A zoom vision-based synchronous tracking type pose measurement method for a marine loader

Abstract

The application discloses a zoom-vision-based synchronous tracking type pose measurement method for a marine loader, and steps are sequentially: a target is fixedly arranged on the tail of the marine loader and vertically; an integrated measurement unit is fixedly arranged at the center of the full operation range of the marine loader; the integrated measurement unit is calibrated before operation, a measurement coordinate system is constructed based on the integrated measurement unit, a body coordinate system is constructed based on the target, and an initial value acquisition mode of a pose transformation matrix between the two is set; an adaptive tracking target mechanism is set for the integrated measurement unit, and coordinates of the center of the target in the measurement coordinate system are acquired in real time; in operation, the initial value of the pose transformation matrix is acquired, an optimal solution of the pose transformation matrix is obtained through iterative solution, and the current pose of the marine loader is obtained; the method can quickly complete real-time tracking and pose measurement of the marine loader in the actual operation process of the marine loader, has high measurement precision, is convenient for realizing conversion of measurement results to operation instructions, and improves the efficiency of ship loading operation.

Description

Technical Field

[0001] This invention relates to the field of marine engineering machinery posture measurement technology, and in particular to a synchronous tracking posture measurement method for marine loaders based on zoom vision. Background Technology

[0002] In bulk cargo hub ports, marine loaders are the core equipment for loading operations, mainly undertaking the task of transferring bulk cargo such as coal and ore between the dock and the ship. These machines need to move frequently within the limited dock space, with operating distances dynamically changing between 10m and 50m. Therefore, dynamic measurement of the 6D posture of marine loaders, including three-dimensional coordinates, pitch angle, roll angle, and heading angle, is a crucial prerequisite for achieving automated loading, avoiding collisions, and improving operational efficiency.

[0003] Currently, the commonly used technologies for attitude measurement of marine loaders in the port industry are mainly divided into three categories: GPS positioning technology, inertial measurement technology, and traditional visual measurement technology. GPS positioning technology relies on satellite signals for global positioning, and its accuracy is significantly affected by obstructions. Numerous warehouses, cranes, and other tall buildings and equipment in ports, and even the metal structures of ship decks, can shield and reflect satellite signals, resulting in a typical GPS positioning accuracy of only 1-3 meters in actual operations. This is far from meeting the ±5mm accuracy requirements of loading operations and cannot provide attitude parameters, making it difficult to stably apply to the high-precision operating environment of ports. Inertial measurement technology, by mounting an inertial measurement unit (IMU) on the loader and calculating attitude in real time based on Newton's laws of motion, is unaffected by external environmental interference. However, attitude errors accumulate significantly over time; after 1 hour of continuous operation, the position error exceeds 5 cm and the attitude error exceeds 0.5°, making long-term stable measurement impossible. Furthermore, it lacks an absolute positioning reference, making it difficult to meet the continuous accuracy requirements of loading operations. Traditional visual measurement techniques typically employ fixed-focal-length cameras and calculate pose based on monocular or binocular vision principles, resulting in projection errors. In practical operations, when the working distance of a marine loader exceeds 20m, the feature points in the image captured by the fixed-focal-length camera become blurred, increasing pixel errors to 3-5 pixels and positional errors to over 10cm. Furthermore, the lack of a dynamic tracking mechanism makes it easy to lose the target when the marine loader moves. Additionally, due to the absence of distance information fusion, the accuracy of pose calculation is greatly affected by image quality, making it unsuitable for dynamic port operation scenarios. Therefore, existing commonly used technologies all suffer from the inability to simultaneously achieve accuracy, stability, and environmental adaptability.

[0004] In recent years, the development of zoom vision technology has made long-distance measurement possible, but its application in the field of marine loader pose measurement still faces four major bottlenecks: 1) Traditional single-focal-length calibration methods can only guarantee accuracy at a fixed focal length, and the fitting error of intrinsic parameters across the entire focal length range is too large, failing to meet the requirements of high-precision measurement; 2) Existing EPNP algorithms rely on iterative solutions using randomly generated initial values, resulting in slow convergence speed and significantly affected solution accuracy by feature point matching errors, leading to insufficient fusion of pose calculation and ranging data; 3) The pose measurement results of marine loaders cannot be directly converted into their positional deviation relative to the dock reference point and the ship's cargo hold, making it difficult to convert the measurement results into operational instructions; 4) Existing integrated measurement units do not consider issues such as port wind vibration and turntable tracking response lag (≥100ms), resulting in poor environmental adaptability and tracking stability that cannot meet the requirements of dynamic operations. Therefore, it is necessary to develop a marine loader pose measurement method based on zoom vision technology that meets the needs of dynamic port operations and possesses both high measurement accuracy and high measurement efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a synchronous tracking pose measurement method for marine loaders based on zoom vision, which solves the above-mentioned technical problems.

[0006] Therefore, the technical solution of the present invention is as follows:

[0007] A synchronous tracking pose measurement method for marine loaders based on zoom vision, comprising the following steps:

[0008] S1. A target is vertically fixed at the rear of the marine loader, with multiple LED modules respectively arranged at the center and on the circumferential edge. The LED module at the center is located outside the plane where the multiple LED modules on the circumferential edge are located. The center of the LED light set in the center of the LED module at the center is defined as the center of the target. An integrated measurement unit is fixed at the center of the entire working range of the marine loader to acquire images containing the target in real time and the distance to the LED module at the center of the target.

[0009] Before the S2 is used in marine loader operations, it completes the full-focal-length parameter calibration of the zoom camera, the angle calibration of the two-dimensional turntable, and the consistency calibration of the laser beam and the camera optical axis of the laser rangefinder. It constructs a measurement coordinate system based on the integrated measurement unit, a fuselage coordinate system based on the target, defines common feature points on the target, and sets the rotation matrix between the measurement coordinate system and the fuselage coordinate system. Translation vector Methods for obtaining initial values;

[0010] S3. An adaptive target tracking mechanism is set for the integrated measurement unit to ensure that the imaging center of the zoom camera is always aligned with the target center range and to obtain the target center coordinates in the measurement coordinate system in real time.

[0011] S4. During the operation of the marine loader, the coordinates of the target center in the measurement coordinate system are obtained in real time, and the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is calculated. The optimal solution of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained by using the iterative solution method, so as to obtain the accurate measurement result of the target's current pose, and use it as the current pose of the marine loader.

[0012] Further, in step S1, the target includes a main frame, corner LED modules, and a central LED module; the main frame includes a regular N-sided frame, which is formed by N connecting beams connected end to end in sequence, and N / 2 support beams are arranged radially inside the frame. One end of each support beam is fixedly connected and located at the center of the main frame, and the other end of each support beam is vertically fixed to the middle of the side wall of the N / 2 connecting beams located at intervals; there are N corner LED modules, which are fixed in the middle of the N connecting beams; there is one central LED module, which is fixed in the center of the main frame by a support frame, so that the central LED module is arranged outside the plane where the N corner LED modules are located; wherein, the value of N is ≥6 and is an even number; the LED lights on each LED module are centrally located; the integrated measurement unit includes a two-dimensional turntable, a zoom camera, and a laser rangefinder arranged in sequence from bottom to top; the lens end of the zoom camera and the laser emission end of the laser rangefinder are arranged in the same direction.

