonar image row by row along the track direction using the output optimal track displacement sequence as the sampling reference; and mapping each row of pixels in the pre-corrected sonar image to new coordinates to obtain the motion-corrected sonar image after motion compensation.

[0009] Furthermore, the process of constructing a joint optimization objective function to perform geometric fine-tuning on motion-corrected sonar images and obtain geometrically fine-tuned sonar images includes: setting a sliding time window and introducing two-dimensional fine-tuning parameters for each frame of motion-corrected sonar image within the sliding time window; constructing a joint optimization objective function based on the optimal track displacement sequence, geographic coordinate mapping relationship, effective overlapping area, and initial depth map; determining the initial geometric transformation of each frame of motion-corrected sonar image based on the optimal track displacement sequence and geographic coordinate mapping relationship, and obtaining the initial geometric fine-tuning parameters for each frame of motion-corrected sonar image; constructing a data fidelity term based on the deviation constraint between the initial geometric fine-tuning parameters and the two-dimensional fine-tuning parameters; and based on the effective overlapping area and motion-corrected sonar image... The process involves: constructing an inter-frame optical flow consistency term; obtaining the water depth value corresponding to each frame of motion-corrected sonar image based on the initial depth map; constructing an inter-frame optical flow consistency term based on the effective overlap area and the water depth value corresponding to each frame of motion-corrected sonar image; constructing a joint optimization objective function based on the obtained data fidelity term and inter-frame optical flow consistency term; obtaining the optimal fine-tuning parameters corresponding to each frame of motion-corrected sonar image within the current sliding time window based on the minimum value output by the joint optimization objective function; shifting the sliding time window forward frame by frame along the time axis; using the optimal fine-tuning parameters obtained from each frame of motion-corrected sonar image within the sliding time window as a benchmark; performing geometric fine-tuning on the motion-corrected sonar image to obtain a geometrically fine-tuned sonar image and the corresponding final depth map.

[0010] Compared with existing technologies, the beneficial effects of this invention are: It collects multi-source underwater observation data packets containing raw sonar image data, platform motion state data, and underwater environment data, and performs preprocessing. By fusing multi-source heterogeneous observation data, it provides a complete and synchronous input foundation for subsequent high-precision geometric correction, avoiding systematic errors caused by missing or asynchronous data. It extracts raw sonar image data from the multi-source underwater observation data packets, performs coarse geometric correction to obtain pre-corrected sonar images, and performs motion compensation on the obtained pre-corrected sonar images to generate motion-corrected sonar images, effectively eliminating track stretching or compression caused by non-uniform platform motion. Distortion is significantly improved, enhancing the fidelity of the internal geometric structure of a single-frame image and providing stable initial conditions for multi-frame joint optimization. A joint optimization objective function is constructed, and a sliding time window strategy is used to perform geometric fine-tuning on motion-corrected sonar images, outputting geometrically fine-corrected sonar images to collaboratively optimize local geometric deviations in multi-frame images. This significantly improves pixel-level alignment accuracy and seabed topography smoothness in overlapping areas of adjacent frames. This invention effectively eliminates geometric distortion caused by non-uniform platform motion, attitude disturbances, and acoustic propagation distortion, significantly improving the inter-frame consistency of sonar images and the continuity of seabed topography. It is suitable for high-precision operation scenarios such as underwater surveying, target recognition, and 3D reconstruction. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0012] Figure 1 is a flowchart of the method of the present invention. Detailed Implementation

[0013] As shown in Figure 1, the sonar image geometric correction method based on image optimization includes the following steps: Step S1: Acquire multi-source underwater observation data packets, which include raw sonar image data, platform motion state data, and underwater environment data, and preprocess the obtained multi-source underwater observation data packets; Step S2: Perform geometric coarse correction on the raw sonar image data to obtain the corresponding pre-corrected sonar image, and perform motion compensation on the obtained pre-corrected sonar image to obtain a motion-corrected sonar image; Step S3: Construct a joint optimization objective function, perform geometric fine-tuning on the motion-corrected sonar image, and obtain a geometrically fine-corrected sonar image.

