A method and apparatus for laser marking tyres

Abstract

The present application relates to the technical field of laser processing, and discloses a tire laser lettering method, comprising the following steps: step 1: collecting a two-dimensional image and a local point cloud of a tire; step 2: converting the local point cloud of each view to a global coordinate system to obtain a complete point cloud; step 3: processing the two-dimensional images of multiple views of the tire to obtain a pure tire tread image, identifying the identification points of the pure tire tread image, obtaining the three-dimensional coordinates of the identification points in the local point cloud, and then mapping to the complete point cloud; step 4: converting the complete point cloud into a plane point cloud, extracting the point cloud data of the inner edge of the tire from the plane point cloud, and obtaining an inner edge fitting straight line; step 5: processing the lettering feature points by using a positioning algorithm to obtain the two-dimensional coordinates of the lettering feature points; and then using a coordinate inverse transformation formula to restore the two-dimensional coordinates of the lettering feature points to three-dimensional coordinates, and printing the printed characters on the tire. Meanwhile, a tire laser lettering device and system are also disclosed.

Description

Technical Field

[0001] This application belongs to the field of laser processing technology, and more specifically, relates to a method and apparatus for laser engraving on tires. Background Technology

[0002] As a core component of vehicles, tires require key information such as product specifications, production batch, and brand logo to be engraved on their sidewalls. In some scenarios, traceability QR codes are also needed for quality tracking. Traditional tire engraving processes mainly use manual engraving or chemical etching, which have many drawbacks: manual engraving is inefficient, labor-intensive, and results in poor engraving accuracy and consistency; chemical etching processes generate pollutants, which do not meet environmental protection requirements, and the engraving durability is poor, with the lettering easily becoming blurred due to wear or aging.

[0003] Secondly, some companies use laser engraving equipment for tire engraving. However, when using laser engraving equipment, it is necessary to locate the engraving area on the tire. The existing positioning method directly calculates the three-dimensional coordinates of the engraving feature points in the global coordinate system. This requires taking into account the complex geometric shape of the tire surface, which involves a large amount of calculation and is easily affected by surface errors. The positioning accuracy is low and problems such as skewed engraving and uneven depth may occur.

[0004] Therefore, the technical problem solved by this application is: how to improve the positioning accuracy of tire markings. Summary of the Invention

[0005] The main objective of this application is to provide a method for laser engraving on tires. By performing a "dimensionality reduction" process on the complete point cloud using polar coordinate expansion, positioning is transformed into simple geometric calculations on a two-dimensional plane. This not only reduces algorithm complexity but also facilitates process parameter adjustment and error troubleshooting through the intuitiveness of the two-dimensional plane. Moreover, relying on the dual reference of "marker point-inner edge," the positioning accuracy is improved by approximately 40% compared to single-reference positioning (such as relying solely on marker points).

[0006] In addition, a tire laser engraving device and system are also provided.

[0007] To achieve the above objectives, the technical solution adopted in this application is as follows:

[0008] A method for laser engraving on tires includes the following steps:

[0009] Step 1: Acquire 2D images and local point clouds of the tire from multiple perspectives;

[0010] Step 2: Obtain calibration extrinsic parameters through global calibration method. Based on the calibration extrinsic parameters, transform the local point cloud of each viewpoint to the global coordinate system to obtain the complete point cloud;

[0011] Step 3: Filter, denoise, binarize, and perform morphological transformation on the two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. Use a feature localization model to identify the marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then map the three-dimensional coordinates of the marker points in the local point cloud to the complete point cloud.

[0012] Step 4: Using coordinate transformation technology, the complete point cloud is converted into a planar point cloud through polar coordinates. The point cloud data of the inner edge of the tire is extracted from the planar point cloud. The point cloud data of the inner edge of the tire is processed by a linear fitting algorithm to obtain the inner edge fitting line.

[0013] Step 5: Based on the alignment of the marking points and the inner edge, a positioning algorithm is used to process the marking feature points to obtain the two-dimensional coordinates of the marking feature points; then, the coordinate inverse transformation formula is used to restore the two-dimensional coordinates of the marking feature points to three-dimensional coordinates. Based on the three-dimensional coordinates, a three-dimensional marking model is generated and the marking characters are engraved on the tire.

[0014] Preferably, step 3 includes the following sub-steps:

[0015] Step A1: Process the complete point cloud using statistical filtering and voxel downsampling methods;

[0016] Step A2: Filter, denoise, binarize, and perform morphological transformation processing on the two-dimensional images of the tire from multiple perspectives, filter the foreground and background, and convert the two-dimensional images of the tire from multiple perspectives to a consistent orientation to obtain a pure tire tread image.

[0017] Step A3: Use a feature localization model to identify marker points in the pure tire tread image; obtain the two-dimensional coordinates of the marker points in the pure tire tread image, and then use a transformation processing method to convert the two-dimensional coordinates of the marker points in the pure tire tread image into the three-dimensional coordinates of the local point cloud.

[0018] Step A4: Based on the calibration extrinsic parameters, map the 3D coordinates of the marker points in the local point cloud to the complete point cloud.

[0019] Preferably, the conversion processing method in step A3 is as follows: based on the distortion coefficient, the two-dimensional coordinates of the marker points in the two-dimensional image are corrected to obtain normalized coordinates; then, based on the normalized coordinates, the depth of the marker points is calculated using the structured light triangulation principle; and finally, based on the normalized coordinates and the depth, the two-dimensional coordinates of the marker points in the pure tire tread image are converted into the three-dimensional coordinates of the local point cloud.

[0020] Preferably, step 4 includes the following sub-steps:

[0021] Step B1: Perform symmetry verification, removal of non-tireside interference points, and rotation axis calibration on the complete point cloud in sequence;

[0022] Step B2: Using coordinate transformation technology, with the tire center axis as the polar axis, the position of the processed complete point cloud in the plane perpendicular to the polar axis is represented by polar coordinates of radial distance - circumferential angle. Then, the circumferential angle is expanded into a linear dimension, so that the processed complete point cloud is converted into a planar point cloud.

[0023] Step B3: Use a multi-stage filtering method to extract the tire inner edge point cloud from the planar point cloud;

[0024] Step B4: Using a piecewise linear fitting method, the inner edge point cloud of the tire is processed by a linear fitting algorithm to obtain the inner edge fitted line.

[0025] Preferably, in step B3, the multi-stage screening includes the following steps:

[0026] Step C1: Based on the preset radial distance range of the inner edge of the tire specification, remove point cloud data in the planar point cloud whose radial distance is not within the radial distance range of the inner edge, and obtain the inner edge point cloud of the candidate tire.

[0027] Step C2: Set a point cloud number threshold, use statistical filtering to calculate the number of points in the K-neighborhood of each candidate point in the inner edge point cloud of the candidate tire, and remove candidate points whose number of points in the K-neighborhood is less than the point cloud number threshold to obtain continuous inner edge point clouds of the tire.

[0028] Step C3: Set the U-axis coordinate threshold, calculate the U-axis coordinate difference between adjacent candidate points in the tire inner edge point cloud chain, and remove candidate points whose U-axis coordinate difference is greater than the horizontal coordinate threshold to obtain the complete tire inner edge point cloud;

[0029] Step C4: Perform morphological closing operation on the complete tire inner edge point cloud, and then fill in the sparse edge area by point cloud interpolation to obtain the tire inner edge point cloud.

[0030] Preferably, in step B4, the piecewise linear fitting includes the following steps:

[0031] Step D1: Divide the point cloud of the inner edge of the tire into multiple segments along the U-axis direction;

[0032] Step D2: Process the inner edge point cloud of each tire segment using a linear fitting algorithm to obtain the inner edge fitting line of each segment. When the engraving feature point is included in the inner edge fitting line of a certain segment, the inner edge fitting line of that segment is used for calculation.

[0033] Step D3: Set the V-coordinate threshold, connect the fitted straight lines of the inner edge of each segment, and judge the V-coordinate difference at the connection point of the fitted straight lines of adjacent segments. If the V-coordinate difference is less than the V-coordinate threshold, the usage requirements are met. If the V-coordinate difference is greater than the V-coordinate threshold, increase the number of segments and repeat steps D1 to D3, or smooth the connection point.

[0034] Preferably, in step 5, the specific formula for the positioning algorithm is as follows:

[0035]

[0036] in, These are the final two-dimensional coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. The azimuth angle of the engraved feature point in the original polar coordinates. Here, a and b are angle compensation values, and both are linear correction coefficients. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. This is the offset compensation value in the vertical direction;

[0037] The inverse coordinate transformation formula is as follows:

[0038]

[0039] in, These are the final coordinates of the engraving feature points in the global coordinate system. Let be the vertical coordinates of the engraved feature points in the cylindrical unfolded plane. The minimum radial distance within the target area. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. This is the reference depth value for the engraving area. This is the depth compensation value.

[0040] Preferably, the method further includes step 6: obtaining the intensity map, depth map and point cloud after the engraving is completed, detecting the quality of the tire engraving characters, and determining that the tire engraving quality is unqualified when any judgment condition is met.

[0041] The judgment conditions include:

[0042] (1) Segment all engraved characters from the intensity diagram, calculate the size of each engraved character, and compare it with the preset standard size. If the size of the engraved character exceeds the preset standard size range, it is deemed unqualified.

[0043] (2) Based on the point cloud, calculate the valley point set of the character image, calculate the average depth difference between the valley point set and the reference plane, and compare it with the preset standard depth. If the depth difference between the valley point and the reference plane is greater than the standard depth, it is judged as unqualified.

[0044] (3) For the circumferentially distributed engraved characters, establish a polar coordinate system with the center of the tire as the origin, extract the smallest bounding rectangle of each engraved character, determine the character center line, calculate the angle between the center line of each character and the polar axis, and statistically obtain the maximum value among all angles. Compare the maximum value among all angles with the preset standard angle. If the maximum value among all angles is greater than the preset standard angle, it is judged as unqualified.

[0045] (4) By edge detection, fit an approximate straight line of the tire edge, calculate the horizontal distance from the innermost pixel of each engraved character to the approximate straight line of the tire edge, and compare it with the standard edge distance. If the horizontal distance exceeds the standard edge distance, it is judged as unqualified.

[0046] Meanwhile, a tire laser marking device is also provided to implement the above-mentioned tire laser marking method, comprising the following units:

[0047] Camera acquisition module: used to acquire two-dimensional images and local point clouds of the tire from multiple perspectives;

[0048] Coordinate transformation module: Used to obtain calibration extrinsic parameters through global calibration method, and based on the calibration extrinsic parameters, transform the local point cloud of each view to the global coordinate system to obtain the complete point cloud;

[0049] The marker point recognition module is used to filter, denoise, binarize, and perform morphological transformation processing on two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. The feature localization model is used to recognize marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then, the three-dimensional coordinates of the marker points in the local point cloud are mapped to the complete point cloud.

