A mold module positioning deviation real-time calibration method based on visual servoing

Abstract

The present application relates to the technical fields of mold assembly and visual servo control, and discloses a mold module positioning deviation real-time calibration method based on visual servo. An industrial camera acquisition module is arranged at a mold installation station to collect the inner circle gray scale diagram of the outer circle and the mold base, the initial boundary points of the inner circle are extracted along multiple radial sampling based on the image coordinate system and according to the gray scale change, the radial direction pairs are constructed in pairs, and the center of the inner circle of the mold base is obtained by averaging the midpoint of the boundary; the inner and outer circle boundaries are resampled with the center, and the pixel and physical size conversion is established combined with the design diameter to obtain the ring gap thickness sequence, the radial thickness difference sequence and the minimum ring gap; further, the eccentricity vector, the eccentricity and the eccentricity direction angle are analyzed to generate the translation calibration instruction to complete the position correction, the installation condition is judged combined with the theoretical residual thickness difference and the process limit value, and the result data is output, so that the quantitative calibration without disassembly and assembly, the anti-interference, the reduction of repeated trial adjustment and the traceability are realized.

Description

Technical Field

[0001] This invention relates to the field of mold assembly and vision servo control technology, specifically a real-time calibration method for mold module positioning deviation based on vision servo. Background Technology

[0002] The assembly process of inserting circular mold modules into mold base sleeves is widely used in the manufacturing and maintenance of stamping dies, injection molds, and precision tooling fixtures. On traditional production lines, the coaxiality of the module and the inner circle of the mold base relies heavily on manual positioning using light, mechanical guide posts, or plug gauges. Operators judge the uniformity of the ring gap and the presence of eccentricity based on experience, then complete the correction through multiple trial assembly steps, grinding, or fine-tuning of the fastening screws. This method is highly dependent on operator experience, results in poor assembly consistency, and is difficult to implement in real-time, quantitative calibration of positioning deviations on automated production lines.

[0003] With the development of industrial vision technology, some existing technologies have begun to introduce overhead industrial cameras to acquire images of the outer circle of the module and the inner circle of the mold base. Methods such as edge detection, binarization segmentation, and circle fitting are used to estimate the center positions of the two circles. Then, based on the deviation of the two center points in the pixel coordinate system, translation commands for the actuator are generated, achieving a certain degree of "visual alignment." However, these solutions typically only focus on the single indicator of the center coordinate difference, simplifying the entire annular gap to a uniform gap between ideal concentric circles. This ignores the multi-point variation of the actual thickness along different radial directions and fails to reflect the multi-directional thickness asymmetry caused by local processing errors, clamping deformation, etc. Existing vision methods also often rely on the empirical approach of "converting pixel deviations proportionally to physical displacements" in terms of control quantity calculation. Some solutions only linearly amplify the center deviation obtained from a single frame fitting before sending it to the actuator, lacking a rigorous calibration process based on the known design diameter of the mold base. This results in an inaccurate mapping relationship between pixel length and actual physical length, easily leading to problems such as deviation amplification, over-correction, or under-correction in scenarios with small mold assembly gaps and high precision requirements. On the other hand, existing technologies generally use the intuitive standard of "the smaller the deviation, the better" as the control target, without establishing a quantitative constraint relationship between eccentricity and minimum annular gap thickness at the algorithm level. They typically cannot provide a clear determination before assembly whether "eccentricity is already less than the minimum annular gap" or "whether the remaining thickness asymmetry is within the process tolerance range." Instead, they can only rely on post-assembly interference, tearing, or jamming phenomena to infer whether calibration was adequate. This method of discovering problems after the fact not only affects mold life but may also lead to the scrapping of the entire mold or materials, increasing production costs and downtime risks.

[0004] Therefore, this case aims to propose a real-time calibration method for mold module positioning deviation based on vision servoing. At the mold installation station, an industrial camera is used to acquire the same field-of-view image of the outer circle of the module and the inner circle of the mold base. The inner and outer circle boundaries are accurately extracted through radial grayscale sampling and boundary mutation analysis. The mapping relationship between pixel size and physical size is established using the design diameter of the mold base. On this basis, the annular gap thickness field and the diameter difference thickness field are constructed along the full circular direction. Then, the thickness asymmetry is converted into a planar eccentricity vector of the module center relative to the inner circle center of the mold base through analytical methods. This generates an accurate translation calibration amount to drive the actuator to complete the centering correction in one go. Finally, the theoretical remaining thickness difference, minimum annular gap thickness and process specification limits are combined to determine whether the module can be pressed into the mold base. Summary of the Invention

[0005] This invention provides a real-time calibration method for positioning deviation of mold modules based on visual servoing, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for real-time calibration of positioning deviation of a mold module based on visual servoing, comprising:

[0007] An industrial camera is placed at the mold installation station to capture a grayscale image that shows the outer circle of the module and the inner circle of the mold base completely in the image and is associated with the design diameter of the inner circle of the mold base and the image coordinate system.

[0008] The grayscale image is sampled along multiple radial sampling directions with the center pixel position of the image as the origin, and the initial boundary points of the inner circle of the mold base are obtained according to the grayscale changes.

[0009] Two radial sampling directions in opposite directions are combined into a pair of radial directions. The midpoint of the boundary is calculated based on the initial boundary point in each pair. The average of the midpoints of the boundary is then used to obtain the coordinates of the center of the inner circle of the mold base.

[0010] Starting from the center of the inner circle of the mold base, resample the outer circle boundary of the module and the inner circle boundary of the mold base in each radial sampling direction, and calculate the conversion ratio coefficient between pixels and physical size;

[0011] By using a conversion scaling factor, the boundary pixel positions in each radial sampling direction are converted into the physical radii of the outer circle of the module and the inner circle of the module base, an annular gap thickness sequence and a diameter-to-diameter annular gap thickness difference sequence are constructed, and the minimum annular gap thickness is obtained.

[0012] Based on the thickness difference sequence of the radial annular gap, an eccentric model of the module center relative to the inner circle center of the mold base is established. The eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base are obtained by weighted projection of the thickness difference of the radial annular gap.

[0013] Based on the components of the eccentric vector in the horizontal and vertical directions, the module generates translation calibration displacement commands in the horizontal and vertical directions, driving the actuator to complete the position correction by translating the module in the plane.

[0014] After the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specification, and the module loading permission conclusion and the result data set of this vision servo real-time calibration are output.

[0015] Optionally, the step of arranging an industrial camera at the mold installation station to acquire a grayscale image that fully displays the outer circle of the module and the inner circle of the mold base in the image and is associated with the designed diameter of the inner circle of the mold base and the image coordinate system specifically includes:

[0016] An industrial camera is fixedly installed directly above the mold installation station, so that the optical axis of the industrial camera is aligned with the normal direction of the mold base end face, and the outer circle contour of the module to be calibrated and the inner circle contour of the mold base are simultaneously within the field of view of the industrial camera.

[0017] Obtain the inner diameter of the mold base from the mold design data;

[0018] An image coordinate system is established in the current image plane captured by the industrial camera. The image coordinate system is set as a two-dimensional rectangular coordinate system. The center pixel position of the image is set as the origin of the coordinate system. The horizontal direction is set as the positive direction of the horizontal axis pointing to the right side of the image, and the vertical direction is set as the positive direction of the vertical axis pointing to the top of the image, so that each pixel position in the image has a unique horizontal coordinate and vertical coordinate.

[0019] Select a total number of radial sampling directions of not less than four and an even number around the origin of the coordinate system, so that each radial sampling direction is evenly distributed within the whole circle at the same angle interval, and set a unique index number and angle parameter between each radial sampling direction and the horizontal direction.

[0020] Before the module is pressed into the sleeve and both the outer circle of the module and the inner circle of the mold base are fully visible in the image, a grayscale image is acquired using an industrial camera. The grayscale value of each pixel in the image is then correlated with the horizontal and vertical coordinates of each pixel in the image coordinate system to form a grayscale data set.

[0021] Optionally, the step of sampling the grayscale image along multiple radial sampling directions with the image center pixel position as the origin, and obtaining the initial boundary points of the inner circle of the mold base in each radial sampling direction based on the grayscale changes, specifically includes:

[0022] In each radial sampling direction, the sampling point is moved outward pixel by pixel from the image center pixel position along the current radial sampling direction. The movement stops when the sampling point is about to exceed the image boundary. The maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective sampling length of the current radial sampling direction.

[0023] In each radial sampling direction, for each radial index position within the effective sampling length range from the first step, the sampling point coordinates are sequentially obtained from the image center pixel position along the current radial sampling direction with a step size of one pixel. The theoretical coordinates are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form the radial grayscale sampling sequence in the current radial sampling direction.

