An underwater visual-inertial positioning method, system, device and medium

Abstract

The present application relates to the technical field of robot positioning, in particular to an underwater visual-inertial positioning method, system, device and medium. The underwater image is preprocessed, and the IMU data between adjacent two underwater images is pre-integrated to obtain the preprocessed underwater image and IMU pre-integration; the preprocessed underwater image is adaptively estimated and feature matched to obtain an adaptive threshold of a visual weight factor and initial pose information of a camera; the initial pose information and the IMU pre-integration are aligned to obtain visual-inertial initialization information; the visual-inertial initialization information is used to obtain a visual re-projection error; the visual re-projection error and the adaptive threshold of the visual weight factor are used to construct a visual dynamic weight factor; a residual equation is constructed according to the visual dynamic weight factor, the visual re-projection error and an IMU residual error; the residual equation is solved to obtain pose information; the coupling information of vision and inertia is fully utilized to improve underwater positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of robot positioning technology, and in particular to an underwater visual inertial positioning method, system, device and medium. Background Technology

[0002] Currently, unmanned underwater vehicles (UUVs) are important tools for humans to conduct underwater research. When UUVs perform tasks in confined environments or at close range, such as underwater cave surveys, subsea pipeline inspections, and ship repairs, a robust and highly accurate UUV positioning technology is required. Cameras can capture environmental information within their field of view and use this information to complete their own positioning.

[0003] However, due to the limitations of the underwater environment, a single visual sensor struggles to estimate accurate position information. Tightly coupled visual-inertial fusion methods can fully leverage the strengths of different sensors, complementing each other's advantages and disadvantages. However, low-quality images significantly impact nonlinear optimization, reducing the accuracy of pose estimation. Summary of the Invention

[0004] The problem this invention aims to solve is that existing technologies in underwater visual-inertial positioning do not consider the impact of low-quality images on visual-inertial optimization, failing to fully utilize the effectiveness of visual and IMU information and ensure the full utilization of visual-inertial information. To address this problem, firstly, this invention proposes an underwater visual-inertial positioning method, the method comprising:

[0005] The underwater images are preprocessed, and the IMU data between two adjacent underwater images is pre-integrated to obtain the preprocessed underwater images and the IMU pre-integrated data.

[0006] Adaptive estimation and feature matching are performed on the preprocessed underwater image to obtain the adaptive threshold of the visual weight factor and the initial pose information of the camera.

[0007] Align the initial pose information with the IMU pre-integration to obtain visual-inertial initialization information;

[0008] Based on the visual inertial initialization information, the visual reprojection error is obtained;

[0009] A visual dynamic weighting factor is constructed using the visual reprojection error and the adaptive threshold of the visual weighting factor.

[0010] A residual equation is constructed based on the visual dynamic weighting factor, visual reprojection error, and IMU residual; the IMU residual is calculated by IMU pre-integration.

[0011] Solve the residual equation to obtain pose information.

[0012] Further, the adaptive estimation and feature matching of the preprocessed underwater image to obtain the adaptive threshold of the visual weight factor and the initial pose information of the camera includes:

[0013] Feature points in the preprocessed underwater image were detected using the FAST algorithm.

[0014] The BRIEF algorithm is used to describe the feature points to obtain descriptors;

[0015] By comparing the similarity of the descriptors, feature point matching is performed using the FLANN algorithm to obtain the total number of matching pairs.

[0016] The RANSAC algorithm is used to remove erroneous matches from the total matching pairs to obtain the removed matching pairs;

[0017] The adaptive threshold of the visual weight factor is calculated based on the eliminated matching pairs and the total number of matching pairs. The adaptive threshold of the visual weight factor is calculated using the following formula:

[0018]

[0019] Subtract the removed matching pairs from the total number of matching pairs to obtain the valid matching pairs;

[0020] Based on the effective matching pairs, the camera motion is estimated, visual initialization is completed, and the initial pose information of the camera is obtained.

[0021] Further, aligning the initial pose information and IMU pre-integration to obtain visual-inertial initialization information includes:

[0022] Align the initial pose information and the IMU pre-integration, solve for the system joint initialization parameters, and obtain the visual-inertial initialization information; the system joint initialization parameters S include:

[0023]

[0024] Where s represents the system scale; is the rotation matrix in the gravity direction in the coordinate system of the first visual keyframe; b represents the acceleration bias and gyroscope bias of the IMU; It is the IMU body speed scaled from keyframe 0 to keyframe k.

