Three-level precision intelligent butt joint installation method for fan tower drum

Abstract

The invention discloses a three-level precision intelligent butt joint installation method for a fan tower drum, which comprises the following steps of: placing a binocular camera in the tower drum needing to be installed in advance, and calibrating before installation to obtain a flange reference circle center coordinate and a fixed physical marking point coordinate under a left camera coordinate system so as to obtain a reference vector under the left camera coordinate system; when the tower drum is installed on the flange, a binocular camera collects stereo image pairs of the bottom of the tower drum to be installed and tower drum bottom identification points in real time; performing preprocessing based on the image pair; based on the preprocessed image pair, the real-time 3D coordinate of the tower bottom identification point and the real-time 3D coordinate of the tower bottom circle center under the left camera coordinate system are solved, and a vector to be docked is obtained; the method has the advantages that the whole pose detection process depends on remote shooting of binocular vision, a person does not need to get close to high-altitude swing or to-be-contacted heavy parts for measurement, and the intrinsically safe distance is achieved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine installation, and in particular to a three-stage precision intelligent docking installation method for wind turbine towers. Background Technology

[0002] As a key load-bearing structure supporting the wind turbine generator, the installation accuracy of the tower directly affects the overall operational safety, stability, and lifespan of the unit. The core of tower installation lies in the precise alignment and connection of the flange at the bottom of the tower section to be installed with the flange at the top of the existing foundation (or the next tower section). Traditional connection methods rely primarily on visual observation by operators, simple measuring tools (such as spirit levels and tape measures), and experience-based judgment, using lifting equipment (main crane) for rough adjustments and trial connections. This method has the following significant limitations:

[0003] Human judgment and manual measurement can have significant discrepancies within the centimeter or even decimeter range, making it difficult to meet the high precision requirements (typically millimeter or even sub-millimeter) for coaxiality and flange parallelism when connecting modern large towers (especially those with a diameter exceeding 5 meters). Operational results heavily rely on the skills and experience of the lifting personnel, resulting in poor consistency.

[0004] To meet the docking requirements, multiple cycles of "lifting-rough adjustment-trial placement-measurement-readjustment" are often required. Especially in the stage of horizontal translation and fine adjustment of rotation angle, the repeated starting and stopping of the main crane and manual adjustment are time-consuming, significantly extending the high-altitude operation window and increasing the uncertainty risks of external environmental factors such as weather and wind.

[0005] Relying on close-range manual observation and command means that personnel must be in a potentially hazardous area at the moment the tower swings or is positioned. Furthermore, repeated touches could damage the flange face or bolt holes.

[0006] Traditional methods cannot provide real-time, continuous three-dimensional spatial deviation data. Operators can only obtain intermittent, qualitative or semi-quantitative feedback, resulting in a discontinuous and inaccurate adjustment process. Summary of the Invention

[0007] Purpose of the invention: The purpose of this invention is to solve the technical problems in the prior art and provide a three-stage precision intelligent docking and installation method for wind turbine towers.

[0008] Technical solution: A three-stage precision intelligent docking installation method for wind turbine towers, comprising:

[0009] Step S1: Place a binocular camera inside the tower where it is to be installed in advance, and calibrate the coordinates of the flange reference center and the coordinates of the fixed physical marker point in the left camera coordinate system before installation to obtain the reference vector in the left camera coordinate system.

[0010] Step S2: When installing the tower on the flange, use a binocular camera to capture real-time stereo images of the bottom of the tower to be installed and the marked points on the bottom of the tower.

[0011] Step S3: Preprocess the image pairs;

[0012] Step S4: Based on the preprocessed image pairs, calculate the real-time 3D coordinates of the marker point at the bottom of the tower and the real-time 3D coordinates of the center of the tower bottom circle in the left camera coordinate system to obtain the docking vector.

[0013] Step S5: Calculate the positional deviation between the real-time 3D coordinates of the bottom center of the tower in the real-time left camera coordinate system and the coordinates of the flange reference center in the left camera coordinate system, and calculate the directional angle between the reference vector and the vector to be docked.