[0024] In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray values ​​of the current radial sampling position and the next radial sampling position is calculated as the first difference, and the difference between the gray values ​​of the current radial sampling position and the next two radial sampling positions and the gray values ​​of the current radial sampling position and the first two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the initial boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation.

[0025] In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose initial boundary mutation amount is not less than the initial boundary mutation amount of the previous radial position and not less than the initial boundary mutation amount of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set of the current radial sampling direction.

[0026] If the first candidate set in any radial sampling direction is empty, output the conclusion that the initial boundary extraction of the inner circle of the mold base failed and terminate the current real-time calibration process.

[0027] In each radial sampling direction where the first candidate set is not empty, the position with the largest radial index in the first candidate set is selected as the radial position of the initial boundary of the inner circle of the mold base in the current radial sampling direction. Based on the radial position of the initial boundary of the inner circle of the mold base and the angle of its corresponding sampling direction, the coordinates of the initial boundary point of the inner circle of the mold base in the current radial sampling direction are calculated in the image coordinate system.

[0028] The coordinates of the initial boundary points of the inner circle of the mold base calculated in all radial sampling directions are combined into a set of initial boundary points of the inner circle of the mold base.

[0029] Optionally, the step of forming a radial direction pair by combining two radial sampling directions in opposite directions, calculating the boundary midpoint based on the initial boundary points in each pair, and averaging the boundary midpoints to obtain the coordinates of the inner circle center of the mold base specifically includes:

[0030] Based on the established radial sampling direction index, all radial sampling directions are divided into pairs according to the index. Each pair consists of two radial sampling directions that are opposite in direction and have an angle difference of 180 degrees on both sides of the pixel position at the center of the image, forming multiple pairs of radial directions.

[0031] In each pair of radial directions, the coordinates of the initial boundary points of the inner circle of the mold base in the corresponding two radial sampling directions are obtained from the initial boundary point set of the inner circle of the mold base. The lateral coordinates of the two points are averaged to obtain the lateral coordinates of the boundary midpoint of the current pair of radial directions. The longitudinal coordinates of the two points are averaged to obtain the longitudinal coordinates of the boundary midpoint of the current pair of radial directions, thus forming the set of boundary midpoint coordinates of all pairs of radial directions.

[0032] The horizontal coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the horizontal coordinates of the midpoints of all boundaries. The vertical coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the vertical coordinates of the midpoints of all boundaries. The two coordinates together constitute the center coordinates of the center of the inner circle of the mold base in the image plane.

[0033] Optionally, the step of resampling the outer circle boundary of the module and the inner circle boundary of the mold base in each radial sampling direction, starting from the center of the inner circle of the mold base, and calculating the conversion ratio coefficient between pixels and physical dimensions, specifically includes:

[0034] After the coordinates of the center of the inner circle of the mold base are determined, in each radial sampling direction, the sampling point is moved outward pixel by pixel from the center of the inner circle of the mold base along the current radial sampling direction. When the sampling point is about to exceed the image boundary, the movement stops and the maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective resampling length of the current radial sampling direction.

[0035] In each radial sampling direction, for each radial index position within the range from the first step to the effective resampling length, the coordinates of the sampling points are sequentially obtained from the center of the inner circle of the mold base along the current radial sampling direction with a step size of one pixel. The calculated coordinate values ​​are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form a radial grayscale resampling sequence starting from the center of the inner circle of the mold base.

[0036] In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray value of the current radial sampling position and the gray value of the next radial sampling position is calculated as the first difference, and the difference between the gray value of the current radial sampling position and the gray value of the next two radial sampling positions and the gray value of the current radial sampling position and the gray value of the previous two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation.

[0037] In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose boundary mutation values ​​are not less than the boundary mutation values ​​of the previous radial position and not less than the boundary mutation values ​​of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set for the current radial sampling direction.

[0038] If the number of candidate radial positions is less than two in the first candidate set in any radial sampling direction, the conclusion that the double boundary resampling boundary extraction based on the inner circle center of the mold base has failed is output, and the real-time calibration process is terminated.

[0039] In each first candidate set, when the number of candidate radial positions is greater than or equal to two, the radial position with the smallest radial index in the candidate set is taken as the radial position of the outer circle boundary of the module in the current radial sampling direction, and the radial position with the smallest index among all radial positions in the candidate set that are greater than the radial index of the outer circle boundary of the module is taken as the radial position of the inner circle boundary of the module in the current radial sampling direction.

[0040] In each pair of radial directions, the radial sampling steps corresponding to the boundary position of the inner circle of the mold base in the two radial sampling directions are obtained respectively. The radial sampling steps in the two directions are added together to obtain the pixel diameter of the inner circle of the mold base in the image in the current pair of radial directions.

[0041] Summing the pixel diameters of all pairs of radial directions, multiplying the sum by two, and dividing by the total number of radial sampling directions yields the average pixel radius of the inner circle of the mold base.

[0042] The scaling factor for converting pixel size to physical size is obtained based on the ratio between the inner circle design diameter of the mold base and the average pixel radius.

[0043] In each radial sampling direction, the number of radial sampling steps corresponding to the outer circle boundary position of the module is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the outer circle of the module in the corresponding radial sampling direction. The number of radial sampling steps corresponding to the inner circle boundary position of the module base is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the inner circle of the module base in the corresponding radial sampling direction.

[0044] Optionally, the step of converting the boundary pixel positions in each radial sampling direction into the physical radii of the module's outer circle and the mold base's inner circle using a conversion scaling factor, constructing annular gap thickness sequences and radial annular gap thickness difference sequences, and obtaining the minimum annular gap thickness specifically includes:

[0045] In each radial sampling direction, the annular gap thickness between the inner circle of the mold base and the outer circle of the module is calculated using the difference between the physical radius of the inner circle of the mold base and the physical radius of the outer circle of the module in the current radial sampling direction, and the annular gap thickness sequence is formed according to the radial sampling direction index order.

[0046] In each pair of radial directions, the annular gap thickness values ​​in the two radial sampling directions are obtained respectively. The annular gap thickness in one direction is subtracted from the annular gap thickness in the other direction that is related to it by the radial direction to obtain the annular gap thickness difference in the current pair of radial directions. The pair of radial directions are then arranged in the index order of the pair of radial directions to form a sequence of annular gap thickness differences.

[0047] Squaring each annular gap thickness difference in the diametrical annular gap thickness difference sequence, and then summing all the squared results, we obtain the total thickness symmetry difference for all diametrical directions.

[0048] The annular gap thickness is compared in all directions in the annular gap thickness sequence, and the smallest annular gap thickness value is selected as the minimum annular gap thickness.

[0049] Optionally, the step of establishing an eccentric model of the module center relative to the inner circle center of the mold base based on the thickness difference sequence of the radial annular gap, and obtaining the eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base through weighted projection of the thickness difference of the radial annular gap, specifically includes:

[0050] An eccentric vector is set in the image plane relative to the center of the inner circle of the mold base. The eccentric vector includes a component in the horizontal direction and a component in the vertical direction. The two components together characterize the planar offset of the module center relative to the center of the inner circle of the mold base.

[0051] In each pair of radial directions, based on the sampling direction angle of the current pair of radial directions and the annular gap thickness difference on the current pair of radial directions, the corresponding annular gap thickness difference is expressed as twice the projection of the eccentric vector on the current pair of radial directions, and a linear relationship model between the annular gap thickness difference and the eccentric vector components is constructed.

[0052] In all radial pairs, the annular gap thickness difference in each pair is multiplied by the cosine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the horizontal direction; the annular gap thickness difference in each pair is multiplied by the sine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the vertical direction.

[0053] Based on the linear model that the relationship between the annular gap thickness difference and the projection of the eccentric vector onto each radial direction is twice, the weighted projections of the annular gap thickness difference in the horizontal and vertical directions are multiplied by two and then divided by the total number of radial sampling directions to calculate the components of the eccentric vector in the horizontal and vertical directions.

[0054] Squaring the components of the eccentric vector in the horizontal and vertical directions respectively, adding the squares together and taking the square root, we obtain the eccentric modulus length of the module center relative to the center of the inner circle of the module base;

[0055] When the eccentricity modulus is zero, the eccentricity direction angle is set to zero degrees. When the eccentricity modulus is not zero, the eccentricity direction angle is obtained by calculating the angle between the horizontal positive direction and the eccentricity vector direction by rotating counterclockwise to the direction of the eccentricity vector using inverse trigonometric functions based on the components of the eccentricity vector in the horizontal and vertical directions.

[0056] In each pair of radial directions, based on the linear relationship between the annular gap thickness difference and the projection of the eccentric vector onto the current radial direction pair, as well as the calculated eccentric vector components, the projection of the eccentric vector onto the current radial direction pair is multiplied by two. The theoretical remaining thickness difference on the current radial direction pair is obtained by subtracting the result of the projection of the eccentric vector onto the current radial direction pair multiplied by two from the annular gap thickness difference corresponding to the current radial direction pair.