[0025] Further, obtaining the visual reprojection error based on the visual inertial initialization information includes:

[0026] Based on the visual inertial initialization information, the pose information at keyframe i is obtained, and the pose information is represented as:

[0027]

[0028] in, and These represent the rotation and translation matrices from the camera coordinate system to the world coordinate system at keyframe i, respectively.

[0029] Obtain the single visual reprojection error of keyframe i and 3D map point j, whereby the single visual reprojection error is expressed as:

[0030]

[0031] Among them, u ij This represents the pixel coordinates of map point j in keyframe i; Represents the three-dimensional coordinates l of a known map point j. j pose information of keyframe i Obtain the pixel coordinates of the projected map point j in keyframe i;

[0032] The total visual reprojection error based on the least squares method is obtained from the single visual reprojection error, and the total visual reprojection error is used as the visual reprojection error.

[0033] The visual reprojection error is expressed as:

[0034] Among them, L i This represents the set of n map points at keyframe i.

[0035] Further, the step of constructing a visual dynamic weighting factor using the visual reprojection error and the adaptive threshold of the visual weighting factor includes:

[0036] The root mean square (RMS) reprojection error is calculated based on the visual reprojection error using the following formula:

[0037]

[0038] A visual dynamic weighting factor is constructed based on the visual reprojection error and the adaptive threshold of the visual weighting factor, and the visual dynamic weighting factor is calculated using the following formula:

[0039]

[0040] Further, the step of constructing the residual equation based on the visual dynamic weighting factor, visual reprojection error, and IMU residual includes:

[0041] The IMU residual and the visual residual with visual dynamic weighting factor are subjected to nonlinear optimization to complete the pose estimation and construct the residual equation.

[0042] The residual equation is constructed using the following formula:

[0043]

[0044] Where η is the variable to be optimized, expressed as η={x0,x1,…,x k ,l0,l1,…,l l},x k Represents the IMU state information at keyframe k; L represents the set of l+1 map points; Kf represents the i+1 keyframes; ρ(·) represents the Huber robust kernel function; EI i,i+1 P represents the IMU residual; V and P I This represents the covariance corresponding to the residual.

[0045] In a second aspect, the present invention provides an underwater visual inertial positioning system, the system comprising:

[0046] The preprocessing module is used to preprocess the underwater images and pre-integrate the IMU data between two adjacent underwater images to obtain the preprocessed underwater images and the IMU pre-integrated data.

[0047] An adaptive estimation and feature matching module is used to perform adaptive estimation and feature matching on the preprocessed underwater image to obtain an adaptive threshold for the visual weight factor and the initial pose information of the camera.

[0048] An initialization module is used to align the initial pose information and IMU pre-integration to obtain visual-inertial initialization information;

[0049] The error calculation module is used to obtain the visual reprojection error based on the visual inertial initialization information.

[0050] The weighting factor construction module is used to construct a visual dynamic weighting factor using the visual reprojection error and the adaptive threshold of the visual weighting factor.

[0051] The residual equation construction module is used to construct residual equations based on the visual dynamic weighting factor, visual reprojection error, and IMU residuals; the IMU residuals are calculated by IMU pre-integration.

[0052] The pose information acquisition module is used to solve the residual equation to obtain pose information.

[0053] Thirdly, the present invention provides a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements an underwater visual inertial positioning method as described above.

[0054] Fourthly, the present invention provides a computer-readable storage medium comprising a stored computer program; wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform an underwater visual inertial positioning method as described above.

[0055] The underwater visual-inertial positioning method, system, device, and medium of this invention have the following advantages: Utilizing the fast CLAHE image processing algorithm, the visibility and quality of underwater images can be improved in real time, enhancing details and effectively adapting to different lighting conditions; ORB feature processing is performed on the underwater image, and the feature point matching is optimized using the RANSAC algorithm. Simultaneously, considering the matching point pair results, the weight factor threshold is adaptively estimated, and a dynamic visual weight factor is constructed using the adaptive threshold and back-end visual reprojection error information; fully leveraging the coupled information of vision and inertia, the underwater positioning accuracy is improved. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of an underwater visual inertial positioning method, system, device, and medium provided in an embodiment of the present invention;