[0014] It should be further explained that, in the specific implementation process, the acquisition of multi-source underwater observation data packets includes: several data acquisition units are deployed on the underwater operation platform, including a side-scan sonar transducer array deployed at the bottom of the underwater operation platform, an inertial navigation unit deployed at the geometric center inside the underwater operation platform, and a temperature, salinity, and depth (TSD) meter deployed outside the platform; raw sonar image data is acquired through the side-scan sonar transducer array, platform motion state data is acquired through the inertial navigation unit, and underwater environmental data is acquired through the TSD meter; the side-scan sonar transducer array has preset sonar geometric parameters, including beam pointing angle, beamwidth, and maximum slant range imaging. The range is used to form a continuous acoustic strip covering both sides of the seabed; the raw sonar image data consists of multiple consecutive raw sonar images formed by side-scan sonar and imaging parameters corresponding to each pixel; the imaging parameters corresponding to each pixel include echo arrival time, initial beam emission angle, range sampling interval, and sonar pulse emission time; the platform motion status data includes the platform position, instantaneous platform velocity, roll angle, pitch angle, and heading angle synchronized with each frame of image; the underwater environment data includes the current water area's temperature profile, salinity profile, and pressure data; the raw sonar image data, platform motion status data, and underwater environment data are uniformly packaged to obtain a multi-source underwater observation data package.

[0015] It should be further explained that, in the specific implementation process, the preprocessing of the obtained multi-source underwater observation data packets includes: labeling each frame of the original sonar image as i, where i = 1, 2, ..., n; extracting the timestamp of each frame of the original sonar image, denoted as... Using the timestamp of each original sonar image frame as a reference, linear interpolation is performed on the platform motion state data and underwater environment data; for example, taking the pitch angle as an example, let... and To surround The two adjacent sampling timestamps correspond to the following pitch angles: and Then the pitch angle after linear interpolation is: The roll angle, heading angle, and underwater environment data were all processed using the same linear interpolation strategy. Based on the linearly interpolated underwater environment data, an empirical formula for sound speed was used to obtain a sound speed profile that varies with depth. Then, a cubic spline interpolation method was used to fit the sound speed profile to a sound speed function. It should be noted that the empirical formula for sound velocity and the cubic spline interpolation method are common techniques used by those skilled in the art, and will not be elaborated upon here.

[0016] It should be further explained that, in the specific implementation process, the process of obtaining the corresponding pre-corrected sonar image by performing geometric coarse correction on the original sonar image data includes: for any pixel in the original sonar image, extracting its corresponding echo arrival time and initial beam emission angle; discretizing the sound velocity profile along the depth direction into several thin layers, with the sound velocity in each layer considered constant; emitting sound rays from the side-scan sonar transducer array at the initial beam emission angle; and performing path integration using a numerical ray tracing method; at each propagation step... Inside, based on the sound speed function at the current depth And Snell's law determines the glancing angle of sound rays. ,Right now: , , ;in, This corresponds to the horizontal displacement. For the corresponding depth increment, This represents the propagation time increment. Since the true depth of the underwater surface is unknown, an iterative solution strategy is employed. Specifically, assuming the underwater surface is flat, the initial depth is estimated based on a constant speed of sound model and the assumption of rectilinear propagation, denoted as... ,Right now: ;in, The preset average sound speed is used; then, sound ray tracing is performed using the initial depth as the bottom boundary to obtain the updated endpoint depth. and horizontal distance , and then As the new underwater boundary, the ray tracing process is repeated to generate the next iteration depth. , And so on; the iteration terminates when the absolute value of the difference between the depths obtained from two adjacent iterations satisfies the preset convergence condition, that is: ;in, The final destination depth, This is the final, true horizontal coordinate of the pixel. The preset convergence threshold is used; based on the above iterative solution strategy, the true horizontal coordinates corresponding to any pixel are obtained. and underwater depth An iterative solution strategy is applied to all pixels in the original sonar image to obtain the complete geographic coordinate mapping relationship; the original pixel grayscale values ​​are resampled to the horizontal distance domain according to their corrected true horizontal coordinates to generate a geometrically coarsely corrected pre-corrected sonar image and the corresponding initial water depth map.

[0017] It should be further explained that, in the specific implementation process, motion compensation is performed on the obtained pre-corrected sonar image. The process of obtaining the motion-corrected sonar image includes: labeling each row of pixels in the pre-corrected sonar image as j, where j=1,2,...,J; and according to the emission time of each sonar pulse... Corresponding platform instantaneous speed Integrate the platform's instantaneous velocity over time to calculate the true track displacement corresponding to the j-th row pixel of the pre-corrected sonar image, denoted as . ,Right now: ;in, These are the transmission times of adjacent sonar pulses. The instantaneous velocity of the platform is denoted as _j_. The actual track displacement refers to the cumulative geographic displacement of the underwater working platform in the track direction corresponding to the pixel in the j-th row of the pre-corrected sonar image, after considering the actual velocity changes of the underwater working platform. This displacement characterizes the actual spatial position of the pixel in the j-th row on the underwater plane. Based on the platform motion state data and sonar geometric parameters, the projection coverage area of ​​the pre-corrected sonar images of adjacent frames on the underwater plane is determined to obtain the effective overlapping area, denoted as _j_. Within the effective overlapping region, the pixel-level optical flow field is calculated to obtain the optical flow observation displacement sequence, i.e.: Using the actual trajectory displacement and the optical flow observation displacement as complementary information, a one-dimensional Kalman filter is constructed for optimal fusion. The filtered state variable is defined as the optimal trajectory displacement, and the state transition equation and observation equation are as follows: ; ;in, For process noise, To eliminate observation noise, the optimal track displacement sequence is output after optimal fusion using Kalman filtering. The optimal trajectory displacement sequence is output. Using the sampling reference, the pre-corrected sonar image Perform line-by-line resampling along the flight path; map the j-th row pixels of the pre-corrected sonar image to the new coordinates. At this location, obtain motion-corrected sonar images after motion compensation. .