[0050] Two-dimensional transformation module: Used to convert the complete point cloud into a planar point cloud through polar coordinates using coordinate transformation technology, extract the point cloud data of the inner edge of the tire from the planar point cloud, and process the point cloud data of the inner edge of the tire through a linear fitting algorithm to obtain the inner edge fitting line.

[0051] The engraving module is used to fit straight lines based on the marking points and inner edges, and to process the engraving feature points using a positioning algorithm to obtain the two-dimensional coordinates of the engraving feature points. Then, the two-dimensional coordinates of the engraving feature points are restored to three-dimensional coordinates using the inverse coordinate transformation formula. Based on the three-dimensional coordinates, a three-dimensional engraving model is generated and the engraved characters are engraved on the tire.

[0052] Meanwhile, a tire laser engraving system is also provided, including a scanning unit, a first tire turning unit, an engraving unit, a vision inspection unit, a second tire turning unit, and a material cutting unit connected in sequence;

[0053] The scanning unit includes a first conveying module, a scanning module, and a first clamping module; two first clamping modules are symmetrically arranged on the first conveying module along the tire conveying direction, the scanning module is connected to the first conveying module and located above the clamping module, and the first conveying module is used to convey the tire to the tire turning unit;

[0054] The first tire turning unit includes a second conveying module, a tire turning module, and a baffle plate; the baffle plate and the tire turning module are sequentially arranged on the second conveying module along the tire conveying direction; the tire turning module is used to turn the tire over; the baffle plate is used to assist the tire turning over; and the second conveying module is used to convey the tire to the marking module.

[0055] The marking unit includes the aforementioned tire laser marking device, a third conveying module, and a second clamping module. The two second clamping modules are symmetrically arranged on the third conveying module along the tire conveying direction, and the tire laser marking device is located above the second clamping module. The third conveying module is used to convey the tire to the vision inspection unit.

[0056] The vision inspection unit includes a fourth conveying module, a vision inspection module, and a third clamping module. Two third clamping modules are symmetrically arranged on the fourth conveying module along the tire conveying direction. The vision inspection module is connected to the fourth conveying module and is located above the third clamping module. The third clamping module is used to clamp the tire and drive the tire to rotate. The vision inspection module is used to detect the characters engraved on the tire. The fourth conveying module is used to convey the tire to the second tire turning unit.

[0057] The second tire-turning unit is used to turn the tire over and transport it to the unloading unit.

[0058] One of the above-mentioned technical solutions in this application has at least one of the following advantages or beneficial effects:

[0059] The tire marking method of this application, through "dimensionality reduction" processing of the complete point cloud by polar coordinate unfolding, transforms the irregular three-dimensional ring edge into a regular two-dimensional straight line, simplifying the radial distance measurement logic: the radial positioning of subsequent marking feature points no longer requires complex calculation of three-dimensional spatial distance, but only the vertical distance from the marking feature point to the fitted straight line in the unfolded image, significantly reducing the complexity of the positioning algorithm. At the same time, the fitted reference straight line can also compensate for the irregular deviation of the inner edge caused by the slight ellipticity of the tire and installation eccentricity, improving the robustness of positioning. Furthermore, by using the marking point as the circumferential reference and the fitted straight line of the inner edge as the radial reference, and through the two key geometric parameters of "angle between the marking feature point and the marking point" and "distance from the marking feature point to the inner edge" preset by the process, the precise positioning of the marking feature point is achieved in the two-dimensional plane after polar coordinate unfolding, further improving the positioning accuracy. Compared with single reference positioning (such as relying solely on the marking point), the positioning accuracy is improved by approximately 40%. Attached Figure Description

[0060] The present application will be further described below with reference to the accompanying drawings and embodiments;

[0061] Figure 1 This is a flowchart of the tire laser engraving method in Example 1;

[0062] Figure 2 This is a schematic diagram of the arrangement of the 3D structured light cameras in Example 1;

[0063] Figure 3 This is a block diagram of the tire laser engraving device of Example 2;

[0064] Figure 4 This is a schematic diagram of the scanning unit in Example 3;

[0065] Figure 5 This is a schematic diagram of the structure of the first clamping module in Embodiment 3;

[0066] Figure 6 This is a schematic diagram of the structure of the first tire-turning unit in Example 3;

[0067] Figure 7 This is a schematic diagram of the engraving unit in Example 3;

[0068] Figure 8 This is a schematic diagram of the structure of the second clamping module in Embodiment 3;

[0069] Figure 9 This is a schematic diagram of the structure of the visual detection unit in Example 3;

[0070] Figure 10 This is a schematic diagram of the feeding unit in Example 3.

[0071] The diagram includes the following labels: barcode scanning unit 1; first tire turning unit 2; engraving unit 3; visual inspection unit 4; unloading unit 5; first conveying module 11; barcode scanning module 12; first clamping module 13; second conveying module 21; tire turning module 22; baffle plate 23; tire laser engraving device 31; third conveying module 32; second clamping module 33; fourth conveying module 41; visual inspection module 42; third clamping module 43; cylinder 131; connecting rod 132; roller 133; first rotating rod 221; second rotating rod 222; third rotating rod 223; tire turning frame 224; tire turning rod 225; motor 331; lead screw 332; clamping wheel 333. Detailed Implementation

[0072] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0073] The following disclosure provides many different implementation methods or examples for different schemes of implementing this application.

[0074] Example 1

[0075] refer to Figure 1-2 A method for laser engraving on tires includes the following steps:

[0076] Step 1: Acquire 2D images and local point clouds of the tire from multiple perspectives;

[0077] In this embodiment, a 3D structured light camera is used to acquire two-dimensional images and local point clouds of the tire from multiple perspectives, specifically arranged as follows: Figure 2 As shown, two 3D structured light cameras are arranged side-by-side vertically, with equal installation heights. Two horizontal points (X1 and X2) are set to trigger image capture, resulting in a total of four image capture stations. These stations are distributed around the tire's centering center at the upper left, upper right, lower left, and lower right, with a 10%–15% overlap in the measurement areas of adjacent stations (preferably 15% in this embodiment). This method allows for the acquisition of two-dimensional images and local point clouds of the tire from four different perspectives.

[0078] Step 2: Obtain calibration extrinsic parameters through global calibration method. Based on the calibration extrinsic parameters, transform the local point cloud of each viewpoint to the global coordinate system to obtain the complete point cloud;

[0079] In this embodiment, the global calibration method specifically involves placing a calibration board in the measurement area to ensure that each 3D structured light camera can capture the feature points of the calibration board. The extrinsic parameters are calculated by solving the correspondence between the feature points in the camera coordinate system and the global coordinate system. These extrinsic parameters include the rotation matrix. Translation vector For the layout of the four photo-taking stations in this engraving station, the calibration accuracy must be guaranteed: rotation angle error ≤ 0.1°, translation error ≤ 0.02mm.

[0080] Based on the calibrated calibration extrinsic parameters (rotation matrix) Translation vector To optimize calibration extrinsic parameters, a closed-condition adjustment method is employed to reduce cumulative errors. Specifically, the closed-condition adjustment method involves setting the coordinates of any calibration target point in the camera's i-coordinate system as P. i Through external parameters (R) i , T i This can be transformed to the global coordinate system P, i.e.: P = R i P i + T i (i=1,2,3,4); Utilizing the closed-loop property, the target point's coordinates, after traveling from camera 1 through cameras 2, 3, and 4 back to camera 1, should correspond to the initial coordinates P. i Consistent, construct a closed equation:

[0081]

[0082] In the formula, Let R be the rotation matrix. i The inverse matrix (because the rotation matrix is ​​an orthogonal matrix), =R i T (transpose matrix); P2 is the measured coordinate of the target point in the camera 2 coordinate system, and the rest are similar.

[0083] Considering the existence of errors in the initial external parameters, the actual closed equation has a closure error v (residual), which is corrected as follows:

[0084]

[0085] Since the external parameter error is small, the closed equation will be adjusted relative to the initial external parameter (R). i , T i Performing a Taylor expansion, ignoring higher-order terms, transforms the equation into a linear error equation:

[0086]

[0087] in:

[0088] ΔX: Extrinsic parameter correction vector, containing the rotation matrix R for each camera.i The three angle corrections and the translation vector T i There are 3 coordinate corrections, and a total of 4 × 6 = 24 unknowns;

[0089] A: The coefficient matrix is ​​composed of the partial derivatives of the closed equation with respect to each unknown external parameter, reflecting the influence of the external parameter correction on the closure error;

[0090] : The constant term is calculated from the initial external parameters and the measured coordinates of the target point, i.e., the negative value of the initial closure error;

[0091] : Residual vector, i.e., the optimized closure error.

[0092] Weights are assigned based on the initial calibration accuracy of the external parameters; the higher the initial calibration accuracy, the larger the weight, and vice versa. The weight matrix P is determined using unit weighted mean square error, satisfying the following condition: ,in, Let be the unit weight mean error, and Q be the covariance matrix of the initial extrinsic parameter error.

[0093] In the context of tire 3D camera calibration, the camera calibration accuracy is consistent, and equal weighting (P is the identity matrix) can be used to simplify calculations and meet accuracy requirements.

[0094] According to the least squares principle, minimize the sum of squared residuals v T Pv, solve for the extrinsic parameter correction ΔX, the optimal solution is:

[0095]

[0096] In the formula, A T Let A be the transpose of the coefficient matrix A, (A T PA) -1 For A T The inverse matrix of PA.

[0097] Optimal extrinsic parameter calculation: The correction amount ΔX obtained from the solution is superimposed on the initial extrinsic parameters to obtain the optimized extrinsic parameters R. i ^opt = R i + ΔR i T i ^opt = T i + ΔT i ;

[0098] Closure error verification: Substitute the optimized extrinsic parameters into the closure equation and calculate the residual v. If the root mean square of the residual (RMS) is ≤0.01mm, it meets the requirements for tire 3D point cloud splicing and marking positioning, and the adjustment is completed.

[0099] Iterative optimization: If the residual does not meet the standard, the optimized extrinsic parameter is used as the new initial value, and the above calculation is repeated until the residual meets the accuracy requirements (generally, convergence can be achieved in 2-3 iterations).

[0100] The local point clouds collected from the four photography stations are uniformly transformed into the global coordinate system. For any local point in the local point cloud of the i-th photography station... By using the external parameter transformation formula, the three-dimensional coordinates of the local point cloud in the global coordinate system can be obtained.

[0101] The formula for converting extrinsic parameters is:

[0102] in, The target point's three-dimensional coordinates in the global coordinate system. For the first Rotation matrix corresponding to each camera For the target point at the 1st Three-dimensional coordinates in a camera coordinate system For the first The translation vector of each camera.

[0103] The expansion of the extrinsic parameter transformation formula is:

[0104]

[0105] in, The target point's three-dimensional coordinates in the global coordinate system. It is an orthogonal matrix that describes the pose of the camera coordinate system relative to the global coordinate system (rotation angles around the X / Y / Z axes). Let be the three-dimensional coordinates of the target point in the i-th camera coordinate system. Let be the translation vector corresponding to the i-th camera.