[0013] Furthermore, in step S2, the parameter calibration method for the zoom camera across the entire focal length is as follows:

[0014] Using a checkerboard calibration plate, several images of the zoom camera were acquired within the full focal length range of 10mm-200mm, and the intrinsic parameters of the zoom camera at each focal length were obtained.

[0015] Constructing with focal length For a mapping model where the independent variable is denoted by and the internal parameter is denoted by , its expression is:

[0016] ,

[0017] ,

[0018] ,

[0019] ,

[0020] ,

[0021] ,

[0022] ,

[0023] ,

[0024] In the formula, ~ Focal length in the x direction The fitting coefficients in the expression, ~ Focal length in the y direction The fitting coefficients in the expression, ~ x-axis coordinates of the principal point The fitting coefficients in the expression, ~ y-axis coordinates of the principal point The fitting coefficients in the expression, ~ The first radial distortion coefficient The fitting coefficients in the expression, ~ The second radial distortion coefficient The fitting coefficients in the expression, ~ The first tangential distortion coefficient The fitting coefficients in the expression, ~ The second tangential distortion coefficient The fitting coefficients in the expression;

[0025] Substituting the intrinsic parameters of the zoom camera at various focal lengths into the mapping model, the fitting coefficients in each expression of the mapping model are obtained by solving the least squares method, so as to acquire the focal length based on the real-time images of the zoom camera. The intrinsic parameters of the zoom camera are obtained using a mapping model.

[0026] Constructing a real-time image acquisition focal length based on zoom camera An image distortion correction model is used to obtain a distortion-corrected image; wherein, the focal length for real-time image acquisition by the zoom camera is... For any pixel coordinate on the image The distortion correction expression is:

[0027] ,

[0028] ,

[0029] In the formula, This represents the x-axis coordinates after pixel distortion correction. The y-axis coordinates after pixel distortion correction; Radial distance, ; Based on focal length The first radial distortion coefficient, Based on focal length The second radial distortion coefficient; Based on focal length The first tangential distortion coefficient, Based on focal length The second tangential distortion coefficient.

[0030] Furthermore, in step S2, the method for calibrating the angle of the two-dimensional turntable is as follows:

[0031] A laser collimator is fixed to the stage of a two-dimensional turntable, and its emitted laser is projected onto the center of the target on the opposite side. Angle encoders are installed on the azimuth and pitch axes of the two-dimensional turntable to obtain the actual rotation angle of the azimuth axis of the two-dimensional turntable. and the actual rotation angle of the pitch axis ;

[0032] The azimuth angle of the two-dimensional turntable was sequentially adjusted from 0° to 360°, and the elevation angle was adjusted from -45° to 90°. The target azimuth axis angle was collected under different azimuth and elevation angle settings. Pitch axis target command angle Actual rotation angle of the azimuth axis and the actual rotation angle of the pitch axis To calculate the test values ​​of azimuth axis rotation deviation respectively and pitch axis rotation deviation test value The expressions for both are:

[0033] , ;

[0034] Construct an angle correction model, the expression of which is:

[0035] , ,

[0036] In the formula, This is the predicted value of the azimuth axis rotation deviation. and These are the azimuth correction factor and the azimuth correction intercept, respectively. This is the predicted value of the pitch axis rotation deviation. and These are the pitch axis correction factor and the pitch axis correction intercept, respectively.

[0037] Test values ​​of azimuth axis rotation deviation obtained under different azimuth and elevation angle settings and pitch axis rotation deviation test value These are respectively used as predicted values ​​of azimuth axis rotation deviation. Predicted values ​​of pitch axis rotation deviation and the corresponding azimuth axis target command angle and pitch axis target command angle Substituting these values ​​into the angle correction model, the azimuth axis correction coefficient and azimuth axis correction intercept, as well as the pitch axis correction coefficient and pitch axis correction intercept, are calculated using a linear fitting method. Furthermore, given the target command angles of the azimuth axis and pitch axis, the actual rotation angles of the two-dimensional turntable's azimuth axis and pitch axis are obtained through the angle correction model.

[0038] Further, in step S2, the laser rangefinder performs the consistency calibration of the laser beam and the camera optical axis as follows: using a collimator with a focal length of 500mm as the optical axis calibration reference, aligning the zoom camera with the outlet of the collimator, and adjusting the camera posture so that its imaging center coincides with the cross center of the collimator target plate; while keeping the zoom camera stationary, adjusting the installation posture of the laser rangefinder so that the laser spot is also projected onto the cross center position of the collimator target plate.

[0039] Furthermore, in step S2, the method also constructs a laser ranging vibration compensation model to compensate for the vibration of the laser ranging value of the laser rangefinder, and replaces the original laser ranging value with the model. The expression for this model is:

[0040] ,

[0041] In the formula, This is the laser ranging compensation value. For vibration acceleration, This is the distance coefficient. This is the original value for laser ranging.

[0042] Further, in step S2,

[0043] The measurement coordinate system is constructed with the rotation center of the two-dimensional turntable of the integrated measurement unit as the origin, the X-axis parallel to the camera optical axis, and the Y and Z axes satisfying the right-hand rectangular coordinate system; the body coordinate system is constructed with the geometric center of the target's main frame as the origin, the axis direction of the LED module at the center as the z-axis, and the x and y axes located on the plane of the main frame.

[0044] Define common feature points on the target, which include: the central common feature point located at the center of the LED light point of the central LED module; and multiple circumferential common feature points located at the centers of the LED light points of multiple LED modules in the circumferential direction.

[0045] The transformation relationship between the common feature points in the measurement coordinate system and the fuselage coordinate system is constructed, and its expression is as follows:

[0046] ,

[0047] In the formula, Let i be the coordinates of the i-th common feature point in the measurement coordinate system; It is a rotation matrix; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; It is a translation vector;

[0048] By obtaining the coordinates of all common feature points in both the measurement coordinate system and the fuselage coordinate system, the rotation matrix can be solved using the singular value decomposition method. Translation vector .

[0049] Furthermore, the specific implementation steps of step S3 are as follows:

[0050] S301. Before the loader operates, with the imaging center of the zoom camera aligned with the center of the target, acquire an initial frame containing an image of the target, and extract all common feature points and their initial pixel coordinates from the image.

[0051] S302. After the loader starts working, it acquires an image containing the target again, and uses the FLANN matching algorithm to extract all common feature points in the newly acquired image containing the target and obtain their real-time pixel coordinates.

[0052] S303. Calculate the geometric center pixel coordinates of all common feature points in the initial frame image containing the target, and use them as the tracking reference point coordinates; calculate the geometric center pixel coordinates of all common feature points in the newly acquired image containing the target, and obtain their pixel deviations in the x-direction based on the tracking reference point coordinates. Pixel deviation in the y direction Determine whether the deviation between the two pixels is ≤5 pixels. If the result is yes, keep the current state of the two-dimensional turntable unchanged and proceed directly to step S306. If the result is no, proceed to step S304 and adjust the current state of the two-dimensional turntable.

