Sample amplification method for gross error in multi-beam sounding data

Abstract

The invention relates to a sample amplification method for gross errors in multi-beam sounding data, which comprises the following steps of: processing original multi-beam sounding data to obtain first data which is kept consistent in numerical value; amplifying four types of multi-beam sounding gross errors for the first data; and setting a threshold value according to a hydrographic measurement standard, comparing the first data after amplification of the four types of multi-beam sounding gross errors, and introducing points with depth offset greater than the threshold value in the first data into a deep learning model to complete sample amplification of gross errors. The sample amplification method for the gross error in the multi-beam sounding data can solve the problems that in the prior art, a gross error detection method depends on manual intervention or parameter adjustment, repeatability is poor, complex environment adaptability is insufficient and the like. The problems that an existing sample generation method is difficult to efficiently simulate the diversity of gross errors in multi-beam sounding data, so that training data of a deep learning model is insufficient, and the accuracy of underwater pile surveying and mapping is affected are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine surveying and mapping engineering, and particularly relates to a method for sample amplification of gross errors in multi-beam sounding data. Background Art

[0002] With its wide coverage ability in the fields of marine surveying and mapping and underwater detection and its ability to finely depict the details of the seabed topography, multi-beam sounding technology has become an indispensable tool, especially playing a core role in the position determination and condition monitoring of underwater piles such as bridge piers and wind power piles. However, the data collected by this technology is often affected by various factors such as equipment noise, environmental interference, and measurement errors, resulting in errors and affecting the accuracy of underwater pile mapping.

[0003] In the prior art, gross error detection methods rely on manual intervention or parameter adjustment, and there are problems such as poor repeatability and insufficient adaptability to complex environments. Existing sample generation methods are difficult to efficiently simulate the diversity of gross errors in multi-beam sounding data, resulting in insufficient training data for deep learning models and affecting the accuracy of underwater pile mapping.

[0004] Therefore, there are defects in the prior art and it needs to be improved and developed. Summary of the Invention

[0005] An embodiment of the invention provides a method for sample amplification of gross errors in multi-beam sounding data, which is used to solve the problems in the prior art that gross error detection methods rely on manual intervention or parameter adjustment, and there are problems such as poor repeatability and insufficient adaptability to complex environments. Furthermore, it solves the problem that existing sample generation methods are difficult to efficiently simulate the diversity of gross errors in multi-beam sounding data, resulting in insufficient training data for deep learning models and affecting the accuracy of underwater pile mapping.

[0006] An embodiment of the present invention provides a method for sample amplification of gross errors in multi-beam sounding data, including the following steps:

[0007] S1. Process the original multi-beam sounding data to obtain first data that is numerically consistent;

[0008] S2. Amplify four types of multi-beam sounding gross errors for the first data, and the multi-beam sounding gross errors include isolated point gross errors, structured gross error clusters far from the seabed, bad pings, and arc loops caused by sidelobe effects;

[0009] S3. Set a threshold according to hydrographic measurement standards, compare the first data after amplifying the four types of multi-beam sounding gross errors, and introduce the points in the first data with a depth offset greater than the threshold into the deep learning model to complete the sample amplification of gross errors.

[0010] Further, the processing of the original multi-beam sounding data to obtain the first data that is numerically consistent includes:

[0011] Calibrate the original multi-beam sounding data, and the calibration method includes at least one of declination detection calibration, sound velocity profile calibration, and tidal residual calibration;

[0012] By visually inspecting the point cloud data, identify and remove abnormal points from the calibrated original multi-beam sounding data. The abnormal points are points in the original multi-beam sounding data points where the value has a gross error compared with the original multi-beam sounding data points of adjacent points, isolated points, and point cloud groups that do not match the terrain during sounding. The forms of gross error include protrusion of the value in the original multi-beam sounding data points compared with the original multi-beam sounding data points of adjacent points and depression of the value in the original multi-beam sounding data points compared with the original multi-beam sounding data points of adjacent points.