[0057] Figure 2 This is a flowchart of an underwater visual-inertial positioning method provided in an embodiment of the present invention;

[0058] Figure 3 This is a flowchart of the dynamic weight factor threshold adaptive estimation process provided in the embodiments of the present invention;

[0059] Figure 4 These are experimental comparison diagrams provided in the embodiments of the present invention;

[0060] Figure 5 This is a structural block diagram of an underwater visual inertial positioning system provided in an embodiment of the present invention;

[0061] Figure 6 This is a structural diagram of the computer device provided in an embodiment of the present invention. Detailed Implementation

[0062] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0063] It should be noted that the step numbers in this document are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are executed. The method provided in this embodiment can be executed by a relevant server, and the following description will use a server as the execution subject.

[0064] To address the impact of low-quality underwater images on visual-inertial fusion systems, such as... Figure 1 As shown in the preferred embodiment of the present invention, an underwater visual-inertial positioning method includes steps S1 to S7:

[0065] Step S1: Preprocess the underwater image and pre-integrate the IMU data between two adjacent underwater images to obtain the preprocessed underwater image and the IMU pre-integrated data.

[0066] This embodiment uses the CLAHE algorithm for underwater image preprocessing. Simultaneously, IMU data between two adjacent underwater image frames (k=m and k=n) can be selected for pre-integration. The IMU pre-integration is obtained using the following formula:

[0067]

[0068] Where, ΔR mn Δv mn Δp mn These represent the rotation, velocity, and translation pre-integration of the IMU between frames m and n, respectively. b represents the raw measurements of angular velocity and acceleration in the k-th frame, respectively; gk n gk and b ak n ak Δt represents the bias and noise of the gyroscope and the bias and noise of the accelerometer in the k-th frame, respectively; Δt is the IMU sampling time interval.

[0069] This embodiment utilizes the fast CLAHE image processing algorithm combined with an IMU inertial measurement unit to improve the visibility and quality of underwater images in real time, enhance details, and effectively adapt to different lighting conditions.

[0070] Step S2: Perform adaptive estimation and feature matching on the preprocessed underwater image to obtain the adaptive threshold of the visual weight factor and the initial pose information of the camera.

[0071] This embodiment obtains feature points of the preprocessed underwater image by performing ORB feature processing. Simultaneously, it utilizes the RANSAC algorithm to eliminate false matches and adaptively estimates the weight factor threshold. A specific example is shown below. Figure 3 As shown below, the above process is described in detail:

[0072] Specifically, in this embodiment, the corner points of the preprocessed underwater image are detected by the FAST algorithm as feature points; based on the position of the corner points, the corner points are described by the BRIEF algorithm to obtain descriptors; by comparing the similarity of the descriptors, the feature points are matched and recorded by the FLANN algorithm to obtain the total matching pairs.

[0073] The RANSAC algorithm is used to remove erroneous matches from the total matching pairs to obtain the removed matching pairs, and the proportion of the removed matching pairs in the total matching pairs is recorded.

[0074] After accumulating the total matching pairs and the eliminated mismatch pairs over multiple frames, the adaptive threshold of the visual weight factor is calculated and obtained.

[0075] The adaptive threshold of the visual weight factor is calculated using the following formula:

[0076]

[0077] Subtract the removed matching pairs from the total number of matching pairs to obtain the valid matching pairs;

[0078] After obtaining valid matching pairs between adjacent image frames, the camera motion is estimated based on the valid matching pairs to complete visual initialization and obtain the initial pose information of the camera.

[0079] This embodiment optimizes feature point matching using the RANSAC algorithm, and adaptively estimates the weight factor threshold while taking into account the matching point pair results.

[0080] Step S3: Align the initial pose information and IMU pre-integration to obtain visual inertial initialization information.

[0081] This embodiment obtains visual-inertial initialization information by aligning the initial pose information of the camera and the pre-integral quantity of the IMU to solve for the joint initialization parameters of the system; the joint initialization parameters S of the system include:

[0082]

[0083] Where s represents the system scale; is the rotation matrix in the gravity direction in the coordinate system of the first visual keyframe; b represents the acceleration bias and gyroscope bias of the IMU; This is the IMU body velocity scaled from keyframe 0 to keyframe k, and its true value is expressed as:

[0084] Step S4: Obtain the visual reprojection error based on the visual inertial initialization information.