[0053] S304. Convert the two pixel deviations obtained in step S303 into azimuth axis angle adjustment amounts. and pitch axis angle adjustment Its expression is:

[0054] , ,

[0055] In the formula, The first proportionality coefficient, This is the second proportionality coefficient; To fit the focal length, , and These are the x-direction focal length and y-direction focal length of the zoom camera when acquiring the initial frame containing the target image;

[0056] S305. Calculate the target angle of the azimuth axis at the current time t to accurately adjust the state of the two-dimensional turntable. and pitch axis target command angle And complete the real-time adjustment of the two-dimensional turntable status;

[0057] S306. Update the horizontal angle of the two-dimensional turntable at the current time t. and pitch angle In order to obtain the central common feature point, that is, the real-time coordinates of the target center in the measurement coordinate system ( , , Its calculation expression is:

[0058] ,

[0059] ,

[0060] ,

[0061] In the formula, This refers to the distance from the integrated measurement unit to the LED module at the center of the target;

[0062] S307. During the operation of the loader, the zoom camera continuously acquires images containing the target at the same time intervals, and processes the newly acquired images containing the target and the images containing the target acquired in the previous frame in the same way as steps S301-S306, so as to continuously adjust the state of the two-dimensional turntable in real time.

[0063] Furthermore, in step S4, the specific method for obtaining the optimal solution of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is as follows:

[0064] S401. Constructing the objective function Its expression is:

[0065] ,

[0066] In the formula, The number of common feature points; This is the intrinsic parameter matrix for a zoom camera; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; For the first The pose transformation matrix between the fuselage coordinate system and the measurement coordinate system in the next iteration; π(·) is the camera reprojection function; Let be the pixel coordinates of the i-th common feature point in the distortion-corrected image; The coordinates of the central common feature point in the fuselage coordinate system; The coordinates of the central common feature point in the measurement coordinate system; λ is the weighting coefficient;

[0067] S402. Based on the target center coordinates obtained in real time in step S3 in the measurement coordinate system, the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained.

[0068] S403. During the iteration process, the least squares method with Levenberg-Marquardt (LM) damping is used to update the rotation matrix and translation vector.

[0069] S404. Set the iteration termination condition as follows: the objective function value converges to the minimum, and the deviation between the coordinates of the central common feature point calculated by the pose transformation matrix in the current iteration and the calculation result in the previous iteration is ≤0.001m; at the same time, the maximum number of iterations is set to 20.

[0070] S405. Based on the settings of steps S401-S404 above, substitute the initial value of the pose transformation matrix at the current moment into the objective function, and with the goal of minimizing the objective function, obtain the pose transformation matrix that satisfies the iteration termination condition through iterative solution, and use it as the optimal solution of the pose transformation matrix at the current moment. .

[0071] Furthermore, in step S4, the method for obtaining the accurate measurement result of the target's current pose is as follows:

[0072] The optimal solution of the pose transformation matrix at the current moment Convert to rotation matrix Translation vector By substituting these values ​​into the target pose output expression, the coordinates of the central common feature point in the measurement coordinate system can be obtained. Pitch angle Roll angle Heading angle The accurate measurement results of the current pose of the target are as follows:

[0073] ,

[0074] ,

[0075] ,

[0076]

[0077] In the formula, The pitch angle, This is the roll angle. For heading angle, Representing the rotation matrix The element in the 3rd row and 1st column, Rotation matrix The element in the 3rd row and 2nd column, Rotation matrix The element in the 3rd row and 3rd column, Rotation matrix The element in the 2nd row and 1st column, Rotation matrix The element in the first row and first column of the middle, The coordinates of the central common feature point in the measurement coordinate system. The coordinates of the central common feature point in the machine base coordinate system.

[0078] Compared with existing technologies, this zoom vision-based synchronous tracking pose measurement method for marine loaders utilizes a target placed at the rear of the marine loader and an integrated measurement unit at the center of the loader's entire operating range. Through pre-calibration of the acquisition equipment, an adaptive tracking target mechanism for the integrated measurement unit, and a rapid iterative solution method using the real-time target position obtained through the adaptive tracking target mechanism as prior information, this method enables rapid real-time tracking and pose measurement of the marine loader during actual operation. It meets the real-time and stability requirements of the dock's pose measurement operations. Furthermore, this method offers high pose measurement accuracy and facilitates the construction of a local observation field navigation system, enabling the conversion of measurement results into operational instructions. It overcomes the shortcomings of traditional measurement methods that cannot guide actual operations, improving loading efficiency by over 30% and providing a scalable solution for dock automation upgrades. Attached Figure Description

[0079] Figure 1 This is a schematic diagram illustrating the setting of the target and integrated measurement unit in the synchronous tracking pose measurement method for marine loaders based on zoom vision in an embodiment of the present invention.

[0080] Figure 2 This is a schematic diagram of the target structure in the synchronous tracking pose measurement method for marine loaders based on zoom vision in an embodiment of the present invention.

[0081] Figure 3 This is a schematic diagram of the integrated measurement unit in the synchronous tracking pose measurement method for marine loaders based on zoom vision in an embodiment of the present invention.

[0082] Figure 4This is a flowchart of a synchronous tracking pose measurement method for marine loaders based on zoom vision, as described in an embodiment of the present invention.

[0083] Figure 5 This is a comparison diagram of the effects of the iterative algorithm in step S4 of the synchronous tracking pose measurement method for marine loaders based on zoom vision in this embodiment of the invention and the traditional EPnP algorithm.

[0084] Figure 6 This is a schematic diagram illustrating an application scenario of the synchronous tracking pose measurement method for marine loaders based on zoom vision, as described in an embodiment of the present invention. Detailed Implementation

[0085] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the following embodiments are by no means intended to limit the present invention.

[0086] See Figure 1 and Figure 4 The specific implementation steps of the synchronous tracking pose measurement method for marine loaders based on zoom vision are described below.

[0087] S1. A target 2 is vertically fixed at the rear of the marine loader 1. Seven LED modules are respectively arranged at the center and on the circumferential edge, and the LED module at the center is arranged outside the plane where the six LED modules on the circumferential side are located. An integrated measurement unit 3 is fixed at the center of the entire working range of the marine loader 1 to acquire images containing the target in real time and the distance from the target to the LED module at the center of the target 2.

[0088] See Figure 2 The target 2 includes a main frame 201 with a regular hexagonal structure, six corner LED modules 202 and a central LED module 203.

[0089] The main frame 201 consists of a regular hexagonal frame and three internal support beams. The regular hexagonal frame is composed of six connecting beams connected end to end. The three internal support beams are arranged radially within the regular hexagonal frame. One end of each beam is fixedly connected and located at the center of the main frame 201, while the other end is vertically fixed to the middle of the three connecting beams located at intervals. This makes the main frame 201 not only have high structural strength and good stability, but also its fully symmetrical structure is conducive to the average transmission and dissipation of vibration energy, so as to adapt to the high-intensity bumps and vibrations during the operation of marine loaders.

[0090] The six corner LED modules 202, i.e., the six LED modules arranged around the target 2, are centrally fixed at the middle of the N connecting beams and located on the outer surface of the target 2. The central LED module 203, i.e., the LED module arranged at the center of the target 2, is fixed to the center of the main frame 201 (i.e., the fixed connection point of the three inner support beams) by the support frame 204 and is also located on the outer surface of the target 2, so that the central LED module 203 protrudes from the arrangement of the six corner LED modules 202, that is, the central LED module is arranged outside the plane where the N corner LED modules are located. The above arrangement of the seven LED modules ensures that the image containing the target acquired by the integrated measurement unit 3 contains LED light spots emitted by the seven LED modules, and meets the requirements of all-weather image acquisition and processing based on the light emission characteristics. On the other hand, the seven LED modules located on different planes facilitate the construction of a three-dimensional coordinate system based on the target.

