A method and system for detecting lamination alignment error

Abstract

This application relates to the field of electronic manufacturing technology, specifically to a method and system for detecting alignment errors in a film coating process. The method includes: acquiring a grayscale image of a coated substrate and a local film thickness gradient along the substrate transport direction; determining the center frequency and notch bandwidth of interference fringes in the image frequency domain based on the local film thickness gradient, and filtering the spectrum of the grayscale image of the coated substrate according to the center frequency and the notch bandwidth to obtain a cleaned image; extracting the gradient amplitude of the cleaned image within the region of interest of a marked point, and determining a calculation window based on the statistical characteristics of the gradient amplitude, calculating the sub-pixel center coordinates of the marked point within the calculation window; performing error calculation based on the sub-pixel center coordinates, and verifying the result by combining a preset alignment tolerance, thereby obtaining the alignment error detection result. This application can adaptively separate interference frequency components and achieve accurate detection and correction of alignment errors.

Description

Technical Field

[0001] This application relates to the field of electronic manufacturing technology, specifically to a method and system for detecting coating alignment error. Background Technology

[0002] In fields such as electronics manufacturing, display panel packaging, and optical component assembly, lamination is a core production process. This process requires the precise lamination of functional films, protective films, or optical films onto designated areas of the substrate. The alignment accuracy directly determines the electrical connectivity, optical uniformity, and final yield of the product. As product integration continues to increase, the alignment tolerance has narrowed to the micrometer level. Traditional manual visual inspection can no longer meet the requirements, making automated visual inspection systems a standard feature of lamination production lines.

[0003] The typical workflow of existing visual inspection solutions involves acquiring images of the coated substrate using an industrial camera, identifying alignment markers pre-placed on the substrate and the film layer, and driving a bonding mechanism to perform correction actions by calculating the coordinate deviation of the markers. To improve the recognition stability of markers in complex backgrounds, image enhancement preprocessing is usually applied before positioning, including histogram equalization, edge sharpening, or Gaussian difference filtering. This is then combined with normalized cross-correlation template matching to complete coarse positioning of the markers, and finally, sub-pixel refinement is performed using the gradient centroid method.

[0004] However, after optically transparent or semi-transparent films are bonded to the substrate, a thin-film interference cavity is formed between the upper and lower surfaces of the film. Incident light reflects off the two interfaces and superimposes to produce interference fringes with a periodic spatial distribution. Since the coating process on the production line cannot guarantee a completely uniform film thickness across the entire substrate, the film thickness exhibits a gradually varying gradient along the substrate direction. The spatial period of the equal-thickness interference fringes is determined by the film thickness gradient, the film refractive index, and the center wavelength of the light source. Because its typical spatial period falls within the range of tens to hundreds of micrometers, it highly overlaps with the main spatial frequencies of the gradient characteristics at the edge of the marker point. If only existing image enhancement methods operating in the spatial domain are used, it is impossible to effectively separate the interference fringe components from the gradient field at the marker point edge in the frequency domain, resulting in the gradient field around the marker point being contaminated by the superposition of interference fringes. During sub-pixel centroid localization, the contaminated gradient weight distribution produces an asymmetric skew, causing a systematic shift in the centroid calculation result relative to the actual marker point center. Ultimately, this leads to the alignment error calculation value deviating from the actual physical error, causing detection inaccuracies in high-precision bonding scenarios. Summary of the Invention

[0005] To address the existing technical problem that the aliasing of optical interference fringes and marker edge gradient features in the spatial frequency domain causes systematic shifts in sub-pixel positioning, resulting in inaccurate alignment error detection, this application provides a method and system capable of adaptively separating interference frequency components and achieving accurate alignment error detection and correction.

[0006] In a first aspect, this application provides a method for detecting alignment error in a film coating, comprising: acquiring a grayscale image of a film-coated substrate and a local film thickness gradient along the substrate transport direction; determining the center frequency and notch bandwidth of interference fringes in the image frequency domain based on the local film thickness gradient, and filtering the spectrum of the grayscale image of the film-coated substrate according to the center frequency and the notch bandwidth to obtain a cleaned image; extracting the gradient amplitude of the cleaned image in the region of interest of a marked point, and determining a calculation window based on the statistical characteristics of the gradient amplitude, calculating the sub-pixel center coordinates of the marked point within the calculation window; performing error calculation based on the sub-pixel center coordinates, and performing compensation verification in conjunction with a preset alignment tolerance to obtain an alignment error detection result.