[0013] Further, amplifying four types of multi-beam sounding gross errors for the first data. The multi-beam sounding gross errors include isolated point gross error, structured gross error group far from the seabed, bad ping, and arc loops caused by sidelobe effects, including:

[0014] According to the Gaussian probability density distribution model, generate a depth deviation z that follows a normal distribution N(u, σ 2 ) :

[0015]

[0016] where u is the average value of the depth change affected by factors;

[0017] σ is the degree of dispersion of the depth deviation z;

[0018] By adding the depth deviation z to the original multi-beam sounding depth value in the first data, obtain the isolated point gross error to complete the amplification of the isolated point gross error.

[0019] Further, the factors affecting u include at least one of sensor failure and cavitation effect.

[0020] Further, amplifying four types of multi-beam sounding gross errors for the first data. The multi-beam sounding gross errors include isolated point gross error, structured gross error group far from the seabed, bad ping, and arc loops caused by sidelobe effects, including:

[0021] Select M points in the first data as reference sounding points, where the value range of M is [3, 10];

[0022] According to the spatial coordinates of the reference sounding points, simulate the planar position of an object far from the seabed that appears in multi-beam sounding;

[0023] Obtain a selection point according to the planar position;

[0024] Add a depth deviation to the selection point to simulate the depth of the object;

[0025] Increase the tilt angle and adjust the attitude of the object so that the selection point can rotate vertically around the object, where the value range of the tilt angle is (0°, 90°);

[0026] According to the planar position of the object and the simulated depth of the object, convert the selection point into three-dimensional space coordinates through the constant sound speed ray tracing method to obtain structured gross error points far from the seabed, so as to complete the amplification of the structured gross error group far from the seabed.

[0027] Further, the obtaining a selection point according to the planar position includes: selecting a point with an azimuth angle between [0°, 360°], a length in meters between [1, 10], and a width in meters between [0.5, 5] as the selection point.

[0028] Further, the adding a depth deviation to the selection point to simulate the depth of the object includes: adding a depth deviation of N meters to the selection point, where the value range of N is [-15, 15].

[0029] Further, the amplifying four types of multibeam sounding gross errors for the first data, the multibeam sounding gross errors including isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arc loops caused by sidelobe effects, includes:

[0030] Arbitrarily select a segment in a ping to obtain the beam in this segment of the ping, where a ping refers to a single acoustic wave pulse emitted by a sonar device, and a beam refers to a beam formed by the acoustic wave emitted by the sonar device;

[0031] According to the Gaussian probability density distribution model, generate a depth deviation z that follows a normal distribution N(u, σ 2 );

[0032] According to the scale factor f on the plane and the depth deviation z, adjust the initial Generate the gross error point Z p，i ' of the bad ping to complete the amplification of the bad ping. The formula for generating the bad ping is as follows:

[0033]

[0034] Where, is the position of sonar device N at beam number i;

[0035] is the position of sonar device N at beam number s;

[0036] N p，i N' is the position of the sonar device N at beam number i after being adjusted based on the scale factor f on the plane;

[0037] is the position of the sonar device E at beam number i;

[0038] is the position of the sonar device E at beam number s;

[0039] E p，i E' is the position of the sonar device E at beam number i after being adjusted based on the scale factor f on the plane;

[0040] i, d, s, t, j are beam numbers, where s is the leftmost beam number in the selected ping, and t is the rightmost beam number in the selected ping;

[0041] j is the beam number corresponding to the position of the sounding point in multi-beam sounding.

[0042] Further, for amplifying the four types of multi-beam sounding gross errors in the first data, the multi-beam sounding gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arc loops caused by sidelobe effects, including:

[0043] According to the Gaussian probability density distribution model, generate a depth deviation z that follows a normal distribution N(u, σ 2 );

[0044] Randomly select a ping, and obtain the beams in this segment of the ping, where a ping refers to a single acoustic wave pulse emitted by the sonar device, and a beam refers to the beam formed by the acoustic wave emitted by the sonar device;

[0045] Add the original depth of the middle beam in this ping to the depth deviation z to obtain the radius r of the arc loop;

[0046] According to the radius r of the arc loop, obtain the gross error points caused by sidelobe effects to complete the amplification of sidelobe effect gross errors, and obtain the gross error points Z p，i ″ The formula is as follows:

[0047]

[0048] where θ is the incident angle of the beam;

[0049] h is the course angle of the beam;

[0050] N transducer is the plane coordinate of the sonar device N;

[0051] Np，i ″ is the position of sonar device N after being adjusted based on the arc at the position with beam number i;

[0052] E transducer is the planar coordinate of sonar device E;

[0053] E p，i ″ is the position of sonar device E after being adjusted based on the arc at the position with beam number i.