[0057] Optionally, the step of generating translation calibration displacement commands in the horizontal and vertical directions based on the components of the eccentric vector in the horizontal and vertical directions, and driving the actuator to complete the position correction in the plane translation module, specifically includes:

[0058] The horizontal and vertical components of the eccentricity vector of the module center relative to the inner circle center of the module base are taken as opposites and used as the horizontal and vertical translation calibration values ​​of the module, thus forming the translation calibration displacement vector of the module in the image plane.

[0059] The actuator is configured to have two mutually orthogonal feed axes, such that the movement direction of one feed axis is parallel to the horizontal direction and the movement direction of the other feed axis is parallel to the vertical direction. Displacement command values ​​corresponding to the horizontal and vertical components of the translation calibration displacement vector are set in the CNC systems of the two feed axes respectively.

[0060] The horizontal and vertical displacement commands are sent to the two feed axes of the actuator, driving the actuator to perform translational movements in the horizontal and vertical directions, so that the module moves in the plane according to the translational calibration displacement vector, thereby correcting the center position of the module.

[0061] Optionally, after the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness, and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specifications, and the module loading permission conclusion and the result data set of this vision servo real-time calibration are output, specifically including:

[0062] For all radial pairs, take the absolute value of the theoretical remaining thickness difference for each pair of radial pairs, compare all the obtained absolute values, and select the largest one as the maximum absolute value of the theoretical remaining thickness difference.

[0063] Compare the eccentric modulus length of the module center relative to the inner circle center of the mold base with the minimum annular gap thickness, and determine at the analytical level whether the eccentric modulus length is less than the minimum annular gap thickness.

[0064] Obtain the maximum allowable limit of theoretical remaining thickness asymmetry given in the process specification, compare the maximum absolute value of the theoretical remaining thickness difference with the maximum allowable limit of theoretical remaining thickness asymmetry, and determine whether the maximum absolute value of the theoretical remaining thickness difference does not exceed the maximum allowable limit of theoretical remaining thickness asymmetry.

[0065] When both conditions are met simultaneously, the module can be installed in the mold base.

[0066] After completing the loading permission determination, the result data set of this vision servo real-time calibration is output. The result data set includes at least the eccentricity components of the module center relative to the inner circle center of the module base in the horizontal and vertical directions, the eccentricity modulus, the eccentricity direction angle, the translation calibration amount of the module in the horizontal direction and the translation calibration amount in the vertical direction in the plane, and the maximum absolute value of the theoretical remaining thickness difference.

[0067] The present invention has the following beneficial effects:

[0068] 1. The geometric datum of the mold installation station is tightly bound to the vision measurement system. By aligning the optical axis of the industrial camera with the normal direction of the mold base end face, the outer circle of the module and the inner circle of the mold base are fully displayed in the same field of view. The design diameter of the inner circle of the mold base is used as the sole dimensional datum, and a rectangular coordinate system with the center pixel of the image is established in the image plane. This avoids the problems of arbitrary camera installation and complex external parameter calibration required in traditional methods, ensuring a clear and stable correspondence between the image coordinate system and the actual coordinate system of the mold in the plane. On the other hand, by directly introducing the design diameter of the mold base into the subsequent pixel scale conversion, the vision measurement results naturally fall into the dimensional system used in the process design, rather than being indirectly converted through additional calibration targets or multiple intermediate transformations.

[0069] 2. Gray-level sequence sampling is performed along multiple uniformly distributed radial directions around the image center pixel. Boundary mutation values ​​are constructed by combining and differencing multiple sampling points. Local maxima locations are then selected as candidates for the inner circle boundary in each direction. Combining the sorting of boundary mutation values ​​and the logic for determining an empty set, a fault-tolerant boundary detection mechanism is formed. Unlike conventional visual methods that simply use full-image edge operators with threshold binarization or directly call circle detection operators to fit the inner circle, this radial one-dimensional gray-level analysis method utilizes the radial monotonicity of the inner circle geometry, reducing the two-dimensional problem to a one-dimensional problem. The algorithm structure is clearer and easier to customize and optimize for on-site interference such as gray-level noise, oil stains, and scratches. Calibration is terminated directly when the candidate set is empty to avoid introducing obviously erroneous boundary results into subsequent calculations; when the candidate set is not empty, the radial position with the largest boundary mutation value is selected to reduce the impact of local false edges on the results. Even if there are local occlusions, uneven lighting, or surface defects in the inner circle of the mold base, a reliable set of boundary points can still be obtained in most radial directions, providing a solid data foundation for subsequent center determination based on the midpoint of the diameter. Compared with the traditional overall circle fitting method, it is less sensitive to local anomalies, and the boundary positioning accuracy and robustness are significantly improved.

[0070] 3. All radial sampling directions are divided into pairs with opposite directions. Each pair consists of two radial directions that differ by a semicircular angle. The midpoint of the boundary points of the inner circles on both sides of each pair is taken as the midpoint of the diameter pair. The coordinates of all the midpoints of the diameter pairs are then averaged to obtain the position of the center of the inner circle of the mold base. This method based on averaging the midpoints of the diameter pairs more directly utilizes the symmetry of the circular structure in the diameter pair direction, geometrically canceling out local errors in pairs. Since each midpoint of the diameter pair is itself the geometric average of the boundary points on both sides, it can naturally suppress the deviation caused by the drift of the detection of one side boundary. Averaging multiple sets of midpoints of the diameter pairs further reduces the influence of local noise and discrete errors. For the inner circle of the mold base with local gaps, wear, or uneven brightness, traditional overall fitting is easily pulled off by abnormal boundaries, and the center position will be systematically offset. By grouping the diameter pairs and obtaining the midpoints, symmetry is embedded as a hard constraint into the algorithm. Even if the boundary points in some radial directions have slight errors, as long as the overall diameter pair structure is still present, the final center coordinates can still maintain high accuracy.

[0071] 4. Using the center of the inner circle of the mold base as the sole starting point, the outer circle boundary of the module and the inner circle boundary of the mold base are simultaneously resampled in each radial direction, constructing paired outer and inner circle radii in the same radial direction. The average pixel radius is calculated statistically from the positions of the inner circle boundaries of the mold base in all radial directions, and a unified conversion coefficient from pixel to physical size is derived in combination with the mold base design diameter. This scheme strictly unifies the sampling starting point, sampling direction, and size conversion, ensuring that the annular gap thickness in each radial direction is obtained by directly subtracting the two boundaries within the same coordinate frame, avoiding the error amplification problem introduced by different fitting centers and different circumscribed geometries. At the same time, using the mold base design diameter as a global constraint, the pixel diameters that fluctuate slightly in various directions in the image are uniformly mapped on the physical scale, achieving not only accurate conversion from pixels to millimeters but also smoothing out detection errors in individual directions through the averaging process. In this way, the entire system can directly establish a stable dimensional bridge between the production image and the design parameters without additional target calibration, ensuring that subsequent annular gap thickness analysis has engineering-usable absolute dimensional accuracy.

[0072] 5. Transforming the previously difficult-to-quantify gap state between the module and the base into a sequence distributed along the entire circle is a crucial step in this scheme, moving from geometric information to eccentric analysis. Specifically, in each radial direction, the annular gap thickness is obtained by subtracting the outer radius of the module from the inner radius of the base, forming a densely sampled annular gap thickness sequence. Then, in each pair of diametrically opposed directions, the difference in diametrically opposed thickness is obtained by subtracting the thickness of the other side from the thickness of one side, forming a sequence of diametrically opposed annular gap thickness differences. Based on this, the minimum annular gap thickness is extracted as a safety assessment indicator. This full-circular sequential representation not only reflects the detailed distribution of gaps in the circumferential direction but also provides structured data input for subsequent analytical models. The diametrically opposed thickness difference naturally contains the projection information of the module center's deviation from the base center in the corresponding direction, allowing complex eccentric geometric relationships to be reconstructed through the combined processing of these differences. At the same time, the extraction of the minimum annular gap thickness provides a direct quantitative basis for collision risk assessment, ensuring calibration accuracy while considering safety boundaries.

[0073] 6. Using the diameter-to-thickness difference sequence as the core, a linear relationship is established between the thickness asymmetry and the module center eccentricity vector. Each set of diameter-to-thickness differences is considered as a multiple of the projection of the eccentricity vector onto the diameter-to-thickness direction. Weighted summation based on direction angles is used to obtain the weighted projection sum of the thickness differences in the horizontal and vertical directions. Then, using the linear relationship, the components of the eccentricity vector in the horizontal and vertical directions are deduced, ultimately determining the magnitude and direction of the eccentricity. This approach employs an analytical solution. Once the diameter-to-thickness difference sequence and corresponding direction angles are obtained, the eccentricity components can be directly calculated through simple weighted summation and scaling, resulting in low computational cost, good real-time performance, and suitability for online control scenarios. Furthermore, this linear relationship originates from geometric symmetry analysis and does not rely on empirical fitting of specific samples, thus exhibiting good transferability to different sizes and module structures. Through this analytical method, the magnitude and direction of the eccentricity are no longer vague qualitative judgments of "too large" or "too small," but rather vector quantities with clear numerical values ​​and orientations. These can be directly used to generate calibration displacements and serve as long-term traceability for quality records.