[0007] By constructing a detection and correction link from interferometric frequency adaptive notch filtering, gradient field purification sub-pixel positioning, multi-point least squares error calculation to compensation verification, the source of contamination of the gradient field of subsequent marker points by interference fringes is effectively eliminated, laying the frequency domain foundation for the recovery of sub-pixel positioning accuracy.

[0008] Preferably, obtaining the local film thickness gradient along the substrate transport direction includes: obtaining a continuously collected measured film thickness sequence and the corresponding sampling interval along the substrate transport direction; and calculating the local film thickness gradient based on the difference between the measured film thickness sequences between adjacent sampling points and the sampling interval.

[0009] Preferably, determining the center frequency of the interference fringes and the notch bandwidth in the image frequency domain based on the local film thickness gradient includes: obtaining the film refractive index of the coating material, the center wavelength of the light source, the pixel equivalent of the camera, and the measured tolerance of the film thickness sensor; calculating the center frequency based on the film refractive index, the local film thickness gradient, the pixel equivalent, and the center wavelength of the light source; and calculating the notch bandwidth based on the film refractive index, the pixel equivalent, the measured tolerance, the center wavelength of the light source, and the sampling interval.

[0010] By incorporating the gradient measured by the online film thickness sensor, device tolerance, and camera pixel into the derivation process of the notch filter center frequency and bandwidth in the frequency domain, the notch filter center frequency and bandwidth are locked to physically measurable objective parameters, thereby improving the stability and consistency of the separation of interference fringes and marker point edge frequencies under different batches of film materials.

[0011] Preferably, the step of filtering the spectrum of the grayscale image of the coated substrate based on the center frequency and the notch bandwidth to obtain a cleaned image includes: performing a two-dimensional discrete Fourier transform on the grayscale image of the coated substrate to obtain a two-dimensional spectrum; constructing an adaptive notch filter with the center frequency as the notch center and the notch bandwidth as the radius; filtering the two-dimensional spectrum using the adaptive notch filter, and performing a two-dimensional inverse discrete Fourier transform on the filtered spectrum to obtain the cleaned image.

[0012] Preferably, extracting the gradient magnitude of the purified image within the region of interest of the marked point includes: calculating a first gradient component in the horizontal direction and a second gradient component in the vertical direction within the region of interest of the marked point; and synthesizing the gradient magnitude based on the first gradient component and the second gradient component.

[0013] Preferably, determining the calculation window based on the statistical characteristics of the gradient magnitude includes: calculating the mean of the gradient magnitude within the region of interest of the marked point, and using the mean as a gradient decay discrimination threshold; expanding outward from the coarse positioning center in concentric circles according to pixel radius, and calculating the ring mean of the gradient magnitude within the corresponding ring; stopping the expansion when the ring mean first decays to less than or equal to the gradient decay discrimination threshold, and using the corresponding pixel radius as the window radius to determine the calculation window.

[0014] By adaptively determining the subpixel centroid calculation window using the mean and peak features of gradient magnitudes within the region of interest, the coverage mismatch of a fixed empirical magnification window under different sized marker points or different contrast scenes is avoided, thus improving the robustness of subpixel localization.

[0015] Preferably, the step of calculating the sub-pixel center coordinates of the marker point within the calculation window includes: within the calculation window, using the gradient magnitude of each pixel as a weight, performing a weighted average of the pixel coordinates to obtain initial sub-pixel coordinates; and substituting the initial sub-pixel coordinates into a pre-established distortion compensation lookup table for mapping to obtain the sub-pixel center coordinates.

[0016] Preferably, the error calculation based on the sub-pixel center coordinates includes: removing outliers from the markers corresponding to the sub-pixel center coordinates based on the median of the residual absolute deviation; and performing least-squares error calculation on the sub-pixel center coordinates after removing the outliers.

