An auxiliary material curved surface fitting deviation monitoring method based on image processing technology

Abstract

The present application relates to the technical field of image processing, in particular to a kind of auxiliary material curved surface fitting deviation monitoring method based on image processing technology, steps include: S1 establishes the compensation relationship function of vibration frequency and image blurring degree;S2 collects product image before pasting and eliminates vibration blur, extracts outer contour shape and curved surface shape data, calculates theoretical projection area reference value;S3 collects vacuum adsorption auxiliary material image and extracts outer contour;S4 compares product and auxiliary material contour data, corrects displacement difference and angle deviation after vacuum fitting;S5 collects image after fitting, extracts composite contour, actual projection area and auxiliary material edge mechanical fringe density gradient, three meet the determination condition when determining qualified.The present application is through the technical framework of "vibration compensation-curved surface three-dimensional reconstruction-multidimensional collaborative determination", the positioning accuracy of curved surface auxiliary material fitting is improved, detection efficiency is improved, and the risk of mechanical failure, such as rework caused by bubble and wrinkle, is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a kind of auxiliary material curved surface fitting deviation monitoring method based on image processing technology. BACKGROUND

[0002] Auxiliary material curved surface fitting refers to the accurate and close attachment of auxiliary materials (auxiliary materials) to the surface of the main material with curved surface shape to realize specific functions or meet design requirements.A process.However, the existing auxiliary material curved surface fitting technology still has the following shortcomings:

[0003] 1.Vibration interference leads to insufficient image acquisition accuracy: industrial production line mechanical vibration (such as 0.5Hz~5Hz low frequency vibration) can cause image blur, and existing technology relies on hardware anti-shake or static shooting, without establishing a quantitative compensation model of vibration frequency and image blur, resulting in large deviation of curved surface outer contour extraction, directly affecting the fitting positioning accuracy;

[0004] 2.Poor adaptability of curved surface shape: traditional planar fitting detection method only compares two-dimensional contour or simply locates coordinates, which cannot accurately calculate the projection area reference value under three-dimensional shape of curved surface, and the area calculation error is large when the product curvature radius is small;

[0005] 3.Limitation of single determination standard: existing technology mostly uses "position deviation" as the only qualified standard, ignoring the mechanical properties of auxiliary material after fitting, such as stress concentration caused by bubbles and wrinkles, resulting in "pseudo-qualified" products flowing into the next process.

[0006] Therefore, a kind of auxiliary material curved surface fitting deviation monitoring method based on image processing technology is proposed. SUMMARY

[0007] The present application aims to provide an auxiliary material curved surface fitting deviation monitoring method based on image processing technology to solve the problems of insufficient image acquisition accuracy caused by vibration interference, poor adaptability of curved surface shape and limitation of single determination standard in existing technology.

[0008] To solve the above technical problems, the present application aims to provide an auxiliary material curved surface fitting deviation monitoring method based on image processing technology, comprising the following steps:

[0009] S1, fix the calibration sheet on the product fixture, apply 0.5Hz to 5Hz mechanical vibration to the fixture, and synchronously collect image sequence of calibration sheet under vibration state; based on pattern edge displacement, establish compensation relationship function of vibration frequency and image blur, wherein image blur refers to pixel displacement deviation value caused by vibration;

[0010] S2, collect the image of the product before mounting, call the compensation relationship function to eliminate the vibration blur of the product before mounting, extract the product outer contour shape data and the curved surface shape data of the product before mounting after eliminating the vibration blur, and calculate the theoretical projection area reference value according to the curved surface shape data;

[0011] S3, collect the image of the auxiliary material in vacuum adsorption, and extract the auxiliary material outer contour shape data;

[0012] S4, compare the product and auxiliary material outer contour shape data, calculate the XY direction displacement difference and angle deviation, and correct the displacement difference and angle deviation to perform vacuum lamination;

[0013] S5, collect the image after lamination, extract the composite outer contour shape data, the actual projection area value and the density gradient of the auxiliary material edge mechanical stripe, and compare them with the judgment condition, and when all three meet the judgment condition, the auxiliary material and the curved surface lamination are qualified.

[0014] As a further improvement of the technical solution, in step S1, the compensation relationship function is established by the following steps:

[0015] S1.1, collect the calibration sheet image under 0.5Hz, 2.5Hz and 5Hz vibration frequency respectively;

[0016] S1.2, measure the pixel displacement δ of the image sequence corresponding to each frequency;

[0017] S1.3, fit to obtain the compensation model δ=k×f², wherein f is the vibration frequency and k is the calibration coefficient.