[0091] In practical applications, multiple fixing plates 205 with screw holes are evenly distributed along the circumferential direction on the outer side of the main frame 201 of the target 2, so as to fix the target 2 to the vehicle body shell at the rear of the marine loader 1 in a vertical manner; the six circumferential LED modules 202 and the central LED module 203 on the target 2 are located on the outside of the target, and their LED lights are set towards the integrated measurement unit 3; the LED lights on each LED module are centered, so that in the image containing the target 2, the seven LED light spots correspond to the center position of the LED module, and these seven LED light spots are also used as seven common feature points in subsequent image processing; the center point of the LED light on the central LED module 203 is defined as the target center.

[0092] See Figure 3 The integrated measurement unit 3 includes a two-dimensional turntable 301, a zoom camera 302, and a laser rangefinder 303 arranged sequentially from bottom to top. The lens of the zoom camera 302 and the laser emitter of the laser rangefinder 303 are aligned in the same direction, and both are fitted with corrosion-resistant protective covers 304 to adapt to the harsh environment of high salt spray and high humidity in ports. In this integrated measurement unit 3, the zoom camera 302 is used to acquire images containing the target, and the laser rangefinder 303 measures the distance to the target center in real time by emitting a laser beam towards the target center. The two-dimensional turntable 301 has an azimuth axis and a pitch axis, which are used to control the horizontal and pitch rotation of the platform of the two-dimensional turntable 301, respectively, to synchronously adjust the attitude of the zoom camera 302 and the laser rangefinder 303 above it. This enables the integrated measurement unit 3 to track the target 2 on the marine loader 1 in real time, continuously acquiring images and distances.

[0093] S2. Before the marine loader operates, complete the full-focal-length parameter calibration of the zoom camera, the angle calibration of the two-dimensional turntable, and the consistency calibration of the laser beam and the camera optical axis of the laser rangefinder; construct a measurement coordinate system based on the integrated measurement unit, construct a fuselage coordinate system based on the target, define common feature points on the target, and set the rotation matrix between the measurement coordinate system and the fuselage coordinate system. Translation vector The method for obtaining initial values.

[0094] The specific implementation steps of step S2 are described below.

[0095] S201. Using a checkerboard calibration board, acquire images of the zoom camera within its full focal length range of 10mm-200mm to complete the parameter calibration of the zoom camera across its entire focal length. The specific calibration steps are as follows:

[0096] 1) A precision lifting platform is set on the opposite side of the zoom camera, and a checkerboard calibration plate is fixed on the precision lifting platform; the checkerboard calibration plate is set vertically, with its center aligned with the zoom camera lens, and the distance between the two is 10m; in this embodiment, the checkerboard calibration plate is a square plate with a side length of 300mm, and the side length of each square on it is 30mm.

[0097] 2) Initially set the focal length of the zoom camera to 10mm, and use a two-dimensional turntable to adjust the zoom camera's pitch angle to 0°, 15°, 30°, 45°, and 60° in sequence so that the zoom camera is set in five different postures in sequence; capture five images in each of the five different postures of the zoom camera, while ensuring that the checkerboard calibration plate fills the field of view.

[0098] 3) Adjust the focal length of the zoom camera from 10mm to 200mm in 5mm increments, and perform the same operation as in step 2) after each focal length adjustment; then, after the above steps 2)-3), a total of 975 images are acquired (39 focal lengths × 5 poses × 5 images / pose).

[0099] 4) Import all images acquired in steps 2)-3) into MATLAB software. First, use the Shi-Tomasi corner detection algorithm to extract the interior corners and their coordinates of the checkerboard pattern in each image. Then, use the least squares method to solve for the intrinsic parameters of the zoom camera at each focal length. The intrinsic parameters are specifically determined by the intrinsic parameter matrix. It consists of distortion coefficients; where the intrinsic parameter matrix is... , The focal length is in the x-direction. The focal length is in the y-direction. The x-axis coordinate of the principal point The y-axis coordinate of the principal point; distortion coefficients include: the first radial distortion coefficient. Second radial distortion coefficient First tangential distortion coefficient Second tangential distortion coefficient ;

[0100] 5) Construct a focal length-based system For a mapping model where the independent variable is denoted by and the internal parameter is denoted by , its expression is:

[0101] ,

[0102] ,

[0103] ,

[0104] ,

[0105] ,

[0106] ,

[0107] ,

[0108] ,

[0109] In the formula, ~ Focal length in the x direction The fitting coefficients in the expression, ~ Focal length in the y direction The fitting coefficients in the expression, ~ x-axis coordinates of the principal point The fitting coefficients in the expression, ~ y-axis coordinates of the principal point The fitting coefficients in the expression, ~ The first radial distortion coefficient The fitting coefficients in the expression, ~ The second radial distortion coefficient The fitting coefficients in the expression, ~ The first tangential distortion coefficient The fitting coefficients in the expression, ~ The second tangential distortion coefficient The fitting coefficients in the expression;

[0110] Substitute the intrinsic parameters of the zoom camera at each focal length obtained in step 4) into the mapping model, and solve for the fitting coefficients in each expression of the mapping model using the least squares method; then, given the known real-time image acquisition focal length of the zoom camera... In this case, the intrinsic parameters of the zoom camera can be directly obtained using the mapping model;

[0111] 6) Construct a real-time image acquisition focal length based on a zoom camera An image distortion correction model is used to obtain a distortion-corrected image; wherein, the focal length for real-time image acquisition by the zoom camera is... For any pixel coordinate on the image The distortion correction expression is:

[0112] ,

[0113] ,

[0114] In the formula, This represents the x-axis coordinates after pixel distortion correction. The y-axis coordinates after pixel distortion correction; Radial distance, ; Based on focal length The first radial distortion coefficient, Based on focal length The second radial distortion coefficient; Based on focal length The first tangential distortion coefficient, Based on focal length The second tangential distortion coefficient.

[0115] It has been verified that in step S201, the goodness of fit R of the zoom camera intrinsic parameter solution method in step 5) above is [missing information]. 2 ≥0.999, goodness of fit R compared to traditional linear fitting 2 Compared to ≤0.95, the fitting error is reduced by more than 85%; and compared to the commonly used simplified radial distortion correction model, the image distortion correction model in step 6) above improves the correction accuracy by 25%, with a pixel-level error of ≤0.1 pixels; in addition, in terms of accuracy, when a verification target is placed at a distance of 20m from the zoom camera and the intrinsic parameter calibration error at different focal lengths is measured using a laser interferometer, it can be tested that the intrinsic parameter repeatability measurement error of the zoom camera is ≤0.01 pixels across the entire focal length range, confirming that the intrinsic parameter calibration method of the zoom camera across the entire focal length meets the requirements for high-precision measurement.