[0074] 7. By directly inverting the horizontal and vertical components of the eccentric vector as the calibration displacement of the two orthogonal feed axes, the eccentricity information on the image plane is seamlessly mapped into the motion commands of the actuator, thus achieving direct coupling between visual measurement and mechanical motion. The actuator is designed with two feed axes parallel to the image coordinate system, allowing the CNC system to directly receive the translation output by the vision algorithm without the need for complex coordinate transformations or interpolation planning at the control layer. Compared to the traditional method of manually fine-tuning the position of each axis by observing the image or indicator lights, this vision servo link truly achieves "measure as much as it moves," reducing human interpretation bias and operational errors, and shortening the module alignment and debugging time. Since the calibration displacement is accurately calculated from the analytical eccentricity, the actuator compensates for the current deviation with each movement, avoiding the inefficiency and potential overcorrection caused by repeated trial-and-error adjustments. In addition, the horizontal and vertical directions of the image coordinate system are still used, transforming the abstract geometric deviation into an intuitive rectangular coordinate translation, which is easy for on-site maintenance personnel to understand and monitor, laying the foundation for deploying automated module assembly in the production line.

[0075] 8. The analytical judgment framework integrates both geometric safety and process consistency: On one hand, by comparing the eccentricity modulus with the minimum annular gap thickness, it checks whether the module still has sufficient clearance after the current correction to avoid interference or jamming during insertion. On the other hand, by comparing the maximum absolute value of the theoretical remaining thickness difference in all diametrical directions with the allowable asymmetry limit given by the process specification, it determines whether the corrected annular gap field meets the design requirements for thickness symmetry overall. This judgment focuses on both the gap at the most dangerous point and the overall thickness distribution balance, avoiding the hidden danger of "the center is aligned but the local area is very thin." The analytical judgment is based on the previously derived eccentricity and theoretical remaining thickness difference, requiring no additional measurement steps. The judgment result can automatically generate a clear conclusion of "allowing insertion" or "prohibiting insertion," and simultaneously output data such as the eccentricity component, eccentricity, eccentricity direction, translation calibration amount, and maximum theoretical remaining thickness difference for this calibration, facilitating quality tracking and process optimization. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the process of the present invention.

[0077] Figure 2 This is a top view schematic diagram of the module-base of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Example, refer to Figure 1 A method for real-time calibration of positioning deviation of mold module based on vision servoing, comprising:

[0080] An industrial camera is placed at the mold installation station to capture a grayscale image that shows the outer circle of the module and the inner circle of the mold base completely in the image and is associated with the design diameter of the inner circle of the mold base and the image coordinate system.

[0081] The grayscale image is sampled along multiple radial sampling directions with the center pixel position of the image as the origin, and the initial boundary points of the inner circle of the mold base are obtained according to the grayscale changes.

[0082] Two radial sampling directions in opposite directions are combined into a pair of radial directions. The midpoint of the boundary is calculated based on the initial boundary point in each pair. The average of the midpoints of the boundary is then used to obtain the coordinates of the center of the inner circle of the mold base.

[0083] Starting from the center of the inner circle of the mold base, resample the outer circle boundary of the module and the inner circle boundary of the mold base in each radial sampling direction, and calculate the conversion ratio coefficient between pixels and physical size;

[0084] By using a conversion scaling factor, the boundary pixel positions in each radial sampling direction are converted into the physical radii of the outer circle of the module and the inner circle of the module base, an annular gap thickness sequence and a diameter-to-diameter annular gap thickness difference sequence are constructed, and the minimum annular gap thickness is obtained.

[0085] Based on the thickness difference sequence of the radial annular gap, an eccentric model of the module center relative to the inner circle center of the mold base is established. The eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base are obtained by weighted projection of the thickness difference of the radial annular gap.

[0086] Based on the components of the eccentric vector in the horizontal and vertical directions, the module generates translation calibration displacement commands in the horizontal and vertical directions, driving the actuator to complete the position correction by translating the module in the plane.

[0087] After the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specification, and the module loading permission conclusion and the result data set of this vision servo real-time calibration are output.

[0088] By deploying an industrial camera at the mold installation station and acquiring grayscale images of the outer circle of the module and the inner circle of the mold base, the geometric relationships that originally required disassembly and assembly and dialing are transformed into image plane processing. By multi-directional radial sampling with the image center as the origin and combining grayscale changes to extract the initial boundary points of the inner circle of the mold base, the problem of unstable boundary recognition caused by discontinuity and unevenness of the inner circle edge under complex on-site lighting conditions is solved. By forming radial direction pairs with opposite directions, calculating the midpoint of the boundary, and then averaging, the center coordinates of the inner circle of the mold base are obtained, solving the problem that traditional overall fitting is easily biased by local contamination points and the center result is unstable. By resampling the outer circle of the module and the inner circle of the mold base with the center of the mold base as the starting point and calculating the pixel to physical size conversion coefficient, the boundary position in the image coordinates is directly linked to the design size, avoiding multiple intermediate calibrations. Furthermore, by constructing annular gap thickness sequences and diameter-to-diameter thickness difference sequences through conversion coefficients and extracting the minimum annular gap thickness, the originally abstract assembly gap situation is quantified into a calculable sequence. By establishing an eccentricity model based on the diameter-to-diameter thickness difference and analytically calculating the eccentricity vector, eccentricity amount, and eccentricity direction angle of the module center relative to the mold base center, and then generating module translation calibration commands through the eccentricity vector components to drive the actuator to complete position correction, the problem of only being able to roughly determine the eccentricity direction and repeatedly adjusting manually has been solved. Finally, by combining the theoretical remaining thickness difference, eccentricity modulus length, minimum annular gap thickness, and the allowable asymmetry limit of the process specification to determine the loading conditions, and outputting a result data set containing key parameters, the entire process from measurement and calibration to release decision is digitized and traceable, which is conducive to improving the safety, efficiency, and consistency of the mold assembly process.

[0089] The step of arranging an industrial camera at the mold installation station to acquire a grayscale image that fully displays the outer circle of the module and the inner circle of the mold base, and is associated with the designed diameter of the inner circle of the mold base and the image coordinate system, specifically includes:

[0090] An industrial camera is fixedly installed directly above the mold installation station, so that the optical axis of the industrial camera is aligned with the normal direction of the mold base end face, and the outer circle contour of the module to be calibrated and the inner circle contour of the mold base are simultaneously within the field of view of the industrial camera.

[0091] Obtain the inner diameter of the mold base from the mold design data;

[0092] An image coordinate system is established in the current image plane captured by the industrial camera. The image coordinate system is set as a two-dimensional rectangular coordinate system. The center pixel position of the image is set as the origin of the coordinate system. The horizontal direction is set as the positive direction of the horizontal axis pointing to the right side of the image, and the vertical direction is set as the positive direction of the vertical axis pointing to the top of the image, so that each pixel position in the image has a unique horizontal coordinate and vertical coordinate.

[0093] Select a total number of radial sampling directions of not less than four and an even number around the origin of the coordinate system, so that each radial sampling direction is evenly distributed within the whole circle at the same angle interval, and set a unique index number and angle parameter between each radial sampling direction and the horizontal direction.

[0094] Before the module is pressed into the sleeve and both the outer circle of the module and the inner circle of the mold base are fully visible in the image, a grayscale image is acquired using an industrial camera. The grayscale value of each pixel in the image is then correlated with the horizontal and vertical coordinates of each pixel in the image coordinate system to form a grayscale data set.

[0095] An industrial camera is placed directly above the mold installation station, with the camera's optical axis aligned with the normal direction of the mold base end, and both the outer circle of the module and the inner circle of the mold base within the camera's field of view.

[0096] Read the design diameter of the inner circle of the mold base, and denot it as... ;

[0097] Establish a two-dimensional rectangular coordinate system in the current image plane. , command point The center point of the image. The positive axis points to the right side of the image. The positive axis points to the top of the image;

[0098] Choose a positive even number that is not less than 4. As the total number of radial sampling directions, calculate the first... Each sampling direction angle is specifically: , ;in, This represents the total number of radial sampling directions; This is the radial sampling direction index; For the first The direction angle of each radial sampling direction;

[0099] A grayscale image is captured before the module is pressed into the sleeve, and both the outer circle of the module and the inner circle of the mold base are fully visible, to obtain the grayscale function. ;in, For image coordinates The grayscale value at the specified location.