[0018] It should be further explained that, in the specific implementation process, the process of constructing a joint optimization objective function and performing geometric fine-tuning on motion-corrected sonar images to obtain geometrically fine-corrected sonar images includes: setting a sliding time window, wherein the sliding time window contains N consecutive frames of motion-corrected sonar images. Where N is an integer greater than or equal to 3; each row of pixels in the motion-corrected sonar image is labeled, denoted as m, where m = 1, 2, ..., N; for the m-th frame of the motion-corrected sonar image within the sliding time window, a two-dimensional fine-tuning parameter is introduced, denoted as... ,Right now: ;in, This represents the local offset in the range direction. The local offset of the trajectory direction is represented; the two-dimensional fine-tuning parameters are used to correct the motion-corrected sonar image; a joint optimization objective function is constructed based on the optimal trajectory displacement sequence, geographic coordinate mapping relationship, effective overlapping area, and initial depth map; specifically: based on the optimal trajectory displacement sequence... Based on the geographic coordinate mapping relationship, the initial geometric transformation of each frame of motion-corrected sonar image is determined, and the initial geometric fine-tuning parameters of each frame of motion-corrected sonar image are obtained, denoted as . Based on the deviation constraints between the initial geometric fine-tuning parameters and the two-dimensional fine-tuning parameters, a data fidelity term is constructed, denoted as... ,Right now: Based on the effective overlap region and motion-corrected sonar images, an inter-frame optical flow consistency term is constructed, denoted as... ,Right now: Based on the initial depth map, the depth value corresponding to each frame of motion-corrected sonar image is obtained and denoted as . Based on the effective overlap region and the water depth value corresponding to each frame of motion-corrected sonar image, an inter-frame optical flow consistency term is constructed, denoted as... ,Right now: Based on the obtained data fidelity term and inter-frame optical flow consistency term, a joint optimization objective function is constructed, denoted as . ,Right now: ;in, , , For empirical weighting coefficients, satisfying , , All are greater than 0, and The relationship is as follows: Based on the minimum value of the joint optimization objective function output, the optimal fine-tuning parameters corresponding to each frame of motion-corrected sonar image within the current sliding time window are obtained; The sliding time window is moved forward frame by frame along the time axis, and the optimal fine-tuning parameters obtained from each frame of motion-corrected sonar image within the sliding time window are used as a benchmark to perform geometric fine-tuning on the motion-corrected sonar image, thereby obtaining the geometrically fine-corrected sonar image and the corresponding final depth map.

[0019] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications or equivalent substitutions made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A sonar image geometric correction method based on image optimization, characterized in that, The process includes the following steps: Step S1: Acquire multi-source underwater observation data packets, which include raw sonar image data, platform motion state data, and underwater environment data, and preprocess the acquired multi-source underwater observation data packets; Step S2: Perform an iterative solution strategy on all pixels in the raw sonar image data to obtain the corresponding geographic coordinate mapping relationship, and then generate the pre-corrected sonar image and the corresponding initial depth map obtained by geometric coarse correction. Based on the platform motion state data and sonar geometric parameters, obtain the effective overlap area of ​​the pre-corrected sonar image on the underwater plane and the corresponding optimal trajectory displacement sequence. Based on the obtained optimal trajectory displacement sequence, perform motion compensation on the pre-corrected sonar image to obtain the motion-corrected sonar image; Step S3: Based on the optimal trajectory displacement sequence and geographic coordinate mapping relationship, determine the initial geometric transformation of each frame of the motion-corrected sonar image to obtain the initial geometric fine-tuning parameters of each frame of the motion-corrected sonar image; A data fidelity term is constructed based on the initial geometric fine-tuning parameters; an inter-frame optical flow consistency term is constructed based on the effective overlapping area and motion-corrected sonar images. Based on the obtained data fidelity term and inter-frame optical flow consistency term, a joint optimization objective function is constructed to perform geometric fine-tuning on motion-corrected sonar images, thereby obtaining geometrically fine-corrected sonar images.