[0017] For the outlier removal process, a robust statistical criterion based on the median of the absolute deviation of the residuals is adopted to avoid the influence of the mean and standard deviation on extreme outliers. This ensures that the least squares error calculation can still output reliable results in scenarios with local contamination of the membrane surface, bubbles, or foreign objects blocking the membrane.

[0018] Preferably, the step of combining the preset alignment tolerance for compensation verification to obtain the alignment error detection result includes: dynamically reading the alignment tolerance in the process file; using the alignment tolerance as the convergence criterion for compensation iteration for compensation verification; and outputting the alignment error detection result.

[0019] In a second aspect, this application provides a coating alignment error detection system, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-described coating alignment error detection method is implemented.

[0020] By adopting the above technical solution, a computer program for detecting the lamination alignment error is generated and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0021] The technical solution provided in this application can adaptively separate interference frequency components and restore the true distribution of the gradient field at the edge of the marker point in a complex optical interference environment. This effectively overcomes the limitations of traditional spatial image enhancement methods in processing periodic interference fringes and ensures the purity of the underlying image features.

[0022] Furthermore, by combining dynamic window calculation based on local statistical features with a robust outlier removal algorithm, this application demonstrates strong anti-interference capabilities in both sub-pixel positioning and multi-point error calculation stages. This not only improves detection accuracy but also ensures reliability during cross-product specification switching by dynamically reading process tolerances, meeting the stringent requirements of modern high-precision coating production lines. Attached Figure Description

[0023] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a schematic flowchart illustrating a coating alignment error detection method and system according to the present invention.

[0024] Figure 2 This is a comparison chart of the positioning accuracy of marker points based on sub-pixel positioning scatter distribution in an embodiment of the present invention and marker points in existing methods. Detailed Implementation

[0025] 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, not all, of the embodiments of the present invention. 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.

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention discloses a method for detecting film alignment error, referring to... Figure 1 This includes steps S1-S4: S1. Obtain the notch frequency of interference fringes based on film thickness gradient adaptive acquisition.

[0028] In an optional embodiment, the system reads the nominal film thickness of the current batch of coating material from the production line process parameter library. Thin film refractive index and the center wavelength of the light source And the film thickness is measured online by a sensor along the substrate conveying direction to sample the spacing. Continuous acquisition of measured film thickness sequences This allows us to obtain the local film thickness gradient between adjacent sampling points:

[0029] In this optional embodiment, there are many other sensors in the art that can achieve online film thickness measurement, including white light interferometers and laser confocal sensors. Those skilled in the art can choose different measurement methods according to the actual production line conditions. The above gradient calculation formula is applicable to the film thickness sequence output by various types of sensors.

[0030] According to the thin film interference optical path difference formula Under paraxial imaging conditions of the camera, take The film thickness difference corresponding to adjacent interference fringes is Since the film thickness gradually changes linearly along space, the spatial period of the uniform thickness stripes on the object plane can be obtained. Further combining the pixel equivalent output during the camera calibration phase... That is, the physical length of the object plane corresponding to each pixel, from which the center frequency of the interference fringes in the image frequency domain is derived, in units of cycles per pixel:

[0031] The above formula directly combines the phase relationship of thin film interference with the spatial film thickness gradient and pixel equivalent. All terms are objective measurement outputs of the sensor and camera, thus transforming the original optical path difference formula into a frequency expression for locating the notch center.

[0032] Next, a two-dimensional discrete Fourier transform is performed on the grayscale image of the coated substrate acquired by the industrial camera to obtain the two-dimensional spectrum. In this spectrum, the interference fringes appear as... The system is centered on a periodic cluster of energy peaks, the direction of which is perpendicular to the direction of the interference fringes. Centered at the notch, with the notch bandwidth at half width. Construct an adaptive notch filter for the radius Due to the inherent measurement tolerances of the film thickness sensor. This tolerance is given directly in the sensor specification sheet, and is determined by... about The linear transfer derivation yields the notch bandwidth half-width:

[0033] because The notch bandwidth is objectively determined by the sensor batch and its operating status, and is updated in real time according to the current measured tolerance of the sensor.