[0014] Step S6: Generate and output a first guidance command based on the position deviation to drive the main crane to perform a translation operation until the position deviation is less than or equal to a first threshold and proceed to the next step.

[0015] Step S7: Generate a second guiding command based on the directional angle to drive the spreader to rotate until the directional angle is less than or equal to the second threshold.

[0016] Preferably, step S1 includes the following steps:

[0017] The 3D calibration of the reference center was simulated in advance when the flange was installed on the base flange, and the coordinates of the flange reference center in the left camera coordinate system were obtained. ;

[0018] Fixed physical markers are set at the pre-defined guide petal positions on the base flange, and these are recorded as point coordinates. , forming a reference vector ,in, equal And it equals ;

[0019] The intrinsic and extrinsic parameter matrices of the left and right cameras of the stereo camera are pre-calibrated, and the distortion coefficients are calibrated. The extrinsic parameter matrix includes the rotation matrix R and the translation matrix T.

[0020] Preferably, step S3 includes the following steps:

[0021] The image pairs are filtered using Gaussian filtering to remove image noise;

[0022] A local adaptive thresholding method is used to perform threshold segmentation on the filtered image in order to initially extract the flange edge region;

[0023] The Canny edge detection algorithm is applied to the thresholded image to obtain a clear flange edge contour curve.

[0024] Preferably, step S4 includes the following steps:

[0025] Obtain the flange profile curve;

[0026] In the left and right camera images, the least squares circle fitting algorithm is applied to the bottom contour point set of the tower respectively to obtain the 2D pixel coordinates of the circle center in the left and right images and directly obtain the 2D pixel coordinates of the marker point.

[0027] Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the tower bottom. And obtain the 3D coordinates of the bottom marker point of the tower in the same way. ;

[0028] Step S404: Obtain the docking vector based on the 3D coordinates of the center of the tower bottom and the 3D coordinates of the marker point. .

[0029] Preferred least-squares circle fitting algorithms include:

[0030] Establish the equation of the circle and define the set of contour points:

[0031] , i is the point set index, the contour point set is the coordinates of each pixel in the image, u i v is the x-axis. i The vertical axis is used as the coordinate.

[0032] Satisfying the general equation of a circle:

[0033] u is the x-coordinate of a point in the coordinate system, and v is the y-coordinate of a point in the coordinate system;

[0034] The coordinates of the center pixel are:

[0035] ;

[0036] The parameters D, E, and F are then solved using least squares optimization to minimize the error function, which is:

[0037] ;

[0038] Solve the system of linear equations: The optimization problem is transformed into a system of linear equations AX=B, where:

[0039] A is a matrix containing the coordinate information of image points, X is a parameter vector containing the coefficients D, E, and F to be optimized, and B is a vector representing the circular error term for each image point.

[0040] Calculate the coordinates of the center of the circle: by solving:

[0041] ;

[0042] The parameters D, E, and F are obtained, and then the 2D pixel coordinates of the circle's center in the left camera image are calculated. and pixel coordinates in the right camera image .

[0043] Preferably, based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower, including:

[0044] The normalized coordinates of the left camera are obtained through the camera intrinsic parameter matrix. It satisfies:

[0045] ;

[0046] K L Let D be the intrinsic parameter matrix of the left camera. L The distortion coefficient of the left camera;

[0047] The normalized coordinates of the right camera are obtained from the camera intrinsic parameter matrix. It satisfies:

[0048] ;

[0049] K R Let D be the intrinsic parameter matrix of the right camera. R The distortion coefficient is for the right camera.

[0050] Preferably, based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower, including:

[0051] Left camera coordinate system center coordinates It satisfies:

[0052] ;

[0053] Right camera coordinate system center coordinates It satisfies:

[0054] ;

[0055] In addition, the coordinates of the center of the circle in the left camera coordinate system The coordinates of the center of the circle in the right camera coordinate system The following relationship exists between them:

[0056] ;

[0057] By minimizing the reprojection error using a simultaneous equation, the coordinates of the center of the circle in the left camera coordinate system can be obtained. To obtain the coordinates of the flange reference center in the left camera coordinate system. Coordinates of the center of the bottom circle of the tower in the left camera coordinate system (within the same coordinate system) ...