[0100] The process of sampling the grayscale image along multiple radial sampling directions with the image center pixel position as the origin, and obtaining the initial boundary points of the inner circle of the mold base in each radial sampling direction based on the grayscale changes, specifically includes:

[0101] In each radial sampling direction, the sampling point is moved outward pixel by pixel from the image center pixel position along the current radial sampling direction. The movement stops when the sampling point is about to exceed the image boundary. The maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective sampling length of the current radial sampling direction.

[0102] In each radial sampling direction, for each radial index position within the effective sampling length range from the first step, the sampling point coordinates are sequentially obtained from the image center pixel position along the current radial sampling direction with a step size of one pixel. The theoretical coordinates are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form the radial grayscale sampling sequence in the current radial sampling direction.

[0103] In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray values ​​of the current radial sampling position and the next radial sampling position is calculated as the first difference, and the difference between the gray values ​​of the current radial sampling position and the next two radial sampling positions and the gray values ​​of the current radial sampling position and the first two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the initial boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation.

[0104] In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose initial boundary mutation amount is not less than the initial boundary mutation amount of the previous radial position and not less than the initial boundary mutation amount of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set of the current radial sampling direction.

[0105] If the first candidate set in any radial sampling direction is empty, output the conclusion that the initial boundary extraction of the inner circle of the mold base failed and terminate the current real-time calibration process.

[0106] In each radial sampling direction where the first candidate set is not empty, the position with the largest radial index in the first candidate set is selected as the radial position of the initial boundary of the inner circle of the mold base in the current radial sampling direction. Based on the radial position of the initial boundary of the inner circle of the mold base and the angle of its corresponding sampling direction, the coordinates of the initial boundary point of the inner circle of the mold base in the current radial sampling direction are calculated in the image coordinate system.

[0107] The coordinates of the initial boundary points of the inner circle of the mold base calculated in all radial sampling directions are combined into a set of initial boundary points of the inner circle of the mold base.

[0108] For each direction Get from point within the image range The maximum radial sampling steps taken at the start are denoted as ;in, In direction Above, with the image center The maximum number of radial sampling steps allowed when starting from [the point of origin];

[0109] For each direction and each radial index Calculate the radial sampling points, specifically: Calculate the corresponding grayscale value: ;in, This is the sequence number of the radial sampling steps along a fixed direction; In the first The radial step number is the nth direction. When walking, with The coordinates of the sampled pixel are obtained from the starting point; For rounding operations; In the first The direction, radial direction sampling points The grayscale value at that location;

[0110] For each direction In radial index The radial boundary abrupt change is calculated within the specified range, specifically as follows: ;in, In the first The direction, radial direction Initial boundary mutation at each location;

[0111] For each direction In the index range The inner selection satisfies: , All indexes This forms a candidate set for that direction;

[0112] If the candidate set is empty, output the conclusion that the initial boundary extraction failed and stop the current calibration process;

[0113] When the candidate set is not empty, the element with the largest radial index in the candidate set is denoted as . The initial boundary points of the inner circle of the mold base are calculated as follows: ;in, In the first Radial index of the inner circle boundary position of the mold base obtained by initial boundary detection in each direction; In the first The coordinates of the inner circle boundary points of the mold base extracted in each direction during the initial stage;

[0114] Create a set of all initial boundary points: ;in, This is the set of all inner circle boundary points of the mold base extracted in the initial stage.

[0115] Reference Figure 2 The process of forming a radial direction pair by combining two radial sampling directions in opposite directions, calculating the midpoint of the boundary based on the initial boundary points in each pair, and averaging the midpoints to obtain the coordinates of the center of the inner circle of the mold base specifically includes:

[0116] Based on the established radial sampling direction index, all radial sampling directions are divided into pairs according to the index. Each pair consists of two radial sampling directions that are opposite in direction and have an angle difference of 180 degrees on both sides of the pixel position at the center of the image, forming multiple pairs of radial directions.

[0117] In each pair of radial directions, the coordinates of the initial boundary points of the inner circle of the mold base in the corresponding two radial sampling directions are obtained from the initial boundary point set of the inner circle of the mold base. The lateral coordinates of the two points are averaged to obtain the lateral coordinates of the boundary midpoint of the current pair of radial directions. The longitudinal coordinates of the two points are averaged to obtain the longitudinal coordinates of the boundary midpoint of the current pair of radial directions, thus forming the set of boundary midpoint coordinates of all pairs of radial directions.

[0118] The horizontal coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the horizontal coordinates of the midpoints of all boundaries. The vertical coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the vertical coordinates of the midpoints of all boundaries. The two coordinates together constitute the center coordinates of the center of the inner circle of the mold base in the image plane.

[0119] For each pair of radial direction index pairs The calculation of the midpoint of the radius is as follows: , ;in, Index for radial direction; For the direction index is The initial stage extracts the inner circle boundary points of the mold base in the direction of the model base; For the first Coordinates of the midpoint of the boundary point of the pair;

[0120] The coordinates of the mold base center are calculated as follows: ;in, The image coordinates of the center of the inner circle of the mold base; , The center of the mold base in the image coordinate system are respectively direction, Direction coordinate components; , Midpoints of Quantity, Quantity.

[0121] The process of resampling the outer circle boundary of the module and the inner circle boundary of the module in each radial sampling direction, starting from the center of the inner circle of the mold base, and calculating the conversion ratio coefficient between pixels and physical dimensions, specifically includes:

[0122] After the coordinates of the center of the inner circle of the mold base are determined, in each radial sampling direction, the sampling point is moved outward pixel by pixel from the center of the inner circle of the mold base along the current radial sampling direction. When the sampling point is about to exceed the image boundary, the movement stops and the maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective resampling length of the current radial sampling direction.

[0123] In each radial sampling direction, for each radial index position within the range from the first step to the effective resampling length, the coordinates of the sampling points are sequentially obtained from the center of the inner circle of the mold base along the current radial sampling direction with a step size of one pixel. The calculated coordinate values ​​are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form a radial grayscale resampling sequence starting from the center of the inner circle of the mold base.

[0124] In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray value of the current radial sampling position and the gray value of the next radial sampling position is calculated as the first difference, and the difference between the gray value of the current radial sampling position and the gray value of the next two radial sampling positions and the gray value of the current radial sampling position and the gray value of the previous two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation.

[0125] In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose boundary mutation values ​​are not less than the boundary mutation values ​​of the previous radial position and not less than the boundary mutation values ​​of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set for the current radial sampling direction.

[0126] If the number of candidate radial positions is less than two in the first candidate set in any radial sampling direction, the conclusion that the double boundary resampling boundary extraction based on the inner circle center of the mold base has failed is output, and the real-time calibration process is terminated.

[0127] In each first candidate set, when the number of candidate radial positions is greater than or equal to two, the radial position with the smallest radial index in the candidate set is taken as the radial position of the outer circle boundary of the module in the current radial sampling direction, and the radial position with the smallest index among all radial positions in the candidate set that are greater than the radial index of the outer circle boundary of the module is taken as the radial position of the inner circle boundary of the module in the current radial sampling direction.

[0128] In each pair of radial directions, the radial sampling steps corresponding to the boundary position of the inner circle of the mold base in the two radial sampling directions are obtained respectively. The radial sampling steps in the two directions are added together to obtain the pixel diameter of the inner circle of the mold base in the image in the current pair of radial directions.

[0129] Summing the pixel diameters of all pairs of radial directions, multiplying the sum by two, and dividing by the total number of radial sampling directions yields the average pixel radius of the inner circle of the mold base.

[0130] The scaling factor for converting pixel size to physical size is obtained based on the ratio between the inner circle design diameter of the mold base and the average pixel radius.

[0131] In each radial sampling direction, the number of radial sampling steps corresponding to the outer circle boundary position of the module is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the outer circle of the module in the corresponding radial sampling direction. The number of radial sampling steps corresponding to the inner circle boundary position of the module base is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the inner circle of the module base in the corresponding radial sampling direction.

