Unmanned ship visual servo stabilization control method and system based on homography matrix

Abstract

The invention relates to the technical field of unmanned ship motion control, in particular to an unmanned ship visual servo stabilization control method and system based on a homography matrix. The method mainly comprises the following steps: resolving a homography matrix from expected image data to current image data; sVD decomposition is carried out on the homography matrix, and a scale factor, a translation vector and a rotation matrix mapped from expected image data to current image data are acquired; obtaining a stabilization control error vector based on the translation vector and the rotation vector; and constructing a pose error vector based on a rotation matrix, dynamically adjusting a pose error, carrying out derivation on a stabilization control error, establishing a monocular vision servo stabilization control system, and continuously outputting a feedback stabilization control parameter in a time varying manner based on a translation error and a rotation error. Feedback signals are provided through the monocular vision sensor, the pose error of the unmanned ship is reconstructed through the homography matrix technology, and accurate stabilization control over the unmanned ship in the complex environment is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned ship motion control, and in particular to an unmanned ship visual servo stabilization control method and system based on a homography matrix. Background Art

[0002] With the rapid progress of artificial intelligence technology and the significant improvement of computer computing power, the visual servo method has shown unprecedented superiority in the field of unmanned ship motion control. Its core relies on the high-precision positioning capability of optical cameras, which provides strong support for the autonomous navigation and precise control of unmanned ships in complex environments.

[0003] Among the many visual servoing methods, monocular visual servoing has attracted wide attention due to its unique advantages. Compared with binocular visual servoing and multi-camera visual servoing, monocular visual servoing processes fewer images, requires less computation, and can better meet the strict real-time requirements of the servo process. This feature is particularly important in the control of high-speed unmanned ships because it is directly related to the response speed and stability of the system.

[0004] However, monocular visual servoing also faces its inherent challenges. Since restoring three-dimensional spatial information using two-dimensional images is essentially an under-constrained problem, existing monocular visual servoing methods often require pre-calibration of camera external parameters or known target model information during relative positioning. This reliance brings many inconveniences in practical applications, limiting the wide application and flexible deployment of monocular visual servoing. Summary of the invention

[0005] According to the technical problem that the monocular visual servo control method proposed above must rely on camera external parameters and target model information, a visual servo stabilization control method for surface unmanned ships based on homography matrix is ​​provided. This method provides feedback signals through monocular vision sensors, reconstructs the position error of the unmanned ship using homography matrix technology, and realizes precise stabilization control of the unmanned ship in complex environments.

[0006] The technical means adopted by the present invention are as follows:

[0007] A visual servo stabilization control method for a surface unmanned ship based on a homography matrix, comprising:

[0008] Randomly select at least four feature points within the shooting range of the monocular camera;

[0009] Pre-acquire expected image data containing four feature points taken by a monocular camera when the unmanned ship is in a stable posture at the current position, and extract the expected two-dimensional coordinates of the feature points in the expected image data; obtain current image data containing four feature points taken by a monocular camera at the current position of the unmanned ship; extract the current two-dimensional coordinates of the feature points in the current image data; perform dimension expansion on the current two-dimensional coordinates and the expected two-dimensional coordinates to generate corresponding current three-dimensional coordinates and expected three-dimensional coordinates;

[0010] Calculate the homography matrix between the expected image data and the current image data based on the current 3D coordinates, the expected 3D coordinates and the monocular camera intrinsic parameter matrix;

[0011] Performing SVD decomposition on the homography matrix, and obtaining, based on the decomposition result, a scale factor for mapping between the desired image data and the current image data, and a translation vector and a rotation error matrix for mapping between the desired image data and the current image data;

[0012] Acquire a stabilization control error vector based on the translation vector and the rotation vector, wherein the stabilization control error vector includes a translation error and a rotation error;

[0013] The pose error vector is constructed based on the rotation matrix, the pose error is dynamically adjusted, the stabilization control error is derived, and a monocular visual servo stabilization control system is established. The stabilization control parameters are fed back based on the continuous time-varying output of the translation error and the rotation error.