[0054] Further, the hydrographic survey standard is one of the special grade standard and the second-class standard.

[0055] Beneficial effects:

[0056] As can be seen from the above technical solutions, the present invention provides a method for sample amplification of gross errors in multi-beam sounding data. Through the simulation of various types of gross errors, the amount of gross error sample data in multi-beam sounding data is enhanced. This method can effectively generate different types of gross error data samples, including isolated point gross errors, structured gross error clusters far from the seabed, bad pings, and arc gross errors caused by sidelobe effects. By amplifying the samples, the training data of the deep learning model is enriched, thereby improving the gross error recognition ability of the model in practical applications. Therefore, the present invention has the following beneficial effects:

[0057] (1) Generate diverse training data: By simulating four types of gross errors, namely isolated point gross errors, structured gross error clusters, bad pings, and arcs caused by sidelobe effects, covering typical error types in multi-beam sounding data, reducing the cost of manual annotation, and improving the generalization ability of the deep learning model.

[0058] (2) Statistical modeling based on actual error characteristics: Design a Gaussian probability density distribution model in combination with actual factors such as sensor failures and cavitation effects to enhance the authenticity and applicability of the simulated data.

[0059] (3) Standardized threshold setting: Set the depth offset threshold according to the hydrographic survey standard of the International Hydrographic Organization to ensure the objectivity and consistency of gross error marking.

[0060] (4) Automated preprocessing process: Improve the initial quality of the original data through declination detection and correction, sound velocity profile correction, and tidal residual correction, providing a reliable basis for gross error amplification.

[0061] It should be understood that all combinations of the foregoing concepts and additional concepts described in more detail below can be regarded as part of the inventive subject matter of the present disclosure as long as such concepts do not conflict with each other.

[0062] The foregoing and other aspects, embodiments, and features of the teachings of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of exemplary embodiments, will be apparent from the following description or will be learned from the practice of specific embodiments in accordance with the teachings of the present invention. Description of the Drawings

[0063] The drawings are not drawn to scale with respect to actual reference objects. In the drawings, each identical or nearly identical component shown in each figure may be denoted by the same reference numeral. For the sake of clarity, not every component is labeled in each figure. Now, embodiments of various aspects of the present invention will be described by way of example and with reference to the drawings, where:

[0064] Figure 1 It is a flowchart of a method for sample amplification of gross errors in multi-beam sounding data in an embodiment of the present application. Detailed Embodiments

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings understood by those of ordinary skill in the art in the field to which the present invention pertains.

[0066] The terms "first", "second", and similar terms used in the specification and claims of this patent application for the present invention do not denote any order, quantity, or importance, but are merely used to distinguish different components. Similarly, unless the context clearly indicates otherwise, singular forms such as "a", "an", or "the" and similar terms do not denote a limitation of quantity, but rather indicate the presence of at least one. The terms "including" or "comprising" and similar terms mean that the elements or items appearing before "including" or "comprising" cover the features, wholes, steps, operations, elements, and / or components listed after "including" or "comprising", and do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations. The terms "up", "down", "left", "right", etc. are only used to indicate relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0067] In the prior art, due to the fact that the gross error detection method relies on manual intervention or parameter adjustment, there are problems such as poor repeatability and insufficient adaptability to complex environments. The existing sample generation methods are difficult to efficiently simulate the diversity of gross errors in multi-beam sounding data, resulting in insufficient training data for deep learning models and affecting the accuracy of underwater pile body mapping.