[0132] For each direction Get from point within the image range Maximum radial sampling steps ;in, In the first In each direction, with the center of the mold base The maximum number of radial sampling steps allowed when starting from [the point of origin];

[0133] For each direction and each radial index The radial sampling points are calculated as follows: Calculate the corresponding grayscale value: ;in, In the first The direction, radial direction The sampling pixel coordinates are obtained at each step length, with the center of the mold base as the starting point; In the first The direction, radial direction sampling points The grayscale value at that location;

[0134] For each direction In radial index The radial boundary abrupt change is calculated within the specified range, specifically as follows: ;in, In the first The direction, radial direction The boundary mutation amount at each location, starting from the center of the mold base;

[0135] For each direction In the index range The inner selection satisfies: , All indexes This forms the first candidate set for that direction, and sorts them by index from smallest to largest;

[0136] If the number of indices in the first candidate set is less than 2, output the conclusion that the resampling boundary extraction failed and stop the current calibration process;

[0137] When the number of indices in the first candidate set is greater than or equal to 2, the element with the smallest index in the candidate set is denoted as... All candidates in the first candidate set that are greater than The smallest one is selected from the index and denoted as . ;in, In the first The index of the outer circle boundary position of the module detected during resampling in each direction, starting from the center of the module base; In the first The index of the inner circle boundary position of the mold base detected when resampling in each direction with the center of the mold base as the starting point;

[0138] For each pair of radial direction index pairs The pixel diameter of the inner circle of the mold base on this pair of diameters is calculated as follows: , ;in, In the first The pixel diameter of the inner circle of the mold base in the group-pair radial direction; The direction index is The direction of the inner circle boundary position index of the mold base;

[0139] The average pixel radius of the inner circle of the mold base is calculated as follows: ;

[0140] The conversion of pixel size to physical quantity is calculated as follows: ;

[0141] And calculate the physical radii of the outer circle of the module and the inner circle of the mold base in each direction, specifically: , , ;in, In the first The physical radius of the module's outer circle in each direction; In the first The physical radius of the inner circle of the mold base in each direction.

[0142] The process involves using a conversion scaling factor to convert the boundary pixel positions in each radial sampling direction into the physical radii of the module's outer circle and the mold base's inner circle, constructing annular gap thickness sequences and radial annular gap thickness difference sequences, and obtaining the minimum annular gap thickness. Specifically, this includes:

[0143] In each radial sampling direction, the annular gap thickness between the inner circle of the mold base and the outer circle of the module is calculated using the difference between the physical radius of the inner circle of the mold base and the physical radius of the outer circle of the module in the current radial sampling direction, and the annular gap thickness sequence is formed according to the radial sampling direction index order.

[0144] In each pair of radial directions, the annular gap thickness values ​​in the two radial sampling directions are obtained respectively. The annular gap thickness in one direction is subtracted from the annular gap thickness in the other direction that is related to it by the radial direction to obtain the annular gap thickness difference in the current pair of radial directions. The pair of radial directions are then arranged in the index order of the pair of radial directions to form a sequence of annular gap thickness differences.

[0145] Squaring each annular gap thickness difference in the diametrical annular gap thickness difference sequence, and then summing all the squared results, we obtain the total thickness symmetry difference for all diametrical directions.

[0146] The annular gap thickness is compared in all directions in the annular gap thickness sequence, and the smallest annular gap thickness value is selected as the minimum annular gap thickness.

[0147] For each direction The annular gap thickness is calculated as follows: , ;in, In the first The radial clearance thickness between the inner circle of the mold base and the outer circle of the module in each direction;

[0148] For each pair of radial direction index pairs The diameter-to-thickness difference is calculated as follows: , ;in, For the first The difference in annular gap thickness between the pairs of radially aligned annular gaps; For the direction index is The thickness of the annular gap in the direction of;

[0149] Calculate the total thickness symmetry difference for all radial directions. ;

[0150] Calculate the minimum annular gap thickness in all directions. .

[0151] The process of establishing an eccentric model of the module center relative to the inner circle center of the mold base based on the thickness difference sequence of the radial annular gap, and obtaining the eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base through weighted projection of the radial annular gap thickness difference, specifically includes:

[0152] An eccentric vector is set in the image plane relative to the center of the inner circle of the mold base. The eccentric vector includes a component in the horizontal direction and a component in the vertical direction. The two components together characterize the planar offset of the module center relative to the center of the inner circle of the mold base.

[0153] In each pair of radial directions, based on the sampling direction angle of the current pair of radial directions and the annular gap thickness difference on the current pair of radial directions, the corresponding annular gap thickness difference is expressed as twice the projection of the eccentric vector on the current pair of radial directions, and a linear relationship model between the annular gap thickness difference and the eccentric vector components is constructed.

[0154] In all radial pairs, the annular gap thickness difference in each pair is multiplied by the cosine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the horizontal direction; the annular gap thickness difference in each pair is multiplied by the sine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the vertical direction.

[0155] Based on the linear model that the relationship between the annular gap thickness difference and the projection of the eccentric vector onto each radial direction is twice, the weighted projections of the annular gap thickness difference in the horizontal and vertical directions are multiplied by two and then divided by the total number of radial sampling directions to calculate the components of the eccentric vector in the horizontal and vertical directions.

[0156] Squaring the components of the eccentric vector in the horizontal and vertical directions respectively, adding the squares together and taking the square root, we obtain the eccentric modulus length of the module center relative to the center of the inner circle of the module base;

[0157] When the eccentricity modulus is zero, the eccentricity direction angle is set to zero degrees. When the eccentricity modulus is not zero, the eccentricity direction angle is obtained by calculating the angle between the horizontal positive direction and the eccentricity vector direction by rotating counterclockwise to the direction of the eccentricity vector using inverse trigonometric functions based on the components of the eccentricity vector in the horizontal and vertical directions.

[0158] In each pair of radial directions, based on the linear relationship between the annular gap thickness difference and the projection of the eccentric vector onto the current radial direction pair, as well as the calculated eccentric vector components, the projection of the eccentric vector onto the current radial direction pair is multiplied by two. The theoretical remaining thickness difference on the current radial direction pair is obtained by subtracting the result of the projection of the eccentric vector onto the current radial direction pair multiplied by two from the annular gap thickness difference corresponding to the current radial direction pair.

[0159] Set the eccentricity vector of the module center relative to the module base center in the image plane as follows: ;in, , The eccentric vectors are respectively in direction, Components in direction;

[0160] For each set of diameter-to-thickness, an eccentric model is constructed, specifically as follows: , ;in, For the first The direction angle of the first direction aligned with the radial direction of the group;

[0161] Calculate the thickness symmetry difference in direction and The projections in the direction are as follows: , ;in, , For all diameter-to-diameter thickness differences direction, Weighted projection sum in the direction;

[0162] Calculate the eccentric vector components separately, specifically: , ;

[0163] The magnitude of the eccentric vector is calculated as follows: ;

[0164] Calculate the eccentricity direction angle as follows: ;

[0165] For each pair of radial direction indexes The theoretical remaining thickness difference is calculated as follows: , ;in, For the first The theoretical remaining thickness difference in the radial direction of the assembly.

[0166] The module that generates translation calibration displacement commands in the horizontal and vertical directions based on the components of the eccentric vector in the horizontal and vertical directions drives the actuator to complete the position correction in the plane translation module, specifically including:

[0167] The horizontal and vertical components of the eccentricity vector of the module center relative to the inner circle center of the module base are taken as opposites and used as the horizontal and vertical translation calibration values ​​of the module, thus forming the translation calibration displacement vector of the module in the image plane.

[0168] The actuator is configured to have two mutually orthogonal feed axes, such that the movement direction of one feed axis is parallel to the horizontal direction and the movement direction of the other feed axis is parallel to the vertical direction. Displacement command values ​​corresponding to the horizontal and vertical components of the translation calibration displacement vector are set in the CNC systems of the two feed axes respectively.

[0169] The horizontal and vertical displacement commands are sent to the two feed axes of the actuator, driving the actuator to perform translational movements in the horizontal and vertical directions, so that the module moves in the plane according to the translational calibration displacement vector, thereby correcting the center position of the module.

[0170] Set the module translation calibration value as follows: Each calibration component is: , ;in, The translational calibration displacement that the module needs to perform in the plane; , The module calibration displacement is respectively in direction, Components in direction;

[0171] The actuator is designed with two feed axes, one of which is perpendicular to the image coordinate system. axis With the axes parallel, the displacement commands for the actuator in the two feed axis directions are set as follows: , ;in, For the executing agency in conjunction with the image The displacement command in the direction of the feed axis parallel to the axis; For the executing agency in relation to images The displacement command in the direction of the feed axis parallel to the axis;

[0172] Displacement command , The data is sent to the two feed axes of the actuator;

[0173] The drive actuator completes the translation of the module in the plane, causing the module center to move along... , Directional movement.

[0174] After the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness, and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specifications. The module loading permission conclusion and the result data set of this real-time visual servo calibration are then output, specifically including:

[0175] For all radial pairs, take the absolute value of the theoretical remaining thickness difference for each pair of radial pairs, compare all the obtained absolute values, and select the largest one as the maximum absolute value of the theoretical remaining thickness difference.

[0176] Compare the eccentric modulus length of the module center relative to the inner circle center of the mold base with the minimum annular gap thickness, and determine at the analytical level whether the eccentric modulus length is less than the minimum annular gap thickness.

[0177] Obtain the maximum allowable limit of theoretical remaining thickness asymmetry given in the process specification, compare the maximum absolute value of the theoretical remaining thickness difference with the maximum allowable limit of theoretical remaining thickness asymmetry, and determine whether the maximum absolute value of the theoretical remaining thickness difference does not exceed the maximum allowable limit of theoretical remaining thickness asymmetry.