[0018] As a further improvement of the technical solution, in step S2, the product outer contour shape data extraction includes:

[0019] S2.1, calculate the gradient amplitude of the product before mounting after eliminating the vibration blur, and preliminarily screen the edge pixels;

[0020] S2.2, Gaussian surface fitting is performed on the neighborhood of the edge pixels to locate the sub-pixel level coordinate points;

[0021] S2.3, connect the coordinate points to generate a closed polygon, calculate the centroid coordinates, and establish the lamination reference position according to the centroid coordinates.

[0022] As a further improvement of the technical solution, in step S2, the theoretical projection area reference value calculation includes:

[0023] S2.5, calibrate the relative spatial position of the camera and the product fixture to establish a three-dimensional coordinate system;

[0024] S2.6, divide the product curved surface into 0.1mm×0.1mm micro units;

[0025] S2.7, calculate the angle θ between the surface normal of each micro-unit and the optical axis of the camera based on the relationship between the relative spatial position of the camera and the product fixture;

[0026] S2.8, accumulate the actual area of each micro-unit multiplied by cosθ to obtain the corrected projection area.

[0027] As a further improvement of the technical solution, in step S5, the determination condition of the composite outer contour shape data is that the shape coincidence degree of the composite outer contour and the product outer contour is greater than or equal to a preset value A.

[0028] The determination condition of the actual projection area value is that the deviation of the actual projection area value relative to the theoretical projection area reference value is less than or equal to a preset value B.

[0029] The determination condition of the density gradient of the auxiliary material edge mechanical stripe is that the density gradient of the auxiliary material edge mechanical stripe is less than or equal to a preset value C.

[0030] As a further improvement of the technical solution, in step S5, the shape coincidence degree calculation includes:

[0031] S5.1, nearest point matching of the polygon vertex coordinates before and after fitting;

[0032] S5.2, using the average distance of the matched point pairs as the coincidence degree evaluation index.

[0033] As a further improvement of the technical solution, in step S5, the density gradient of the auxiliary material edge mechanical stripe is defined by the following method:

[0034] Along the normal direction of the auxiliary material edge, the spatial derivative value of the number of gray value peak-valley changes per unit length is calculated in a strip-shaped detection area with a width W, and the value range of W is 0.1mm≤W≤0.5mm.

[0035] As a further improvement of the technical solution, in step S5, the determination result further includes:

[0036] If the composite outer contour shape data or the actual projection area value does not meet the determination condition, the determination result is that the fitting position is offset.

[0037] If the density gradient of the auxiliary material edge mechanical stripe does not meet the determination condition, the determination result is that the mechanical limit risk is exceeded.

[0038] Compared with the prior art, the beneficial effects of the present application are:

[0039] 1、The auxiliary material curved surface fitting offset monitoring method based on image processing technology, through the technical framework of "vibration compensation-curved surface three-dimensional reconstruction-multi-dimensional collaborative judgment", the positioning accuracy of the curved surface auxiliary material fitting is improved, the detection efficiency is improved, and the mechanical failure risk such as rework caused by bubbles and wrinkles is reduced.

[0040] 2、The auxiliary material curved surface fitting offset monitoring method based on image processing technology, by establishing a compensation model, the image blur caused by 0.5Hz-5Hz vibration can be eliminated in real time, the edge extraction accuracy is improved, and the data foundation for subsequent contour matching is laid.

[0041] 3、The auxiliary material curved surface fitting offset monitoring method based on image processing technology, by using the "micro-unit division-normal angle correction" algorithm, the error of curved surface projection area calculation is reduced, which is especially suitable for complex curved surface scenes with a curvature radius of 3mm-10mm.

[0042] 4、The auxiliary material curved surface fitting offset monitoring method based on image processing technology, by fusing "shape coincidence degree, projection area deviation, and mechanical stripe density gradient", the misjudgment rate is reduced compared with single position judgment standard, and the two failure modes of "position deviation" and "mechanical overrun" can be distinguished. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0045] Auxiliary material curved surface fitting refers to precisely and closely attaching auxiliary materials (auxiliary materials) to the surface of main materials with curved surface shape to realize specific functions or meet design requirements. However, the existing auxiliary material curved surface fitting technology still has deficiencies in "vibration interference leading to insufficient image acquisition accuracy, poor curved surface form adaptability, and limitations of single judgment standard";

[0046] In view of this, as Figure 1 shown, the present application aims to provide an auxiliary material curved surface fitting offset monitoring method based on image processing technology, comprising the following steps:

[0047] S1, fix the calibration sheet on the product fixture, apply mechanical vibration of 0.5Hz to 5Hz to the fixture, and synchronously collect image sequences of the calibration sheet in the vibration state; establish a compensation relationship function between vibration frequency and image blurring degree based on the pattern edge displacement amount, wherein the image blurring degree refers to the pixel displacement deviation value caused by vibration;

[0048] In view of the fact that in the automated fitting production of auxiliary materials and curved products, the image acquisition system is easily disturbed by environmental mechanical vibration, the periodic vibration of the product fixture is caused by the operation of the equipment in the production workshop (such as the driving of the conveyor belt, the movement of the mechanical arm, the micro-vibration of the fixture fixing mechanism, etc.), and when the calibration sheet, the product to be fitted or the auxiliary material vibrates with the fixture, relative motion is formed with the camera. This relative motion will cause the object edge in the image to produce "smearing" within the exposure time, that is, the same pixel position corresponds to different parts of the object at different times, which appears as image blurring, and the essence is the pixel displacement deviation (δ) caused by vibration. This blurring will directly damage the accuracy of subsequent image processing, for example, when the product outer contour is extracted, the blurred edge may cause the sub-pixel positioning error to increase (usually up to 2-5 pixels); when the XY direction displacement difference of the auxiliary material and the product is calculated, the contour shift caused by the blurring may be misjudged as the actual fitting deviation; and even when the curved shape data is extracted, the blurring will cover the true normal direction of the micro-unit, causing the calculation of the theoretical projection area reference value to be distorted. The vibration frequency of the fixture in production is often unstable (usually fluctuates in the range of 0.5Hz to 5Hz, which covers the frequency characteristics of common vibration sources such as small motor-driven equipment and pneumatic components), and if a quantitative compensation relationship is not established for vibration of different frequencies, it is difficult to eliminate dynamic blurring simply by image filtering, which will eventually lead to a decrease in the accuracy of fitting quality determination. However, the traditional technology only mentions that "vibration affects image quality", and does not disclose the mathematical relationship between 0.5Hz-5Hz low-frequency vibration (common vibration range of production line) and pixel displacement amount δ, which makes the compensation model unable to be engineered and landed, therefore, in step S1, the systematic technical scheme of "vibration parameter calibration, dynamic image acquisition and mathematical modeling" is adopted, specifically:

[0049] The calibration sheet is fixed on the product fixture (fixture material is 45# steel, flatness ≤0.02mm) by vacuum adsorption (adsorption force 0.08MPa±0.01MPa), and a electromagnetic vibration table (model: LDSV406) is used to apply sinusoidal mechanical vibration (amplitude 0.1mm, acceleration 0.5g) of 0.5Hz to 5Hz to the fixture;

[0050] An industrial camera (resolution 5 million pixels, frame rate 120fps, lens focal length 25mm) is used to synchronously collect image sequences (100 frames of images are collected for each frequency point) of the calibration sheet (containing a 10x10 array of 0.1mmx0.1mm black and white checkerboard) in the vibration state;

[0051] A compensation relationship function between the vibration frequency and the image blurring degree is established based on the pattern edge displacement amount, wherein the image blurring degree refers to a pixel displacement deviation value caused by vibration (calculated by a sub-pixel level edge detection algorithm, with a precision of 0.01 pixels);

[0052] The compensation relationship function is established by the following steps:

[0053] S1.1, images of the calibration sheet are collected at 0.5 Hz, 2.5 Hz and 5 Hz vibration frequencies respectively (the basis for selecting the three frequency points: covering 90% of the energy distribution interval of the actual production line vibration spectrum, and meeting the Nyquist sampling theorem, and when the frequency interval is 2 Hz, spectral aliasing can be avoided);

[0054] S1.2, a sub-pixel edge detection algorithm based on gray gradient (Canny operator + quadratic curve fitting) is used to measure the pixel displacement amount δ of the image sequence corresponding to each frequency (the average displacement amount of 100 images is calculated, and the standard deviation is ≤0.05 pixels);

[0055] S1.3, based on the simple harmonic vibration principle (vibration displacement is related to acceleration, and acceleration is proportional to the square of frequency), the δ value of the preset number (such as 3) of frequency points is fitted with the frequency f to obtain a compensation model of δ=k×f² (wherein f is the vibration frequency, unit Hz; k is the calibration coefficient, calculated by the least square method, and the fitting error is ≤0.1 pixels);