[0014] Further, based on the current three-dimensional coordinates, the expected three-dimensional coordinates and the monocular camera intrinsic parameter matrix, a homography matrix between the expected image data and the current image data is solved, including solving the homography matrix according to the following steps:

[0015]

[0016] in, Indicates the current three-dimensional coordinates, represents the expected three-dimensional coordinates, H I represents the essential matrix;

[0017] Solve the homography matrix between the expected image data and the current image data based on the following formula:

[0018] H=K -1 H I K (14)

[0019] Among them, K represents the intrinsic parameter matrix of the monocular camera, H represents the homography matrix between the expected image data and the current image data, and

[0020] Further, the homography matrix is ​​subjected to SVD decomposition, and a scale factor for mapping the desired image data to the current image data is obtained based on the decomposition result, including:

[0021] Calculate and obtain the singular values ​​of the homography matrix, wherein the data amount of the singular values ​​is the smaller of the number of rows and the number of columns of the homography matrix;

[0022] The largest singular value is selected as the scale factor d.

[0023] Further, the homography matrix is ​​subjected to SVD decomposition, and a translation vector and a rotation matrix for mapping the desired image data to the current image data are obtained based on the decomposition result, including:

[0024] Get the current three-dimensional coordinates of the four feature points on the current image data and the desired 3D coordinates on the desired image data i=1,2,3,4, construct the following matrix based on the coordinates of the four points:

[0025]

[0026] The homography matrix is ​​decomposed by SVD according to the following formula:

[0027] A=USV T (16)

[0028] Among them, U and V are orthogonal matrices, S is a diagonal matrix, and the elements on the diagonal are the singular values ​​of H;

[0029] The rotation error matrix is ​​obtained according to the following formula:

[0030]

[0031] Where R(θ) represents the rotation error matrix between the inertial coordinate system and the attached coordinate system, and θ represents the rotation error angle between the inertial coordinate system and the attached coordinate system;

[0032] Get the translation vector according to the following formula:

[0033]

[0034] Among them, t represents the translation vector and d represents the scale factor.

[0035] Further, obtaining a stabilization control error vector based on the translation vector and the rotation matrix includes:

[0036] The translation error is obtained according to the following formula:

[0037] Translation error (e x , e y )for

[0038] e x =t(1),e y =t(2) (19)

[0039] Among them, (e x , e y ) represents the translation error;

[0040] The rotation error is obtained according to the following formula:

[0041] e ψ =atan2(R(2,1),R(1,1)) (20)

[0042] Among them, e ψ represents the rotation error;

[0043] Based on the translation error and rotation error, a stabilization control error vector is constructed as e=[e x ,e y ,e ψ ] T .

[0044] Further, constructing a posture error vector based on the rotation matrix includes constructing a posture error vector according to the following formula:

[0045] e′=R T (ψ)(η d -η) (21)

[0046] in, Represents the rotation matrix between the inertial coordinate system and the attached coordinate system, η=[x,y,ψ] T represents the current posture vector, (x, y, ψ) is the position and yaw angle of the unmanned ship in the inertial coordinate system, η d =[x d ,y d , ψ d ] T represents the desired pose vector, (x d ,y d , ψ d ) is the expected position and expected bow angle of the unmanned ship in the inertial coordinate system.

[0047] Furthermore, the posture error is dynamically adjusted, the stabilization control error is derived, and a monocular visual servo stabilization control system is established. The stabilization control parameters are fed back based on the continuous time-varying output of the translation error and the rotation error, including:

[0048] The general kinematic and dynamic model for building the stabilization control of unmanned ships is:

[0049]

[0050] Among them, the current pose vector η=[x,y,ψ] T Where (x, y, ψ) is the position and pitch angle of the unmanned ship in the inertial coordinate system; the velocity vector ν = [u, v, r] T where (u, v, r) represents the longitudinal / sway velocity and bow angular rate of the unmanned ship in the appendage coordinate system; τ = [τ u , 0, τ r ] T is the system control input vector; is the rotation matrix between the inertial coordinate system and the attached coordinate system; Μ=diag(m 11 , m 22 , m 33 )、D(ν)=diag(d 11 , d 22 , d 33 )and They are the inertial mass matrix, the system damping matrix and the Coriolis centripetal matrix respectively;

[0051] Combining the kinematic and dynamic general models of unmanned ship stabilization control to derive the stabilization control error, a monocular visual servo stabilization control system is established as follows:

[0052]

[0053] When the image depth, motion speed and model parameters are all unknown, the continuous time-varying output feedback stabilization controller is designed as:

[0054]

[0055] Among them, k1, k2, k3, k4 are feedback gain parameters.