[0034] Furthermore, in the notch bandwidth The spectral energy is set to zero inside the bandwidth, while the original spectrum remains unchanged outside the bandwidth, resulting in the filtered spectrum. Perform a two-dimensional inverse discrete Fourier transform on the filtered spectrum to restore the image to the spatial domain, resulting in a cleaned image with suppressed interference fringe components. .

[0035] Thus, by jointly deriving the image domain center frequency of the interference fringes using film thickness gradient, refractive index, light source wavelength, and pixel equivalent, and determining the notch bandwidth using sensor tolerance, the frequency components of the interference fringes under different batches of film materials and different film thickness distributions were located and suppressed. This eliminated the source of contamination of the gradient field of subsequent marker points by the interference fringes, laying the frequency domain foundation for the recovery of sub-pixel positioning accuracy.

[0036] S2, Subpixel marker positioning after gradient field purification.

[0037] In an optional embodiment, the purified image can be... Calculate the horizontal and vertical gradient components within the region of interest of the marked point. and Synthesized gradient magnitude map:

[0038] Since the interference fringe components have been separated in the aforementioned notch sieving stage, the main energy in the gradient field is now concentrated at the true edge of the marked point, and the symmetry of the gradient magnitude distribution is restored, providing an unbiased weighting basis for subsequent weighted centroid calculation.

[0039] Next, the calculation window size for the center coordinates of the sub-pixel markers is determined using a derivation method based on the statistical characteristics of the gradient magnitude in the region of interest. First, the peak value of the gradient magnitude is calculated within the region of interest. and mean The gradient decay threshold is obtained from the statistical characteristics of the region of interest itself. Then, based on the coarse positioning center, proceed outwards according to pixel radius. Expanding in circles, for a radius of The average gradient magnitude within the ring band is taken. ,when First decay to less than or equal to Stop expanding at that time, and As the window radius This decision locks the window boundaries at positions where the edge gradient response has decayed to a level comparable to the background average, using a specific threshold. It is given by the average gradient magnitude of the current region of interest.

[0040] Furthermore, in a radius of Within the window, the gradient magnitude of each pixel is used. Using the pixel coordinates as weights, a weighted average is performed to obtain the center coordinates of the marker point with sub-pixel precision. :

[0041] To eliminate inherent distortions in the optical system and systematic positioning deviations introduced by camera mounting eccentricity, the system performs an offline calibration process during the production line initialization phase: acquiring images of the standard calibration board in an uncoated state, extracting the mapping relationship between the real physical coordinates and image coordinates of each marker point, and establishing a distortion compensation lookup table. During online detection, the sub-pixel coordinates of each marker point are... Substitute into the above distortion compensation lookup table Complete distortion compensation and output the corrected coordinates after compensation. This serves as the input for subsequent error calculation. It is worth noting that there are many other ways to achieve distortion compensation in this field. Those skilled in the art can choose different methods such as polynomial distortion models or lookup table interpolation based on actual calibration conditions. The above offline calibration process is applicable to all types of compensation methods.

[0042] Thus, by using the mean gradient magnitude of the region of interest as the stopping criterion for window expansion, the centroid calculation window converges with the actual edge response range of the marker point, further eliminating the impact of window coverage mismatch on sub-pixel positioning accuracy on the basis of purifying the gradient field.

[0043] S3. Solve the alignment error of three degrees of freedom by solving multiple marker points simultaneously.

[0044] In an optional embodiment, the coating alignment error typically comprises three independent components: translational error along the X-axis. Translation error along the Y-axis and rotational error around the center of contact The system arranges at least three non-collinear marker points on the substrate, and corrects the coordinates using the calculated marker points. Its corresponding design reference coordinates Construct a system of residual equations and solve for the three error components simultaneously using the least squares method:

[0045] in Let be a rotation matrix. The total number of markers participating in the solution. For the marker point number; when When the value is greater than 3, the system of equations is overdetermined, and the least squares solution further reduces the impact of random errors in the positioning of each marker point on the solution results in a statistical sense.