[0056] The core benefit of this technical solution is:

[0057] Firstly, through the calibration of the characteristic frequency points, the abstract relationship between "vibration and blurring" is converted into a calculable mathematical model, and the specific values of the pixel displacement deviation at different frequencies are determined (for example, δ is about 2.3 pixels at 5 Hz vibration, and about 0.6 pixels at 2.5 Hz), which provides accurate quantitative basis for subsequent image correction;

[0058] Secondly, by selecting 0.5 Hz (low-speed equipment vibration), 2.5 Hz (medium-speed motor) and 5 Hz (high-frequency pneumatic component) three frequency points, the vibration interval commonly seen in production is covered, and the compensation accuracy of any frequency in the range of 0.5-5 Hz is ensured through three-point fitting (interpolation error ≤0.05 pixels), covering the vibration range in actual production;

[0059] Thirdly, the quadratic function model has small calculation amount (single calculation time ≤1 ms), which can be integrated into the image preprocessing module in the system. When collecting product or auxiliary material images subsequently, the δ value can be quickly calculated by substituting the jig vibration frequency (such as installing a frequency sensor) into the model, and then the blurring is eliminated through the pixel offset correction algorithm, so that the image edge sharpness is improved, and real-time dynamic compensation is realized;

[0060] Fourthly, the edge positioning error of the compensated image is controllable, ensuring the precision of subsequent product outer contour extraction, auxiliary material position calibration, projection area calculation and other steps, so that the accuracy of the fitting deviation is improved, and the misjudgment rate is reduced.

[0061] S2, collect the image before the product is attached, call the compensation function to eliminate the vibration blur of the image before the product is attached, extract the product outer contour shape data and the curved surface shape data of the image after the vibration blur is eliminated, and calculate the theoretical projection area reference value according to the curved surface shape data;

[0062] Considering that the image quality before the product is attached directly determines the accuracy of subsequent fitting positioning and quality judgment: on the one hand, even after the vibration compensation in step S1, the residual slight vibration blur may still cause the product edge pixel positioning deviation, and if the contour is directly extracted, the edge may be "jagged" or the key feature points may be lost, thereby affecting the positioning accuracy of the auxiliary material and the product; on the other hand, there is a natural difference between the actual area of the curved surface product and the plane projection area, the angles between the parts of the curved surface and the camera optical axis are different, and the projection sizes of the same actual area region in the image are different, and if the pixel area in the plane image is directly taken as the reference, it will lead to the distortion of the judgment of the "actual projection area deviation" after the fitting, for example, the projection area of the convex part of the curved surface will be underestimated, and the concave part will be overestimated, and the two problems will jointly cause the inaccuracy of the fitting reference position and the area judgment, and finally cause the auxiliary material deviation or misjudgment, therefore, in step S2, a hierarchical technical solution of "vibration blur elimination, high-precision contour extraction, and curved surface projection correction" is adopted, specifically:

[0063] Vibration blur preprocessing: first, the compensation function established in S1 is called, the corresponding pixel displacement deviation δ is calculated according to the real-time collected jig vibration frequency, the vibration blur of the image before the product is attached is eliminated through the image pixel offset correction algorithm such as bilinear interpolation remapping, and the edge of the image is ensured to be clear;

[0064] Product outer contour extraction technology:

[0065] S2.1, gradient amplitude calculation (based on Sobel operator or Prewitt operator) is adopted, the change rate (gradient amplitude) of the pixel gray value in the image is calculated, the edge pixels with gray mutation (pixels with a gradient amplitude exceeding a preset threshold are marked as candidate edges) are preliminarily screened out, and the interference of non-edge regions is quickly excluded;

[0066] S2.2, Gaussian surface fitting is performed on the 3*3 or 5*5 neighborhood of the candidate edge pixels (a two-dimensional Gaussian function is used to fit the neighborhood gray distribution), the sub-pixel level coordinates are positioned by finding the extreme points of the Gaussian surface (the precision can reach within 0.1 pixels), and the "discretization error" of the traditional pixel level positioning is solved.

