Horizontal attitude extraction method, system and device and medium

Abstract

The invention discloses a horizontal attitude extraction method, system and device and a medium. The method comprises the following steps: determining a first coordinate vector of a zenith vector in a carrier coordinate system according to an estimated value of a central pixel coordinate of a Snell window, a lens model and a first conversion coefficient; the estimated value of the coordinates of the central pixel of the Snell window is determined according to propagation evaluation of incident light; determining a first coordinate function of the zenith vector in a carrier coordinate system according to the horizontal attitude rotation coefficient; and determining a horizontal attitude parameter according to the first coordinate vector and the first coordinate function. The embodiment of the invention can reduce the calculation amount, improve the precision and the anti-interference capability, and can be widely applied to the technical field of navigation.

Description

Technical Field

[0001] The present invention relates to the field of navigation technology, and in particular to a method, system, device and medium for extracting horizontal posture. Background Art

[0002] Polarized light navigation is an important type of bionic navigation technology. In recent years, equipment such as AUV (autonomous underwater vehicle) has played an increasingly important role in ocean exploration. There are many studies on the use of polarized light navigation for AUV. The key technology of underwater polarized light navigation is the study of the polarization (or polarization) pattern of underwater light. It is difficult for an underwater vehicle to keep its horizontal attitude angle (roll angle and pitch angle) at zero all the time during navigation. Therefore, a three-dimensional attitude solution algorithm is required.

[0003] Most of the existing methods for calculating the horizontal attitude angle in polarized light navigation use dual-time (the same sensor observes at two times) or dual-position (the same sensor observes at two different attitude angles at two times) or dual sensors (two sensors observe at the same time) observations, which have the disadvantages of complex measurement process and large amount of calculation of attitude angle solution algorithm. In addition, most of the current horizontal attitude angle solution algorithms based on underwater polarized light apply the relevant algorithms based on polarized light in the air, without considering the refraction of light waves by the water surface. Therefore, the error of horizontal attitude angle solution is large and the anti-interference ability is weak. Summary of the invention

[0004] In view of this, in order to solve one of the above problems, an embodiment of the present invention aims to provide a method, system, device and medium for extracting horizontal posture, which can reduce the amount of calculation and improve the accuracy and anti-interference ability.

[0005] On the one hand, an embodiment of the present invention provides a method for extracting a horizontal posture, comprising:

[0006] Determine a first coordinate vector of the zenith vector in the carrier coordinate system according to an estimated value of the center pixel coordinate of the Snell window, a lens model and a first conversion coefficient; the estimated value of the center pixel coordinate of the Snell window is determined according to a propagation evaluation of the incident light;

[0007] Determine the first coordinate function of the zenith vector in the carrier coordinate system according to the horizontal attitude rotation coefficient;

[0008] A horizontal posture parameter is determined according to the first coordinate vector and the first coordinate function.

[0009] Optionally, determining a first coordinate vector of the zenith vector in the carrier coordinate system according to an estimated value of the center pixel coordinate of the Snell window, a lens model and a first conversion coefficient comprises:

[0010] Determine a second coordinate vector of the zenith vector in the camera coordinate system according to the estimated value of the coordinate of the center pixel of the Snell window and the lens model;

[0011] A first coordinate vector of the zenith vector in the carrier coordinate system is determined according to the second coordinate vector and the first conversion coefficient.

[0012] Optionally, the estimated value of the center pixel coordinate of the Snell window is determined by the following method:

[0013] Determine a second coordinate parameter set of a plurality of incident light rays in the image sensor coordinate system according to the inverse lens model;

[0014] Determining a plurality of Hough space values ​​according to the second coordinate parameter set and a preset half cone angle range;

[0015] An estimated value of the coordinate of the center pixel of the Snell window is determined according to a preset number of larger values ​​among the plurality of Hough space values.

[0016] Optionally, a second coordinate parameter of an incident light ray in the image sensor coordinate system is determined by the following method:

[0017] Determine the third coordinate parameter of the incident light in the carrier coordinate system according to the second conversion coefficient;

[0018] Determine a fourth coordinate parameter of the incident light in the camera coordinate system according to the third coordinate parameter and the first conversion coefficient;

[0019] The second coordinate parameter is determined according to the fourth coordinate parameter and the inverse lens model.

[0020] Optionally, a second coordinate parameter determines a plurality of Hough space values ​​by the following method, including:

[0021] Determining a plurality of half cone angles according to the preset half cone angle range and angle step;

[0022] A plurality of Hough space values ​​are determined according to the second coordinate parameter and the plurality of the semi-cone angles.