[0178] When both conditions are met simultaneously, the module can be installed in the mold base.

[0179] After completing the loading permission determination, the result data set of this vision servo real-time calibration is output. The result data set includes at least the eccentricity components of the module center relative to the inner circle center of the module base in the horizontal and vertical directions, the eccentricity modulus, the eccentricity direction angle, the translation calibration amount of the module in the horizontal direction and the translation calibration amount in the vertical direction in the plane, and the maximum absolute value of the theoretical remaining thickness difference.

[0180] Calculate the maximum absolute value of the theoretical remaining thickness difference for centering in all radial directions. Specifically: ;

[0181] Determine the relationship between the magnitude of eccentricity and the minimum annular gap thickness, and calculate the inequality results. ;

[0182] The maximum theoretical remaining thickness asymmetry allowed by the process is denoted as: Determine the inequality ;

[0183] In inequalities and inequalities When the condition is met, output the conclusion that allows loading.

[0184] exist or Output the conclusion that loading is prohibited;

[0185] The output dataset of the real-time visual servo calibration results is as follows:

[0186] .

[0187] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0188] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for real-time calibration of positioning deviation of a mold module based on vision servoing, characterized in that, include: An industrial camera is placed at the mold installation station to capture a grayscale image that shows the outer circle of the module and the inner circle of the mold base completely in the image and is associated with the design diameter of the inner circle of the mold base and the image coordinate system. The grayscale image is sampled along multiple radial sampling directions with the center pixel position of the image as the origin, and the initial boundary points of the inner circle of the mold base are obtained according to the grayscale changes. Two radial sampling directions in opposite directions are combined into a pair of radial directions. The midpoint of the boundary is calculated based on the initial boundary point in each pair. The average of the midpoints of the boundary is then used to obtain the coordinates of the center of the inner circle of the mold base. Starting from the center of the inner circle of the mold base, resample the outer circle boundary of the module and the inner circle boundary of the mold base in each radial sampling direction, and calculate the conversion ratio coefficient between pixels and physical size; By using a conversion scaling factor, the boundary pixel positions in each radial sampling direction are converted into the physical radii of the outer circle of the module and the inner circle of the module base, an annular gap thickness sequence and a diameter-to-diameter annular gap thickness difference sequence are constructed, and the minimum annular gap thickness is obtained. Based on the thickness difference sequence of the radial annular gap, an eccentric model of the module center relative to the inner circle center of the mold base is established. The eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base are obtained by weighted projection of the thickness difference of the radial annular gap. Based on the components of the eccentric vector in the horizontal and vertical directions, the module generates translation calibration displacement commands in the horizontal and vertical directions, driving the actuator to complete the position correction by translating the module in the plane. After the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specification, and the module loading permission conclusion and the result data set of this vision servo real-time calibration are output.

2. The method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 1, characterized in that, The step of arranging an industrial camera at the mold installation station to acquire a grayscale image that fully displays the outer circle of the module and the inner circle of the mold base, and is associated with the designed diameter of the inner circle of the mold base and the image coordinate system, specifically includes: An industrial camera is fixedly installed directly above the mold installation station, so that the optical axis of the industrial camera is aligned with the normal direction of the mold base end face, and the outer circle contour of the module to be calibrated and the inner circle contour of the mold base are simultaneously within the field of view of the industrial camera. Obtain the inner diameter of the mold base from the mold design data; An image coordinate system is established in the current image plane captured by the industrial camera. The image coordinate system is set as a two-dimensional rectangular coordinate system. The center pixel position of the image is set as the origin of the coordinate system. The horizontal direction is set as the positive direction of the horizontal axis pointing to the right side of the image, and the vertical direction is set as the positive direction of the vertical axis pointing to the top of the image, so that each pixel position in the image has a unique horizontal coordinate and vertical coordinate. Select a total number of radial sampling directions of not less than four and an even number around the origin of the coordinate system, so that each radial sampling direction is evenly distributed within the whole circle at the same angle interval, and set a unique index number and angle parameter between each radial sampling direction and the horizontal direction. Before the module is pressed into the sleeve and both the outer circle of the module and the inner circle of the mold base are fully visible in the image, a grayscale image is acquired using an industrial camera. The grayscale value of each pixel in the image is then correlated with the horizontal and vertical coordinates of each pixel in the image coordinate system to form a grayscale data set.

3. The method for real-time calibration of mold module positioning deviation based on vision servoing according to claim 2, characterized in that, The process of sampling the grayscale image along multiple radial sampling directions with the image center pixel position as the origin, and obtaining the initial boundary points of the inner circle of the mold base in each radial sampling direction based on the grayscale changes, specifically includes: In each radial sampling direction, the sampling point is moved outward pixel by pixel from the image center pixel position along the current radial sampling direction. The movement stops when the sampling point is about to exceed the image boundary. The maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective sampling length of the current radial sampling direction. In each radial sampling direction, for each radial index position within the effective sampling length range from the first step, the sampling point coordinates are sequentially obtained from the image center pixel position along the current radial sampling direction with a step size of one pixel. The theoretical coordinates are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form the radial grayscale sampling sequence in the current radial sampling direction. In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray values ​​of the current radial sampling position and the next radial sampling position is calculated as the first difference, and the difference between the gray values ​​of the current radial sampling position and the next two radial sampling positions and the gray values ​​of the current radial sampling position and the first two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the initial boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation. In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose initial boundary mutation amount is not less than the initial boundary mutation amount of the previous radial position and not less than the initial boundary mutation amount of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set of the current radial sampling direction. If the first candidate set in any radial sampling direction is empty, output the conclusion that the initial boundary extraction of the inner circle of the mold base failed and terminate the current real-time calibration process. In each radial sampling direction where the first candidate set is not empty, the position with the largest radial index in the first candidate set is selected as the radial position of the initial boundary of the inner circle of the mold base in the current radial sampling direction. Based on the radial position of the initial boundary of the inner circle of the mold base and the angle of its corresponding sampling direction, the coordinates of the initial boundary point of the inner circle of the mold base in the current radial sampling direction are calculated in the image coordinate system. The coordinates of the initial boundary points of the inner circle of the mold base calculated in all radial sampling directions are combined into a set of initial boundary points of the inner circle of the mold base.

4. The method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 3, characterized in that, The process of forming a radial direction pair from two opposing radial sampling directions, calculating the boundary midpoint based on the initial boundary points in each pair, and averaging the boundary midpoints to obtain the coordinates of the inner circle center of the mold base specifically includes: Based on the established radial sampling direction index, all radial sampling directions are divided into pairs according to the index. Each pair consists of two radial sampling directions that are opposite in direction and have an angle difference of 180 degrees on both sides of the pixel position at the center of the image, forming multiple pairs of radial directions. In each pair of radial directions, the coordinates of the initial boundary points of the inner circle of the mold base in the corresponding two radial sampling directions are obtained from the initial boundary point set of the inner circle of the mold base. The lateral coordinates of the two points are averaged to obtain the lateral coordinates of the boundary midpoint of the current pair of radial directions. The longitudinal coordinates of the two points are averaged to obtain the longitudinal coordinates of the boundary midpoint of the current pair of radial directions, thus forming the set of boundary midpoint coordinates of all pairs of radial directions. The horizontal coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the horizontal coordinates of the midpoints of all boundaries. The vertical coordinates of the center of the inner circle of the mold base in the image coordinate system are obtained by averaging the vertical coordinates of the midpoints of all boundaries. The two coordinates together constitute the center coordinates of the center of the inner circle of the mold base in the image plane.