[0116] S202. A high-precision angle encoder and a laser collimator are used in conjunction to calibrate the deviation between the two-dimensional turntable angle command value and the actual output value, and a linear correction model is constructed to compensate for the backlash error of the mechanical transmission. The specific calibration steps are as follows:

[0117] 1) Without installing the zoom camera and laser rangefinder, fix the laser collimator on the platform of the 2D turntable, and project the laser emitted by the laser collimator onto the center of target 2, which is located 50m away on the opposite side; at the same time, install angle encoders (accuracy 0.001°) on the azimuth and pitch axes of the 2D turntable to obtain the actual rotation angle of the azimuth axis of the 2D turntable. and the actual rotation angle of the pitch axis ;

[0118] 2) Initially set the azimuth angle of the 2D turntable to 0°, and then adjust the pitch angle of the 2D turntable sequentially from -45° to 90° at 15° intervals. At each pitch angle, collect the target command angle of the azimuth axis of the 2D turntable. Pitch axis target command angle Actual rotation angle of the azimuth axis and the actual rotation angle of the pitch axis To calculate the test values ​​of azimuth axis rotation deviation respectively and pitch axis rotation deviation test value The expressions for both are:

[0119] ,

[0120] ;

[0121] 3) Adjust the azimuth angle of the two-dimensional turntable from 0° to 360° in 10° intervals, and perform the same operation as in step 2) after each azimuth angle adjustment;

[0122] Furthermore, through steps 2) to 3) above, a series of angle deviations under different azimuth and pitch angle conditions are obtained; wherein, under each specified pitch and azimuth angle condition, it is preferable to repeat the test through multiple sets (e.g., 5 sets), and calculate the rotation deviation test value from the collected data and take the average.

[0123] 4) Construct the angle correction model, whose expression is:

[0124] ,

[0125] ,

[0126] In the formula, This is the predicted value of the azimuth axis rotation deviation. and These are the azimuth correction factor and the azimuth correction intercept, respectively. This is the predicted value of the pitch axis rotation deviation. and These are the pitch axis correction factor and the pitch axis correction intercept, respectively.

[0127] The azimuth target command angles obtained from steps 2) and 3) under different azimuth and elevation angle settings will be... Pitch axis target command angle Azimuth axis rotation deviation test value and pitch axis rotation deviation test values Substitute these values ​​into the angle correction model and use linear fitting to calculate the azimuth axis correction coefficients. and azimuth axis corrected intercept and pitch axis correction factor and pitch axis correction intercept .

[0128] In practical applications, by using the angle correction model, the target command angle can be determined based on the known azimuth axis. and pitch axis target command angle At that time, the actual rotation angle of the azimuth axis of the two-dimensional turntable is obtained. and the actual rotation angle of the pitch axis .

[0129] Verification showed that the angle correction model constructed in step S202 can improve the accuracy of the corrected angle measurement error to ≤0.001°, which is 10 times higher than the uncorrected angle measurement error of ≤0.01°.

[0130] S203. The laser rangefinder undergoes a coaxial calibration of its laser beam and the camera's optical axis. This ensures that, in practical use, when the zoom camera's imaging center is aligned with the target center, the laser spot emitted by the laser rangefinder also falls on the target center. The specific calibration steps are as follows: Use a collimator with a focal length of 500mm to provide an infinity target as a calibration reference for the optical axis; align the zoom camera with the collimator's output port and adjust the camera's orientation so that its imaging center coincides with the crosshair center of the collimator's target plate, thus determining the zoom camera's optical axis; while keeping the zoom camera stationary, adjust the laser rangefinder's mounting orientation so that the laser spot is projected onto the crosshair center of the collimator's target plate, completing the coaxial calibration of the laser rangefinder's optical axis and the zoom camera's optical axis. Therefore, in subsequent use, when the zoom camera is aligned with the target center, it means that the laser rangefinder is also aligned with the target center.

[0131] After the laser beam and camera optical axis are calibrated in step S203, the coaxiality deviation between the laser beam and the camera optical axis is ≤0.1° within the measurement range of 10m-50m, and the corresponding error between the ranging data and the image data is ≤1 pixel.

[0132] As a preferred technical solution in this embodiment, considering that the target vibrates during the operation of the marine loader, which can cause vibration errors in the laser ranging value of the laser rangefinder, a laser ranging vibration compensation model can be constructed to compensate for the vibration of the laser ranging value in practical applications and replace the original laser ranging value. Specifically, the expression of the laser ranging vibration compensation model is as follows:

[0133] ,

[0134] In the formula, This is the laser ranging compensation value. For vibration acceleration, This is the distance coefficient. This is the original value for laser ranging.

[0135] S204. Construct a measurement coordinate system based on the integrated measurement unit and a fuselage coordinate system based on the target; define common feature points on the target, and use a calibration method based on multiple common feature points to set the rotation matrix between the measurement coordinate system and the fuselage coordinate system. Translation vector The method for obtaining initial values.

[0136] The specific implementation steps of step S204 are as follows:

[0137] 1) Construct a measurement coordinate system with the origin O at the rotation center of the two-dimensional turntable of the integrated measurement unit 3, the X-axis parallel to the camera optical axis, and the Y-axis and Z-axis satisfying the right-hand rectangular coordinate system;

[0138] 2) Construct a fuselage coordinate system with the geometric center of the main frame 201 of the target 2 as the origin o, the axis of the central LED module 203 (that is, the direction perpendicular to the plane of the main frame 201) as the z-axis, and the x-axis and y-axis located on the plane of the main frame 201.

[0139] 3) Define the center of the LED light point of the central LED module on target 2 (i.e., the target center) as the central common feature point, and the centers of the LED light points of the six circumferential LED modules as circumferential common feature points, for a total of seven common feature points; the coordinates of these seven common feature points in the fuselage coordinate system are represented as follows: (x i , y i , z i), i∈[1,7], which can be directly obtained through the pre-calibration of the target; the coordinates of these seven common feature points in the measurement coordinate system are represented as (X i ,Y i Z i ), i∈[1,7], which can be obtained by real-time ranging of the target center by the laser rangefinder and combined with the real-time azimuth and pitch angle of the two-dimensional turntable;

[0140] 4) Construct the transformation relationship between the common feature points in the measurement coordinate system and the fuselage coordinate system, expressed as:

[0141] ,

[0142] in, Let i be the coordinates of the i-th common feature point in the measurement coordinate system; It is a rotation matrix, specifically a 3×3 rotation matrix; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; It is a translation vector, specifically a 3×1 translation vector;

[0143] Furthermore, based on the coordinates of the seven common feature points in the fuselage coordinate system... and coordinates in the measurement coordinate system The rotation matrix can be solved using the Singular Value Decomposition (SVD) method. Translation vector .

[0144] S3. An adaptive target tracking mechanism is set for the integrated measurement unit to ensure that the imaging center of the zoom camera is always aligned with the target center and to obtain the coordinates of the target center in the measurement coordinate system in real time.

[0145] Specifically, the operation steps of step S3 are as follows:

[0146] S301. Before the marine loader operates, with the imaging center of the zoom camera aligned with the center of the target, the zoom camera acquires an initial frame containing an image of the target; by extracting seven common feature points from the image, the initial pixel coordinates of the common feature points are obtained.

[0147] S302. After the marine loader starts working, the zoom camera will re-acquire an image containing the target and use the FLANN matching algorithm to match the newly acquired image containing the target with the image containing the target in the initial frame in order to extract seven common feature points in the newly acquired image containing the target and obtain the real-time pixel coordinates of the seven common feature points.

[0148] S303. Calculate the geometric center pixel coordinates of the seven common feature points in the initial frame image containing the target. The coordinates of the target are used as the tracking reference point; the geometric center pixel coordinates of the seven common feature points in the newly acquired image containing the target are calculated. And based on the coordinates of the tracking reference point, the pixel deviation in the x-direction is obtained. Pixel deviation in the y direction Determine whether the deviation between the two pixels is ≤5 pixels. If the result is yes, keep the current state of the two-dimensional turntable unchanged and proceed directly to step S306. If the result is no, proceed to step S304 and adjust the current state of the two-dimensional turntable.