[0046] After the solution is completed, the system calculates the positioning residual for each marker point. This refers to the remaining deviation between the measured coordinates of the marker point and the reference coordinates after error compensation. To objectively identify outlier marker points, a robust statistical criterion based on the median of the residual absolute deviation can be used:

[0047] Based on existing conclusions in robust statistics, a robust estimate of the standard deviation of residuals under a normal distribution is: The system takes the discrimination threshold. This corresponds to a 95% confidence interval. All conditions that meet this requirement... The marked points are identified as outliers affected by local contamination and removed from the solution set. The least squares solution is then re-executed using the remaining valid marked points. and All of these are existing mathematical constants for robust estimates of the normal distribution and 95% confidence quantiles.

[0048] In this way, by solving the least squares problem using the corrected coordinates of at least three non-collinear marker points, and then removing abnormal marker points affected by local contamination using the median of the residual absolute deviation criterion, the solution results of the three components of translation error and rotation error still have statistical robustness in the case of local contamination of the membrane surface, effectively avoiding the distortion caused by single-point anomalies to the overall error solution.

[0049] S4. Perform error compensation and verify convergence.

[0050] In an optional embodiment, the already calculated error components are... The motion controller of the input fitting mechanism converts the error in the image coordinate system into the compensating motion of the mechanical actuation axis according to the kinematic model of the mechanism. Translation error and Compensation displacement directly mapped to the XY platform, rotation error The compensation angle mapped to the rotation axis, and the conversion coefficients of the compensation amounts for each axis are determined by the camera pixel calibration values, ensuring an accurate correspondence between image coordinates and physical motion quantities.

[0051] In response to the compensation action performed by the mechanism, the system triggers the camera to re-acquire the current bonding state image, repeating the aforementioned detection process of interference notch filtering, gradient field purification positioning, and multi-point error calculation, and calculating the corrected residual error. When the absolute value of the residual error is lower than the alignment tolerance threshold required by the product specification, the bonding is deemed qualified, the system outputs a qualification signal, and records the error data of this detection to the quality traceability database. The alignment tolerance threshold is specified by the product process document, and the system dynamically reads it from the process parameter library, thereby ensuring the real-time consistency between the detection criteria and the product specifications when switching between product specifications.

[0052] In cases where a single compensation is insufficient, if the residual error still exceeds the tolerance threshold, the system uses this residual error as the input error for a new round, driving the mechanism to perform compensation again until the error converges within the tolerance range or reaches the preset maximum number of iterations. The maximum number of iterations is configured in the process parameter library and determined based on the production line cycle time and the mechanism response time; for example, the maximum number of iterations can be set to 3. When the maximum number of iterations is exceeded, the system triggers a re-inspection alarm to prevent the system from falling into an invalid loop due to abnormal operating conditions, thus ensuring the safe operation of the production line.

[0053] Furthermore, after each iteration of compensation, the system simultaneously writes the residual error, compensation amount, and iteration number into the quality traceability database. Through statistical analysis of historical iteration data, it identifies situations where the compensation convergence speed is abnormally slow in a specific batch of film material or within a specific time period, providing objective data support for the periodic optimization of production line process parameters.

[0054] Figure 2 This is a comparison of the positioning accuracy of marker points based on sub-pixel positioning scatter distribution in an embodiment of the present invention with that of marker points in existing methods. It can be seen that the positioning scatter points of the existing methods exhibit obvious asymmetric diffusion distribution around the reference circle, with a large radial dispersion deviating from the reference circle, reflecting that the interference fringes contaminate the gradient weight field, causing a systematic shift in the centroid calculation. In contrast, the positioning scatter points of the method in this application are closely clustered near the reference circle, with a significantly narrowed radial dispersion, indicating that the gradient field symmetry is restored after adaptive notch filtering, and the sub-pixel positioning accuracy is effectively improved.

[0055] In this way, by converting the calculated three-degree-of-freedom error components into compensation values ​​for each axis through a kinematic model, the mechanism is driven to execute. The alignment tolerance dynamically read from the process document is used as the convergence criterion for iterative verification, thus realizing a complete loop from error detection to physical correction. When a single compensation is insufficient, it converges robustly in a progressive iterative manner. At the same time, the maximum number of iterations is set to prevent invalid loops under abnormal working conditions, ensuring the engineering reliability of compensation execution in the high-precision coating production line.