[0068] In view of this, referring to Figure 1 , an embodiment of the present invention provides a sample amplification method for gross errors in multi-beam sounding data, including the following steps:

[0069] Step S1: Process the original multi-beam sounding data to obtain first data that is numerically consistent. Before performing sample amplification, the primary step is to comprehensively preprocess the original multi-beam sounding data. This step covers key steps such as declination detection, sound velocity profile correction, and tidal residual correction to ensure the accuracy of the data. Subsequently, by performing data consistency checks, check whether the data points are numerically consistent with the surrounding data points. If the value of a certain data point is significantly higher or lower compared to the neighboring points, it may be regarded as a gross error. By visually inspecting the point cloud data, identify obvious abnormal points, such as isolated points or point cloud groups that do not match the surrounding terrain, and identify and remove the gross errors existing in the data to provide a clean data basis for subsequent operations.

[0070] Step S2: Amplify four types of multi-beam sounding gross errors for the first data. The multi-beam sounding gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arc loops caused by sidelobe effects.

[0071] Step S3: Set a threshold according to hydrographic measurement standards, compare the first data after amplifying the four types of multi-beam sounding gross errors, and introduce the points in the first data with a depth offset greater than the threshold into the deep learning model to complete the sample amplification of gross errors.

[0072] In some embodiments, processing the original multi-beam sounding data to obtain first data that is numerically consistent includes:

[0073] Correct the original multi-beam sounding data, and the correction method includes at least one of declination detection correction, sound velocity profile correction, and tidal residual correction;

[0074] Visually inspect the point cloud data, for example, identify and remove abnormal points from the corrected original multibeam sounding data using a three-dimensional point cloud visualization tool. Abnormal points are those in the original multibeam sounding data points where there are gross errors in the values compared to adjacent original multibeam sounding data points, isolated points, and point cloud groups that do not match the terrain during sounding. The forms of gross errors include protrusions in the values of the original multibeam sounding data points compared to adjacent original multibeam sounding data points and depressions in the values of the original multibeam sounding data points compared to adjacent original multibeam sounding data points.

[0075] In some embodiments, four types of multibeam sounding gross errors are amplified for the first data. The multibeam sounding gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, including:

[0076] According to the Gaussian probability density distribution model, generate a depth deviation z that follows a normal distribution N(u, σ 2 )

[0077]

[0078] where u is the average value of the depth change affected by factors;

[0079] σ is the degree of dispersion of the depth deviation z; usually σ is the standard deviation;

[0080] Obtain the isolated point gross error by adding the depth deviation z to the original multibeam sounding depth value in the first data to complete the amplification of the isolated point gross error.

[0081] In some embodiments, the factors affecting u include at least one of sensor failure and cavitation effect. The factors affecting u also include induction element failure; operation conflicts of other devices in the same frequency band; signal multipath reflection; sidelobe interference phenomenon; low signal quality under extreme climate conditions; seabed identification error; interference from other objects in the water area, such as fish swimming, algae floating, bubble plumes, etc.

[0082] In some embodiments, four types of multibeam sounding gross errors are amplified for the first data. The multibeam sounding gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, including:

[0083] Select M points in the first data as reference sounding points, where the value range of M is [3, 10];

[0084] According to the spatial coordinates of the reference sounding points, simulate the planar position of an object far from the seabed that appears in multibeam sounding;

[0085] The following methods are usually used for the simulation of planar positions: Least squares method: For reference points that do not exactly conform to the expected positions, the least squares method can be used for fitting, and the planar position of the object is determined by minimizing the error; Triangulation: If the reference points can form a triangle, triangulation techniques can be used to locate the planar position of the object based on the calculation of angles and distances; Interpolation method: Based on the selected reference points, interpolation algorithms such as Kriging interpolation and inverse distance weighting can be used to infer the position of the object.

[0086] Based on the planar position, selection points are obtained;

[0087] A depth deviation is added to the selection points to simulate the depth of the object;

[0088] An inclination angle is increased to adjust the attitude of the object so that the selection points can rotate vertically around the object, where the value range of the inclination angle is (0°, 90°);

[0089] The purpose of increasing the inclination angle is to change the angle of the object, thus facilitating the simulation of gross error points in different forms. For example, assume that the reference point of the simulated object is the position of a certain point on the seabed, and the inclination angle is the rotation angle of the object around that point. When the inclination angle is set to a specific angle, the gross error points of the object will rotate vertically around the reference point of the object. If the object is originally horizontal, by setting a certain inclination angle, the object is rotated into an inclined state, thus simulating the deviation or gross error that may occur in underwater measurement.