[0023] Optionally, the preset semi-cone angle range is determined by the following method:

[0024] Determining the Snell angle based on the refractive index of the first medium and the refractive index of the second medium;

[0025] The preset semi-cone angle range is determined according to the Snell angle and the preset redundancy angle.

[0026] Optionally, the preset number is determined by the following method:

[0027] Image data from an image sensor is acquired, and the image data is input into a trained neural network model to obtain the preset number.

[0028] On the other hand, an embodiment of the present invention provides a horizontal posture extraction system, including:

[0029] A first module is used to determine a first coordinate vector of a zenith vector in a carrier coordinate system according to an estimated value of a central pixel coordinate of a Snell window, a lens model and a first conversion coefficient; the estimated value of the central pixel coordinate of the Snell window is determined according to a propagation evaluation of an incident light ray;

[0030] The second module is used to determine the first coordinate function of the zenith vector in the carrier coordinate system according to the horizontal attitude rotation coefficient;

[0031] The third module is used to determine the horizontal posture parameter according to the first coordinate vector and the first coordinate function.

[0032] On the other hand, an embodiment of the present invention provides a computer-readable storage medium, wherein a program executable by a processor is stored, and the program executable by the processor is used to execute the above-mentioned horizontal posture extraction method when executed by the processor.

[0033] On the other hand, an embodiment of the present invention provides a device for extracting a horizontal posture, comprising an underwater vehicle and a computer device connected to the underwater vehicle; wherein:

[0034] The underwater vehicle comprises a camera, an image sensor and a carrier, and the incident light passes through the camera and forms an image on the image sensor;

[0035] The computer device comprises:

[0036] at least one processor;

[0037] at least one memory for storing at least one program;

[0038] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned method for extracting the horizontal posture.

[0039] Implementation of the embodiments of the present invention includes the following beneficial effects: In the embodiment, first, the first coordinate vector of the zenith vector in the carrier coordinate system is determined according to the estimated value of the coordinate of the center pixel of the Snell window, the lens model and the first conversion coefficient. The estimated value of the coordinate of the center pixel of the Snell window is determined according to the propagation evaluation of the incident light, which is closer to the real scene and has strong anti-interference ability. The coordinate vector of the center pixel of the Snell window in the carrier coordinate system is established. Then, the first coordinate function of the zenith vector in the carrier coordinate system is determined according to the horizontal attitude rotation coefficient, and the coordinate function of the zenith vector in the carrier coordinate system is established. Finally, the horizontal attitude parameters are determined according to the first coordinate vector and the first coordinate function, and the amount of calculation is small, thereby reducing the amount of calculation and improving the accuracy and anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a schematic diagram of a coordinate system of a horizontal posture extraction system provided by an embodiment of the present invention;

[0041] Figure 2 It is a schematic diagram of the steps of a method for extracting a horizontal posture provided by an embodiment of the present invention;

[0042] Figure 3 is a coordinate schematic diagram of a Snell window provided by an embodiment of the present invention;

[0043] Figure 4 is a structural block diagram of a horizontal posture extraction system provided by an embodiment of the present invention;

[0044] Figure 5 It is a structural block diagram of a horizontal posture extraction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only provided for the convenience of explanation and description, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0046] The underwater attitude of an underwater vehicle is determined relative to the water surface, and a bright circular area called the Snell window can be seen when observing the water surface from bottom to top. Utilizing the characteristics of the Snell window coinciding with the water surface, the image taken by the underwater camera is used to deduce the observation vector of the 3D reconstruction of light through the imaging of the Snell window, and the horizontal attitude of the carrier is solved by establishing a real-time light intensity field. In view of the complex environment in the ocean, the Snell window image taken has great uncertainty. The extraction of the Snell window is difficult due to biological occlusion, dense light spots caused by sea waves, or light influence.

[0047] See also Figure 1 The whole model establishes the following five coordinate systems: image sensor coordinate system (horizontal and vertical axes are u and v respectively), carrier coordinate system b system (right-front-up), camera coordinate system c system (Z c The axis is along the camera optical axis, from the camera lens to the image sensor), the horizontal coordinate system h system (X h Axis and X n The angle between the axes is the heading angle), the light coordinate system l system (a local coordinate system, Z l The axis is outward along the incident light, X l The axis points downward along a tangent to the meridian at the origin).