[0106] Furthermore, the parameter transformation formula is a rigid body transformation (only rotation and translation, without scaling or deformation), and the rotation matrix is ​​obtained through calibration using a 3D structured light camera. Translation vector This allows the local point cloud measured by each 3D structured light camera to be uniformly transformed into a global coordinate system, enabling the stitching of local point clouds from multiple perspectives and global positioning.

[0107] Step 3: Filter, denoise, binarize, and perform morphological transformation on the two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. Use a feature localization model to identify the marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then map the three-dimensional coordinates of the marker points in the local point cloud to the complete point cloud.

[0108] Preferably, step 3 includes the following sub-steps:

[0109] Step A1: Process the complete point cloud using statistical filtering and voxel downsampling methods;

[0110] In this embodiment, the stitched complete point cloud is denoised and optimized: a statistical filtering algorithm is used to remove discrete noise points other than 3σ (σ is the standard deviation of the distance mean of the complete point cloud), and then the point cloud density of the complete point cloud is reduced by a voxel downsampling method (to improve computational efficiency while retaining key geometric features).

[0111] It should be noted that:

[0112] The specific process of the statistical filtering algorithm is as follows: traverse each point P in the complete point cloud. i Retrieve all neighboring points within its k-neighborhood (for tire point clouds, a value of k of 50-100 is recommended to match the point cloud density characteristics of tires). Calculate point P. i The Euclidean distances to all its neighbors are used to obtain the distance set. The mean μ and standard deviation σ of the distance set are calculated using the following formula: ; Set the threshold to μ+3σ (3σ criterion), if point P i Points whose average neighborhood distance is greater than the threshold are identified as discrete noise points and removed; points whose average distance is less than or equal to the threshold are retained to complete point cloud denoising.

[0113] To address the isolated noise that can easily occur in the point cloud on the black surface of the tire, the neighborhood k can be adjusted to 80 to ensure that point clouds in key areas such as the tire sidewall and inner edge are not mistakenly removed.

[0114] The specific process of the voxel downsampling method is as follows: Divide the three-dimensional space of the complete point cloud into cubic voxel meshes of equal size. The side length of the voxels is set according to the point cloud density of the tire (0.5-1mm is recommended to preserve key geometric features such as tire markings and markers). Traverse all point clouds, mapping each point to the corresponding voxel mesh, and record the set of point clouds contained in each voxel. For each non-empty voxel, calculate the geometric center (centroid) of all points within it, using the following formula: Where n is the number of points within the voxel. Let be the coordinates of the i-th point within a voxel, and use this centroid point to replace all points within the voxel. Collect the centroid points of all voxels to form a complete point cloud of the tire after density reduction. While preserving key features such as the tread, inner edge, and marker points, the number of points in the point cloud is reduced by 60%–80%, improving the computational efficiency of subsequent marker point recognition and coordinate mapping. For the point cloud along the tire circumference, the voxel mesh is uniformly divided along the circumference to avoid loss of features of the voxels containing etched points and marker points.

[0115] Step A2: Filter, denoise, binarize, and perform morphological transformation processing on the two-dimensional images of the tire from multiple perspectives, filter the foreground and background, and convert the two-dimensional images of the tire from multiple perspectives to a consistent orientation to obtain a pure tire tread image.

[0116] In this embodiment, the specific process of filtering and denoising is as follows: the two-dimensional images of the tire from multiple perspectives are converted to grayscale, and the color images are converted into single-channel grayscale images; Gaussian filtering is used to smooth the grayscale images, suppressing Gaussian noise and high-frequency interference generated during image acquisition, and preserving key edge information such as tire contours and marking points; then the size of the filtering kernel is adaptively adjusted according to the noise intensity of the two-dimensional image, so as to ensure that the features of the target area are not blurred while removing noise.

[0117] The specific process of binarization is as follows: calculate the grayscale histogram of the grayscale image and determine the grayscale segmentation threshold between the foreground and the background; use adaptive thresholding or Otsu's method to achieve automatic binarization, dividing the image into foreground (tire area) and background (non-tire area); set pixels greater than the threshold to white (foreground) and pixels less than the threshold to black (background) to obtain a black and white binary image, highlighting the features of the tire area and the marker point.

[0118] The specific process of morphological transformation processing is as follows: First, morphological closing operations are performed on the binary image, involving dilation followed by erosion to fill in small holes, gaps, and breaks within the tire region. Then, morphological opening operations are performed, involving erosion followed by dilation to remove isolated noise, burrs, and small interference blocks from the image. Morphological processing ensures the tire region has a complete outline and smooth edges, achieving thorough separation of foreground and background, resulting in a clean tire tread image. Based on the camera's external orientation, the tire images from the four stations are rotated to a standard orientation, ensuring consistent target point positions and angles across different viewpoints, thus improving the stability and success rate of marker feature localization. This method filters out the foreground and background in the 2D image, and then converts the different orientations of the 2D images from the four shooting stations to a consistent orientation. For example, the tire orientation in the first 2D image is 0°, in the second it is 90°, in the third it is 180°, and in the fourth it is 270°. Rotating the second, third, and fourth 2D images ensures that the orientation of the first, second, third, and fourth images is all 0°. This method yields pure tire tread images with consistent angle orientation and filtered background. The consistency of invariant features of target points in each two-dimensional image is improved after processing, effectively increasing the success rate of feature localization.

[0119] Step A3: Use a feature localization model to identify marker points in the pure tire tread image; obtain the two-dimensional coordinates of the marker points in the pure tire tread image, and then use a transformation processing method to convert the two-dimensional coordinates of the marker points in the pure tire tread image into the three-dimensional coordinates of the local point cloud.

[0120] In industrial settings, marker points are typically designed with regular geometric shapes (such as circular or square protrusions) to provide high contrast and low noise, making them suitable for point cloud data characteristics measured by structured light.

[0121] In this embodiment, to address the issue of varying tire tread patterns, markings, and characters, a feature localization model based on deep learning is employed, such as PointNet, 3D-JEPA, or Mech-Mind's proprietary deep learning algorithm. In this embodiment, the feature localization model uses Mech-Mind's proprietary deep learning algorithm, and historical two-dimensional images are used to train Mech-Mind's proprietary deep learning algorithm with two-dimensional image annotation.

[0122] The pure tire tread image is input into the feature localization model to identify the two-dimensional coordinates of the marker points in the pure tire tread image. Then, a transformation processing method is used to obtain the three-dimensional coordinates of the marker points in the local point cloud, specifically:

[0123] First, using the distortion coefficients in the intrinsic parameter matrix K of the 3D structured light camera ( , The radial distortion coefficient is... , (where tangential distortion coefficient is used) to correct the two-dimensional coordinates of marker points in the two-dimensional image, obtaining normalized coordinates. The specific formula used is as follows:

[0124]

[0125]

[0126] in, The focal length on the x-axis is a camera intrinsic parameter. The focal length along the y-axis is a camera intrinsic parameter. This represents the x-coordinate of the principal point in the camera's intrinsic parameters, i.e., the x-coordinate offset of the image coordinate system origin from the camera sensor. This represents the ordinate of the principal point in the camera's intrinsic parameters, i.e., the offset of the ordinate of the image coordinate system origin from the camera sensor. The x-coordinate of the corrected marker point. The ordinate of the corrected marker point. To identify the original x-coordinate of the point in the 2D image, To identify the original ordinate of the point in the two-dimensional image, , The radial distortion coefficient is... The distance from the original coordinates of the marker point to the principal point. , The tangential distortion coefficient is... , Normalized coordinates;

[0127] Based on normalized coordinates, the depth of the marker point is calculated using the structured light triangulation principle. The specific formula is as follows:

[0128]

[0129] in, The depth of the marker point, The baseline distance between the projector and the camera in a structured light camera is the straight-line distance between the optical centers of the projector and the camera, obtained through camera calibration. The x-axis focal length in the camera's intrinsic parameters is related to the intrinsic parameter matrix K mentioned earlier. Consistent, The normalized x-coordinate of the marker point on the left graph. The normalized x-coordinate of the marker point on the right graph;

[0130] Then, based on normalized coordinates and depth, the two-dimensional coordinates of the marker points in the pure tire tread image are converted into three-dimensional coordinates of the local point cloud. The specific formula is as follows:

[0131]

[0132] in, , To identify the XY coordinates of a point in the camera coordinate system, , To identify the normalized coordinates of the points on the left graph, The depth of the marker point.

[0133] After identifying the marker point, the positioning accuracy of the marker point is verified. Specifically, multiple measurements are performed on the same marker point, and the coordinate deviation between the multiple measurements is detected. If the coordinate deviation is greater than the deviation threshold (in this embodiment, the deviation threshold is 0.02mm), steps A1 to A3 are repeated until the coordinate deviation is less than the deviation threshold.

[0134] Step A4: Based on the calibration extrinsic parameters, map the 3D coordinates of the marker points in the local point cloud to the complete point cloud.

[0135] Using the extrinsic parameter transformation formula in step 2, the three-dimensional coordinates of the marker points in the local point cloud are mapped to the complete point cloud.

[0136] Step 4: Using coordinate transformation technology, the complete point cloud is converted into a planar point cloud through polar coordinates. The point cloud data of the inner edge of the tire is extracted from the planar point cloud. The point cloud data of the inner edge of the tire is processed by a linear fitting algorithm to obtain the inner edge fitting line.

[0137] Coordinate system definition: 3D global coordinate system O-XYZ: The coordinate system in which the complete point cloud is located. The Z-axis coincides with the tire center axis (vertical direction), the XY plane is a horizontal plane, and the origin O is the intersection of the tire center axis and the XY plane. The global coordinate system is consistent with the global measurement coordinate system when the four photo-taking stations are collecting data, to ensure the continuity of the point cloud coordinates.

[0138] 2D polar coordinate system O''-UV: The unfolded rectangular coordinate system, with the U axis in the horizontal direction (corresponding to the tire circumference angle θ, which, after unfolding, linearly maps the radian range of [0, 2π) to a continuous length range), and the V axis in the vertical direction (corresponding to the tire radial distance r, i.e., the distance from the point to the central axis). The origin O'' corresponds to the intersection of the tire circumference 0° angle and the minimum radial distance (inner edge radius).