[0090] Based on the planar position of the object and the simulated depth of the object, the selection points are converted into three-dimensional space coordinates through the constant sound speed ray tracing method to obtain structured gross error points far from the seabed, so as to complete the amplification of the structured gross error group far from the seabed.

[0091] By completing the amplification of the structured gross error group far from the seabed, random depth changes are added to the simulated gross error points, which can make the gross error group far from the seabed more realistic and can reflect the typical errors in multi-beam sounding data.

[0092] Large objects in the water often cause structured gross errors in multi-beam sounding data, and these gross errors are formed by the reflection of multiple continuous and aggregated beams. The appearance, position, size, shape and attitude of large objects are all random, and the closer they are to the transducer, the more beam reflection points there are. Therefore, according to the statistical principle, such gross errors can be simulated, and the multi-path effect needs to be considered. In actual underwater measurement, sound waves may be reflected due to the multi-path effect, which will cause some gross error points to appear at different depths on the seabed. To increase the authenticity of the simulation, the multi-path effect is simulated through the above steps, and random depth changes are added to the simulated gross error points. This can make the gross error group far from the seabed more realistic and can reflect the typical errors in multi-beam sounding data.

[0093] In some embodiments, obtaining a selection point according to the planar position includes: selecting a point with an azimuth angle between [0°, 360°], a length in meters between [1, 10], and a width in meters between [0.5, 5] as the selection point.

[0094] In some embodiments, adding a depth deviation to the selection point to simulate the depth of an object includes: adding a depth deviation of N meters to the selection point, where the value range of N is [-15, 15].

[0095] In some embodiments, amplifying four types of multibeam sounding gross errors for the first data, where the multibeam sounding gross errors include isolated point gross errors, structured gross error clusters far from the seabed, bad pings, and arc loops caused by sidelobe effects, includes:

[0096] Arbitrarily select a segment in a ping, and obtain the beam in this segment of the ping, where a ping refers to a single acoustic wave pulse emitted by a sonar device, and a beam refers to a beam formed by the acoustic wave emitted by the sonar device;

[0097] According to the Gaussian probability density distribution model, generate a depth deviation z that follows a normal distribution N(u, σ 2 )

[0098] According to the scale factor f on the plane and the depth deviation z, adjust the initial Generate the gross error point Z p，i ' of the bad ping to complete the amplification of the bad ping. The formula for generating the bad ping is as follows:

[0099]

[0100] Where is the position of sonar device N at beam number i;

[0101] is the position of sonar device N at beam number s;

[0102] N p，i ' is the position of sonar device N at beam number i after being adjusted based on the scale factor f on the plane;

[0103] is the position of sonar device E at beam number i;

[0104] is the position of sonar device E at beam number s;

[0105] E p，i ' is the position of sonar device E at beam number i after being adjusted based on the scale factor f on the plane;

[0106] In some embodiments, the transducer and the sonar device are closely related. The transducer is one of the key components of the sonar device. Therefore, N and E can also represent the location of the transducer.

[0107] i, d, s, t, j are beam numbers, where s is the leftmost beam number in the selected ping, and t is the rightmost beam number in the selected ping;

[0108] j is the beam number corresponding to the location of the sounding point in multi-beam sounding.

[0109] In some embodiments, four types of multi-beam sounding gross errors are amplified for the first data. The multi-beam sounding gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arc loops caused by sidelobe effects, including:

[0110] According to the Gaussian probability density distribution model, a depth deviation z that follows a normal distribution N(u, σ 2 ) is generated;

[0111] Arbitrarily select a ping and obtain the beams in this segment of the ping. Here, a ping refers to a single acoustic wave pulse emitted by the sonar device, and a beam refers to the acoustic wave beam formed by the sonar device.

[0112] Add the original depth of the middle beam in the ping to the depth deviation z to obtain the radius r of the arc loop.

[0113] According to the radius r of the arc loop, obtain the gross error points caused by the sidelobe effect to complete the amplification of the sidelobe effect gross error and obtain the gross error points Z p，i ″ The formula is as follows:

[0114]

[0115] where θ is the incident angle of the beam;

[0116] h is the course angle of the beam;

[0117] N transducer is the planar coordinate of sonar device N;

[0118] N p，i ″ is the position of sonar device N adjusted based on the arc loop at the position of beam number i;

[0119] E transducer is the planar coordinate of sonar device E;

[0120] E p，i ″ is the position of sonar device E adjusted based on the arc loop at the position of beam number i.