[0067] S2.3, generate a closed polygon (fit the product outer contour) by connecting sub-pixel coordinate points through least squares method, and calculate the centroid coordinates (x0, y0) of the polygon as the reference position for the auxiliary material fitting (the centroid has translational invariance, which can avoid the influence of local contour deformation on the reference);

[0068] Theoretical projection area reference value calculation technology:

[0069] S2.5, accurately calibrate the relative spatial position of the camera and the product fixture (including camera intrinsic parameters such as focal length, principal point, and extrinsic parameters such as rotation matrix and translation vector) through Zhang Zhengyou calibration method or laser tracker, and establish a three-dimensional coordinate system with the fixture as the origin (X axis along the length direction of the fixture, Y axis along the width direction, Z axis perpendicular to the fixture plane);

[0070] S2.6, divide the product curved surface into 0.1mm x 0.1mm micro-units (the size is much smaller than the accuracy requirement of auxiliary material fitting, which ensures that the details of the curved surface form are retained);

[0071] S2.7, based on the pose relationship between the camera and the fixture, calculate the normal vector of the surface of each micro-unit (obtained by surface fitting and derivation), and then calculate the angle θ between the normal and the camera optical axis (along the Z axis of the camera coordinate system);

[0072] S2.8, according to the projection geometry principle (plane projection area = actual area x cosθ), accumulate the "actual area x cosθ" of all micro-units to obtain the corrected theoretical projection area reference value (eliminate the influence of curved surface angle on projection);

[0073] The core benefits of this technical solution are:

[0074] First, through the combination of gradient amplitude preliminary screening and Gaussian surface fitting, the positioning accuracy of the product outer contour is improved from pixel level (about 0.02-0.1mm, depending on the camera resolution) to sub-pixel level (≤0.01mm), which ensures that the alignment deviation between the auxiliary material and the product is within the allowable range, and improves the contour accuracy;

[0075] Second, using the contour centroid as the fitting reference avoids the interference of local edge defects (such as edge mutations caused by product surface scratches) on the reference, making the calculation of subsequent XY direction displacement difference and angle deviation more stable;

[0076] Third, through micro-unit division and θ angle calculation, the deviation from "actual area to projection area" of the curved product is reduced, providing an accurate reference for the judgment of "actual projection area deviation" after fitting, for example, the traditional plane projection area error of a mobile phone back cover with large curved surface radius can reach 8%, and after correction, the error is ≤0.5%;

[0077] Fourth, from vibration blur elimination to contour extraction and area calculation, the technologies of each link are interconnected to ensure that the product's shape data (contour, area, reference position) can truly reflect its spatial state, providing "true value" reference data for the bonding calibration in step S4 and the quality judgment in step S5, and reducing the bonding failure rate caused by data distortion.

[0078] S3. Collect images of the excipients under vacuum adsorption and extract the outer contour shape data of the excipients;

[0079] Considering that auxiliary materials are prone to morphological changes under vacuum adsorption (e.g., thin auxiliary materials may experience edge warping or wrinkling due to uneven adsorption force, or positional shift due to minute displacement during adsorption), and that the auxiliary materials themselves may have transparent, semi-transparent, or reflective properties, direct image acquisition is prone to problems such as blurred edges and low contrast. If the outer contour shape data cannot be accurately extracted, it will lead to deviations in subsequent comparisons with the product's outer contour, thus affecting the calculation accuracy of displacement difference and angle deviation in the XY directions, ultimately resulting in inaccurate bonding and positioning. Therefore, in step S3, a technical solution of "targeted image acquisition and adaptive contour extraction" is adopted, specifically:

[0080] Synchronous image acquisition under vacuum: A high-resolution industrial camera (resolution ≥ 20 million pixels) is used in conjunction with a coaxial light source or a backlight source (selected according to the material of the auxiliary material; backlight is used to enhance edge contrast for transparent auxiliary materials, and coaxial light is used to suppress surface reflection for opaque auxiliary materials). The acquisition is triggered the instant the auxiliary material is stably adsorbed by the vacuum adsorption device (the adsorption pressure is stabilized by a sensor to ensure that the shape of the auxiliary material is fixed at the time of acquisition), avoiding the influence of dynamic deformation during adsorption on the image.

[0081] Adaptive edge extraction algorithm: First, the acquired image is preprocessed (including Gaussian filtering to remove noise and local threshold segmentation to enhance the gray-level difference between the edge and the background). Then, an improved Canny edge detection algorithm is used (by dynamically adjusting high and low thresholds, lowering the threshold in blurred areas to retain weak edges and raising the threshold in high-contrast areas to filter noise) to extract the initial edge. Finally, morphological operations (such as erosion and dilation) are used to remove edge burrs, and a contour tracking algorithm (such as contour connection based on chain code) is used to generate a closed outer contour curve of the auxiliary material to eliminate interference from non-contour areas (such as small stains or scratches on the surface of the auxiliary material).