[0085] The process of obtaining visual reprojection error is described in detail below:

[0086] In this embodiment, the pose information in keyframe i is obtained through the joint initialization parameter S of the visual-inertial initialization information obtained above. The pose information is represented as follows:

[0087]

[0088] in, and These represent the rotation and translation matrices from the camera coordinate system to the world coordinate system at keyframe i, respectively.

[0089] Obtain the single visual reprojection error of keyframe i and 3D map point j, whereby the single visual reprojection error is expressed as:

[0090]

[0091] Among them, u ij This represents the pixel coordinates of map point j in keyframe i; Represents the three-dimensional coordinates l of a known map point j. j pose information of keyframe i Obtain the pixel coordinates of the projected map point j in keyframe i;

[0092] This embodiment obtains the total visual reprojection error based on the least squares method through the single visual reprojection error, and uses this total visual reprojection error as the visual reprojection error; the visual reprojection error is expressed as... Among them, L i This represents the set of n map points at keyframe i.

[0093] The purpose of obtaining visual reprojection error is to construct a visual dynamic weighting factor and add it to the visual-inertial coupling system, thereby achieving accurate positioning of underwater images.

[0094] Step S5: Construct a visual dynamic weighting factor using the visual reprojection error and the adaptive threshold of the visual weighting factor.

[0095] In this embodiment, the root mean square (RMS) reprojection error is calculated based on the visual reprojection error, using the following formula:

[0096]

[0097] The visual dynamic weighting factor is calculated using the following formula:

[0098]

[0099] This embodiment constructs a visual dynamic weight factor by combining the adaptively estimated weight factor threshold with the back-end visual reprojection error information. Under this premise, the adaptive dynamic weight factor is then added to the visual-inertial tightly coupled system, which can effectively reduce the impact of low-quality underwater images on the fusion system.

[0100] Steps S6 to S7 involve constructing a residual equation based on the visual dynamic weighting factor, visual reprojection error, and IMU residual; the IMU residual is calculated by IMU pre-integration; and solving the residual equation yields pose information.

[0101] This embodiment completes pose estimation by performing nonlinear optimization processing on the IMU residual and the visual residual with visual dynamic weighting factors;

[0102] Specifically, in this embodiment, within a local window, there are i+1 keyframes, denoted by Kf; a set of l+1 map points, which can be represented by L; and η is set as the variable to be optimized, expressed as η={x0,x1,…,x k l0, l1, ..., l l},in, This represents the IMU status information under keyframe k;

[0103] The IMU residuals are obtained by calculating the IMU predicted values ​​in this embodiment using EI. i，i+1 The IMU residuals are represented by the following formula, which is used to calculate the IMU residuals:

[0104]

[0105]

[0106] in, IMU pre-integration from keyframe i to keyframe i+1;

[0107] The least squares residual equation is constructed and expressed by the following formula:

[0108]

[0109] Where ρ(·) represents the Huber robust kernel function; P V and P I This represents the covariance corresponding to the residual;

[0110] The pose information of the keyframe can be obtained by solving the least squares residual equation using a nonlinear optimization method.

[0111] This invention can demonstrate the technical effects of the above steps through an experimental method. For example, based on an ORB-SLAM3 system, using data measured by a monocular camera and IMU, an experiment can be conducted on a port area and on port sequences 01, 02, and 03, comparing systems with and without visual dynamic weighting factors. To ensure more accurate data results, multiple experiments can be performed and recorded. In this embodiment, five experiments were conducted, and the results, i.e., the root mean square error (RMSE) comparison chart, are shown below. Figure 4As shown.

[0112] This embodiment incorporates an adaptive visual dynamic weighting factor into a visual-inertial tightly coupled system and solves it using a nonlinear optimization method by constructing a least-squares residual equation, thereby obtaining pose information. An example is shown below. Figure 2 As shown, this method can effectively reduce the impact of low-quality underwater images on the fusion system, give full play to the role of visual information and IMU information in pose optimization, and improve the accuracy of underwater visual inertial positioning. In addition, this invention can also be applied to other multi-sensor fusion systems that include camera-IMU information with appropriate modifications, and has a wide range of applications and promising development prospects.