[0056] The present invention also discloses a visual servo stabilization control system for an unmanned surface ship based on a homography matrix, which is used to implement the above method, and includes:

[0057] A feature point acquisition unit, which is used to arbitrarily select at least four feature points within the shooting range of the monocular camera;

[0058] An image feature coordinate acquisition unit is used to pre-acquire expected image data containing four feature points taken by a monocular camera when the unmanned ship is in a stable posture at a current position, and extract expected two-dimensional coordinates of the feature points in the expected image data; acquire current image data containing four feature points taken by a monocular camera at the current position of the unmanned ship; extract current two-dimensional coordinates of the feature points in the current image data; the current two-dimensional coordinates and the expected two-dimensional coordinates are both acquired through a target recognition algorithm, and the current two-dimensional coordinates and the expected two-dimensional coordinates are dimensionally expanded to generate corresponding current three-dimensional coordinates and expected three-dimensional coordinates;

[0059] A homography matrix acquisition unit, which is used to calculate the homography matrix between the expected image data and the current image data based on the current three-dimensional coordinates, the expected three-dimensional coordinates and the monocular camera intrinsic parameter matrix;

[0060] An SVD decomposition unit, which is used to perform SVD decomposition on the homography matrix, and based on the decomposition result, obtain a scale factor for mapping between the desired image data and the current image data, and obtain a translation vector and a rotation error matrix for mapping between the desired image data and the current image data;

[0061] A stabilization control error vector acquisition unit, which is used to acquire a stabilization control error vector based on the translation vector and the rotation error matrix, wherein the stabilization control error vector includes a translation error and a rotation error;

[0062] The stabilization control unit is used to construct a posture error vector based on the rotation matrix, dynamically adjust the posture error, derive the stabilization control error, establish a monocular visual servo stabilization control system, and output feedback stabilization control parameters based on the continuous time-varying translation error and rotation error.

[0063] Compared with the prior art, the present invention has the following advantages:

[0064] The present invention proposes a monocular visual servo stabilization control method and system for an underactuated unmanned ship based on a homography matrix. By cleverly using the homography decomposition technology, the pose error containing the unknown scale factor is directly parsed from the current and expected images, thereby completely eliminating the reliance on the precise calibration of the camera's external parameters and the prior information of the target object. In view of the inherent constraints of the underactuated system, the present invention incorporates a periodic function based on the continuous excitation of the bow roll angle into the design of the continuous time-varying output feedback stabilization controller. This design enables the underactuated unmanned ship to achieve precise stabilization control even when the image depth information, motion speed and model parameters are unknown. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0066] Figure 1 The present invention is a flowchart of a visual servoing stabilization control method for an unmanned surface ship based on a homography matrix in an embodiment of the present invention.

[0067] Figure 2In the embodiment of the present invention, the inertial coordinate system E-XYZ and the unmanned ship attached coordinate system BX b Y b Z b 、Camera coordinate system CX c Y c Z c And the image coordinate system OX o Y o Relationship diagram.

[0068] Figure 3 It is the pose error related to the unknown image depth obtained by decomposing the homography matrix in an embodiment of the present invention.

[0069] Figure 4 Schematic diagram of control input of a continuous time-varying output feedback stabilization controller in an embodiment of the present invention.

[0070] Figure 5 The figure is a schematic diagram of the pitch / swing speed and the bow angular rate of the unmanned ship under the control input in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0072] The present invention provides a visual servo stabilization control method for a surface unmanned ship based on a homography matrix, which mainly comprises the following steps:

[0073] S1: Feature point selection.