[0058] Preferably, step S5 includes the following steps:

[0059] Obtain the coordinates of the center of the tower bottom in the reference left camera coordinate system. Left camera coordinate system lower flange reference center coordinates Vector to be docked and reference vector

[0060] ;

[0061] The formula for calculating positional deviation is as follows:

[0062] ;

[0063] The formula for calculating the directional deviation angle is as follows:

[0064] .

[0065] Preferably, step S6 includes the following steps:

[0066] Adjust the crane driving the tower to move horizontally and vertically;

[0067] Determine X t The sign of the difference between X0 and X0 is determined by the left or right direction of the horizontal movement;

[0068] Determine Y t The sign of the difference between Y and Y0 indicates the direction of vertical movement (up or down).

[0069] Continue until the position deviation is less than or equal to the first threshold to proceed to the next step.

[0070] Preferably, step S7 includes the following steps:

[0071] Repeat steps S2-S5 to obtain the new direction angle, which is the final direction angle.

[0072] The crane that drives the tower rotates the tower by adjusting the rotation angle of the final direction.

[0073] Determining the direction of rotation includes:

[0074] Calculate the vector to be docked and reference vector The cross product yields:

[0075] ;

[0076] To obtain its Z component ,judge >0 indicates counterclockwise rotation; A value less than 0 indicates clockwise rotation.

[0077] The rotation step size is split and executed, as shown in the following formula:

[0078] ;

[0079] The steps for implementing feedback and correction are as follows:

[0080] The new direction angle θ is recalculated at each rotation step. i And to obtain the actual rotation angle α i The following formula is used to determine the value of the rotation angle. If the formula is true, the subsequent rotation angle is dynamically adjusted.

[0081] ;

[0082] When the angle between the vectors collected for a preset number of consecutive times is less than the preset second threshold angle, the tower angle calibration is completed.

[0083] Beneficial effects:

[0084] 1. By employing edge detection and least-squares circle fitting, even in complex lighting conditions or partially obscured industrial environments, the 2D pixel coordinates of the flange center can be extracted from images with high precision. Combined with calibrated binocular camera parameters, a binocular triangulation algorithm reliably converts these 2D coordinates into 3D world coordinates between the marker point and the tower center. Finally, the spatial deviation from the reference center is directly calculated using vector subtraction. This closed-loop algorithm eliminates reading and estimation errors inherent in manual measurements.

[0085] 2. Based on the obtained coordinates of the center of the circle, the scheme calculates the angle θ by calculating the dot product and magnitude of the reference vector and the real-time vector. The vector cross product not only determines the direction of rotation (clockwise / counterclockwise), but its magnitude can also serve as a reference for the magnitude of the deviation. Closed-loop step-by-step control breaks down the total adjustment angle θ into small steps that conform to the response characteristics of the spreader. After each step, the system recalculates the current deviation and determines the next action, forming a closed loop of "perception-decision-execution-feedback" to dynamically eliminate overshoot and residual error until it converges to the target accuracy.

[0086] 3. The entire pose detection process relies on remote binocular vision imaging, eliminating the need for personnel to approach heavy components that are swinging at height or about to come into contact with them for measurement, thus achieving an inherently safe distance. High-precision guidance allows the tower to be adjusted into position smoothly in a "one-time" or "very few" attempts, greatly reducing the risk of rigid contact and impacts during trial docking, protecting the flange sealing surface and bolt threads. Gaussian filtering and adaptive threshold segmentation technologies in preprocessing enhance the system's resistance to interference from on-site noise and lighting changes. Closed-loop control logic ensures stable convergence during the adjustment process, avoiding repeated oscillations or over-adjustments caused by human error.