[0113] In summary, this invention provides an underwater visual-inertial localization method that takes into account the impact of low-quality underwater images on the tight coupling of visual and inertial systems. It utilizes the fast CLAHE image processing algorithm to improve the visibility and quality of underwater images in real time, enhance details, and effectively adapt to different lighting conditions. The method performs ORB feature processing on the underwater image and optimizes feature point matching using the RANSAC algorithm. Simultaneously, it adaptively estimates the weight factor threshold based on the matching point pair results. Finally, it constructs a dynamic visual weight factor using the adaptive threshold and backend visual reprojection error information.

[0114] Based on the above-described underwater visual-inertial positioning method, embodiments of the present invention also provide an underwater visual-inertial positioning system, such as... Figure 5 As shown, the system includes S21 to S27:

[0115] The preprocessing module S21 is used to preprocess the underwater image and pre-integrate the IMU data between two adjacent underwater images to obtain the preprocessed underwater image and the IMU pre-integrated data.

[0116] The adaptive estimation and feature matching module S22 is used to perform adaptive estimation and feature matching on the preprocessed underwater image to obtain the adaptive threshold of the visual weight factor and the initial pose information of the camera.

[0117] The initialization module S23 is used to align the initial pose information and the IMU pre-integration to obtain visual inertial initialization information;

[0118] Error calculation module S24 is used to obtain visual reprojection error based on the visual inertial initialization information;

[0119] The weight factor construction module S25 is used to construct a visual dynamic weight factor using the visual reprojection error and the adaptive threshold of the visual weight factor.

[0120] The residual equation construction module S26 is used to construct residual equations based on the visual dynamic weighting factor, visual reprojection error, and IMU residuals; the IMU residuals are obtained by IMU pre-integration calculation.

[0121] The pose information acquisition module S27 is used to solve the residual equation to obtain pose information.

[0122] The technical features and effects of the underwater visual-inertial positioning system proposed in this invention are the same as those of the underwater visual-inertial positioning method proposed in this invention, and will not be repeated here. Each module in the above-mentioned underwater visual-inertial positioning system can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0123] This invention also provides a computer-readable storage medium comprising a stored computer program; wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform an underwater visual-inertial positioning method as described above.

[0124] like Figure 6 As shown in the figure, this invention also provides a computer device. The figure is a structural block diagram of a preferred embodiment of the computer device provided by this invention. The computer device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an underwater visual inertial positioning method as described above.

[0125] Preferably, the computer program can be divided into one or more modules / units (such as computer program 1, computer program 2, ...), and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0126] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines.

[0127] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., while the data storage area can store related data, etc. Furthermore, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard drive, a SmartMedia Card (SMC), a Secure Digital (SD) card, and a Flash Card, or other volatile solid-state storage devices.

[0128] It should be noted that the aforementioned terminal devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 6 The block diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown in the diagram, or combine certain components, or use different components.

[0129] The above description is only a preferred embodiment of the present invention. It should be noted that for ordinary counters in the art, several improvements and substitutions can be made without departing from the counting principle of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. An underwater visual-inertial positioning method, characterized in that, The method includes: The underwater images are preprocessed, and the IMU data between two adjacent underwater images is pre-integrated to obtain the preprocessed underwater images and the IMU pre-integrated data. Adaptive estimation and feature matching are performed on the preprocessed underwater image to obtain the adaptive threshold of the visual weight factor and the initial pose information of the camera; specifically, the feature points of the preprocessed underwater image are detected using the FAST algorithm. The BRIEF algorithm is used to describe the feature points to obtain descriptors; By comparing the similarity of the descriptors, feature point matching is performed using the FLANN algorithm to obtain the total number of matching pairs. The RANSAC algorithm is used to remove erroneous matches from the total matching pairs to obtain the removed matching pairs; The adaptive threshold of the visual weight factor is calculated based on the eliminated matching pairs and the total number of matching pairs. The adaptive threshold of the visual weight factor is calculated using the following formula: Subtract the removed matching pairs from the total number of matching pairs to obtain the valid matching pairs; Based on the effective matching pairs, estimate the camera motion, complete visual initialization, and obtain the initial pose information of the camera; align the initial pose information with the IMU pre-integration to obtain visual inertial initialization information; Based on the visual inertial initialization information, the visual reprojection error is obtained; A visual dynamic weighting factor is constructed using the visual reprojection error and the adaptive threshold of the visual weighting factor. A residual equation is constructed based on the visual dynamic weighting factor, visual reprojection error, and IMU residual; the IMU residual is calculated by IMU pre-integration. Solve the residual equation to obtain pose information.