[0082] The advantages of this technical solution are:

[0083] First, through synchronous collection after stable vacuum adsorption, the extracted outer contour reflects the actual shape before the auxiliary material is attached (rather than the shape in the natural state), avoiding contour errors caused by adsorption deformation (for example, the edge of a certain thin film auxiliary material is straight in the natural state, but slightly warped after adsorption. Synchronous collection can accurately capture this deformation and ensure the pertinence of subsequent comparison), thereby capturing the true shape;

[0084] Second, targeted light source selection and adaptive edge detection algorithm solve the problem of edge blur caused by transparent / reflective auxiliary materials, improving the completeness of contour extraction, thereby adapting to the characteristics of auxiliary materials;

[0085] Third, through preprocessing and morphological operation, noise, stains and other interference are effectively filtered, making the extracted outer contour edge smooth and continuous (reducing burr rate), providing high-precision basic data for the comparison of product and auxiliary material outer contour shape data in step S4.

[0086] S4, compare product and auxiliary material outer contour shape data, calculate XY direction displacement difference and angle deviation, and correct displacement difference and angle deviation before vacuum lamination;

[0087] Considering that the actual position and angle of the auxiliary material after vacuum adsorption may deviate from the preset lamination position of the product, this deviation may be caused by multiple factors: slight sliding of the auxiliary material during vacuum adsorption (especially thin auxiliary materials are prone to translation due to uneven adsorption force), mechanical tolerance of jig positioning (usually within ±0.05mm), and incomplete alignment of coordinate systems when extracting product and auxiliary material contours (differences in viewing angle during image acquisition may cause implicit angle deviation). If lamination is performed directly without correcting these deviations, it will cause the edge of the auxiliary material to be misaligned with the edge of the product (XY direction deviation) or rotated as a whole (angle deviation), ultimately resulting in lamination deviation, incomplete coverage and other quality problems. Especially for curved products, slight positional deviation may be amplified due to the curvature of the curved surface, resulting in local warping or wrinkling. Therefore, in step S4, the technical solution of "contour feature matching, deviation quantification calculation, and precise driving correction" is adopted, which is specifically:

[0088] Outer contour feature matching: the product outer contour (closed polygon) extracted in step S2 and the auxiliary material outer contour (closed polygon) extracted in step S3 are imported into the same coordinate system (with the product centroid as the origin), and the Iterative Closest Point (ICP) algorithm is used for contour matching. By calculating the Euclidean distance of corresponding feature points (such as vertices, curvature extreme points) on the two contours, the matching relationship that minimizes the overall distance is found, excluding the interference of local edge noise;

[0089] Deviation quantification: Based on the matching results, calculate the XY direction displacement difference (taking the product centroid as the reference, the coordinate difference Δx, Δy of the auxiliary material centroid on the X axis and Y axis), and the angle deviation (fit the main direction vectors of the two contours through the least squares method, calculate the included angle Δθ between the vectors);

[0090] Precise driving correction: Convert Δx, Δy, Δθ into driving signals to control the multi-axis servo platform (positioning accuracy ≤0.001mm, angle control accuracy ≤0.01°) to drive the auxiliary material adsorption device to adjust translation and rotation until the deviation value is reduced to within the preset threshold (usually Δx, Δy ≤0.02mm, Δθ ≤0.1°), and then trigger the vacuum lamination action;

[0091] The benefits of this technical solution are:

[0092] Firstly, by contour feature matching instead of single feature point comparison, asymmetric deviations caused by local deformation (such as auxiliary material edge micro-curl) can be captured, avoiding the limitations of traditional "point-to-point" comparison (which may ignore overall angle or translation deviation);

[0093] Secondly, displacement and angle deviation are converted into specific numerical values (Δx, Δy, Δθ), making the correction process quantifiable and traceable, ensuring consistency in each correction (uniform deviation correction standards for different batches of auxiliary materials);

[0094] Thirdly, precise driving correction can compensate for the "cumulative deviation" of curved products, even if the product curve has a slight radius, through joint correction of XY direction and angle, it can ensure that the auxiliary material edge is accurately aligned with the target lamination line of the product curve, avoiding local deviation caused by curve inclination.

[0095] S5, collect the lamination image, extract the composite outer contour shape data, actual projection area value and auxiliary material edge mechanical stripe density gradient, and compare with the judgment condition, when all three meet the judgment condition, the auxiliary material and the curved surface lamination is qualified.