[0048] The light intensity I is first pre-processed by filtering or binarization, and then the edge detection algorithm is used to generate the binary image I of the edge. e , and then through 3D cone Hough transform (3D-CoHT) from I e The Snell window is extracted from the 3D image. 3D-CoHT uses a two-stage mapping process: in the first stage, the image is first converted to the real 3D space; in the second stage, it is mapped to the Hough space through the inverse transformation. The cone is obtained from the real 3D space. The l system is an important coordinate system used to expand the edge pixels of the 2D circle into 3D generatrix vectors. After obtaining the Snell window, 3D optical reconstruction needs to be realized based on the calibrated camera to provide a basis for attitude solution.

[0049] See also Figure 2 , an embodiment of the present invention provides a method for extracting a horizontal posture, comprising:

[0050] S100, determining a first coordinate vector of the zenith vector in a carrier coordinate system according to an estimated value of the center pixel coordinate of the Snell window, a lens model and a first conversion coefficient; the estimated value of the center pixel coordinate of the Snell window is determined according to a propagation evaluation of an incident light ray.

[0051] S200. Determine a first coordinate function of a zenith vector in a carrier coordinate system according to a horizontal attitude rotation coefficient.

[0052] S300, determining a horizontal posture parameter according to a first coordinate vector and a first coordinate function.

[0053] The lens model represents the optical transmission relationship from the image sensor coordinate system to the camera coordinate system. The first conversion coefficient refers to the conversion coefficient from the camera coordinate system to the carrier coordinate system. The horizontal posture rotation coefficient refers to the rotation coefficient from the horizontal coordinate system to the carrier coordinate system.

[0054] Specifically, first, the estimated value of the Snell window center pixel coordinates is determined based on the propagation evaluation of the incident light, and then, the estimated value of the Snell window center pixel coordinates is converted from the image sensor coordinate system to the carrier coordinate system according to the lens model and the first conversion coefficient. At the same time, according to the horizontal attitude rotation coefficient, the zenith vector is converted from the horizontal coordinate system to the carrier coordinate system, and then, the parameters obtained by the two different path conversions are associated to determine the horizontal attitude parameters.

[0055] Optionally, determining a first coordinate vector of the zenith vector in the carrier coordinate system according to an estimated value of the center pixel coordinate of the Snell window, a lens model and a first conversion coefficient includes:

[0056] S110, determining a second coordinate vector of the zenith vector in the camera coordinate system according to the estimated value of the center pixel coordinate of the Snell window and the lens model;

[0057] S120. Determine a first coordinate vector of the zenith vector in the carrier coordinate system according to the second coordinate vector and the first conversion coefficient.

[0058] Specifically, an operation is performed based on the estimated value of the center pixel coordinate of the Snell window and the lens model, the zenith vector is converted from the image sensor coordinate system to the camera coordinate system to obtain a second coordinate vector, and then an operation is performed based on the second coordinate vector and the first conversion coefficient to obtain a first coordinate vector of the zenith vector in the carrier coordinate system.

[0059] In a specific embodiment, the zenith vector is represented as z in the horizontal coordinate system (h coordinate system). h =[0 0 1] T , the zenith vector in the carrier coordinate system (b coordinate system) is z b Can be The calculation results are:

[0060]

[0061] in, represents the estimated value of the center pixel coordinate of the Snell window; Θ represents the lens model; represents the first conversion coefficient, that is, the conversion coefficient from the c series to the b series; z b Represents the first coordinate vector of the zenith vector in the b coordinate system,

[0062] The horizontal attitude rotation coefficient from the h system to the b system is:

[0063]

[0064] In formula (2), represents the conversion coefficient from h system to b system, θ represents the pitch angle, and γ represents the roll angle. The zenith vector z in h system h=[0 0 1] T , so z b It can also be expressed as:

[0065]

[0066] Therefore, θ and γ can be obtained by the following formula:

[0067]

[0068] The z in the formula b (i) (i=1,2,3) represents the vector z b The ith element of .

[0069] See also Figure 3 , Figure 3 Schematic diagram showing the relationship between Snell window coordinates, Ω i,j is a cone, α0 is the Snell angle, α j is the semi-cone angle of the cone, g i,j,k is the generatrix of the cone, and τ is the step angle of the numerical calculation.

[0070] Optionally, the preset semi-cone angle range is determined by the following method:

[0071] S1121, determining the Snell angle according to the refractive index of the first medium and the refractive index of the second medium;

[0072] S1122. Determine a preset range of semi-cone angles according to the Snell angle and a preset redundant angle.