2. The underwater visual-inertial positioning method according to claim 1, characterized in that, The step of aligning the initial pose information and IMU pre-integration to obtain visual-inertial initialization information includes: Align the initial pose information and the IMU pre-integration, solve for the system joint initialization parameters, and obtain the visual-inertial initialization information; the system joint initialization parameters include: ； in, Indicates the scale of the system; This is the rotation matrix for the gravity direction in the coordinate system of the first visual keyframe; This indicates the IMU's acceleration bias and gyroscope bias; From keyframe 0 to the The IMU body speed is scaled by the keyframe.

3. The underwater visual-inertial positioning method according to claim 2, characterized in that, The step of obtaining the visual reprojection error based on the visual inertial initialization information includes: Based on the visual inertial initialization information, obtain the keyframes. The pose information is as follows: in, and These represent the keyframes respectively. Below, the rotation and translation matrices from the camera coordinate system to the world coordinate system; Get keyframes and 3D map points The single visual reprojection error, expressed as: in, Represents map points In keyframe The pixel coordinates below; This indicates that for a known map point 3D coordinates and keyframes pose information To obtain the projected map points In keyframe The pixel coordinates below; The total visual reprojection error based on the least squares method is obtained from the single visual reprojection error, and the total visual reprojection error is used as the visual reprojection error. The visual reprojection error is expressed as: , in, Indicates in keyframe Down, A set of map points.

4. The underwater visual-inertial positioning method according to claim 3, characterized in that, The step of constructing a visual dynamic weight factor using the visual reprojection error and the adaptive threshold of the visual weight factor includes: The root mean square (RMS) reprojection error is calculated based on the visual reprojection error using the following formula: ； A visual dynamic weighting factor is constructed based on the visual reprojection error and the adaptive threshold of the visual weighting factor, and the visual dynamic weighting factor is calculated using the following formula: 。 5. The underwater visual-inertial positioning method according to claim 4, characterized in that, The step of constructing the residual equation based on the visual dynamic weighting factor, visual reprojection error, and IMU residual includes: The IMU residual and the visual residual with visual dynamic weighting factor are subjected to nonlinear optimization to complete the pose estimation and construct the residual equation. The residual equation is constructed using the following formula: in, Let be the variable to be optimized, denoted as , express IMU status information in keyframes; express A set of map points; express One keyframe; This represents the Huber robust kernel function; Indicates the IMU residual; and This represents the covariance corresponding to the residual.

6. An underwater visual inertial positioning system, characterized in that, The system includes: The preprocessing module is used to preprocess the underwater images and pre-integrate the IMU data between two adjacent underwater images to obtain the preprocessed underwater images and the IMU pre-integrated data. An adaptive estimation and feature matching module is used to perform adaptive estimation and feature matching on the preprocessed underwater image to obtain an adaptive threshold for the visual weight factor and the initial pose information of the camera. An initialization module is used to align the initial pose information and IMU pre-integration to obtain visual-inertial initialization information; specifically, the FAST algorithm is used to detect feature points of the preprocessed underwater image. The BRIEF algorithm is used to describe the feature points to obtain descriptors; By comparing the similarity of the descriptors, feature point matching is performed using the FLANN algorithm to obtain the total number of matching pairs. The RANSAC algorithm is used to remove erroneous matches from the total matching pairs to obtain the removed matching pairs; The adaptive threshold of the visual weight factor is calculated based on the eliminated matching pairs and the total number of matching pairs. The adaptive threshold of the visual weight factor is calculated using the following formula: Subtract the removed matching pairs from the total number of matching pairs to obtain the valid matching pairs; The camera motion is estimated based on the effective matching pairs, and visual initialization is completed to obtain the initial pose information of the camera; the error calculation module is used to obtain the visual reprojection error based on the visual inertial initialization information. The weighting factor construction module is used to construct a visual dynamic weighting factor using the visual reprojection error and the adaptive threshold of the visual weighting factor. The residual equation construction module is used to construct residual equations based on the visual dynamic weighting factor, visual reprojection error, and IMU residuals; the IMU residuals are calculated by IMU pre-integration. The pose information acquisition module is used to solve the residual equation to obtain pose information.

7. A computer device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements an underwater visual-inertial positioning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform an underwater visual inertial positioning method as described in any one of claims 1 to 5.

Images