[0096] Considering that after the auxiliary material is laminated with the product, it is difficult to fully reflect the lamination quality by a single indicator (such as contour alignment), there may be situations such as "contour alignment but insufficient area coverage" (such as auxiliary material local lamination leading to small projection area) and "area meets the standard but edge mechanics is unstable" (such as auxiliary material edge has hidden curl or wrinkles, contour alignment in the short term but easy to fall off in the long term) in actual production. If only rely on a single indicator for judgment, it may lead to potential quality risk missed detection (such as mechanically unstable products may fail in subsequent processes) or misjudgment of qualified products (such as local contour micro-deviation due to curve characteristics but actual lamination is good), therefore, multi-dimensional indicators need to be used for joint determination to ensure the reliability of lamination quality, therefore, in step S5, the technical solution of "multi-dimensional feature extraction, quantitative comparison and grading judgment" is adopted, specifically:

[0097] First, image acquisition and feature extraction after bonding: use the same parameters as steps S2 and S3 to collect the image after bonding (ensure imaging consistency), separate the composite area of the product and the auxiliary material through image segmentation technology, and extract three core features:

[0098] Composite outer contour shape data: use the same sub-pixel edge detection algorithm as step S2 to extract the overall outer contour (composite contour) after bonding of the product and the auxiliary material;

[0099] Actual projection area value: based on the pixel area within the composite contour, combined with the three-dimensional coordinate system and micro-unit division established in step S2, calculate the actual projection area after bonding according to the same "actual area x cos θ" method as step S2.8;

[0100] Auxiliary material edge mechanical fringe density gradient: set a 0.1-0.5mm wide strip-shaped detection area (width W is selected according to the thickness of the auxiliary material, thick auxiliary material takes a large value) along the normal direction of the auxiliary material edge, calculate the number of gray value peak-valley changes per unit length (reflecting the density of edge wrinkles or lifting), and then calculate the spatial derivative value of the number (reflecting the degree of mechanical change);

[0101] Second, quantitative comparison and judgment logic:

[0102] Shape coincidence degree judgment: through step S5.1, match the composite outer contour with the polygon vertices of the product outer contour extracted in step S2 (ensure one-to-one correspondence of corresponding vertices), step S5.2 calculates the average distance of the matched point pairs, when the average distance ≤ preset threshold A (i.e. coincidence degree ≥ A, such as A=95%), the contour alignment is qualified;

[0103] Area deviation judgment: calculate the relative deviation of the actual projection area value and the theoretical projection area reference value of step S2 (| actual - reference | / reference x 100%), when the deviation ≤ preset value B (such as B=3%), the area coverage is qualified;

[0104] Mechanical fringe judgment: when the density gradient of the auxiliary material edge mechanical fringe ≤ preset value C (such as C=5 times / mm²), the edge mechanical stability is qualified;

[0105] Third, output of grading results:

[0106] If the contour or area is not up to standard, it is judged as "bonding position offset" (the correction parameters of step S4 need to be adjusted);

[0107] If the mechanical fringe is not up to standard, it is judged as "mechanical overrun risk" (the vacuum suction force or bonding pressure needs to be optimized);

[0108] All three standards are met, then it is determined to be qualified;

[0109] The technical scheme has the advantages that:

[0110] Firstly, the three indicators are respectively evaluated from the three dimensions of "position accuracy" (contour coincidence degree), "coverage integrity" (area deviation), and "mechanical stability" (stripe gradient), avoiding the limitation of a single indicator (such as the traditional detection of contour may miss the edge implicit lifting), so that the comprehensiveness of quality determination is improved;

[0111] Secondly, by setting the determination threshold (A, B, C) of the numerical coincidence degree, deviation rate and gradient value, the traditional subjective judgment of manual visual inspection is replaced, so that the determination standards of different batches and different operators are consistent, and the misjudgment rate is reduced;

[0112] Thirdly, the grading determination result can directly point to the root cause of the quality problem - the insufficient correction accuracy of "position deviation" corresponding to S4, and the unreasonable process parameters (such as adsorption force and pressure) corresponding to "mechanical over-limit risk", which provides a clear direction for production optimization and reduces the rework rate;

[0113] Fourthly, through the area calculation and edge normal direction stripe detection related to the three-dimensional coordinate system, the "projection distortion" and "edge stress concentration" characteristics of curved surface products are specially designed, solving the applicability problem of the planar bonding determination method in the curved surface scene (such as the area deviation calculation of curved surface edge is more accurate), ensuring the bonding quality controllable of complex curved surface products, and adapting to the characteristics of curved surface bonding.