[0149] S304. Convert the two pixel deviations obtained in step S303 into azimuth axis angle adjustment amounts. and pitch axis angle adjustment Its expression is:

[0150] ,

[0151] ,

[0152] In the formula, The first proportionality coefficient, This is the second proportionality coefficient. ; To fit the focal length, , and These are the x-direction focal length and y-direction focal length of the zoom camera when acquiring the initial frame containing the target image, respectively. Specifically, they can be calculated using the mapping model constructed in step S201.

[0153] S305. Adjust the azimuth axis angle according to the amount obtained in step S304. and pitch axis angle adjustment Based on the angle correction model obtained in step S202, the azimuth axis target command angle for accurately adjusting the state of the two-dimensional turntable at the current time t is calculated. and pitch axis target command angle And complete the real-time adjustment of the two-dimensional turntable status;

[0154] Specifically, at the current time t, the azimuth axis target angle for precisely adjusting the state of the two-dimensional turntable is achieved. and pitch axis target command angle The calculation expression is:

[0155] ,

[0156] .

[0157] S306. Update the horizontal angle of the two-dimensional turntable at the current time t. and pitch angle To obtain the central common feature points That is, the real-time coordinates of the target center in the measurement coordinate system. , , Its calculation expression is:

[0158] ,

[0159] ,

[0160] ,

[0161] In the formula, To measure the distance from the integrated measurement unit to the target center, the specific value is the original laser ranging value obtained from the laser rangefinder in step S2, or the laser ranging compensation value obtained after vibration compensation.

[0162] S307. During the operation of the marine loader, the zoom camera continuously acquires images containing the target at the same time intervals, and processes each newly acquired image containing the target and the image containing the target acquired at the previous moment in the same way as steps S301-S306, so as to continuously adjust the state of the two-dimensional turntable in real time.

[0163] The tracking mechanism and target center real-time coordinate acquisition method in step S3 solve the problem of scale uncertainty in monocular vision. It has been verified that the tracking mechanism achieves a target center measurement error of ≤2mm in the measurement coordinates, providing accurate distance prior information for the optimization algorithm in the subsequent step S4.

[0164] S4. During the operation of the marine loader, the coordinates of the target center in the measurement coordinate system are obtained in real time, and the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is calculated. The optimal solution of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained by using the iterative solution method, so as to obtain the accurate measurement result of the target's current pose, and use it as the current pose of the marine loader.

[0165] The specific implementation steps of step S4 are described below.

[0166] S401. Constructing the objective function Its expression is:

[0167] ,

[0168] In the formula, In this embodiment, the number of common feature points is... ; This is the intrinsic parameter matrix of the zoom camera, which is calculated by the method in step S201 based on the focal length of the image acquisition at the current moment; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; For the first The pose transformation matrix between the fuselage coordinate system and the measurement coordinate system in the next iteration, corresponding to... This represents the coordinates of the i-th common feature point in the measurement coordinate system; π(·) is the camera reprojection function, which is used to project the coordinates of the i-th common feature point in the measurement coordinate system into pixel coordinates in the image coordinate system; Let be the pixel coordinates of the i-th common feature point in the distortion-corrected image; The coordinates of the central common feature point in the fuselage coordinate system; The coordinates of the central common feature point in the measurement coordinate system are λ; λ is the weighting coefficient, with a value range of 0.7-0.9, used to balance the influence of reprojection error and distance prior constraint. In this embodiment, λ=0.8.

[0169] The objective function constructed in this step middle, The reprojection error term is used to construct an error constraint between the coordinates of common feature points in the fuselage coordinate system after pose transformation and camera projection, and the actual pixel coordinates extracted in the image. By minimizing this error constraint, the optimal solution for the pose of the marine loader can be achieved. As a constraint term, its function is to limit the central spatial position of the target, thereby reducing the impact of common feature point matching errors on the solution results from the source.

[0170] S402. To avoid the slow convergence problem caused by random initial values ​​in traditional algorithms, the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained based on the target center coordinates obtained in real time in step S3. The specific method for obtaining this value is as follows:

[0171] 1) Since the relative positions of the central common feature point and the six circumferential common feature points are fixed, the coordinates of the six circumferential common feature points in the measurement coordinate system can be calculated based on the coordinates of the target center (central common feature point) in the measurement coordinate system at the current moment.

[0172] 2) Substitute the coordinates of the seven common feature points in the base coordinate system, and the real-time coordinates of the seven common feature points in the measurement coordinate system determined in step 1), into the transformation relationship between the measurement coordinate system and the fuselage coordinate system determined in step S204, and the rotation matrix between the measurement coordinate system and the fuselage coordinate system at the current moment can be calculated. Translation vector The initial value of the pose transformation matrix is ​​then formed. :

[0173] ;

[0174] S403. During the iteration process, the least squares method with Levenberg-Marquardt (LM) damping is used to update the rotation matrix and translation vector. This is to achieve a balance between local convergence accuracy and global numerical stability by dynamically adjusting the damping factor, and to successively correct the pose parameters to reduce the error of the objective function.

[0175] S404. Set the iteration termination condition as follows: the objective function value converges to the minimum, and the deviation between the coordinates of the central common feature point calculated by the pose transformation matrix in the current iteration and the calculation result in the previous iteration is ≤0.001m; at the same time, the maximum number of iterations is set to 20.

[0176] S405. Based on the settings in steps S401-S404 above, set the initial value of the pose transformation matrix at the current moment. Substituting the objective function and minimizing it, we obtain the pose transformation matrix that satisfies the iteration termination condition, which is then used as the optimal solution for the pose transformation matrix at the current time step. ;

[0177] S406. Obtain the optimal solution of the pose transformation matrix at the current time obtained from step S405. Convert to rotation matrix Translation vector Substitute these values ​​into the target pose output expression to obtain the coordinates of the central common feature point in the measurement coordinate system. Pitch angle Roll angle Heading angle The accurate measurement results of the current pose of the target are as follows:

[0178] ,

[0179] ,

[0180] ,

[0181]

[0182] In the formula, The pitch angle, This is the roll angle. For heading angle, Representing the rotation matrix The element in the 3rd row and 1st column, Rotation matrix The element in the 3rd row and 2nd column, Rotation matrix The element in the 3rd row and 3rd column, Rotation matrix The element in the 2nd row and 1st column, Rotation matrix The element in the first row and first column of the middle, The coordinates of the central common feature point in the measurement coordinate system. The coordinates of the central common feature point in the machine base coordinate system.

[0183] Since the target is fixed on the marine loader, the relative position between the two is fixed. Therefore, the accurate measurement result of the target's current pose can be directly used as the accurate measurement result of the marine loader's current pose.

[0184] like Figure 5 The figure shows a comparison of the reprojection error iteration curves of the iterative algorithm in step S4 of the synchronous tracking pose measurement method for marine loaders based on zoom vision in this embodiment with those of the traditional EPnP algorithm; wherein, part (a) in the figure is the reprojection error iteration curve of the traditional EPnP algorithm, and part (b) in the figure is the reprojection error iteration curve of the iterative algorithm in step S4 of this invention.

[0185] from Figure 5 The comparison of the reprojection error iteration curves of the two methods shows that the initial reprojection error of the traditional EPNP algorithm is relatively high, with a maximum of about 4.7 pixels, and the error decreases slowly, requiring about 15-16 iterations to converge to a stable value. In contrast, the initial reprojection error of the method of this invention is significantly reduced, with a maximum of about 2.0 pixels, and the error decreases much faster, requiring only about 9-10 iterations to converge to a stable value.