[0121] In some embodiments, the hydrographic survey standard is one of the special grade standard and the second grade standard. In order to enable the deep learning model to better learn the difference between gross errors and normal sounding points during training, a threshold is set during the comprehensive sample amplification. If the depth offset is greater than the threshold, the point is marked as a gross error; otherwise, no change is made. The setting of the threshold can refer to the hydrographic survey standards of the International Hydrographic Organization, and select the special grade standard or the second grade standard according to actual needs.

[0122] In summary, a method for amplifying samples of gross errors in multi-beam sounding data provided by the present invention utilizes the main sources of multi-beam sounding errors and proposes a specific method for sample amplification to simulate gross errors in multi-beam sounding data and provide training data for the deep learning model. This method can effectively generate different types of gross error data samples, including isolated point gross errors, structured gross error clusters far from the seabed, bad pings, and arc-shaped gross errors caused by sidelobe effects. By amplifying the samples, the training data of the deep learning model is enriched, thereby improving the gross error recognition ability of the model in practical applications. Therefore, a method for amplifying samples of gross errors in multi-beam sounding data provided by the present invention has the following beneficial effects: generating diverse training data: by simulating four types of gross errors, namely isolated point gross errors, structured gross error clusters, bad pings, and arcs caused by sidelobe effects, covering typical error types in multi-beam sounding data, reducing the cost of manual annotation, and enhancing the generalization ability of the deep learning model. Statistical modeling based on actual error characteristics: designing a Gaussian probability density distribution model in combination with actual factors such as sensor failures and cavitation effects to enhance the authenticity and applicability of the simulated data. Standardized threshold setting: setting the depth offset threshold according to the hydrographic survey standards of the International Hydrographic Organization to ensure the objectivity and consistency of gross error marking. Automated preprocessing process: improving the initial quality of the original data through declination detection and correction, sound velocity profile correction, and tidal residual correction, providing a reliable basis for gross error amplification.

[0123] Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Those with ordinary knowledge in the technical field to which the present invention pertains can make various modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to what is defined by the claims.

Claims

1. A sample augmentation method for gross errors in multi-beam bathymetric data, characterized in that: The following steps are involved: S1. Processing the original multi-beam bathymetric data to obtain first data that is numerically consistent; S2, amplifying four types of multi-beam bathymetric gross errors for the first data, wherein the multi-beam bathymetric gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects; S3. Set a threshold according to the hydrological measurement standard, compare and amplify the first data after the four types of multi-beam bathymetric gross errors, and introduce the points in the first data whose depth offset is greater than the threshold into the deep learning model to complete the sample amplification of the gross errors.

2. A sample augmentation method for gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The processing of the original multi-beam bathymetric data to obtain first data that is numerically consistent includes: Correcting the original multi-beam bathymetric data, wherein the correction method includes at least one of angle detection correction, sound velocity profile correction, and tidal residual correction; By visually inspecting the point cloud data, abnormal points are identified and removed from the corrected original multibeam bathymetry data. The abnormal points are points where there are gross differences between the values ​​of the original multibeam bathymetry data points and those of the neighboring points, isolated points, and point cloud groups that do not match the terrain during bathymetry. The forms of gross differences include protrusions between the values ​​of the original multibeam bathymetry data points and those of the neighboring points, and depressions between the values ​​of the original multibeam bathymetry data points and those of the neighboring points.

3. A sample augmentation method for gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The amplification of the first data into four types of multi-beam bathymetric gross errors, wherein the multi-beam bathymetric gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, includes: According to the Gaussian probability density distribution model, generate a normal distribution N(u, σ 2 )’s depth deviation z: Where u is the average value of the depth variation affected by the factor; σ is the discreteness of the depth deviation z; The isolated point gross error is obtained by adding the depth deviation z to the original depth value of the multi-beam sounding in the first data, so as to complete the amplification of the isolated point gross error.