[0073] The first medium and the second medium are related according to the propagation path of the light. For example, when the light is emitted from air into water, the first medium represents air and the second medium represents water. The redundancy angle represents the magnitude of the change of the Snell angle in a certain direction.

[0074] In a specific embodiment, the Snell angle α0 ray formula is calculated as:

[0075] α0=arcsin(n a / n w ) (5)

[0076] Among them, n a represents the refractive index of air, n w Represents the refractive index of water.

[0077] Optionally, a second coordinate parameter of an incident light ray in the image sensor coordinate system is determined by the following method:

[0078] S1111, determining a third coordinate parameter of the incident light in the carrier coordinate system according to the second conversion coefficient;

[0079] S1112, determining a fourth coordinate parameter of the incident light in the camera coordinate system according to the third coordinate parameter and the first conversion coefficient;

[0080] S1113. Determine the second coordinate parameter according to the fourth coordinate parameter and the inverse lens model.

[0081] The second conversion coefficient represents the conversion parameter from the light coordinate system to the carrier coordinate system, and the inverse lens model represents the optical transmission relationship from the camera coordinate system to the image sensor coordinate system.

[0082] Specifically, first, the third coordinate parameter is calculated based on the incident light and the second conversion coefficient, then, the first conversion coefficient is inversely calculated to obtain the conversion coefficient of the b coordinate system to the c coordinate system, and the fourth coordinate parameter of the incident light in the camera coordinate system is determined based on the third coordinate parameter and the conversion coefficient of the b coordinate system to the c coordinate system, and finally, the second coordinate parameter is determined based on the fourth coordinate parameter and the inverse lens model.

[0083] Optionally, a second coordinate parameter determines a plurality of Hough space values ​​by the following method, including:

[0084] S1123, determining a plurality of half cone angles according to a preset half cone angle range and angle step;

[0085] S1124. Determine a plurality of Hough space values ​​according to the second coordinate parameter and a plurality of half cone angles.

[0086] Specifically, the second coordinate parameter refers to the coordinate of the incident light in the image sensor coordinate system. Specifically, the preset semi-cone angle range is divided according to the angle step to obtain a plurality of semi-cone angles, and statistical calculations are determined according to the second coordinate parameter and the plurality of semi-cone angles to obtain a plurality of Hough space values.

[0087] Optionally, an estimate of the coordinates of the center pixel of the Snell window is determined by:

[0088] S111, determining a second coordinate parameter set of a plurality of incident light rays in an image sensor coordinate system according to an inverse lens model;

[0089] S112, determining a plurality of Hough space values ​​according to the second coordinate parameter set and a preset semi-cone angle range;

[0090] S113. Determine an estimated value of the center pixel coordinate of the Snell window according to a preset number of larger values ​​among a plurality of Hough space values.

[0091] Specifically, first, several incident light rays and inverse lens models are calculated respectively to obtain a second coordinate parameter set of the incident light rays in the image sensor coordinate system, then, several Hough space values ​​are determined according to each second coordinate parameter and a preset semi-cone angle range, and several Hough space values ​​corresponding to the second coordinate parameter are determined, and then, an estimated value of the central pixel coordinate of the Snell window is determined according to a preset number of larger values ​​of the several Hough space values.

[0092] The following is a specific example to illustrate the calculation process of the estimated value of the center pixel coordinate of the Snell window.

[0093] α j The expression is: j =α0-Δα+jδ, (jδ≤2Δα), where Δα and δ are the redundant angle of the Snell angle and the angle step of the numerical calculation respectively. The value range of α is [α0-Δα, α0+Δα].

[0094] Binary image I e The non-zero pixels in are denoted as p i , in the plane pixel coordinate system p i The coordinates are expressed as (u i ,v i ). In the camera position coordinate system c, it corresponds to p i The observation vector v i c , along the generatrix of the Snell window cone (such as Figure 3 Calculated by the lens model: Among them, Θ is the lens operator, which is used to calculate the pixel p i The observation vector v i c , which can be obtained from the OCamCalib toolbox. The expression of the observation vector in the carrier coordinate system b is: in, is a constant matrix. Figure 1 The incident light propagating along the Snell window cone is the generatrix of the cone. For any incident light beam, we establish l i Department, z li is the reverse direction vector of the incident direction of any light. Construct a li The axis is the center, the semi-cone angle is α j The cone Ω i,j , cone Ω i,j The generatrix vector is denoted as The expression of this vector in the carrier coordinate system b can be obtained according to and Get. Among them, τ is the step angle of the numerical calculation. All elements in can be represented by the vector Using the inverse lens model Θ -1 Calculate the incident light along g i,j,k The projection point (i.e. the edge point of the Snell window):