[0087] 4. The solution encodes optimal alignment logic and judgment criteria into algorithms and processes, enabling any operator with basic training to reach the level of a top-tier technician with system assistance, thus solving the problem of difficulty in transferring and replicating experience. All operational instructions originate from data collected and processed in real time, making the decision-making process transparent and traceable. Attached Figure Description

[0088] Figure 1 A schematic diagram of the method framework for this invention is provided;

[0089] Figure 2 A schematic diagram of the overall structure of this invention is provided;

[0090] Figure 3 Provided for the present invention Figure 2 Enlarged diagram of position A in the middle;

[0091] Figure 4 This invention provides a schematic diagram of the structure at the location of the tower to be installed. Detailed Implementation

[0092] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0093] Example 1

[0094] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed following the word and its equivalents, but do not exclude other elements or objects.

[0095] The tower 1 mentioned in this application, such as Figure 2 As shown, the tower 1 is used to install on the already fixed flange 2, wherein, as Figure 3 As shown, a physical marking point 3 is provided on the base flange 3 for physical marking, such as... Figure 4 As shown, the binocular camera 4 is installed on the bottom side inside the tower.

[0096] In response to the problems existing in the current technology, such as Figure 1 As shown, a three-stage precision intelligent docking installation method for wind turbine towers is proposed, including:

[0097] Step S1: Place a binocular camera inside the tower where it is to be installed in advance, and calibrate the coordinates of the flange reference center and the coordinates of the fixed physical marker point in the left camera coordinate system before installation to obtain the reference vector in the left camera coordinate system.

[0098] In some specific embodiments, step S1 includes the following steps:

[0099] The 3D calibration of the reference center was simulated in advance when the flange was installed on the base flange, and the coordinates of the flange reference center in the left camera coordinate system were obtained. ;

[0100] Fixed physical markers are set at the pre-defined guide petal positions on the base flange, and these are recorded as point coordinates. , forming a reference vector ,in, equal And it equals (With the plane of the base flange as the XY plane, and the direction perpendicular to the flange surface upwards as the positive Z-axis).

[0101] The intrinsic and extrinsic parameter matrices of the left and right cameras of the stereo camera are pre-calibrated, and the distortion coefficients are calibrated. The extrinsic parameter matrix includes the rotation matrix R and the translation matrix T.

[0102] And calibrate the distortion coefficient of the left camera D L Right camera D R Left camera C L The intrinsic parameter matrix is ​​K L Right camera C R The intrinsic parameter matrix is ​​K R .

[0103] Step S2: When installing the tower on the flange, a stereo image pair of the bottom of the tower to be installed and the marked point on the bottom of the tower is acquired in real time using a binocular camera. The image pair is the image taken by the right camera and the image taken by the right camera.

[0104] Step S3: Preprocess the image pairs;

[0105] In some specific embodiments, step S103 includes the following steps:

[0106] The image pairs (including the left and right images captured by the left and right eyes) are filtered using Gaussian filtering to remove image noise. The kernel size is 5×5 and the standard deviation is 1.2.

[0107] The local adaptive thresholding method is used to perform threshold segmentation on the filtered image to initially extract the flange edge region. The threshold T(x,y) of each pixel (x,y) is calculated based on the mean of its neighboring pixels. The calculation formula is: T(x,y) = mean(I(x,y,r))-C, where the neighborhood radius r=3 and the offset C=2.

[0108] The Canny edge detection algorithm is applied to the thresholded image to obtain a clear flange edge contour curve, with a low threshold of 50 and a high threshold of 150.

[0109] Step S4: Based on the preprocessed image pairs, calculate the real-time 3D coordinates of the marker point at the bottom of the tower and the real-time 3D coordinates of the center of the tower bottom circle in the left camera coordinate system to obtain the docking vector.