[0074] Select at least four non-collinear feature points p within the field of view of the monocular camera. i , each feature point should have high stability and identifiability. Among them, these four points form a target plane Ω. In this embodiment, four feature points are used for solution, i=1,2,3,4.

[0075] S2: Intrinsic calibration of shipborne monocular camera.

[0076] Use the calibration board to calibrate the shipborne monocular camera using Zhang Zhengyou's calibration method, obtain the intrinsic parameter matrix K of the monocular camera, and obtain the focal length a of the camera in the corresponding direction x and a y , where K is defined as follows:

[0077]

[0078] in, are the coordinates of the image center, which are usually determined based on the resolution of the image.

[0079] S3: Image acquisition and feature point recognition.

[0080] Start the ship-borne monocular camera to obtain images of the unmanned ship in the initial state and the expected state, and obtain the feature point p through the general target recognition algorithm. i The current image 2D coordinates and the expected image 2D coordinates. The image in the expected state is an image containing feature points extracted by the ship-borne monocular camera after the ship is stabilized after the ship is driven to the current position in advance. In order to facilitate subsequent calculations, the current 2D coordinates and the expected 2D coordinates are dimensionally expanded to generate the corresponding current 3D coordinates and the expected 3D coordinates. and The definition is as follows:

[0081]

[0082] S4: homography matrix calculation;

[0083] According to the principle of epipolar geometry, and The following relationship exists:

[0084]

[0085] By using formula (3), the homography matrix H between the two images can be directly calculated using four pairs of corresponding feature points: I Among them, H I is the essential matrix.

[0086] Combined with the internal parameter matrix K, we can get:

[0087] H=K -1 H I K (28)

[0088] Among them, H is the Euclidean homography matrix, which is generally a 3×3 matrix and has the following form:

[0089]

[0090] The Euclidean homography matrix is ​​often used to describe linear transformations from one plane to another, including operations such as rotation, translation, and scaling.

[0091] S5: Homography matrix decomposition and stabilization control error calculation.

[0092] H can be decomposed into a translation vector and a rotation vector, which can be specifically expressed as

[0093]

[0094] Among them, R(θ) is the rotation error matrix between the inertial coordinate system and the attached coordinate system, t is the translation vector, and d is the scale factor, that is, The perpendicular distance to the target plane Ω, representing the inherently unknown image depth. r is the unit normal vector of the target plane Ω.

[0095] From formula (6), it can be seen that H contains the error information of unmanned ship stabilization control e = [e x ,e y ,e ψ ] T .

[0096] Furthermore, the singular value decomposition (SVD) method of the matrix is ​​used to decompose H into the translation error (e x , e y ), rotation error e ψ And the scale factor d is used to construct the unmanned ship stabilization control error e.

[0097] The specific process is as follows:

[0098] Get the current three-dimensional coordinates of the four feature points on the current image data and the desired 3D coordinates on the desired image data i=1,2,3,4, construct the following matrix based on the coordinates of the four points:

[0099]

[0100] The homography matrix is ​​decomposed by SVD according to the following formula:

[0101] A=USV T (32)

[0102] Among them, U and V are orthogonal matrices, S is a diagonal matrix, and the elements on the diagonal are the singular values ​​of H.

[0103] The scale factor d is further extracted, which is usually taken from the largest singular value of S.

[0104] Finally, the rotation error matrix R(θ) and the translation vector t are calculated. The specific calculation process is as follows:

[0105]

[0106] Wherein, R(θ) represents the rotation error matrix between the inertial coordinate system and the attached body coordinate system, and θ represents the rotation error angle between the inertial coordinate system and the attached body coordinate system.

[0107]

[0108] Among them, t represents the translation vector and d represents the scale factor.

[0109] Stabilization control error calculation:

[0110] Translation error (e x , e y )for

[0111] e x =t(1),e y =t(2) (35)

[0112] Rotation error e ψ for

[0113] e ψ =atan2(R(2,1),R(1,1)) (36)

[0114] S6: Stabilization controller design.