[0139] Preferably, step 4 includes the following sub-steps:

[0140] Step B1: Perform symmetry verification, removal of non-tireside interference points, and rotation axis calibration on the complete point cloud in sequence;

[0141] Before converting a complete point cloud to a planar point cloud using polar coordinates, the complete point cloud needs to be processed, specifically including:

[0142] ① Symmetry verification: The rotational symmetry characteristics of the tire point cloud are confirmed through complete point cloud analysis to ensure that the point cloud density around the central axis is uniform and the symmetry error is ≤0.5%, so as to avoid unfolding deformation due to asymmetry;

[0143] ② Point cloud preprocessing: Non-sideway interference points such as wheel hubs and ground are removed through region segmentation, retaining only the effective point cloud of the tire sidewall (radial distance range: r_inner~r_outer, where r_inner is the inner edge radius and r_outer is the outer edge radius); at the same time, the tire sidewall point cloud is denoised to remove the influence of discrete noise points on the unfolding accuracy. The denoising process is as follows: neighborhood statistical analysis is performed on the segmented effective point cloud of the tire sidewall, and the number of neighborhood search points is set to k=20 (the density adaptation value of the tire sidewall point cloud). The average Euclidean distance from each point to its k nearest neighbors is calculated; based on the average distance of all points, the overall mean μ and standard deviation σ are calculated, and discrete points with an average distance exceeding the range of μ+2σ are removed (these points are global noise points far away from the main body of the tire sidewall, such as environmental noise points and scanning reflection pseudo points). Local neighborhood verification was performed on the statistically filtered point cloud. A search radius of R = 0.5 mm was set (to match the tire sidewall scanning accuracy while considering point cloud density), and a minimum number of neighboring points N = 5. Each point was traversed, and the number of neighboring points within its search radius was counted. Isolated points with fewer than N neighboring points were removed (these points represent sporadic noise in the sidewall point cloud, such as scan jump points, lacking continuous sidewall surface support). Gaussian smoothing (Gaussian kernel radius 0.3 mm, smoothing times 1) was applied to the denoised sidewall point cloud to eliminate minor local jitter and ensure the continuity of the sidewall point cloud surface. If minor voids (not due to defects in the sidewall structure itself) existed after denoising, neighborhood interpolation was used to fill them, filling the voids based on the coordinate trends of surrounding valid points to avoid point cloud discontinuities affecting subsequent polar coordinate transformations.

[0144] ③ Rotation axis calibration: Accurately determine the position of the tire's central axis. The center can be obtained by fitting the smallest circumcircle of the tire's inner edge point cloud, and then the central axis is generated along the vertical direction. The positioning error of the central axis must be ≤0.02mm, otherwise it will cause the point cloud to shift as a whole after unfolding.

[0145] Step B2: Using coordinate transformation technology, with the tire's central axis as the polar axis, the position of the processed complete point cloud in the plane perpendicular to the polar axis is represented by polar coordinates of radial distance minus circumferential angle. Then, the circumferential angle is expanded into a linear dimension, converting the processed complete point cloud into a planar point cloud. The planar point cloud is calculated using the following formula:

[0146] Polar coordinate radius calculation: ;

[0147] Polar coordinate angle calculation: ;

[0148] Cylindrical expansion scaling factor: ;

[0149] The horizontal coordinates of the unfolded plane: ;

[0150] The vertical coordinates of the unfolded plane: ;

[0151] From the above formula, we can obtain: ;

[0152] Where r is the vertical distance from the target point in the XY plane to the central axis (Z-axis), and X and Y are the horizontal and vertical coordinates of the target point in the global coordinate system. Let be the azimuth angle of the target point in the XY plane, with a value in the range [0, 2π). The scaling factor from the angle to the unfolded plane. The total horizontal length of the cylindrical target after unfolding. This coefficient represents the radian value corresponding to a complete circle. An angle in the range [0, 2π) can be mapped to a horizontal length in the range [0, L). θ represents the horizontal coordinates of the target point in the cylindrical unfolded plane, which is the result after scaling angle θ by a factor of k. arctan2(Y,X) is the built-in four-quadrant arctangent function, which can directly output angles in the range [-π,π], replacing the previous piecewise function and simplifying the calculation. V represents the vertical coordinates of the target point in the cylindrical unfolded plane. The minimum radial distance to all points within the target area can be subtracted to map the radial distance to a vertical coordinate starting from 0, thus achieving planarization of the cylindrical surface.

[0153] This set of formulas is used to convert the three-dimensional point cloud of a cylindrical surface into a two-dimensional planar image through polar coordinate mapping, so that traditional vision algorithms can be used to process the features that were originally on the curved surface.

[0154] From a technical perspective, the core function of this method is to eliminate the curvature effect of cylindrical surfaces: the point cloud originally distributed on the circumferential surface is transformed into a horizontal straight line in the circumferential direction after unfolding, and the radial thickness direction is transformed into a vertical direction in the 2D plane. This simplifies the originally complex 3D surface positioning problem into an intuitive 2D plane positioning problem. This transformation process does not change the key geometric features of the point cloud (such as the relative distance and angle relationship between points), but only changes the coordinate representation, reducing the algorithm complexity for subsequent operations such as inner edge fitting and lettering feature point positioning.

[0155] From a data processing perspective, the principle of polar coordinate unfolding relies on two core prerequisites: first, the target point cloud must have a clear axis of rotational symmetry to ensure the effectiveness of polar coordinate parameterization; second, the point cloud data must completely cover the rotationally symmetric region to avoid information gaps after unfolding due to missing data. For local point clouds acquired at four camera stations, since they are distributed in a circle around the central axis, they perfectly meet the applicable conditions for polar coordinate unfolding, thus enabling high-quality unfolding results through this method.

[0156] Compared with other point cloud projection technologies such as spherical projection and orthogonal projection, the unique advantage of the polar coordinate unfolding principle lies in its adaptability to rotationally symmetric targets: spherical projection is suitable for unfolding spherical targets, orthogonal projection is prone to stretching and deformation of the point cloud at the edge of cylindrical targets, while polar coordinate unfolding can preserve the local geometric accuracy of rotationally symmetric targets to the greatest extent, and is especially suitable for industrial scenarios such as tire markings where local positional accuracy is extremely important.

[0157] Step B3: Use a multi-stage filtering method to extract the tire inner edge point cloud from the planar point cloud;

[0158] The multi-stage screening includes the following steps:

[0159] Step C1: Based on the preset radial distance range of the inner edge of the tire specification, remove point cloud data in the planar point cloud whose radial distance is not within the radial distance range of the inner edge, and obtain the inner edge point cloud of the candidate tire.

[0160] In this embodiment, the preset inner edge radial distance range of the tire specification is [0, 4]. Point cloud data whose radial distance (V coordinate) is not within the inner edge radial distance range are removed from the planar point cloud, and point cloud data whose radial distance (V coordinate) is within the range are retained. Non-inner edge points such as the middle and outer edge of the tire sidewall are initially removed to obtain the candidate tire inner edge point cloud.

[0161] Step C2: Set a point cloud number threshold, use statistical filtering to calculate the number of points in the K-neighborhood of each candidate point in the inner edge point cloud of the candidate tire, and remove candidate points whose number of points in the K-neighborhood is less than the point cloud number threshold to obtain continuous inner edge point clouds of the tire.

[0162] Because the inner edge point cloud is a continuous ring structure, its point cloud density is uniform and continuous, while interference points (such as noise points caused by oil stains or dust) are mostly isolated and have extremely low density. A statistical filtering algorithm is used to calculate the number of points in the k-neighborhood (k is 50-100, adapted to the density of structured light point clouds) of each candidate point in the inner edge point cloud of the candidate tire. Isolated points with a neighboring point cloud number less than the point cloud number threshold are removed, and continuous inner edge point clouds are retained to obtain continuous inner edge point clouds of the tire.

[0163] Step C3: Set the U-axis coordinate threshold, calculate the U-axis coordinate difference between adjacent candidate points in the tire inner edge point cloud chain, and remove candidate points whose U-axis coordinate difference is greater than the horizontal coordinate threshold to obtain the complete tire inner edge point cloud;

[0164] The inner edge point cloud is continuously distributed along the U-axis (circumferential direction) in the unfolded image. The difference in U-coordinates between adjacent points should be less than the U-axis coordinate threshold (generally ≤0.5mm, corresponding to approximately 0.1° in the tire circumferential direction). By traversing the continuous inner edge point cloud of the tire, the difference in U-coordinates between adjacent points is checked, and breakpoints and abrupt changes with differences greater than the U-axis coordinate threshold are removed to ensure the integrity of the inner edge point cloud chain, thus obtaining the complete inner edge point cloud of the tire.

[0165] Step C4: Perform morphological closing operation on the complete tire inner edge point cloud, and then fill in the sparse edge area by point cloud interpolation to obtain the tire inner edge point cloud.

[0166] A morphological closing operation (dilation followed by erosion) is performed on the complete tire inner edge point cloud to fill in minor breaks caused by missing points and smooth the edge contour. Finally, point cloud interpolation is used to complete the sparse edge regions, ensuring uniform point cloud density (≥300 points / mm) to obtain the tire inner edge point cloud, providing sufficient and effective data for subsequent fitting. Specifically, the point cloud interpolation method involves: circumferentially ordering the tire inner edge point cloud after the morphological closing operation (rearranging the point cloud coordinates from smallest to largest based on the tire's rotation axis), ensuring continuous distribution of the point cloud along the inner edge circumferential contour; traversing the ordered point cloud, calculating and accumulating the Euclidean distance between adjacent points to obtain the cumulative arc length of the inner edge point cloud, and determining the arc length range of the sparse region (regions with adjacent point spacing >0.0033mm, corresponding to a density <300 points / mm). For sparse regions, using a target density of 300 points / mm as a benchmark, the number of interpolation points required between adjacent points is calculated (number of interpolation points = arc length of adjacent points (1 / 300) - 1). Linear interpolation is performed between the three-dimensional coordinates of adjacent points, generating interpolation points at equal intervals according to arc length, quickly filling the point cloud gaps in sparse regions, ensuring that the point cloud density meets the standard, and simultaneously conforming to the basic linear contour of the inner edge. Local neighborhood optimization is performed on the linearly interpolated point cloud, taking the 8 effective neighboring edge points of each interpolation point (the original point cloud after morphological closing operation), and using a distance-weighted average method to correct the coordinates of the interpolation points (the closer the neighboring points are, the greater their weight, and the weight coefficient is inversely proportional to the Euclidean distance), eliminating local deviations in linear interpolation, making the interpolation points better conform to the actual curved surface contour of the inner edge of the tire, and avoiding edge deformation caused by pure linear interpolation. The overall density of the inner edge point cloud after completion is recalculated to confirm that the global density is ≥300 points / mm. The circular ordered rearrangement is performed again to check the continuity and contour smoothness of the point cloud. Abnormal points (points with a deviation from neighboring points >0.001mm) that may occur during the interpolation process are removed to ensure the consistency of the inner edge point cloud of the tire.

[0167] Step B4: Using a piecewise linear fitting method, the inner edge point cloud of the tire is processed by a linear fitting algorithm to obtain the inner edge fitted line.

[0168] The linear fitting algorithm is as follows:

[0169]

[0170] in, Let x be the abscissa of the point cloud at the inner edge of the tire, representing the position coordinates of the point cloud along the tire's circumference. Let be the planar ordinate of the point cloud at the inner edge of the tire, representing the position coordinates of the point cloud along the radial direction of the tire. The slope of the line (characterizing the inclination of the inner edge; ideally, a=0). It is the line intercept (the reference V-coordinate representing the inner edge, which is the core reference value for radial positioning);

[0171] It should be noted that, The calculation process is as follows:

[0172]

[0173]

[0174] in, The sum of the products of the deviations from the mean of the U and V coordinates. Let U be the sum of squared deviations from the mean of the coordinates. The mean of the V-coordinates of the point cloud at the inner edge of the tire. The mean of the U-coordinates of the point cloud at the inner edge of the tire. Let be the horizontal coordinates of the i-th point in the unfolded plane. Let be the vertical coordinate of the i-th point in the unfolded plane. Let V be the sum of squared deviations from the mean of the coordinates;

[0175] The line parameters are obtained by solving the above formula. and This allows us to obtain a straight line fitted to the inner edge. For example, if a≈0 and b=3mm, then the inner edge is approximately a horizontal straight line, and the reference radial distance corresponds to V=3mm.