[0186] from Figure 5The comparison of the reprojection error iteration curves of the two methods also shows that the error curves of all feature points exhibit a smooth decrease without abnormal fluctuations, indicating that the reprojection error levels after final convergence are similar for both methods. This means that in terms of the solution results, both algorithms can ultimately converge to similar accuracy levels. This proves that the method of this invention significantly accelerates the convergence speed and computational efficiency without sacrificing solution accuracy. Applying this method to the dynamic operation scenario of marine loaders meets their real-time and stability requirements for pose measurement. Verification shows that the method in step S4 reduces the impact of feature point matching errors from the source by utilizing the objective function. Compared to the traditional EPNP algorithm, the convergence speed is improved by 30%, the number of iterations is ≤10, and the solution time is ≤20ms. It can quickly complete the tracking and real-time pose measurement of marine loaders in actual operation, meeting the operational requirements. Regarding the pose measurement results, verification shows that the method has a position error ≤5mm and an angle error ≤0.01° at a distance of 50m, representing a 40% improvement in accuracy compared to the traditional EPNP algorithm.

[0187] See Figure 6 In actual operation, a local observation field 4 covering the entire operating range of the marine loader 1 is constructed, and an observation field coordinate system with the same origin as the measurement coordinate system is constructed. By establishing a rigid transformation relationship between the observation field coordinate system and the fixed dock coordinate system constructed with the dock reference pile 5 as the origin, the precise alignment of the local observation field 4 with the dock operating area is ensured. Furthermore, during the operation of the marine loader 1, by converting the precise measurement results of the current pose of the marine loader obtained by the method of this invention to the fixed dock coordinate system, it is possible to judge in real time whether the operating pose of the marine loader has deviated or exceeded the safety boundary of the local observation field 4 according to the operation specifications, thereby sending alarm information in a timely manner. The local observation field 4 is generally constructed as an area with a radius of at least 50m and the integrated measurement unit as the origin, to cover the entire operating range of the marine loader from 10 to 50m. The observation field coordinate system is constructed with the rotation center of the two-dimensional turntable as the origin, using a right-handed coordinate system, with the X-axis pointing towards the dock shoreline, the Y-axis perpendicular to the shoreline and pointing towards the sea surface, and the Z-axis vertically upward.

Claims

1. A synchronous tracking pose measurement method for marine loaders based on zoom vision, characterized in that, The steps are as follows: S1. A target is vertically fixed at the rear of the marine loader, with multiple LED modules respectively arranged at the center and on the circumferential edge. The LED module at the center is located outside the plane where the multiple LED modules on the circumferential edge are located. The center of the LED light set in the center of the LED module at the center is defined as the center of the target. An integrated measurement unit is fixed at the center of the entire working range of the marine loader to acquire images containing the target in real time and the distance to the LED module at the center of the target. The target comprises a main frame, corner LED modules, and a central LED module. The main frame consists of a regular N-sided frame formed by N connecting beams connected end to end. N / 2 supporting beams are arranged radially within the frame. One end of each supporting beam is fixedly connected to and located at the center of the main frame, while the other end is vertically fixed to the middle of the side wall of each of the N / 2 connecting beams located at intervals. There are N corner LED modules, each centered and fixed in the middle of one of the N connecting beams. There is one central LED module, fixed to the center of the main frame by a support frame, so that the central LED module is positioned outside the plane containing the N corner LED modules. N is ≥6 and is an even number. The LEDs on each LED module are centered. The integrated measurement unit comprises a two-dimensional turntable, a zoom camera, and a laser rangefinder, arranged sequentially from bottom to top. The lens of the zoom camera and the laser emitter of the laser rangefinder are aligned in the same direction. S2. Before the marine loader operates, complete the full-focal-length parameter calibration of the zoom camera, the angle calibration of the two-dimensional turntable, and the consistency calibration of the laser beam and the camera optical axis of the laser rangefinder; construct a measurement coordinate system based on the integrated measurement unit, construct a fuselage coordinate system based on the target, define common feature points on the target, and set the rotation matrix between the measurement coordinate system and the fuselage coordinate system. Translation vector Methods for obtaining initial values; S3. An adaptive target tracking mechanism is set for the integrated measurement unit to ensure that the imaging center of the zoom camera is always aligned with the target center range and to obtain the target center coordinates in the measurement coordinate system in real time. S4. During the operation of the marine loader, the coordinates of the target center in the measurement coordinate system are obtained in real time, and the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is calculated. The optimal solution of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained by using the iterative solution method, so as to obtain the accurate measurement result of the target's current pose, and use it as the current pose of the marine loader.

2. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 1, characterized in that, In step S2, the parameter calibration method for the zoom camera across the entire focal length is as follows: Using a checkerboard calibration plate, several images of the zoom camera were acquired within the full focal length range of 10mm-200mm, and the intrinsic parameters of the zoom camera at each focal length were obtained. Constructing with focal length For a mapping model where the independent variable is denoted by and the internal parameter is denoted by , its expression is: ， ， ， ， ， ， ， ， In the formula, ~ Focal length in the x direction The fitting coefficients in the expression, ~ Focal length in the y direction The fitting coefficients in the expression, ~ x-axis coordinates of the principal point The fitting coefficients in the expression, ~ y-axis coordinates of the principal point The fitting coefficients in the expression, ~ The first radial distortion coefficient The fitting coefficients in the expression, ~ The second radial distortion coefficient The fitting coefficients in the expression, ~ The first tangential distortion coefficient The fitting coefficients in the expression, ~ The second tangential distortion coefficient The fitting coefficients in the expression; Substituting the intrinsic parameters of the zoom camera at various focal lengths into the mapping model, the fitting coefficients in each expression of the mapping model are obtained by solving the least squares method, so as to acquire the focal length based on the real-time images of the zoom camera. The intrinsic parameters of the zoom camera are obtained using a mapping model. Constructing a real-time image acquisition focal length based on zoom camera An image distortion correction model is used to obtain a distortion-corrected image; wherein, the focal length for real-time image acquisition by the zoom camera is... For any pixel coordinate on the image The distortion correction expression is: ， ， In the formula, The x-axis coordinates are after pixel distortion correction. The y-axis coordinates after pixel distortion correction; Radial distance, ; Based on focal length The first radial distortion coefficient, Based on focal length The second radial distortion coefficient; Based on focal length The first tangential distortion coefficient, Based on focal length The second tangential distortion coefficient.

3. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 1, characterized in that, In step S2, the method for calibrating the angle of the two-dimensional turntable is as follows: A laser collimator is fixed to the stage of a two-dimensional turntable, and its emitted laser is projected onto the center of the target on the opposite side. Angle encoders are installed on the azimuth and pitch axes of the two-dimensional turntable to obtain the actual rotation angle of the azimuth axis of the two-dimensional turntable. and the actual rotation angle of the pitch axis ; The azimuth angle of the two-dimensional turntable was sequentially adjusted from 0° to 360°, and the elevation angle was adjusted from -45° to 90°. The target azimuth axis angle was collected under different azimuth and elevation angle settings. Pitch axis target command angle Actual rotation angle of the azimuth axis and the actual rotation angle of the pitch axis To calculate the test values ​​of azimuth axis rotation deviation respectively and pitch axis rotation deviation test value The expressions for both are: ， ； Construct an angle correction model, the expression of which is: ， ， In the formula, This is the predicted value of the azimuth axis rotation deviation. and These are the azimuth correction factor and the azimuth correction intercept, respectively. This is the predicted value of the pitch axis rotation deviation. and These are the pitch axis correction factor and the pitch axis correction intercept, respectively. Test values ​​of azimuth axis rotation deviation obtained under different azimuth and elevation angle settings and pitch axis rotation deviation test value These are respectively used as predicted values ​​of azimuth axis rotation deviation. Predicted values ​​of pitch axis rotation deviation and the corresponding azimuth axis target command angle and pitch axis target command angle Substitute these values ​​into the angle correction model and use the linear fitting method to calculate the azimuth correction coefficient and azimuth correction intercept, as well as the pitch correction coefficient and pitch correction intercept. Furthermore, given the target command angles of the azimuth axis and pitch axis, the actual rotation angles of the azimuth axis and pitch axis of the two-dimensional turntable are obtained through the angle correction model.

4. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 1, characterized in that, In step S2, the laser rangefinder performs the following calibration method to ensure the consistency between the laser beam and the camera optical axis: using a collimator with a focal length of 500mm as the optical axis calibration reference, aligning the zoom camera with the outlet of the collimator, and adjusting the camera's orientation so that its imaging center coincides with the crosshair center of the collimator target plate; while keeping the zoom camera stationary, adjusting the installation orientation of the laser rangefinder so that the laser spot is also projected onto the crosshair center of the collimator target plate.

5. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 1, characterized in that, In step S2, a laser ranging vibration compensation model is constructed to compensate for the vibration of the laser ranging value of the laser rangefinder and replace the original laser ranging value. The expression for this model is: ， In the formula, This is the laser ranging compensation value. For vibration acceleration, This is the distance coefficient. This is the original value for laser ranging.

6. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 1, characterized in that, In step S2, The measurement coordinate system is constructed with the rotation center of the two-dimensional turntable of the integrated measurement unit as the origin, the X-axis parallel to the camera optical axis, and the Y and Z axes satisfying the right-hand rectangular coordinate system; the body coordinate system is constructed with the geometric center of the target's main frame as the origin, the axis direction of the LED module at the center as the z-axis, and the x and y axes located on the plane of the main frame. Define common feature points on the target, which include: the central common feature point located at the center of the LED light point of the central LED module; and multiple circumferential common feature points located at the centers of the LED light points of multiple LED modules in the circumferential direction. The transformation relationship between the common feature points in the measurement coordinate system and the fuselage coordinate system is constructed, and its expression is as follows: ， In the formula, Let i be the coordinates of the i-th common feature point in the measurement coordinate system; It is a rotation matrix; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; It is a translation vector; By obtaining the coordinates of all common feature points in both the measurement coordinate system and the fuselage coordinate system, the rotation matrix can be solved using the singular value decomposition method. Translation vector .

7. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 6, characterized in that, The specific implementation steps of step S3 are as follows: S301. Before the loader operates, with the imaging center of the zoom camera aligned with the center of the target, acquire an initial frame containing an image of the target, and extract all common feature points and their initial pixel coordinates from the image. S302. After the loader starts working, it will collect images containing the target again and use the FLANN matching algorithm to extract all common feature points in the newly collected images containing the target and obtain their real-time pixel coordinates. S303. Calculate the geometric center pixel coordinates of all common feature points in the initial frame image containing the target, and use them as the tracking reference point coordinates; calculate the geometric center pixel coordinates of all common feature points in the newly acquired image containing the target, and obtain their pixel deviations in the x-direction based on the tracking reference point coordinates. Pixel deviation in the y direction Determine whether the deviation between the two pixels is ≤5 pixels. If the result is yes, keep the current state of the two-dimensional turntable unchanged and proceed directly to step S306. If the result is no, proceed to step S304 and adjust the current state of the two-dimensional turntable. S304. Convert the two pixel deviations obtained in step S303 into azimuth axis angle adjustment amounts. and pitch axis angle adjustment Its expression is: ， ， In the formula, The first proportionality coefficient, This is the second proportionality coefficient; To fit the focal length, , and These are the x-direction focal length and y-direction focal length of the zoom camera when acquiring the initial frame containing the target image; S305. Calculate the target angle of the azimuth axis at the current time t to accurately adjust the state of the two-dimensional turntable. and pitch axis target command angle And complete the real-time adjustment of the two-dimensional turntable status; S306. Update the horizontal angle of the two-dimensional turntable at the current time t. and pitch angle In order to obtain the central common feature point, that is, the real-time coordinates of the target center in the measurement coordinate system ( , , Its calculation expression is: ， ， ， In the formula, This refers to the distance from the integrated measurement unit to the LED module at the center of the target; S307. During the operation of the loader, the zoom camera continuously acquires images containing the target at the same time intervals, and processes the newly acquired images containing the target and the images containing the target acquired in the previous frame in the same way as steps S301-S306, so as to continuously adjust the state of the two-dimensional turntable in real time.

8. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 7, characterized in that, In step S4, the specific method for obtaining the optimal solution of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is as follows: S401. Constructing the objective function Its expression is: ， In the formula, The number of common feature points; This is the intrinsic parameter matrix for a zoom camera; Let be the coordinates of the i-th common feature point in the fuselage coordinate system; For the first The pose transformation matrix between the fuselage coordinate system and the measurement coordinate system in the next iteration; π(·) is the camera reprojection function; Let be the pixel coordinates of the i-th common feature point in the distortion-corrected image; The coordinates of the central common feature point in the fuselage coordinate system; The coordinates of the central common feature point in the measurement coordinate system; λ is the weighting coefficient; S402. Based on the target center coordinates obtained in real time in step S3 in the measurement coordinate system, the initial value of the pose transformation matrix between the fuselage coordinate system and the measurement coordinate system at the current moment is obtained. S403. During the iteration process, the least squares method with Levenberg-Marquardt (LM) damping is used to update the rotation matrix and translation vector. S404. Set the iteration termination condition as follows: the objective function value converges to the minimum, and the deviation between the coordinates of the central common feature point calculated by the pose transformation matrix in the current iteration and the calculation result in the previous iteration is ≤0.001m; at the same time, the maximum number of iterations is set to 20. S405. Based on the settings of steps S401-S404 above, substitute the initial value of the pose transformation matrix at the current moment into the objective function, and with the goal of minimizing the objective function, obtain the pose transformation matrix that satisfies the iteration termination condition through iterative solution, and use it as the optimal solution of the pose transformation matrix at the current moment. .

9. The method for synchronous tracking pose measurement of marine loaders based on zoom vision according to claim 8, characterized in that, In step S4, the method for obtaining the accurate measurement result of the target's current pose is as follows: The optimal solution of the pose transformation matrix at the current moment Convert to rotation matrix Translation vector By substituting these values ​​into the target pose output expression, the coordinates of the central common feature point in the measurement coordinate system can be obtained. Pitch angle Roll angle Heading angle The accurate measurement results of the current pose of the target are as follows: ， ， ， In the formula, The pitch angle, This is the roll angle. For heading angle, Representing the rotation matrix The element in the 3rd row and 1st column, Rotation matrix The element in the 3rd row and 2nd column, Rotation matrix The element in the 3rd row and 3rd column, Rotation matrix The element in the 2nd row and 1st column, Rotation matrix The element in the first row and first column of the middle, The coordinates of the central common feature point in the measurement coordinate system. The coordinates of the central common feature point in the machine base coordinate system.

Images