[0115] This example shows Figure 1 The inertial coordinate system E-XYZ and the unmanned ship body coordinate system BX shown in b Y b Z b 、Camera coordinate system CX c Y c Z c And the image coordinate system OX o Y o Relationship diagram. The optical axis X of the camera optical center C c Towards the front of the bow. The appendage coordinate system is located directly below the camera coordinate system. E-XYZ is selected as the north-east-down coordinate system. In addition, the origin O in the two-dimensional image coordinate system is located at the upper left corner of the image, and X o The axis is horizontal to the right, Y o The axis points vertically downward.

[0116] The kinematic and dynamic mathematical models of the general unmanned ship stabilization control can be described as:

[0117]

[0118] Among them, the current pose vector η=[x,y,ψ] T Where (x, y, ψ) is the position and pitch angle of the unmanned ship in the inertial coordinate system; the velocity vector ν = [u, v, r] T where (u, v, r) represents the longitudinal / sway velocity and bow angular rate of the unmanned ship in the appendage coordinate system; τ = [τ u , 0, τ r ] Tis the system control input vector; is the rotation matrix; Μ=diag(m 11 , m 22 , m 33 )、D(ν)=diag(d 11 , d 22 , d 33 )and They are the inertial mass matrix, the system damping matrix and the Coriolis centripetal matrix respectively.

[0119] Define the pose error vector e′:

[0120] e′=R T (ψ)(η d -η) (38)

[0121] Among them, e′ is used to adjust the unmanned ship from the current posture to the desired posture.

[0122] Combined with formula (13), the orientation error dynamics can be obtained by derivation of e′ as follows:

[0123]

[0124] Where: e′: = [e1, e2, e3] T , η d =[x d ,y d , ψ d ] T .

[0125] The unmanned ship kinematic and dynamic model (13) considered in the present invention is a generalized form. At the same time, considering the actual situation, the internal unmodeled dynamics and external disturbances can be ignored. Here, the assumption is given: Assume that the unmanned ship dynamic model parameters (m 11 , m 22 , m 33 )>0 and (d 11 , d 22 , d 33 )>0 means it is unrecognized; the feature point is always in the field of view, and the image depth and motion speed (u, v, r) are unknown.

[0126] Combined with the scale factor d decomposed from H, we can further obtain:

[0127]

[0128] Furthermore, the posture error is adjusted dynamically, and the stabilization control error is derived by combining formula (13-16), and the monocular visual servo stabilization control system is established as follows:

[0129]

[0130] In order to make the unmanned ship monocular visual servo system asymptotically stable when the image depth, motion speed and model parameters are all unknown, the continuous time-varying output feedback stabilization controller is designed as follows:

[0131]

[0132] Among them, k1, k2, k3, and k4 are feedback gain parameters, and stable stabilization control is achieved through real-time adjustment.

[0133] Figure 2 It is a flow chart of a monocular visual servo stabilization control system for an underactuated unmanned ship based on a homography matrix disclosed by the present invention;

[0134] In order to verify the superior performance of the invented continuous finite-time monocular visual servo control system, the famous CyberShip I standard model is used to conduct simulation experiments to verify the effectiveness of the continuous time-varying output feedback stabilization controller. 11 , m 22 , m 33 ) is (19, 35.2, 4.2), (d 11 , d 22 , d 33 ) is (4, 1, 10).

[0135] The coordinates of the four feature points are (0.9, -1, 0), (0.9, 1, 0), (0.9, 1, -2) and (0.9, -1, -2), and the feature plane Ω with combination d = 0.9. [x, y, ψ, u, v, r] T =[-10,-10,1.5,0,0,0] T is the initial position and velocity vector of the unmanned ship. Camera internal parameters The image resolution is 1024×1024. In addition, the custom feedback gain parameters are set to: k1=0.06, k2=0.06, k3=0.06, k4=0.2. The initial error of the visual servo obtained by homography matrix decomposition is e x (0) = 12.25, e y (0) = -10.46, e ψ (0) = -1.6.