[0176] While ideally, the inner edge of a tire should be a horizontal straight line when unfolded, in reality, due to tire manufacturing errors (such as slight ellipticity), installation eccentricity, elastic deformation, and other factors, the inner edge may be slightly curved (appearing as a curve when unfolded). In this case, using a single straight line for fitting will result in a large error (RMS > 0.05mm), and a "segmented straight line fitting" strategy needs to be used for correction. The core principle is "local linear approximation", that is, the curved inner edge is divided into multiple regions of locally approximate straight lines, and multiple segments of the inner edge fitting straight lines are fitted separately.

[0177] The piecewise linear fitting includes the following steps:

[0178] Step D1: Divide the point cloud of the inner edge of the tire into multiple segments along the U-axis direction;

[0179] In this embodiment, the point cloud at the inner edge of the tire is divided into multiple segments along the U-axis direction (circumferential direction). The length of each segment is determined according to the degree of curvature. Generally, each segment corresponds to 5°-10° of the tire circumference (corresponding to a U-axis length of about 4-8mm, adapted according to the tire diameter), ensuring that the point cloud in each segment is approximately linearly distributed.

[0180] Step D2: Process the inner edge point cloud of each tire segment using a linear fitting algorithm to obtain the inner edge fitting line of each segment. When the engraving feature point is included in the inner edge fitting line of a certain segment, the inner edge fitting line of that segment is used for calculation.

[0181] For each segment of the tire's inner edge point cloud, a single least squares method is used to fit a straight line, obtaining the fitted straight line parameters (a1, b1), (a2, b2), ..., (a...). k b k (k is the number of segments), thus obtaining the inner edge fitting line for each segment. When locating subsequent lettering feature points, first, based on the U-shape of the lettering feature points... m The coordinates determine the segmented region to which it belongs, and then the inner edge fitting line corresponding to that region is used for positioning, achieving "regional positioning and precise adaptation". For example, if the engraving feature point U m =15mm, corresponding to the fitting region of the third segment, then V is calculated by calling a3 and b3 of the third segment. m .

[0182] Step D3: Set the V-coordinate threshold, connect the fitted straight lines of the inner edge of each segment, and judge the V-coordinate difference at the connection point of the fitted straight lines of adjacent segments. If the V-coordinate difference is less than the V-coordinate threshold, the usage requirements are met. If the V-coordinate difference is greater than the V-coordinate threshold, increase the number of segments and repeat steps D1 to D3, or smooth the connection point.

[0183] In this embodiment, the V-coordinate threshold is 0.02 mm. By judging the difference in V-coordinates at the junction of the fitted straight lines of two adjacent segments, positioning faults at the segmentation points are avoided.

[0184] Increase the number of segments and repeat steps D1 to D3. That is, when performing step D1 again, the number of segments should be more than the number of segments in the previous step.

[0185] The specific process for smoothing the junction is as follows: Let the two adjacent fitted straight lines be the first segment. The latter part The junction is the common U-shaped boundary point of the two segments. Extract the core fitting parameters (slope) of the two straight lines respectively. ,intercept ), and the V coordinate value at the junction , Confirm the V-coordinate difference at the junction. (ΔV > 0.02 mm). At the joint... Centered on the target segment, extend 5% of the single-segment U-direction length forward (L1) and backward (L2) to define a smooth transition zone. (ΔU is 5% of the circumferential length of a single segment, adapted to the circumferential unfolding characteristics of the inner edge of the tire, avoiding an excessively wide transition interval that would affect the overall fitting accuracy); based on the intercept of two straight lines. The intercept fusion coefficient k within the transition interval is calculated using the distance-weighted method: for any point U within the transition interval... x , Where k∈[0,1] (k approaches 1 on the first side and k approaches 0 on the second side); determine the smooth fitting formula within the transition interval: This process achieves a linear gradient of slope and intercept, eliminating abrupt coordinate changes at the transition points. The original tire inner edge point cloud within the transition zone is extracted, and the least squares method is used to refit this point cloud using a smoothing fitting formula, resulting in a smoothed fitting line L3 for the transition segment. This ensures the smoothed line closely matches the original point cloud. The transition segment L3 replaces the original transition sections L1 and L2, forming a continuous contour of "front non-transition area L1 + transition segment L3 + rear non-transition area L2." The non-transition area retains the parameters of the original fitted line, maintaining its positioning accuracy. Verification of the difference at the junction: Recalculate the V coordinate difference between the transition segment L3 and the non-transition zone between the preceding and following segments. The new difference should be ≤0.02mm to confirm the elimination of the fault. Verification of the overall fitting accuracy: Calculate the root mean square error (RMSE) of the fitting residual of the transition segment L3 to the point cloud. The result should be ≤0.002mm, consistent with the original segmental fitting accuracy, to avoid contour distortion caused by smoothing. Iterative processing: If the difference at the junction is still >0.02mm after smoothing, appropriately expand the transition interval (not exceeding 10% of the length of the single segment in the U direction), recalculate the fusion parameters and fit again until the accuracy requirements are met.

[0186] From a geometric perspective, the inner edge of the tire is a ring-shaped structure (approximately circular) around its central axis in three-dimensional space. After being unfolded in polar coordinates, each point on the ring corresponds to a unique coordinate of "circumferential angle θ - radial distance r" in the unfolded diagram. Since the radial distance r of the inner edge of the tire is basically constant (characteristics of an ideal cylindrical surface), the V coordinates (corresponding to radial distance r) of all inner edge points are basically the same after unfolding. Therefore, in the 2D unfolded system O''-UV, it appears as an approximately horizontal straight line (ideally, V is a fixed value, and the slope of the line a=0).

[0187] From a technical perspective, its core function is to transform irregular three-dimensional ring edges into regular two-dimensional straight lines, simplifying the radial distance measurement logic: subsequent radial positioning of engraved feature points no longer requires complex calculations of three-dimensional spatial distances; it only requires calculating the perpendicular distance from the engraved feature points in the unfolded image to the fitted straight line, significantly reducing the complexity of the positioning algorithm. Simultaneously, the fitted straight line can also compensate for minor tire ellipticity and irregular deviations in the inner edge caused by installation eccentricity, improving the robustness of the positioning.

[0188] Differences from general planar straight line fitting: Tire inner edge straight line fitting is a fitting application of "targeted adaptation of polar coordinate unfolded point cloud". It is necessary to focus on the angle mapping error and point cloud stretching deviation that may be introduced during the unfolding process. Therefore, before fitting, the unfolded inner edge point cloud needs to be preprocessed to ensure the effectiveness of the fitting model.

[0189] Step 5: Based on the alignment of the marking points and the inner edge, a positioning algorithm is used to process the marking feature points to obtain the two-dimensional coordinates of the marking feature points; then, the coordinate inverse transformation formula is used to restore the two-dimensional coordinates of the marking feature points to three-dimensional coordinates. Based on the three-dimensional coordinates, a three-dimensional marking model is generated and the marking characters are engraved on the tire.

[0190] In this embodiment, the marking requirement parameters of the corresponding part number are read from the MES system. The marking positioning parameters mainly include the angle α between the marking target area and the marking point, and the distance d between the marking target area and the inner edge of the tire.

[0191] The localization algorithm is used to process the engraved feature points to obtain their two-dimensional coordinates. The specific formula is as follows:

[0192] Calculation of the horizontal coordinates of the unfolded plane: ;

[0193] Vertical coordinate correction of the unfolded plane: ;

[0194] Angle correction: ;

[0195] Calculation of the original angle of the engraving feature points: ;

[0196] The combined form of the complete coordinate mapping is:

[0197] in, These are the final two-dimensional coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. The azimuth angle of the engraved feature point in the original polar coordinates. Here, a and b are angle compensation values, and both are linear correction coefficients. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. This is the offset compensation value in the vertical direction; Let be the vertical coordinates of the engraved feature points in the cylindrical unfolded plane. This is the final angle after offset compensation. These are the X and Y coordinates of the engraved feature points in the local coordinate system.

[0198] This set of formulas converts the engraving feature points on the tire surface into two-dimensional coordinates on the unfolded plane through polar coordinate mapping and linear correction, thereby eliminating the visual distortion caused by the curvature of the surface and making engraving detection more accurate.

[0199] After completing the two-dimensional positioning of the engraving feature points, the coordinate inverse transformation formula is used to restore the 2D coordinates on the unfolded plane back to the three-dimensional coordinates, thereby achieving the precise positioning of the engraving feature points in the global coordinate system.

[0200] The inverse coordinate transformation formula is as follows:

[0201] Calculated from the perspective of the horizontal coordinates of the unfolded plane: ;

[0202] Calculate the radial distance from the vertical coordinates of the unfolded plane: ;

[0203] Polar coordinates to 2D planar coordinates: ;

[0204] Depth coordinate correction: ;

[0205] The combined form of complete 3D coordinates is:

[0206]

[0207] in, These are the final coordinates of the engraving feature points in the global coordinate system. Let be the vertical coordinates of the engraved feature points in the cylindrical unfolded plane. The minimum radial distance within the target area. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. This is the reference depth value for the engraving area. This is the depth compensation value; The azimuth angle of the inscription feature point in polar coordinates is the horizontal coordinate of the unfolded plane. The inverse transform result, Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. The radial distance from the engraving feature point to the central axis of the cylinder. , The horizontal and vertical coordinates of the engraved feature points on the XY plane represent a standard polar coordinate to rectangular coordinate conversion, expressed as radial distance. and azimuth Restore the planar position. The depth coordinates of the engraved feature points in three-dimensional space.

[0208] The laser algorithm automatically generates a 3D lettering model that conforms to the tire tread shape based on the calculated 3D point cloud coordinates and engraves the characters on the tire. Laser algorithms are existing technology; this application uses the built-in laser algorithm of Han's Sensor Technology's laser 3D dynamic scanning system to generate the 3D lettering model. Other existing algorithms can also be used to generate the 3D lettering model.

[0209] Using the marking point as the circumferential reference and the fitted line of the inner edge as the radial reference, the system achieves precise positioning of the marking point in a two-dimensional plane after polar coordinate unfolding by employing two key geometric parameters preset in the process: the angle between the marking point and the marking point, and the distance from the marking point to the inner edge. Essentially, it leverages the "linearization" advantage of the tire's geometric features after polar coordinate unfolding to transform the complex positioning problem in three-dimensional space into an angle and distance constraint problem in a two-dimensional plane, significantly reducing positioning difficulty while ensuring accuracy.