5. The method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 4, characterized in that, The process of resampling the outer circle boundary of the module and the inner circle boundary of the module in each radial sampling direction, starting from the center of the inner circle of the mold base, and calculating the conversion ratio coefficient between pixels and physical dimensions, specifically includes: After the coordinates of the center of the inner circle of the mold base are determined, in each radial sampling direction, the sampling point is moved outward pixel by pixel from the center of the inner circle of the mold base along the current radial sampling direction. When the sampling point is about to exceed the image boundary, the movement stops and the maximum number of radial sampling steps allowed in the current radial sampling direction is recorded as the effective resampling length of the current radial sampling direction. In each radial sampling direction, for each radial index position within the range from the first step to the effective resampling length, the coordinates of the sampling points are sequentially obtained from the center of the inner circle of the mold base along the current radial sampling direction with a step size of one pixel. The calculated coordinate values ​​are mapped to the nearest pixel coordinates by rounding, and the gray value of the corresponding pixel position is read in the grayscale image to form a radial grayscale resampling sequence starting from the center of the inner circle of the mold base. In each radial sampling direction, for each radial sampling position other than the first two and the last two, the difference between the gray value of the current radial sampling position and the gray value of the next radial sampling position is calculated as the first difference, and the difference between the gray value of the current radial sampling position and the gray value of the next two radial sampling positions and the gray value of the current radial sampling position and the gray value of the previous two radial sampling positions is calculated as the second difference. The first difference and the second difference are added together to obtain the boundary change amount of the current radial sampling position. For radial positions close to the start or end point, when there are less than two sampling points before and after, the actual sampling point closest to the start or end point is used to replace the missing sampling point in the difference calculation. In each radial sampling direction, after removing the first two and last two radial sampling positions, within the remaining radial index range, select all radial positions whose boundary mutation values ​​are not less than the boundary mutation values ​​of the previous radial position and not less than the boundary mutation values ​​of the next radial position. Sort these radial positions in ascending order of index to form the first candidate set for the current radial sampling direction. If the number of candidate radial positions is less than two in the first candidate set in any radial sampling direction, the conclusion that the double boundary resampling boundary extraction based on the inner circle center of the mold base has failed is output, and the real-time calibration process is terminated. In each first candidate set, when the number of candidate radial positions is greater than or equal to two, the radial position with the smallest radial index in the candidate set is taken as the radial position of the outer circle boundary of the module in the current radial sampling direction, and the radial position with the smallest index among all radial positions in the candidate set that are greater than the radial index of the outer circle boundary of the module is taken as the radial position of the inner circle boundary of the module in the current radial sampling direction. In each pair of radial directions, the radial sampling steps corresponding to the boundary position of the inner circle of the mold base in the two radial sampling directions are obtained respectively. The radial sampling steps in the two directions are added together to obtain the pixel diameter of the inner circle of the mold base in the image in the current pair of radial directions. Summing the pixel diameters of all pairs of radial directions, multiplying the sum by two, and dividing by the total number of radial sampling directions yields the average pixel radius of the inner circle of the mold base. The scaling factor for converting pixel size to physical size is obtained based on the ratio between the inner circle design diameter of the mold base and the average pixel radius. In each radial sampling direction, the number of radial sampling steps corresponding to the outer circle boundary position of the module is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the outer circle of the module in the corresponding radial sampling direction. The number of radial sampling steps corresponding to the inner circle boundary position of the module base is multiplied by the scaling factor for converting pixel size to physical size to obtain the physical radius of the inner circle of the module base in the corresponding radial sampling direction.

6. The method for real-time calibration of mold module positioning deviation based on vision servoing according to claim 5, characterized in that, The process involves using a conversion scaling factor to convert the boundary pixel positions in each radial sampling direction into the physical radii of the module's outer circle and the mold base's inner circle, constructing annular gap thickness sequences and radial annular gap thickness difference sequences, and obtaining the minimum annular gap thickness. Specifically, this includes: In each radial sampling direction, the annular gap thickness between the inner circle of the mold base and the outer circle of the module is calculated using the difference between the physical radius of the inner circle of the mold base and the physical radius of the outer circle of the module in the current radial sampling direction, and the annular gap thickness sequence is formed according to the radial sampling direction index order. In each pair of radial directions, the annular gap thickness values ​​in the two radial sampling directions are obtained respectively. The annular gap thickness in one direction is subtracted from the annular gap thickness in the other direction that is related to it by the radial direction to obtain the annular gap thickness difference in the current pair of radial directions. The pair of radial directions are then arranged in the index order of the pair of radial directions to form a sequence of annular gap thickness differences. Squaring each annular gap thickness difference in the diametrical annular gap thickness difference sequence, and then summing all the squared results, we obtain the total thickness symmetry difference for all diametrical directions. The annular gap thickness is compared in all directions in the annular gap thickness sequence, and the smallest annular gap thickness value is selected as the minimum annular gap thickness.

7. The method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 6, characterized in that, The process of establishing an eccentric model of the module center relative to the inner circle center of the mold base based on the thickness difference sequence of the radial annular gap, and obtaining the eccentric vector, eccentricity modulus, and eccentricity direction angle of the module center relative to the inner circle center of the mold base through weighted projection of the radial annular gap thickness difference, specifically includes: An eccentric vector is set in the image plane relative to the center of the inner circle of the mold base. The eccentric vector includes a component in the horizontal direction and a component in the vertical direction. The two components together characterize the planar offset of the module center relative to the center of the inner circle of the mold base. In each pair of radial directions, based on the sampling direction angle of the current pair of radial directions and the annular gap thickness difference on the current pair of radial directions, the corresponding annular gap thickness difference is expressed as twice the projection of the eccentric vector on the current pair of radial directions, and a linear relationship model between the annular gap thickness difference and the eccentric vector components is constructed. In all radial pairs, the annular gap thickness difference in each pair is multiplied by the cosine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the horizontal direction; the annular gap thickness difference in each pair is multiplied by the sine of the angle between the corresponding radial pair and the horizontal direction, and all products are added together to obtain the weighted projection sum of the annular gap thickness difference in the vertical direction. Based on the linear model that the relationship between the annular gap thickness difference and the projection of the eccentric vector onto each radial direction is twice, the weighted projections of the annular gap thickness difference in the horizontal and vertical directions are multiplied by two and then divided by the total number of radial sampling directions to calculate the components of the eccentric vector in the horizontal and vertical directions. Squaring the components of the eccentric vector in the horizontal and vertical directions respectively, adding the squares together and taking the square root, we obtain the eccentric modulus length of the module center relative to the center of the inner circle of the module base; When the eccentricity modulus is zero, the eccentricity direction angle is set to zero degrees. When the eccentricity modulus is not zero, the eccentricity direction angle is obtained by calculating the angle between the horizontal positive direction and the eccentricity vector direction by rotating counterclockwise to the direction of the eccentricity vector using inverse trigonometric functions based on the components of the eccentricity vector in the horizontal and vertical directions. In each pair of radial directions, based on the linear relationship between the annular gap thickness difference and the projection of the eccentric vector onto the current radial direction pair, as well as the calculated eccentric vector components, the projection of the eccentric vector onto the current radial direction pair is multiplied by two. The theoretical remaining thickness difference on the current radial direction pair is obtained by subtracting the result of the projection of the eccentric vector onto the current radial direction pair multiplied by two from the annular gap thickness difference corresponding to the current radial direction pair.

8. The method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 7, characterized in that, The module that generates translation calibration displacement commands in the horizontal and vertical directions based on the components of the eccentric vector in the horizontal and vertical directions drives the actuator to complete the position correction in the plane translation module, specifically including: The horizontal and vertical components of the eccentricity vector of the module center relative to the inner circle center of the module base are taken as opposites and used as the horizontal and vertical translation calibration values ​​of the module, thus forming the translation calibration displacement vector of the module in the image plane. The actuator is configured to have two mutually orthogonal feed axes, such that the movement direction of one feed axis is parallel to the horizontal direction and the movement direction of the other feed axis is parallel to the vertical direction. Displacement command values ​​corresponding to the horizontal and vertical components of the translation calibration displacement vector are set in the CNC systems of the two feed axes respectively. The horizontal and vertical displacement commands are sent to the two feed axes of the actuator, driving the actuator to perform translational movements in the horizontal and vertical directions, so that the module moves in the plane according to the translational calibration displacement vector, thereby correcting the center position of the module.

9. A method for real-time calibration of mold module positioning deviation based on visual servoing according to claim 8, characterized in that, After the module position is corrected, the loading conditions are determined by combining the theoretical remaining thickness difference, eccentric modulus length, minimum annular gap thickness, and the maximum allowable limit of theoretical remaining thickness asymmetry given by the process specifications. The module loading permission conclusion and the result data set of this real-time visual servo calibration are then output, specifically including: For all radial pairs, take the absolute value of the theoretical remaining thickness difference for each pair of radial pairs, compare all the obtained absolute values, and select the largest one as the maximum absolute value of the theoretical remaining thickness difference. Compare the eccentric modulus length of the module center relative to the inner circle center of the mold base with the minimum annular gap thickness, and determine at the analytical level whether the eccentric modulus length is less than the minimum annular gap thickness. Obtain the maximum allowable limit of theoretical remaining thickness asymmetry given in the process specification, compare the maximum absolute value of the theoretical remaining thickness difference with the maximum allowable limit of theoretical remaining thickness asymmetry, and determine whether the maximum absolute value of the theoretical remaining thickness difference does not exceed the maximum allowable limit of theoretical remaining thickness asymmetry. When both conditions are met simultaneously, the module can be installed in the mold base. After completing the loading permission determination, the result data set of this vision servo real-time calibration is output. The result data set includes at least the eccentricity components of the module center relative to the inner circle center of the module base in the horizontal and vertical directions, the eccentricity modulus, the eccentricity direction angle, the translation calibration amount of the module in the horizontal direction and the translation calibration amount in the vertical direction in the plane, and the maximum absolute value of the theoretical remaining thickness difference.

Images