[0114] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An auxiliary material curved surface fitting deviation monitoring method based on image processing technology, characterized in that, Includes the following steps: S1. Fix the calibration plate to the product fixture, apply mechanical vibration of 0.5Hz to 5Hz to the fixture, and simultaneously acquire image sequences of the calibration plate under vibration; establish a compensation relationship function between vibration frequency and image blur based on the pattern edge displacement, where image blur refers to the pixel displacement deviation caused by vibration; the compensation relationship function is established through the following steps: S1.1 Acquire calibration plate images at vibration frequencies of 0.5Hz, 2.5Hz, and 5Hz respectively; S1.2 Measure the pixel displacement δ of the image sequence corresponding to each frequency; S1.

3. The compensation model δ=k×f² is obtained by fitting, where f is the vibration frequency and k is the calibration coefficient; S2. Collect the product image before mounting, call the compensation relationship function to eliminate the vibration blur of the product image before mounting, extract the product outer contour shape data and surface morphology data from the product image before mounting after vibration blur is eliminated, and calculate the theoretical projection area benchmark value based on the surface morphology data. S3. Collect images of the excipients under vacuum adsorption and extract the outer contour shape data of the excipients; S4. Compare the outer contour shape data of the product and the auxiliary materials, calculate the displacement difference and angle deviation in the XY direction, and perform vacuum bonding after correcting the displacement difference and angle deviation. S5. Collect images after bonding, extract composite outer contour shape data, actual projected area value and density gradient of mechanical stripes on the edge of the auxiliary material, and compare them with the judgment conditions. If all three are met, the auxiliary material is judged to be bonded to the curved surface.

2. The image processing technology-based method for monitoring the misalignment of the drape curve according to claim 1, characterized in that, In step S2, the extraction of product outer contour shape data includes: S2.1 Calculate the gradient amplitude of the product image before mounting to eliminate vibration blur and initially screen edge pixels; S2.

2. Perform Gaussian surface fitting on the neighborhood of edge pixels to locate sub-pixel level coordinate points; S2.3 Connect the coordinate points to generate a closed polygon, calculate its centroid coordinates, and establish a fitting reference position based on the centroid coordinates.

3. The image processing technology-based method for monitoring the misalignment of the drape curve according to claim 1, characterized in that, In step S2, the calculation of the theoretical projected area reference value includes: S2.

5. Calibrate the relative spatial position of the camera and the product fixture, and establish a three-dimensional coordinate system; S2.6 Divide the product surface into micro-units of 0.1mm × 0.1mm; S2.

7. Based on the relative spatial position relationship between the calibration camera and the product fixture, calculate the angle θ between the surface normal of each micro-unit and the optical axis of the camera; S2.

8. Accumulate the actual area of ​​each micro-unit and multiply it by cosθ to obtain the corrected projected area.

4. The image processing technology-based method of monitoring the misalignment of the drape on the curved surface according to claim 1, characterized in that, In step S5, the determination condition for the composite outer contour shape data is: the overlap between the composite outer contour and the product outer contour is ≥ preset value A. The criterion for determining the actual projected area value is: the deviation of the actual projected area value from the theoretical projected area reference value is ≤ preset value B; The criterion for determining the density gradient of the mechanical stripes at the edge of the auxiliary material is: the density gradient of the mechanical stripes at the edge of the auxiliary material ≤ the preset value C.

5. The image processing technology-based method for monitoring the misalignment of the drape curve according to claim 4, characterized in that, In step S5, the shape overlap calculation includes: S5.1 Match the nearest points of the vertices of the polygons before and after the fit; S5.

2. The average distance between matching point pairs is used as the overlap evaluation index.

6. The image processing technology-based method of monitoring the misregistration of a web of auxiliary material according to claim 1, wherein, In step S5, the density gradient of the mechanical stripes at the edge of the auxiliary material is defined in the following way: The spatial derivative value of the number of times of peak-valley changes of the gray value per unit length is calculated in a strip-shaped detection region with a width W along the normal direction of the edge of the auxiliary material, and the value of W is in the range of 0.1mm≤W≤0.5mm.

7. The image processing technology-based method for monitoring the misplacement of the drape on the curved surface according to claim 4, wherein, The determination result in the step S5 further includes: If the composite outer contour shape data or the actual projection area value does not meet the determination condition, the determination result is that the fitting position is deviated; If the density gradient of the mechanical fringe of the edge of the auxiliary material does not meet the determination condition, the determination result is that the mechanical limit is exceeded.

Images