[0110] In some specific embodiments, the flange profile curve is obtained;

[0111] In the left and right camera images, the least squares circle fitting algorithm is applied to the bottom contour point set of the tower to obtain the 2D pixel coordinates of the circle center in the left and right images and the 2D pixel coordinates of the marker point. The circle center needs to be calculated because it is not a point that is directly detected, but the marker point has been determined at the marked position and does not need to be calculated directly.

[0112] Least squares circle fitting algorithms include:

[0113] Establish the equation of the circle and the coordinate system for image recognition. Define the contour point set: the circular curve is denoted as L.

[0114] , i is the point set index, the contour point set is the coordinates of each pixel in the image, u i v is the x-axis. i The vertical axis is used as the coordinate.

[0115] Satisfying the general equation of a circle:

[0116] u is the x-coordinate of a point in the coordinate system, and v is the y-coordinate of a point in the coordinate system;

[0117] The coordinates of the center pixel are:

[0118] ;

[0119] The parameters D, E, and F are then solved using least squares optimization to minimize the error function, which is:

[0120] ;

[0121] Solve the system of linear equations: The optimization problem is transformed into a system of linear equations AX=B, where:

[0122] A is a matrix containing the coordinate information of image points, X is a parameter vector containing the coefficients D, E, and F to be optimized, and B is a vector representing the circular error term for each image point.

[0123] Calculate the coordinates of the center of the circle: by solving:

[0124] ;

[0125] Parameters D, E, and F are obtained, and then different circular curves are calculated through image recognition using the left and right cameras respectively. The 2D pixel coordinates of the circle's center in the left camera image are then calculated for each curve. and pixel coordinates in the right camera image .

[0126] Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the tower bottom. And obtain the 3D coordinates of the bottom marker point of the tower in the same way. The following steps convert the coordinates of the center of the circle obtained by the left and right cameras into the coordinates of the center of the circle in the reference coordinates. Similarly, when the 2D coordinates of the point are obtained directly by the left and right cameras, the 3D coordinates of the point in the reference coordinates are calculated in the same way as the following steps. Therefore, they will not be described in detail.

[0127] Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower. That is, the center coordinates of the image are transformed into center coordinates in a reference coordinate system, including:

[0128] The normalized coordinates of the left camera are obtained through the camera intrinsic parameter matrix. It satisfies:

[0129] ;

[0130] K L Let D be the intrinsic parameter matrix of the left camera. L The distortion coefficient of the left camera;

[0131] The normalized coordinates of the right camera are obtained from the camera intrinsic parameter matrix. It satisfies:

[0132] ;

[0133] K R Let D be the intrinsic parameter matrix of the right camera. R The distortion coefficient is for the right camera.

[0134] Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower, including:

[0135] Left camera coordinate system center coordinates It satisfies:

[0136] ;

[0137] Right camera coordinate system center coordinates It satisfies:

[0138] ;

[0139] In addition, the coordinates of the center of the bottom of the tower in the left camera coordinate system The coordinates of the center of the circle at the bottom of the right camera tower The following relationship exists between them:

[0140] ;

[0141] By minimizing the reprojection error using a simultaneous equation, the coordinates of the center of the circle in the left camera coordinate system can be obtained. To obtain the coordinates of the flange reference center in the left camera coordinate system. Coordinates of the center of the bottom circle of the tower in the left camera coordinate system (within the same coordinate system) , Recorded as .

[0142] Based on the 3D coordinates of the center of the tower's bottom circle in the left camera coordinate system (reference coordinate system) and the 3D coordinates of the marked points, the docking vector is obtained. The calculation process follows standard mathematical methods and will not be elaborated upon here.

[0143] Step S5: Calculate the positional deviation between the real-time 3D coordinates of the bottom center of the tower in the real-time left camera coordinate system and the coordinates of the flange reference center in the left camera coordinate system, and calculate the directional angle between the reference vector and the vector to be docked.

[0144] Step S5 includes the following steps:

[0145] Obtain the coordinates of the center of the tower bottom in the reference left camera coordinate system. Left camera coordinate system lower flange reference center coordinates Vector to be docked and reference vector

[0146] ;

[0147] The formula for calculating positional deviation is as follows:

[0148] ;

[0149] The formula for calculating the directional deviation angle is as follows:

[0150]

[0151] in,

[0152] ;

[0153] ;

[0154] .