[0136] Figure 3 It is the pose error related to the unknown image depth obtained by decomposing the homography matrix. Figure 4 is the control input of the continuous time-varying output feedback stabilization controller. Figure 5 Under the action of control input, the pitch / swing speed and bow roll angular rate of the unmanned ship. It can be seen that the posture error of the unmanned ship converges to the origin asymptotically.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A visual servo stabilization control method for a surface unmanned ship based on a homography matrix, characterized in that: include: Randomly select at least four feature points within the shooting range of the monocular camera; Pre-acquire expected image data containing four feature points taken by a monocular camera when the unmanned ship is in a stable posture at the current position, and extract the expected two-dimensional coordinates of the feature points in the expected image data; obtain current image data containing four feature points taken by a monocular camera at the current position of the unmanned ship; extract the current two-dimensional coordinates of the feature points in the current image data; perform dimension expansion on the current two-dimensional coordinates and the expected two-dimensional coordinates to generate corresponding current three-dimensional coordinates and expected three-dimensional coordinates; Calculate the homography matrix between the expected image data and the current image data based on the current 3D coordinates, the expected 3D coordinates and the monocular camera intrinsic parameter matrix; Performing SVD decomposition on the homography matrix, and obtaining, based on the decomposition result, a scale factor for mapping between the desired image data and the current image data, and a translation vector and a rotation error matrix for mapping between the desired image data and the current image data; Acquire a stabilization control error vector based on the translation vector and the rotation error matrix, wherein the stabilization control error vector includes a translation error and a rotation error; The pose error vector is constructed based on the rotation matrix, the pose error is dynamically adjusted, the stabilization control error is derived, and a monocular visual servo stabilization control system is established. The stabilization control parameters are fed back based on the continuous time-varying output of the translation error and the rotation error.

2. According to the method of visual servo stabilization control of a surface unmanned ship based on homography matrix in claim 1, it is characterized in that: The homography matrix between the expected image data and the current image data is calculated based on the current three-dimensional coordinates, the expected three-dimensional coordinates, and the monocular camera intrinsic parameter matrix, including solving the homography matrix according to the following steps: in, Indicates the current three-dimensional coordinates, represents the expected three-dimensional coordinates, H I represents the essential matrix; Solve the homography matrix between the expected image data and the current image data based on the following formula: H=K -1 H I K (2) Among them, K represents the intrinsic parameter matrix of the monocular camera, H represents the homography matrix between the expected image data and the current image data, and 3. According to the method of visual servo stabilization control of a surface unmanned ship based on homography matrix in claim 1, it is characterized in that: Performing SVD decomposition on the homography matrix, and obtaining a scale factor for mapping between the desired image data and the current image data based on the decomposition result, including: The singular values ​​of the homography matrix are calculated and the largest one among the singular values ​​is selected as the scale factor d.

4. According to the method of visual servo stabilization control of a surface unmanned ship based on homography matrix in claim 1, it is characterized in that: Performing SVD decomposition on the homography matrix, and obtaining a translation vector and a rotation error matrix for mapping the desired image data to the current image data based on the decomposition result, including: Get the current three-dimensional coordinates of the four feature points on the current image data and the desired 3D coordinates on the desired image data i=1,2,3,4, construct the following matrix based on the coordinates of the four points: The homography matrix is ​​decomposed by SVD according to the following formula: A=USV T (4) Among them, U and V are orthogonal matrices, S is a diagonal matrix, and the elements on the diagonal are the singular values ​​of H; The rotation error matrix is ​​obtained according to the following formula: Where R(θ) represents the rotation error matrix between the inertial coordinate system and the attached coordinate system, and θ represents the rotation error angle between the inertial coordinate system and the attached coordinate system; Get the translation vector according to the following formula: Among them, t represents the translation vector and d represents the scale factor.

5. The method for visual servoing stabilization control of an unmanned surface ship based on a homography matrix according to claim 4 is characterized in that: Obtaining a stabilization control error vector based on the translation vector and the rotation error matrix includes: The translation error is obtained according to the following formula: e x =t(1),e y =t(2) (7) Among them, (e x , e y ) represents the translation error; The rotation error is obtained according to the following formula: e ψ = atan2(R(2,1),R(1,1)) (8) Based on the translation error and rotation error, the stabilization control error vector is constructed as e=[e x ,e y ,e ψ ] T .