4. A sample augmentation method for gross errors in multi-beam bathymetric data according to claim 3, characterized in that: u is affected by at least one of the factors of sensor failure and cavitation effect.

5. The method for sample augmentation of gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The amplification of the first data into four types of multi-beam bathymetric gross errors, wherein the multi-beam bathymetric gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, includes: Selecting M points in the first data as reference sounding points, where the value range of M is [3, 10]; Simulating the plane position of an object far from the seabed appearing in the multi-beam bathymetry according to the spatial coordinates of the reference bathymetry point; According to the plane position, a selection point is obtained; Adding a depth deviation to the selected point to simulate the depth of the object; Adding a tilt angle to adjust the posture of the object so that the selected point can rotate vertically around the object, wherein the tilt angle has a value range of (0°, 90°); According to the plane position of the object and the depth of the simulated object, the selected points are converted into three-dimensional space coordinates through the constant sound velocity ray tracking method to obtain the structured gross error points far away from the seabed, so as to complete the expansion of the structured gross error group far away from the seabed.

6. A sample augmentation method for gross errors in multi-beam bathymetric data according to claim 5, characterized in that: The obtaining of the selected point according to the plane position includes: selecting a point whose azimuth angle is between [0°, 360°], whose length in meters is between [1, 10], and whose width in meters is between [0.5, 5] as the selected point.

7. The method for sample augmentation of gross errors in multi-beam bathymetric data according to claim 5, characterized in that: The adding of the depth deviation to the selected point to simulate the depth of the object includes: adding a depth deviation of N meters to the selected point, wherein the value range of N is [-15, 15].

8. The method for sample augmentation of gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The amplification of the first data into four types of multi-beam bathymetric gross errors, wherein the multi-beam bathymetric gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, includes: Randomly select a segment of ping and obtain the beam in the segment of ping, where ping refers to a single sound wave pulse emitted by the sonar device, and beam refers to the beam formed by the sound waves emitted by the sonar device; According to the Gaussian probability density distribution model, generate a normal distribution N(u, σ 2 )’s depth deviation z; According to the scale factor f and depth deviation z on the plane, adjust the initial The worst point Z that generates bad ping p，i ', to complete the amplification of bad ping, the formula for generating bad ping is as follows: i={d|d∈N * ,d∈[s,t]} in, It is the position of sonar device N at beam number i; It is the position of sonar device N at beam number s; N p，i′ It is the position of beam number i of sonar device N adjusted based on the scale factor f on the plane; It is the position of sonar device E at beam number i; It is the position of sonar device E at beam number s; E p，i ′ is the position of the sonar device E at beam number i after adjustment based on the scale factor f on the plane; i, d, s, t, j are beam numbers, where s is the leftmost beam number in the selected ping, and t is the rightmost beam number in the selected ping; j is the beam number corresponding to the location of the multi-beam sounding point.

9. The method for sample augmentation of gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The amplification of the first data into four types of multi-beam bathymetric gross errors, wherein the multi-beam bathymetric gross errors include isolated point gross errors, structured gross error groups far from the seabed, bad pings, and arcs caused by sidelobe effects, includes: According to the Gaussian probability density distribution model, a normal distribution N(u, σ 2 )’s depth deviation z; Choose any ping and get the beam in that ping, where ping refers to a single sound wave pulse emitted by the sonar device, and beam refers to the beam formed by the sound waves emitted by the sonar device; Add the original depth of the middle beam in the ping to the depth deviation z to obtain the radius r of the arc; According to the radius r of the arc, the gross error point caused by the side lobe effect is obtained to complete the amplification of the gross error of the side lobe effect and obtain the gross error point Z caused by the side lobe effect p，i The formula of ′ is as follows: Wherein, θ is the incident angle of the beam; h is the heading angle of the beam; N transducer is the plane coordinate of the sonar device N; N p，i ″ is the position of the sonar device N at beam number i after adjustment based on the arc circle; E transducer is the plane coordinate of the sonar device E; E p，i ″ is the position of the sonar device E at beam number i after adjustment based on the arc circle.

10. The method for sample augmentation of gross errors in multi-beam bathymetric data according to claim 1, characterized in that: The hydrological measurement standard is one of the special standard and the second-class standard.

Images