2. The sonar image geometric correction method based on image optimization according to claim 1, characterized in that, The process of acquiring multi-source underwater observation data packets includes: deploying several data acquisition units on an underwater operating platform, each including a side-scan sonar transducer array deployed at the bottom of the platform, an inertial navigation unit deployed at the geometric center of the platform, and a temperature, salinity, and depth (TDM) meter deployed outside the platform; acquiring raw sonar image data through the side-scan sonar transducer array, acquiring platform motion state data through the inertial navigation unit, and acquiring underwater environmental data through the TDM meter; the raw sonar image data consists of multiple consecutive frames of raw sonar images formed by the side-scan sonar and imaging parameters corresponding to each pixel; the imaging parameters corresponding to each pixel include echo arrival time, initial beam emission angle, range sampling interval, and sonar pulse emission time; the platform motion state data includes platform position, instantaneous platform velocity, roll angle, pitch angle, and heading angle synchronized with each frame of image; the underwater environmental data includes the current water temperature profile, salinity profile, and pressure data; and unifying and encapsulating the raw sonar image data, platform motion state data, and underwater environmental data to obtain a multi-source underwater observation data packet.

3. The sonar image geometric correction method based on image optimization according to claim 2, characterized in that, The preprocessing of the acquired multi-source underwater observation data package includes: extracting the timestamp of each frame of the original sonar image; using the timestamp of each frame of the original sonar image as a reference, performing linear interpolation calculations on the platform motion state data and underwater environment data; based on the linearly interpolated underwater environment data, using the empirical formula for sound speed, obtaining the sound speed profile that varies with depth; and then using the cubic spline interpolation method to fit the sound speed profile into a sound speed function.

4. The sonar image geometric correction method based on image optimization according to claim 3, characterized in that, The process of extracting raw sonar image data from multi-source underwater observation data packets and performing geometric coarse correction includes: for any pixel in the raw sonar image, extracting its corresponding echo arrival time and initial beam emission angle; discretizing the sound velocity profile along the depth direction into several sound velocity layers, emitting sound rays from the side-scan sonar transducer array at the initial beam emission angle, and performing path integration; within each propagation step, determining the grazing angle of the sound ray based on the sound velocity function at the current depth and Snell's law; obtaining the true horizontal coordinates and bottom depth corresponding to any pixel according to the iterative solution strategy, performing the iterative solution strategy on all pixels in the raw sonar image to obtain the complete geographic coordinate mapping relationship; resampling the original pixel grayscale values ​​to the horizontal distance domain according to their corrected true horizontal coordinates to generate a geometrically coarsely corrected pre-corrected sonar image and the corresponding initial depth map.

5. The sonar image geometric correction method based on image optimization according to claim 4, characterized in that, The process of motion compensation for pre-corrected sonar images includes: obtaining the true track displacement corresponding to each row of pixels in the pre-corrected sonar image based on the platform's instantaneous velocity at the time of each sonar pulse transmission; determining the projection coverage area of ​​adjacent frames of pre-corrected sonar images on the underwater plane based on platform motion state data and sonar geometric parameters to obtain an effective overlapping area; calculating the pixel-level optical flow field within the effective overlapping area to obtain an optical flow observation displacement sequence; constructing a one-dimensional Kalman filter for optimal fusion using the true track displacement and optical flow observation displacement as complementary information, defining the filter state variable as the optimal track displacement; outputting the optimal track displacement sequence after optimal fusion via Kalman filtering; resampling the pre-corrected sonar image row by row along the track direction using the output optimal track displacement sequence as the sampling reference; and mapping each row of pixels in the pre-corrected sonar image to new coordinates to obtain the motion-corrected sonar image after motion compensation.

6. The sonar image geometric correction method based on image optimization according to claim 5, characterized in that, The process of constructing a joint optimization objective function to perform geometric fine-tuning on motion-corrected sonar images and obtain geometrically fine-tuned sonar images includes: setting a sliding time window; introducing two-dimensional fine-tuning parameters for each frame of motion-corrected sonar image within the sliding time window; the data fidelity term is obtained based on the deviation constraint between the initial geometric fine-tuning parameters and the two-dimensional fine-tuning parameters; the inter-frame optical flow consistency term is obtained by obtaining the water depth value corresponding to each frame of motion-corrected sonar image based on the initial water depth map; obtaining the optimal fine-tuning parameters corresponding to each frame of motion-corrected sonar image within the current sliding time window based on the minimum value output by the joint optimization objective function; shifting the sliding time window forward frame by frame along the time axis; and using the optimal fine-tuning parameters obtained from each frame of motion-corrected sonar image within the sliding time window as a benchmark to perform geometric fine-tuning on the motion-corrected sonar image to obtain the geometrically fine-tuned sonar image and the corresponding final water depth map.

Images