[0155] Overcoming the errors and risks of manual contact measurement, the system can complete all pose detections at locations far from the docking area, ensuring personnel safety. The system continuously outputs the three-dimensional positional deviation between the tower center and the reference center, as well as the directional deviation angle θ, providing precise input for automatic control. Through a combination of image preprocessing techniques, including Gaussian filtering, adaptive threshold segmentation, and Canny edge detection, noise from ambient lighting changes, dust, and localized occlusion is effectively suppressed, improving system robustness. The entire image-to-coordinate conversion process is automated by the algorithm, eliminating human reading errors, and all data is recordable and reviewable, enhancing process quality control capabilities.

[0156] Step S6: Generate and output a first guidance command based on the position deviation to drive the main crane to perform a translation operation until the position deviation is less than or equal to a first threshold and proceed to the next step.

[0157] Step S6 includes the following steps:

[0158] Adjust the crane driving the tower to move horizontally and vertically;

[0159] Determine X t The sign of the difference between X0 and X0 is determined by the left or right direction of the horizontal movement;

[0160] Determine Y t The sign of the difference between Y and Y0 indicates the direction of vertical movement (up or down).

[0161] Continue until the position deviation is less than or equal to the first threshold to proceed to the next step.

[0162] The first threshold can be set to 5cm. The first guiding instruction is generated by the position deviation and the positive and negative judgment relationship, and is adjusted accordingly according to the preset strategy, such as "left arm 30cm", "right arm 20cm", "arm up 15cm", "arm down 25cm".

[0163] Step S7: Generate a second guiding command based on the directional angle to drive the spreader to rotate until the directional angle is less than or equal to the second threshold.

[0164] Step S7 includes the following steps:

[0165] Repeat steps S2-S5 to obtain the new direction angle, which is the final direction angle θ. final ;

[0166] The crane that drives the tower rotates the tower by adjusting the rotation angle of the final direction.

[0167] Determining the direction of rotation includes:

[0168] Calculate the vector to be docked and reference vector The cross product yields:

[0169] ;

[0170] ;

[0171] To obtain its Z component ,judge >0 indicates counterclockwise rotation; A value less than 0 indicates clockwise rotation.

[0172] The rotation step size is split and executed, as shown in the following formula:

[0173] ;

[0174] α is the target rotation angle; it is divided according to the minimum step size of 0.01 degrees;

[0175] For example, when α is 0.35 degrees, it is divided into 35 steps, each step being 0.01 degrees. The rotation is performed step by step according to the step size, resulting in higher precision.

[0176] The steps for implementing feedback and correction are as follows:

[0177] The new direction angle θ is recalculated at each rotation step. i And to obtain the actual rotation angle α i The following formula is used to determine the value of the rotation angle. If the formula is true, the subsequent rotation angle is dynamically adjusted.

[0178] ;

[0179] The rotation angle of subsequent steps is dynamically corrected to ensure that the total deviation is less than 0.01 degrees;

[0180] When the included angle of the vectors collected for a preset number of consecutive times is less than a preset second threshold angle, the tower angle calibration is completed, and a rotation alignment completion command is issued. For example, when the included angle of the vector directions collected for three consecutive times is less than a preset second threshold angle, the tower angle calibration is completed.