[0210] From a positioning logic perspective, this principle relies on the synergistic effect of two core benchmarks: First, the "angular benchmark" (marker point). The marker point has a clear U-coordinate (corresponding to a circumferential angle) in the polar coordinate unfolded diagram. By presetting the included angle, the circumferential offset of the engraved feature point relative to the marker point can be determined, thus locking the U-coordinate of the engraved feature point. Second, the "radial benchmark" (inner edge fitted line). The fitted line has a clear V-coordinate benchmark in the unfolded diagram. By presetting the radial distance, the radial offset of the engraved feature point relative to the inner edge can be determined, thus locking the V-coordinate of the engraved feature point. These two benchmarks work together to form a dual constraint of "angle-distance," ensuring the uniqueness and accuracy of the engraved feature point positioning.

[0211] From a technical compatibility perspective, this method perfectly meets the industrial requirements of tire marking: on the one hand, after the tire is unfolded in polar coordinates, the geometric relationship between the circumferential and radial directions has been linearized, and the calculation of angles and distances does not need to consider the influence of surface curvature. The calculation logic is simple and efficient, and it meets the real-time requirements of industry (the positioning time of a single marking feature point is ≤0.1ms); on the other hand, the marking point and the inner edge are both stable geometric features on the tire, which are less affected by tire deformation and installation errors. Using these two as a reference can improve the robustness of positioning and avoid positioning deviations caused by the failure of a single reference.

[0212] Preferably, the method further includes step 6: obtaining the intensity map, depth map and point cloud after the engraving is completed by a contour measuring instrument, detecting the quality of the engraved characters on the tire, and determining that the tire engraving quality is unqualified when any one of the judgment conditions is met.

[0213] The judgment conditions include:

[0214] (1) Segment all engraved characters from the intensity diagram, calculate the size of each engraved character, and compare it with the preset standard size. If the size of the engraved character exceeds the preset standard size range, it is deemed unqualified.

[0215] In this embodiment, the intensity map is first denoised using Gaussian filtering, and then deep learning threshold segmentation is used to distinguish the engraved characters from the tire sidewall background, thus separating individual engraved characters. Based on all the segmented engraved characters, the number of pixels in the horizontal and vertical directions of each engraved character is counted. Combined with camera calibration parameters (pixel equivalent: 1 pixel = 0.01 mm), the actual width and height are calculated. That is, the width and height of the character are measured at the top, middle and bottom three sections, and the average value is calculated as the final width and height (the size of each engraved character). This is compared with the preset standard size (standard width and marked height). If it exceeds the preset standard size range, it is judged as unqualified; if it is within the preset standard size range, it is judged as qualified.

[0216] (2) Based on the point cloud, calculate the valley point set of the character image, calculate the average depth difference between the valley point set and the reference plane, and compare it with the preset standard depth. If the depth difference between the valley point and the reference plane is greater than the standard depth, it is judged as unqualified.

[0217] Extract the 3D point cloud of the engraved area from the point cloud, fit it to the tire sidewall reference plane (excluding the influence of surface curvature); locate the valley point set of each character, calculate the average depth difference with the reference plane, and compare it with the standard depth (in this embodiment, the standard depth is 0.4mm±0.1mm). If it exceeds the standard depth range, it is judged as unqualified; if it is within the standard depth range, it is judged as qualified.

[0218] (3) For the circumferentially distributed engraved characters, establish a polar coordinate system with the center of the tire as the origin, extract the smallest bounding rectangle of each engraved character, determine the character center line, calculate the angle between the center line of each character and the polar axis, and statistically obtain the maximum value among all angles. Compare the maximum value among all angles with the preset standard angle. If the maximum value among all angles is greater than the preset standard angle, it is judged as unqualified.

[0219] Based on the point cloud, a polar coordinate system is established with the tire center as the origin. The deviation between the character centerline and the preset standard angle is detected. Specifically, the smallest bounding rectangle of a single character is extracted to determine the character centerline. The angle between the centerline and the polar axis (preset reference direction) is calculated. The angle deviation of all characters is then counted, and the maximum value is taken as the detection result. It is compared with the preset standard angle. If it exceeds the preset standard angle range, it is judged as unqualified. If it is within the preset standard angle range, it is judged as qualified.

[0220] (4) By edge detection, fit an approximate straight line of the tire edge, calculate the horizontal distance from the innermost pixel of each engraved character to the approximate straight line of the tire edge, and compare it with the standard edge distance. If the horizontal distance exceeds the standard edge distance, it is judged as unqualified; if the horizontal distance is less than the standard edge distance, it is judged as qualified.

[0221] It should be noted that the preset standard dimensions, standard depth, preset standard angle, and standard edge distance are all obtained from the production information of tire marking, which is the processing accuracy requirement of tire marking.

[0222] The specific process of fitting an approximate straight line for the tire edge through edge detection is as follows: Set two thresholds for the Canny algorithm: a high threshold of 80 and a low threshold of 30 (to adapt to the gray-level contrast characteristics of the tire sidewall intensity map and avoid missing tire edges and falsely detecting fine edges of engraved characters); perform gradient calculation and non-maximum suppression on the filtered intensity map, retain the gradient peak points of edge pixels, and then filter and connect the edges through the dual thresholds to obtain a global edge pixel set that includes the tire edge, engraved character edges, and background clutter edges; based on the production information of the tire engraving (the circumferential distribution range of characters on the tire sidewall), delineate the region of interest (ROI) for the tire edge (avoiding the engraved character area and taking the fixed gray-level band area of ​​the inner edge of the tire sidewall), extract the pixels in this area from the global edge pixel set as candidate pixels for the tire edge, and remove irrelevant edge interference such as characters and background.

[0223] A neighborhood-based statistical filter is used, and the effective pixel determination condition of the 4-neighborhood (top, bottom, left, right) of a candidate pixel is set: if the number of edge pixels in the 4-neighborhood of a candidate pixel is less than 2, it is determined as an outlier and removed. After traversing all candidate pixels of the tire edge and removing outliers, a clean set of effective pixels of the tire edge is obtained. This set of pixels has a continuous linear distribution and fits the actual contour of the tire edge.

[0224] Based on the camera calibration parameters of the contour measuring instrument (pixel equivalent: 1 pixel = 0.01 mm), and combined with the transformation matrix between the image coordinate system and the physical detection coordinate system, the mapping from pixel coordinates to physical coordinates is completed. The transformation formula is as follows: (in (The coordinates are: X = X, Y = Y; ... This provides a unified physical dimension benchmark for subsequent linear fitting.

[0225] The physical coordinate fitting formula for assuming an approximately straight tire edge is: (Where X is the horizontal physical coordinate, Y is the vertical physical coordinate, k is the slope of the line, and c is the intercept; ideally, the tire edge is a horizontal straight line, and k≈0); Calculate the statistics of the coordinate set: number of effective pixels n, mean of horizontal coordinates. Mean of vertical coordinates , , Substituting the values ​​into the core formula of the least squares method to solve for the slope k and intercept c, we can ensure that the sum of the squared vertical distances from all valid pixels to the fitted line is minimized. The formula is as follows: Substituting the obtained k and c into the fitting formula, we obtain an approximate straight line at the edge of the tire, which serves as the reference straight line for measuring the horizontal distance of the innermost pixel of the engraved character.

[0226] The accuracy of the fitted approximate straight line of the tire edge is verified to avoid distortion of the fitted line due to pixel extraction deviation. The verification criteria are as follows: calculate the average vertical distance from all effective pixels of the tire edge to the fitted line, requiring the average distance to be ≤0.03mm and the root mean square error (RMSE) of the fitting residual to be ≤0.05mm. If the accuracy requirements are not met, the ROI area of ​​the tire edge is expanded again, additional candidate pixels of the edge are extracted, and the above process is repeated to fit again until the accuracy requirements are met.

[0227] Example 2

[0228] refer to Figure 3 A tire laser marking device for implementing the above-mentioned tire laser marking method includes the following units:

[0229] Camera acquisition module: used to acquire two-dimensional images and local point clouds of the tire from multiple perspectives;

[0230] Coordinate transformation module: Used to obtain calibration extrinsic parameters through global calibration method, and based on the calibration extrinsic parameters, transform the local point cloud of each view to the global coordinate system to obtain the complete point cloud;

[0231] The marker point recognition module is used to filter, denoise, binarize, and perform morphological transformation processing on two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. The feature localization model is used to recognize marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then, the three-dimensional coordinates of the marker points in the local point cloud are mapped to the complete point cloud.

[0232] Two-dimensional transformation module: Used to convert the complete point cloud into a planar point cloud through polar coordinates using coordinate transformation technology, extract the point cloud data of the inner edge of the tire from the planar point cloud, and process the point cloud data of the inner edge of the tire through a linear fitting algorithm to obtain the inner edge fitting line.

[0233] The engraving module is used to fit straight lines based on the marking points and inner edges, and to process the engraving feature points using a positioning algorithm to obtain the two-dimensional coordinates of the engraving feature points. Then, the two-dimensional coordinates of the engraving feature points are restored to three-dimensional coordinates using the inverse coordinate transformation formula. Based on the three-dimensional coordinates, a three-dimensional engraving model is generated and the engraved characters are engraved on the tire.

[0234] In this embodiment, the specific workflow of the tire laser marking device is as follows: The camera acquisition module acquires two-dimensional images and local point clouds of the tire from multiple perspectives, then sends the local point clouds to the coordinate transformation module and the two-dimensional images to the marker point recognition module; the coordinate transformation module transforms the four local point clouds into the global coordinate system to obtain a complete point cloud, and then sends the complete point cloud to the marker point recognition module; the marker point recognition module processes the two-dimensional images to obtain a pure tire tread image, identifies the two-dimensional coordinates of the marker points from the pure tire tread image, converts the two-dimensional coordinates into the three-dimensional coordinates of the local point clouds, and then maps the three-dimensional coordinates of the local point clouds to the complete point cloud, thereby obtaining a complete point cloud with the three-dimensional coordinates of the marker points. The point cloud is then processed, and the complete point cloud with the 3D coordinates of the marker points is sent to the 2D conversion module. The 2D conversion module converts the complete point cloud into a planar point cloud using polar coordinates. It then extracts the point cloud data of the inner edge of the tire from the planar point cloud and processes the point cloud data of the inner edge of the tire using a linear fitting algorithm to obtain the inner edge fitting line. The inner edge fitting line is then sent to the engraving module. Based on the dual positioning of the marker points and the inner edge fitting line, the engraving module obtains the 2D coordinates of the engraving feature points in the planar point cloud. Then, using the inverse coordinate transformation formula, it restores the 2D coordinates of the engraving feature points to 3D coordinates and generates a 3D engraving model. Based on the 3D engraving model, it drives the laser to engrave the engraved characters on the tire.

[0235] Example 3

[0236] refer to Figure 4-10 A tire laser engraving system includes a scanning unit 1, a first tire turning unit 2, an engraving unit 3, a vision inspection unit 4, a second tire turning unit, and a material cutting unit 5 connected in sequence.