6. The method for visual servoing stabilization control of an unmanned surface ship based on a homography matrix according to claim 1 is characterized in that: Constructing a pose error vector based on the rotation matrix includes constructing a pose error vector according to the following formula: e′=R T (ψ)(η) d -h) (9) in, Represents the rotation matrix between the inertial coordinate system and the attached coordinate system, η=[x,y,ψ] T represents the current posture vector, (x, y, ψ) is the position and yaw angle of the unmanned ship in the inertial coordinate system, η d =[x d ,y d , ψ d ] T represents the desired pose vector, (x d ,y d , ψ d ) is the expected position and expected bow angle of the unmanned ship in the inertial coordinate system.

7. The method for visual servoing stabilization control of an unmanned surface ship based on a homography matrix according to claim 1 is characterized in that: Dynamically adjust the posture error, derive the stabilization control error, establish a monocular visual servo stabilization control system, and output feedback stabilization control parameters based on the continuous time-varying translation error and rotation error, including: The general kinematic and dynamic model for building the stabilization control of unmanned ships is: Among them, the current pose vector η=[x,y,ψ] T Where (x, y, ψ) is the position and pitch angle of the unmanned ship in the inertial coordinate system; the velocity vector ν = [u, v, r] T where (u, v, r) represents the longitudinal / sway velocity and bow angular rate of the unmanned ship in the appendage coordinate system; τ = [τ u , 0, τ r ] T is the system control input vector; is the rotation matrix between the inertial coordinate system and the attached coordinate system; Μ=diag(m 11 , m 22 , m 33 )、D(ν)=diag(d 11 , d 22 , d 33 )and They are the inertial mass matrix, the system damping matrix and the Coriolis centripetal matrix respectively; Combining the kinematic and dynamic general models of unmanned ship stabilization control to derive the stabilization control error, a monocular visual servo stabilization control system is established as follows: When the image depth, motion speed and model parameters are all unknown, the continuous time-varying output feedback stabilization controller is designed as: Among them, k1, k2, k3, k4 are feedback gain parameters.

8. A visual servo stabilization control system for an unmanned surface ship based on a homography matrix, used to implement the method as claimed in claim 1, characterized in that: include: A feature point acquisition unit, which is used to arbitrarily select at least four feature points within the shooting range of the monocular camera; An image feature coordinate acquisition unit is used to pre-acquire expected image data containing four feature points taken by a monocular camera when the unmanned ship is in a stable posture at a current position, and extract expected two-dimensional coordinates of the feature points in the expected image data; acquire current image data containing four feature points taken by a monocular camera at the current position of the unmanned ship; extract current two-dimensional coordinates of the feature points in the current image data; the current two-dimensional coordinates and the expected two-dimensional coordinates are both acquired through a target recognition algorithm, and the current two-dimensional coordinates and the expected two-dimensional coordinates are dimensionally expanded to generate corresponding current three-dimensional coordinates and expected three-dimensional coordinates; A homography matrix acquisition unit, which is used to calculate the homography matrix between the expected image data and the current image data based on the current three-dimensional coordinates, the expected three-dimensional coordinates and the monocular camera intrinsic parameter matrix; An SVD decomposition unit, which is used to perform SVD decomposition on the homography matrix, and based on the decomposition result, obtain a scale factor for mapping between the desired image data and the current image data, and obtain a translation vector and a rotation error matrix for mapping between the desired image data and the current image data; A stabilization control error vector acquisition unit, which is used to acquire a stabilization control error vector based on the translation vector and the rotation error matrix, wherein the stabilization control error vector includes a translation error and a rotation error; The stabilization control unit is used to construct a posture error vector based on a rotation error matrix, dynamically adjust the posture error, derive the stabilization control error, establish a monocular visual servo stabilization control system, and output feedback stabilization control parameters based on the continuous time-varying translation error and rotation error.

Images