[0095]

[0096] Among them, round(·) is the rounding function. By solving the Hough space H(m,n,α) of the combination of multiple incident light beams, the estimated value of the central pixel coordinate of the Snell window can be obtained. and an estimate of the Snell window angle

[0097] Hough space H(m,n,α) is a multi-layer structure, each α j Each layer contains several elements with horizontal and vertical coordinates m and n respectively. Whenever an element is determined by formula (6), the weight of the element increases by one unit. The weight is recorded as H q (q traverses the entire Hough space). and It is obtained by searching the first N global maxima in H(m,n,α). These N global maxima are denoted as: H1(m1,n1,α1), H2(m2,n2,α2), …, H N (m N ,n N ,α N ). The specific calculation formula is: The value of N is usually set based on experience. When the Snell window distortion is severe, the value of N should be appropriately larger.

[0098] Optionally, the preset quantity is determined by:

[0099] S1131. Acquire image data from an image sensor, and input the image data into a trained neural network model to obtain a preset number.

[0100] It is also difficult to find the optimal value of the preset number N by empirically setting the range of values. The embodiment of the present invention introduces a convolutional neural network (CNN) model. Before the neural network model predicts, it needs to be trained. The training sample data includes image samples of the image sensor and a preset number of samples. For a certain input, the optimal N is obtained through the neural network model prediction. and Calculation.

[0101] It can be understood by those skilled in the art that a group of four gratings with angles between the polarizer transmission axis and the heading of the AUV of 0, π / 4, π / 2, and 3π / 4 are selected, and the results of the simultaneous exposure measurements are used to solve the non-polarized light intensity map of the Snell window. The specific calculation process is: add the four polarized light images and divide by 2. In addition, in order to ensure the reliability of the image data, all the measurement data are subjected to anti-interference processing.

[0102] The experimental process of the embodiment of the present invention is carried out in two steps: the first step is to place a customized underwater camera equipped with a multi-layer metal polarization array into a water pool, collect a large amount of data, and perform simulation tests on the algorithm. The accuracy and reliability of the algorithm are verified by large-scale calculation samples, and the various parameters of the algorithm are optimized; the second step is to carry the camera on an AUV for an underwater actual flight test, and verify the accuracy, calculation speed and reliability of the algorithm on the AUV platform through actual flight. At the same time, all parameters in the algorithm can be further fine-tuned to improve accuracy.

[0103] Generally, when an underwater camera is equipped with a wide-angle lens, the Snell window will appear in the field of view when the field of view is greater than 97.5°. The control signal and data transmission of the underwater camera rely on data line (cable) transmission.

[0104] Implementation of the embodiments of the present invention includes the following beneficial effects: In the embodiment, first, the first coordinate vector of the zenith vector in the carrier coordinate system is determined according to the estimated value of the coordinate of the center pixel of the Snell window, the lens model and the first conversion coefficient. The estimated value of the coordinate of the center pixel of the Snell window is determined according to the propagation evaluation of the incident light, which is closer to the real scene and has strong anti-interference ability. The coordinate vector of the center pixel of the Snell window in the carrier coordinate system is established. Then, the first coordinate function of the zenith vector in the carrier coordinate system is determined according to the horizontal attitude rotation coefficient, and the coordinate function of the zenith vector in the carrier coordinate system is established. Finally, the horizontal attitude parameters are determined according to the first coordinate vector and the first coordinate function, and the amount of calculation is small, thereby reducing the amount of calculation and improving the accuracy and anti-interference ability.

[0105] See also Figure 4 , an embodiment of the present invention provides a horizontal posture extraction system, comprising:

[0106] The first module is used to determine the first coordinate vector of the zenith vector in the carrier coordinate system according to the estimated value of the center pixel coordinate of the Snell window, the lens model and the first conversion coefficient; the estimated value of the center pixel coordinate of the Snell window is determined according to the propagation evaluation of the incident light;

[0107] The second module is used to determine the first coordinate function of the zenith vector in the carrier coordinate system according to the horizontal attitude rotation coefficient;

[0108] The third module is used to determine the horizontal posture parameter according to the first coordinate vector and the first coordinate function.