[0237] The scanning unit 1 includes a first conveying module 11, a scanning module 12, and a first clamping module 13; two first clamping modules 13 are symmetrically arranged on the first conveying module 11 along the tire conveying direction, the scanning module 12 is connected to the first conveying module 11 and is located above the clamping module, and the first conveying module 11 is used to convey the tire to the tire turning unit.

[0238] In this embodiment, the first conveying module 11 is a Mecanum wheel conveyor, and the scanning module 12 includes a barcode scanner and a Z-axis linear module. The Z-axis linear module is connected to the first conveying module 11, and the barcode scanner is connected to the Z-axis linear module. The barcode scanner is a Delijie Matrix smart barcode reader. 410N700-010, the linear module is used to move the barcode scanner closer to or away from the tire. The first clamping module 13 includes a cylinder 131 and symmetrically arranged connecting rods 132. The connecting rods are provided with symmetrically arranged rollers 133. The axial direction of the rollers 133 extends vertically. The end of the cylinder 131 is rotatably connected to the end of one of the connecting rods 132. The two connecting rods 132 are meshed by gears. The power output end of the cylinder 131 extends out, driving one of the connecting rods 132 to rotate. The connecting rod 132 drives the other connecting rod 132 to rotate through gears, thereby realizing that the rollers 133 on the two connecting rods 132 move closer to or away from each other, thereby realizing the clamping or releasing of the tire by the two first clamping modules 13.

[0239] The first tire turning unit 2 includes a second conveying module 21, a tire turning module 22, and a baffle plate 23; the baffle plate 23 and the tire turning module 22 are sequentially arranged on the second conveying module 21 along the tire conveying direction. The tire turning module 22 is used to turn the tire over, the baffle plate 23 is used to assist the tire turning over, and the second conveying module 21 is used to convey the tire to the marking module.

[0240] The first tire-turning unit 2 also includes a positioning sensor, specifically model SICK GSE6-N1212. When the tire arrives at the tire-turning station, the positioning sensor sends a signal to drive the tire-turning module 22 to perform the tire-turning operation.

[0241] In this embodiment, the second conveying module 21 is a roller conveyor line, and the baffle plate 23 is connected to the second conveying module 21. The baffle plate 23 has an angle of 45° with the horizontal direction. Of course, the angle of the baffle plate 23 can also be changed according to actual needs. The tire turning module 22 includes a motor, a first rotating rod 221, a second rotating rod 222, a third rotating rod 223, and a tire turning frame 224. The tire turning frame 224 is rotatably connected to the second conveying module 21. The tire turning frame 224 is provided with a tire turning rod 225. The initial position of the tire turning rod 225 is in the gap of the second conveying module 21 to avoid affecting the tire conveying. The motor is connected to the first rotating rod 221 through gears. One end of each of the two second rotating rods 222 is rotatably connected to the two ends of the first rotating rod 221. The other end of the second rotating rod 222 is rotatably connected to one end of the third rotating rod 223. The other end of the third rotating rod 223 is rotatably connected to the tire turning frame 224. The motor drives the first rotating rod 221 to rotate, which in turn drives the second rotating rod 222 to rotate. The rotation of the second rotating rod 222 then drives the third rotating rod 223 to move, thereby causing the tire-turning frame 224 to rotate around the second conveying module 21. This, in turn, causes the tire-turning rod 225 to turn the tire over. After turning, the tire first hits the baffle plate 23 and then falls onto the second conveying module 21, thus reducing the impact of the tire on the second conveying module 21.

[0242] The marking unit 3 includes the aforementioned tire laser marking device 31, a third conveying module 32, and a second clamping module 33. The two second clamping modules 33 are symmetrically arranged on the third conveying module 32 along the tire conveying direction. The tire laser marking device 31 is located above the second clamping module 33. The third conveying module 32 is used to convey the tire to the vision inspection unit 4.

[0243] In this embodiment, the third conveying module 32 is a Mecanum wheel conveyor, and the second clamping module 33 includes a motor 331, two lead screws 332 and two sets of clamping wheels 333. The two lead screws 332 are connected together by a coupling (equivalent to a double-ended screw). The two sets of clamping wheels 333 are threadedly connected to the two lead screws 332 respectively. The motor 331 drives the two lead screws 332 to rotate, thereby driving the two sets of clamping wheels 333 to move closer or further apart, thereby achieving the clamping of the tire.

[0244] It should be noted that the 3D structured light camera used in the camera acquisition module is model Mech-Eye PRO S, and the laser used in the engraving module is a CO2 laser, model Cx-10L (10.6μm), with an output power of 120W, water-cooled, and equipped with a three-axis linear module. The laser is connected to the three-axis linear module, and the three-axis linear module drives the laser to move.

[0245] The visual inspection unit 4 includes a fourth conveying module 41, a visual inspection module 42, and a third clamping module 43. Two third clamping modules 43 are symmetrically arranged on the fourth conveying module 41 along the tire conveying direction. The visual inspection module 42 is connected to the fourth conveying module 41 and is located above the third clamping modules 43. The third clamping modules 43 are used to clamp the tire and drive the tire to rotate. The visual inspection module 42 is used to detect the characters engraved on the tire. The fourth conveying module 41 is used to convey the tire to the second tire retreading unit.

[0246] The fourth conveying module 41 is a Mecanum wheel conveyor. The vision inspection module 42 includes a laser profile measuring instrument and a YZ two-axis linear module. The YZ two-axis linear module is connected to the fourth conveying module 41, and the laser profile measuring instrument is connected to the two-axis linear module. The two-axis linear module drives the laser profile measuring instrument to move up and down and left and right (defining the tire conveying direction as the front and back direction), thereby adapting to tires of different specifications. The laser profile measuring instrument is a Mech-Eye LNX-8300 3D line laser profile measuring instrument.

[0247] It should be noted that in this embodiment, the X-axis is the front-to-back direction (tire conveying direction), the Y-axis is the left-to-right direction, and the Z-axis is the up-to-down direction.

[0248] The structure of the third clamping module 43 is basically the same as that of the second clamping module 33. The difference is that one of the two sets of clamping wheels in the third clamping module 43 is an electric clamping wheel. After the third clamping module 43 clamps the tire, the electric clamping wheel drives the tire to rotate, so that the laser profile measuring instrument can detect the entire tire tread.

[0249] The second tire-turning unit is used to turn the tire over and transport it to the unloading unit 5.

[0250] The second retreading unit has the same structure as the first retreading unit 2, and will not be described in detail here. For the specific structure, please refer to the structure of the first retreading unit 2.

[0251] The feeding unit 5 is a roller conveyor belt.

[0252] The tire laser engraving system also includes a control unit, which is connected to the scanning unit 1, the first tire turning unit 2, the engraving unit 3, the vision inspection unit 4, the second tire turning unit, and the unloading unit 5. The control unit is used to drive the scanning unit 1, the first tire turning unit 2, the engraving unit 3, the vision inspection unit 4, the second tire turning unit, and the unloading unit 5 to operate sequentially. The control unit uses a Mitsubishi FX5U-80MT / ES PLC with FX5-16EX / ES and FX5-16EYT / ES expansion modules, as well as a GS2512-WXTBD human-machine interface touch screen.

[0253] The overall workflow of the tire laser engraving system is as follows: Taking the engraving of the front or back of a tire as an example, the tire is placed on the first conveying module 11. The first conveying module 11 moves the tire to the position of the first clamping module 13. The cylinder 131 of the first clamping module 13 is activated, causing the two first clamping modules 13 to clamp the tire from the side. Then, the Mecanum wheels on both sides of the first conveying module 11 rotate in opposite directions to make the tire rotate stably. The Z-axis linear module drives the barcode scanner to descend to 280mm from the top surface of the tire for real-time acquisition. When the tire barcode rotates into the effective field of view of the barcode scanner, the barcode area is identified through the built-in algorithm of the barcode scanner, and decoding is performed to read the correct barcode number and obtain the tire's production information. The tire's production information includes whether the tire needs to be retreaded, the location of the engraving area, and the engraving style.

[0254] The first conveying module 11 conveys the tire to the second conveying module 21, which then conveys the tire to the tire-turning station of the first tire-turning unit 2. It determines whether the tire needs to be turned. If the tire does not need to be turned, the second conveying module 21 conveys the tire to the third conveying module 32. If the tire needs to be turned, the motor drives the tire-turning frame 224 to rotate, causing the tire-turning rod 225 on the tire-turning frame 224 to rotate one end of the tire, thus turning the tire over. After turning, the second conveying module 21 conveys the tire to the third conveying module 32, and the motor then drives the tire-turning frame 224 to rotate, returning the tire-turning frame 224 to its initial position.

[0255] The third conveying module 32 moves the tire to the position of the second clamping module 33. The motor 331 of the second clamping module 33 starts, causing the two sets of clamping wheels 333 to move toward the tire, thereby clamping the tire. Then, the tire laser engraving device 31 positions the engraving area of ​​the tire and controls the three-axis linear module to move the laser to the engraving position. The laser engraves the characters on the tire surface. After engraving is completed, the third conveying module 32 conveys the engraved tire to the fourth conveying module 41. The fourth conveying module 41 moves the tire to the third clamping module 43, which clamps the tire. The electric clamping wheel drives the tire to rotate, and the laser profile measuring instrument scans the tire tread. After the electric clamping wheel rotates a preset number of times (specifically set according to the tire specifications), the electric clamping wheel stops rotating, the laser profile measuring instrument stops scanning and performs calculations, and outputs an intensity map, a depth map, and a point cloud. The intensity map, depth map, and point cloud are compared with the engraving area position and engraving style in the production information. If the engraving area position and engraving style in the production information are met, the tire is deemed qualified. If either the engraving area position or engraving style in the production information is not met, the tire is deemed unqualified. The fourth conveying module 41 conveys the tire to the second tire-turning unit. The second tire-turning unit determines whether the tire needs to be turned over. When the tire does not need to be turned over, the second tire-turning unit conveys the tire to the unloading unit 5. When the tire needs to be turned over, the tire is turned over. After completion, the second tire-turning unit conveys the tire to the unloading unit 5.

[0256] After the tire is conveyed to the unloading unit 5, it waits for the vision inspection unit 4 to complete the calculation and send the result to the external sorting system before the tire is conveyed to the next station.

[0257] It should be noted that since there are three scenarios for tire marking: marking the front (the side with the barcode), marking the back, and marking both sides, by setting up a first tire-turning unit 2 and a second tire-turning unit, when marking the front is required, neither the first nor the second tire-turning unit performs a tire-turning operation. This allows the tire to be conveyed to each process with the front side facing up, completing the front marking. When marking the back is required, both the first and second tire-turning units operate. The first tire-turning unit 2 flips the tire, turning it from front to back. When the tire moves to the marking unit 3, the back of the tire can be marked. The second tire-turning unit... The purpose of the tire flipping operation is to ensure that the tire barcode faces upward after the engraving is completed, so that subsequent processes can read the barcode for sorting. When double-sided engraving is required, the first tire flipping unit 2 does not perform the tire flipping operation. After the engraving unit 3 completes the engraving on the front of the tire, the second tire flipping unit performs the tire flipping operation, turning the tire from front to back. The second tire flipping unit and the vision inspection unit 4 transport the tire back to the engraving unit 3 for engraving on the back of the tire. After the engraving is completed, the tire is transported back to the second tire flipping unit, which performs the tire flipping operation again, so that the front of the tire faces upward, so that subsequent processes can read the barcode for sorting.