[0056] This invention also discloses a coating alignment error detection system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a coating alignment error detection method according to the present invention.

[0057] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for detecting lamination alignment error, characterized in that, include: Acquire grayscale images of the coated substrate and local film thickness gradients along the substrate transport direction; The method for obtaining the local film thickness gradient along the substrate transport direction includes: obtaining a continuously acquired measured film thickness sequence along the substrate transport direction and the corresponding sampling interval; and calculating the local film thickness gradient based on the difference between the measured film thickness sequences of adjacent sampling points and the sampling interval. Determining the center frequency and notch bandwidth of interference fringes in the image frequency domain based on the local film thickness gradient includes: obtaining the film refractive index of the coating material, the center wavelength of the light source, the pixel equivalent of the camera, and the measured tolerance of the film thickness sensor; calculating the center frequency based on the film refractive index, the local film thickness gradient, the pixel equivalent, and the center wavelength of the light source; and calculating the notch bandwidth based on the film refractive index, the pixel equivalent, the measured tolerance, the center wavelength of the light source, and the sampling interval. The spectrum of the grayscale image of the coated substrate is filtered according to the center frequency and the notch bandwidth to obtain a cleaned image; Extract the gradient magnitude of the cleaned image within the region of interest of the marked point, and determine the calculation window based on the statistical characteristics of the gradient magnitude. Calculate the sub-pixel center coordinates of the marked point within the calculation window. Error calculation is performed based on the sub-pixel center coordinates, and compensation verification is performed in conjunction with the preset alignment tolerance to obtain the alignment error detection result.

2. The method according to claim 1, characterized in that, The step of filtering the spectrum of the grayscale image of the coated substrate based on the center frequency and the notch bandwidth to obtain a cleaned image includes: Perform a two-dimensional discrete Fourier transform on the grayscale image of the coated substrate to obtain a two-dimensional spectrum; An adaptive notch filter is constructed with the center frequency as the notch center and the notch bandwidth as the radius. The two-dimensional spectrum is filtered using the adaptive notch filter, and the filtered spectrum is subjected to a two-dimensional discrete Fourier inverse transform to obtain the purified image.

3. The method according to claim 1, characterized in that, Extracting the gradient magnitude of the cleaned image within the region of interest of the marked points includes: Calculate the first gradient component in the horizontal direction and the second gradient component in the vertical direction within the region of interest of the marked point; The gradient magnitude is obtained by synthesizing the first gradient component and the second gradient component.

4. The method according to claim 1, characterized in that, The determination of the calculation window based on the statistical characteristics of the gradient magnitude includes: Calculate the mean value of the gradient magnitude within the region of interest of the marked point, and use the mean value as the gradient decay discrimination threshold; Expand outward from the coarse positioning center in concentric rings according to the pixel radius, and calculate the ring average value of the gradient magnitude within the corresponding ring. When the mean value of the annular band first decays to less than or equal to the gradient decay discrimination threshold, the expansion stops, and the corresponding pixel radius is used as the window radius to determine the calculation window.

5. The method according to claim 1, characterized in that, The calculation of the sub-pixel center coordinates of the marker point within the calculation window includes: Within the calculation window, the pixel coordinates are weighted and averaged using the gradient magnitude of each pixel as the weight to obtain the initial sub-pixel coordinates. The initial subpixel coordinates are substituted into a pre-established distortion compensation lookup table for mapping to obtain the subpixel center coordinates.

6. The method according to claim 1, characterized in that, The error calculation based on the sub-pixel center coordinates includes: Outlier points are removed from the marker points corresponding to the sub-pixel center coordinates based on the median of the residual absolute deviation. The least squares error is calculated for the sub-pixel center coordinates after removing the abnormal markers.

7. The method according to claim 1, characterized in that, The process of compensating and verifying by combining preset alignment tolerances to obtain alignment error detection results includes: Dynamically read the alignment tolerances mentioned in the process document; The alignment tolerance is used as the convergence criterion for compensation iteration to perform compensation verification, and the alignment error detection result is output.

8. A film alignment error detection system, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute a computer program stored in the memory, implements the coating alignment error detection method according to any one of claims 1 to 7.

Images