[0109] It can be seen that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0110] like Figure 5 As shown, an embodiment of the present invention further provides a device for extracting a horizontal posture, comprising:

[0111] at least one processor;

[0112] at least one memory for storing at least one program;

[0113] When at least one program is executed by at least one processor, the at least one processor implements the horizontal posture extraction method steps of the above method embodiment.

[0114] Among them, the memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. The memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a remote memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0115] It can be seen that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0116] On the other hand, an embodiment of the present invention provides a horizontal posture extraction device, including an underwater vehicle and a computer device connected to the underwater vehicle; wherein:

[0117] The underwater vehicle includes a camera, an image sensor and a carrier, and the incident light passes through the camera to form an image on the image sensor;

[0118] Computer equipment includes:

[0119] at least one processor;

[0120] at least one memory for storing at least one program;

[0121] When at least one program is executed by at least one processor, the at least one processor implements the above-mentioned method for extracting horizontal posture.

[0122] The computer device may be different types of electronic devices, including but not limited to desktop computers, laptops and other terminals.

[0123] It can be seen that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0124] The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for extracting horizontal posture, characterized in that: include: Determine a first coordinate vector of the zenith vector in the carrier coordinate system according to an estimated value of the center pixel coordinate of the Snell window, a lens model and a first conversion coefficient; the estimated value of the center pixel coordinate of the Snell window is determined according to a propagation evaluation of the incident light; Determine the first coordinate function of the zenith vector in the carrier coordinate system according to the horizontal attitude rotation coefficient; A horizontal posture parameter is determined according to the first coordinate vector and the first coordinate function.

2. The method according to claim 1, characterized in that The method of determining the first coordinate vector of the zenith vector in the carrier coordinate system according to the estimated value of the center pixel coordinate of the Snell window, the lens model and the first conversion coefficient comprises: Determine a second coordinate vector of the zenith vector in the camera coordinate system according to the estimated value of the coordinate of the center pixel of the Snell window and the lens model; A first coordinate vector of the zenith vector in the carrier coordinate system is determined according to the second coordinate vector and the first conversion coefficient.

3. The method according to claim 2, characterized in that The estimated value of the center pixel coordinate of the Snell window is determined by the following method: Determine a second coordinate parameter set of a plurality of incident light rays in the image sensor coordinate system according to the inverse lens model; Determining a plurality of Hough space values ​​according to the second coordinate parameter set and a preset half cone angle range; An estimated value of the coordinate of the center pixel of the Snell window is determined according to a preset number of larger values ​​among the plurality of Hough space values.

4. The method according to claim 3, characterized in that The second coordinate parameter of an incident light ray in the image sensor coordinate system is determined by the following method: Determine the third coordinate parameter of the incident light in the carrier coordinate system according to the second conversion coefficient; Determine a fourth coordinate parameter of the incident light in the camera coordinate system according to the third coordinate parameter and the first conversion coefficient; The second coordinate parameter is determined according to the fourth coordinate parameter and the inverse lens model.

5. The method according to claim 3, characterized in that: A second coordinate parameter determines several Hough space values ​​by the following methods, including: Determining a plurality of half cone angles according to the preset half cone angle range and angle step; A plurality of Hough space values ​​are determined according to the second coordinate parameter and the plurality of the semi-cone angles.

6. The method according to claim 3, characterized in that The preset semi-cone angle range is determined by the following method: Determining the Snell angle based on the refractive index of the first medium and the refractive index of the second medium; The preset semi-cone angle range is determined according to the Snell angle and the preset redundancy angle.

7. The method according to claim 3, characterized in that The preset number is determined by the following method: Image data from an image sensor is acquired, and the image data is input into a trained neural network model to obtain the preset number.

8. A horizontal posture extraction system, characterized in that: include: A first module is used to determine a first coordinate vector of a zenith vector in a carrier coordinate system according to an estimated value of a central pixel coordinate of a Snell window, a lens model and a first conversion coefficient; the estimated value of the central pixel coordinate of the Snell window is determined according to a propagation evaluation of an incident light ray; The second module is used to determine the first coordinate function of the zenith vector in the carrier coordinate system according to the horizontal attitude rotation coefficient; The third module is used to determine the horizontal posture parameter according to the first coordinate vector and the first coordinate function.

9. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 7 when executed by the processor.

10. A device for extracting horizontal posture, characterized in that: The invention comprises an underwater vehicle and a computer device connected to the underwater vehicle; wherein: The underwater vehicle comprises a camera, an image sensor and a carrier, and the incident light passes through the camera and forms an image on the image sensor; The computer device comprises: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

Images