[0258] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for laser engraving on tires, characterized in that, Includes the following steps: Step 1: Acquire 2D images and local point clouds of the tire from multiple perspectives; Step 2: Obtain calibration extrinsic parameters through global calibration method. Based on the calibration extrinsic parameters, transform the local point cloud of each viewpoint to the global coordinate system to obtain the complete point cloud; Step 3: Filter, denoise, binarize, and perform morphological transformation on the two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. Use a feature localization model to identify the marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then map the three-dimensional coordinates of the marker points in the local point cloud to the complete point cloud. Step 4: Using coordinate transformation technology, the complete point cloud is converted into a planar point cloud through polar coordinates. The point cloud data of the inner edge of the tire is extracted from the planar point cloud. The point cloud data of the inner edge of the tire is processed by a linear fitting algorithm to obtain the inner edge fitting line. Step 5: Based on the alignment of the marking points and the inner edge, a positioning algorithm is used to process the marking feature points to obtain the two-dimensional coordinates of the marking feature points; then, the coordinate inverse transformation formula is used to restore the two-dimensional coordinates of the marking feature points to three-dimensional coordinates. Based on the three-dimensional coordinates, a three-dimensional marking model is generated and the marking characters are engraved on the tire.

2. The tire laser engraving method according to claim 1, characterized in that, Step 3 includes the following sub-steps: Step A1: Process the complete point cloud using statistical filtering and voxel downsampling methods; Step A2: Filter, denoise, binarize, and perform morphological transformation processing on the two-dimensional images of the tire from multiple perspectives, filter the foreground and background, and convert the two-dimensional images of the tire from multiple perspectives to a consistent orientation to obtain a pure tire tread image. Step A3: Use a feature localization model to identify marker points in the pure tire tread image; obtain the two-dimensional coordinates of the marker points in the pure tire tread image, and then use a transformation processing method to convert the two-dimensional coordinates of the marker points in the pure tire tread image into the three-dimensional coordinates of the local point cloud. Step A4: Based on the calibration extrinsic parameters, map the 3D coordinates of the marker points in the local point cloud to the complete point cloud.

3. The tire laser engraving method according to claim 2, characterized in that, The conversion processing method in step A3 is as follows: based on the distortion coefficient, the two-dimensional coordinates of the marker points in the two-dimensional image are corrected to obtain normalized coordinates. Then, based on the normalized coordinates, the depth of the marker points is calculated using the structured light triangulation principle. Finally, based on the normalized coordinates and the depth, the two-dimensional coordinates of the marker points in the pure tire tread image are converted into the three-dimensional coordinates of the local point cloud.

4. The tire laser engraving method according to claim 1, characterized in that, Step 4 includes the following sub-steps: Step B1: Perform symmetry verification, removal of non-tireside interference points, and rotation axis calibration on the complete point cloud in sequence; Step B2: Using coordinate transformation technology, with the tire center axis as the polar axis, the position of the processed complete point cloud in the plane perpendicular to the polar axis is represented by polar coordinates of radial distance - circumferential angle. Then, the circumferential angle is expanded into a linear dimension, so that the processed complete point cloud is converted into a planar point cloud. Step B3: Use a multi-stage filtering method to extract the tire inner edge point cloud from the planar point cloud; Step B4: Using a piecewise linear fitting method, the inner edge point cloud of the tire is processed by a linear fitting algorithm to obtain the inner edge fitted line.

5. The tire laser engraving method according to claim 4, characterized in that, In step B3, the multi-stage screening includes the following steps: Step C1: Based on the preset radial distance range of the inner edge of the tire specification, remove point cloud data in the planar point cloud whose radial distance is not within the radial distance range of the inner edge, and obtain the inner edge point cloud of the candidate tire. Step C2: Set a point cloud number threshold, use statistical filtering to calculate the number of points in the K-neighborhood of each candidate point in the inner edge point cloud of the candidate tire, and remove candidate points whose number of points in the K-neighborhood is less than the point cloud number threshold to obtain continuous inner edge point clouds of the tire. Step C3: Set the U-axis coordinate threshold, calculate the U-axis coordinate difference between adjacent candidate points in the tire inner edge point cloud chain, and remove candidate points whose U-axis coordinate difference is greater than the horizontal coordinate threshold to obtain the complete tire inner edge point cloud; Step C4: Perform morphological closing operation on the complete tire inner edge point cloud, and then fill in the sparse edge area by point cloud interpolation to obtain the tire inner edge point cloud.

6. The tire laser engraving method according to claim 4, characterized in that, In step B4, the piecewise linear fitting includes the following steps: Step D1: Divide the point cloud of the inner edge of the tire into multiple segments along the U-axis direction; Step D2: Process the inner edge point cloud of each tire segment using a linear fitting algorithm to obtain the inner edge fitting line of each segment. When the engraving feature point is included in the inner edge fitting line of a certain segment, the inner edge fitting line of that segment is used for calculation. Step D3: Set the V-coordinate threshold, connect the fitted straight lines of the inner edge of each segment, and judge the V-coordinate difference at the connection point of the fitted straight lines of adjacent segments. If the V-coordinate difference is less than the V-coordinate threshold, the usage requirements are met. If the V-coordinate difference is greater than the V-coordinate threshold, increase the number of segments and repeat steps D1 to D3, or smooth the connection point.

7. The tire laser engraving method according to claim 1, characterized in that, In step 5, the specific formula for the positioning algorithm is as follows: ； in, These are the final two-dimensional coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. The azimuth angle of the engraved feature point in the original polar coordinates. Here, a and b are angle compensation values, and both are linear correction coefficients. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. This is the offset compensation value in the vertical direction; The inverse coordinate transformation formula is as follows: ； in, These are the final coordinates of the engraving feature points in the global coordinate system. Let be the vertical coordinates of the engraved feature points in the cylindrical unfolded plane. The minimum radial distance within the target area. Let be the horizontal coordinates of the engraved feature points in the cylindrical unfolded plane. The scaling factor from the angle to the unfolded plane. This is the reference depth value for the engraving area. This is the depth compensation value.

8. The tire laser engraving method according to claim 1, characterized in that, It also includes step 6: obtaining the intensity map, depth map and point cloud after the engraving is completed, and detecting the quality of the tire engraving characters. If any judgment condition is met, the tire engraving quality is determined to be unqualified. The judgment conditions include: (1) Segment all engraved characters from the intensity diagram, calculate the size of each engraved character, and compare it with the preset standard size. If the size of the engraved character exceeds the preset standard size range, it is deemed unqualified. (2) Based on the point cloud, calculate the valley point set of the character image, calculate the average depth difference between the valley point set and the reference plane, and compare it with the preset standard depth. If the depth difference between the valley point and the reference plane is greater than the standard depth, it is judged as unqualified. (3) For the circumferentially distributed engraved characters, establish a polar coordinate system with the center of the tire as the origin, extract the smallest bounding rectangle of each engraved character, determine the character center line, calculate the angle between the center line of each character and the polar axis, and statistically obtain the maximum value among all angles. Compare the maximum value among all angles with the preset standard angle. If the maximum value among all angles is greater than the preset standard angle, it is judged as unqualified. (4) By edge detection, fit an approximate straight line of the tire edge, calculate the horizontal distance from the innermost pixel of each engraved character to the approximate straight line of the tire edge, and compare it with the standard edge distance. If the horizontal distance exceeds the standard edge distance, it is judged as unqualified.

9. A tire laser engraving device, characterized in that, A tire laser engraving method according to any one of claims 1-8 comprises the following units: Camera acquisition module: used to acquire two-dimensional images and local point clouds of the tire from multiple perspectives; Coordinate transformation module: Used to obtain calibration extrinsic parameters through global calibration method, and based on the calibration extrinsic parameters, transform the local point cloud of each view to the global coordinate system to obtain the complete point cloud; The marker point recognition module is used to filter, denoise, binarize, and perform morphological transformation processing on two-dimensional images of the tire from multiple perspectives to obtain a pure tire tread image. The feature localization model is used to recognize marker points in the pure tire tread image to obtain the three-dimensional coordinates of the marker points in the local point cloud. Then, the three-dimensional coordinates of the marker points in the local point cloud are mapped to the complete point cloud. Two-dimensional transformation module: Used to convert the complete point cloud into a planar point cloud through polar coordinates using coordinate transformation technology, extract the point cloud data of the inner edge of the tire from the planar point cloud, and process the point cloud data of the inner edge of the tire through a linear fitting algorithm to obtain the inner edge fitting line. The engraving module is used to fit straight lines based on the marking points and inner edges, and to process the engraving feature points using a positioning algorithm to obtain the two-dimensional coordinates of the engraving feature points. Then, the two-dimensional coordinates of the engraving feature points are restored to three-dimensional coordinates using the inverse coordinate transformation formula. Based on the three-dimensional coordinates, a three-dimensional engraving model is generated and the engraved characters are engraved on the tire.

10. A tire laser engraving system, characterized in that, It includes a barcode scanning unit, a first tire turning unit, an engraving unit, a vision inspection unit, a second tire turning unit, and a material feeding unit connected in sequence; The scanning unit includes a first conveying module, a scanning module, and a first clamping module; two first clamping modules are symmetrically arranged on the first conveying module along the tire conveying direction, the scanning module is connected to the first conveying module and located above the clamping module, and the first conveying module is used to convey the tire to the tire turning unit; The first tire turning unit includes a second conveying module, a tire turning module, and a baffle plate; the baffle plate and the tire turning module are sequentially arranged on the second conveying module along the tire conveying direction; the tire turning module is used to turn the tire over; the baffle plate is used to assist the tire turning over; and the second conveying module is used to convey the tire to the marking unit. The marking unit includes the tire laser marking device as described in claim 9, a third conveying module, and a second clamping module. The two second clamping modules are symmetrically arranged on the third conveying module along the tire conveying direction, and the tire laser marking device is located above the second clamping module. The third conveying module is used to convey the tire to the vision inspection unit. The vision inspection unit includes a fourth conveying module, a vision inspection module, and a third clamping module. Two third clamping modules are symmetrically arranged on the fourth conveying module along the tire conveying direction. The vision inspection module is connected to the fourth conveying module and is located above the third clamping module. The third clamping module is used to clamp the tire and drive the tire to rotate. The vision inspection module is used to detect the characters engraved on the tire. The fourth conveying module is used to convey the tire to the second tire turning unit. The second tire-turning unit is used to turn the tire over and transport it to the unloading unit.

Images