[0181] Achieving a smooth transition from coarse to fine adjustment, the tiered strategy aligns with the actual physical processes of hoisting operations. It first addresses large-scale positional deviations before eliminating minor directional deviations, avoiding system oscillations or overshoot that might occur with a single, direct approach. The rotation direction is automatically determined by the sign of the Z-component of the vector cross product, replacing manual judgment. Large-angle rotations are decomposed into microsteps, with deviations remeasured after each step, and subsequent step lengths dynamically adjusted. This effectively compensates for factors such as spreader transmission errors and wind load interference, ensuring that the final directional accuracy consistently converges to extremely high requirements (≤0.01°). The system provides continuous and explicit guidance instructions, eliminating the need for operators to repeatedly attempt and judge. Both translation and rotation processes proceed towards clearly quantified objectives, significantly reducing the number of adjustments and time spent in the air, achieving successful docking on the first attempt. The optimal docking logic is embedded in the algorithm, providing operational suggestions directly via voice or commands, enabling ordinary operators to complete high-precision docking tasks after basic training, reducing reliance on experienced specialized technicians.

[0182] The system forms a complete closed loop of perception-decision-execution, achieving full digitization and automation of the tower docking process, fundamentally changing the traditional manual-dependent model. The entire process is controlled by algorithms and programs, avoiding uncertainties introduced by human fatigue and emotions, maintaining the same high standard of precision for each docking operation. The core algorithm features a modular design, allowing for parameter adjustments (such as filter kernel size, threshold, and control step size) to adapt to different tower diameters, different types of lifting equipment, and varying on-site environments. The standardized, digital command interface output by this solution can directly interface with more advanced automatic control systems (such as fully automatic main cranes), representing a key step towards fully unmanned and intelligent lifting operations.

[0183] The above description is merely a specific implementation of the embodiments of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of the present invention should be covered within the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention should be determined by the protection scope of the claims.

Claims

1. A three-level precision intelligent butt joint installation method for a fan tower, characterized in that, include: Step S1: Place a binocular camera inside the tower where it is to be installed in advance, and calibrate the coordinates of the flange reference center and the coordinates of the fixed physical marker point in the left camera coordinate system before installation to obtain the reference vector in the left camera coordinate system. Step S2: When installing the tower on the flange, use a binocular camera to capture real-time stereo images of the bottom of the tower to be installed and the marked points on the bottom of the tower. Step S3: Preprocess the image pairs; Step S4: Based on the preprocessed image pairs, calculate the real-time 3D coordinates of the marker point at the bottom of the tower and the real-time 3D coordinates of the center of the tower bottom circle in the left camera coordinate system to obtain the docking vector. Step S5: Calculate the positional deviation between the real-time 3D coordinates of the bottom center of the tower in the real-time left camera coordinate system and the coordinates of the flange reference center in the left camera coordinate system, and calculate the directional angle between the reference vector and the vector to be docked. Step S6: Generate and output a first guidance command based on the position deviation to drive the main crane to perform a translation operation until the position deviation is less than or equal to a first threshold and proceed to the next step. Step S7: Generate a second guiding command based on the directional angle to drive the spreader to rotate until the directional angle is less than or equal to the second threshold.

2. The method of claim 1, wherein, Step S1 includes the following steps: Simulate the 3D calibration of the reference circle center when installed on the base flange in advance, obtain the flange reference circle center coordinates in the left camera coordinate system ; A fixed physical mark is set at the guide petal position on the base flange, denoted as point coordinates , a reference vector is formed , wherein is equal to , and it is equal to ; The intrinsic and extrinsic parameter matrices of the left and right cameras of the stereo camera are pre-calibrated, and the distortion coefficients are calibrated. The extrinsic parameter matrix includes the rotation matrix R and the translation matrix T.

3. The method of claim 1, wherein, Step S3 includes the following steps: The image pairs are filtered using Gaussian filtering to remove image noise; A local adaptive thresholding method is used to perform threshold segmentation on the filtered image in order to initially extract the flange edge region; The Canny edge detection algorithm is applied to the thresholded image to obtain a clear flange edge contour curve.

4. The method of claim 2, wherein the method further comprises: Step S4 includes the following steps: Obtain the flange profile curve; In the left and right camera images, the least squares circle fitting algorithm is applied to the bottom contour point set of the tower respectively to obtain the 2D pixel coordinates of the circle center in the left and right images and directly obtain the 2D pixel coordinates of the marker point. Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into 3D coordinates of the center of the tower bottom circle , and the 3D coordinates of the tower bottom marker points are obtained in the same way ; Step S404: Obtain the docking vector based on the 3D coordinates of the center of the tower bottom and the 3D coordinates of the marker point. .

5. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 4, characterized in that, Least squares circle fitting algorithms include: Establish the equation of the circle and define the set of contour points: , i is the point set index, the contour point set is the coordinates of each pixel in the image, u i v is the x-axis. i The vertical axis is used as the coordinate. Satisfying the general equation of a circle: u is the x-coordinate of a point in the coordinate system, and v is the y-coordinate of a point in the coordinate system; The coordinates of the center pixel are: ； The parameters D, E, and F are then solved using least squares optimization to minimize the error function, which is: ； Solve the system of linear equations: The optimization problem is transformed into a system of linear equations AX=B, where: A is a matrix containing the coordinate information of image points, X is a parameter vector containing the coefficients D, E, and F to be optimized, and B is a vector representing the circular error term for each image point. Calculate the coordinates of the center of the circle: by solving: ； The parameters D, E, and F are obtained, and then the 2D pixel coordinates of the circle's center in the left camera image are calculated. and pixel coordinates in the right camera image .

6. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 5, characterized in that, Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower, including: The normalized coordinates of the left camera are obtained from the camera intrinsic parameter matrix. It satisfies: ； K L is the intrinsic matrix of the left camera, D L is the distortion coefficient of the left camera; The normalized coordinates of the right camera are obtained from the camera intrinsic parameter matrix. It satisfies: ； K R is the intrinsic matrix of the right camera, D R is the distortion coefficient of the right camera.

7. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 6, characterized in that, Based on the principle of binocular vision triangulation and camera parameters, the 2D pixel coordinates are calculated into the 3D coordinates of the center of the bottom circle of the tower, including: Left camera coordinate system center coordinates It satisfies: ； Right camera coordinate system center coordinates It satisfies: ； In addition, the coordinates of the center of the circle in the left camera coordinate system The coordinates of the center of the circle in the right camera coordinate system The following relationship exists between them: ； By minimizing the reprojection error using a simultaneous equation, the coordinates of the center of the circle in the left camera coordinate system can be obtained. To obtain the coordinates of the flange reference center in the left camera coordinate system. Coordinates of the center of the bottom circle of the tower in the left camera coordinate system (within the same coordinate system) .

8. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 1, characterized in that, Step S5 includes the following steps: Obtain the coordinates of the center of the tower bottom in the reference left camera coordinate system. Left camera coordinate system lower flange reference center coordinates Vector to be docked and reference vector ； The formula for calculating positional deviation is as follows: ； The formula for calculating the directional deviation angle is as follows: 。 9. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 8, characterized in that, Step S6 includes the following steps: The crane driving the tower is adjusted to move horizontally and vertically based on the positional deviation; determining X t the positive or negative difference between X and X0 in the horizontal direction determination Y t the positive or negative determination of the difference between Y0 and Y is in the vertical direction of movement Continue until the position deviation is less than or equal to the first threshold to proceed to the next step.

10. The method for intelligent docking and installation of a wind turbine tower with three-stage precision according to claim 9, characterized in that, Step S7 includes the following steps: Repeat steps S2-S5 to obtain the new direction angle, which is the final direction angle. The crane that drives the tower rotates the tower by adjusting the rotation angle of the final direction. Determining the direction of rotation includes: Calculate the vector to be docked and reference vector The cross product yields: ； To obtain its Z component ,judge >0 indicates counterclockwise rotation; A value less than 0 indicates clockwise rotation. The rotation step size is split and executed, as shown in the following formula: α is the target rotation angle; The steps to implement feedback and correction are as follows: The new direction angle θ is recalculated at each rotation step i The actual rotation angle αi is obtained, and the following formula is judged. If it is true, the subsequent rotation angle is dynamically corrected. ； When the angle between the vectors collected for a preset number of consecutive times is less than the preset second threshold angle, the tower angle calibration is completed.

Images