Visual detection method and system for coating quality of spin-coated optical filter

Abstract

The invention relates to the technical field of image processing, in particular to a coating quality visual detection method and system for a spin-coated optical filter. The method comprises the following steps: acquiring gray level images of a plurality of transmission images of an optical filter after light source transmission under at least three incidence angles in a plurality of orientations; taking a difference image between each gray-scale image and the background gray-scale image as a residual image; performing binarization processing on the residual image to obtain a corresponding binary image; performing connected domain analysis on each binary image to obtain a suspected defect region of each binary image; carrying out trend test on each suspected defect area to obtain trend direction statistics of corresponding directions; in response to the fact that the trend statistics in all the directions are consistent, determining that the coating film uniformity is abnormal in the suspected area; on the contrary, the suspected defect area is interfered by inconsistent tensity. According to the scheme, the uniformity of the coating layer on the optical filter can be accurately detected so as to evaluate whether the coating quality of the surface of the optical filter is qualified or not.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method and system for visual inspection of coating quality for spin-coated filters. Background Technology

[0002] As a key component in optical systems, the uniformity of the coating layer on spin-coated filters directly affects optical performance and image quality. Therefore, accurate and efficient automated testing of filter coating quality is a core aspect of ensuring product quality.

[0003] Currently, machine vision-based inspection methods are widely used due to their high efficiency and non-contact advantages. However, in the actual production and inspection of filters, the physical causes of suspected defective areas detected in images are quite complex. These defects include not only uneven coating thickness directly caused by coating process issues, but also minute surface undulations (commonly known as "orange peel") caused by inconsistent tension of the filter substrate during fixing or transport.

[0004] Traditional visual inspection methods typically extract abnormal regions through image differencing, threshold segmentation, and other methods. However, these methods are difficult to effectively distinguish between the two types of defects with completely different causes. They often misjudge false defects caused by inconsistent tension as coating quality problems, resulting in a high false detection rate and affecting production efficiency.

[0005] Therefore, accurately distinguishing between real defects and pseudo-defects to improve the accuracy of quality inspection of spin-coated filter coatings is of paramount importance. Summary of the Invention

[0006] The purpose of this invention is to provide a visual inspection method and system for the coating quality of spin-coated filters, in order to solve the problem that the non-uniformity of the coating layer on the surface of the filter cannot be accurately detected in the prior art; to this end, the present invention provides solutions in the following two aspects.

[0007] In the first aspect, a visual inspection method for coating quality of spin-coated filters includes... Acquire grayscale images of multiple transmitted images of the light source after transmission through the filter at at least three incident angles in multiple orientations; The difference between each grayscale image and its corresponding standard normal image is used as the residual image; The residual map is binarized to obtain the corresponding binary map; If the average gray value of pixels at the same location in all binary images under at least three orientations is greater than or equal to the first threshold, then the corresponding pixel is marked as an outlier. Obtain the gray value sequence of all grayscale images of the suspected anomaly point in each orientation, perform a trend test on each gray value sequence, and obtain the trend significance of the corresponding gray value sequence; if the trend significance is less than the trend threshold in at least two orientations, the suspected anomaly point belongs to the coating uniformity anomaly; otherwise, the coating uniformity of the suspected anomaly point is disturbed by the tension inconsistency.

[0008] Optionally, the binarization of the residual image includes: setting pixels with gray values ​​greater than an adaptive threshold to 1, and pixels with gray values ​​less than or equal to the adaptive threshold to 0; the adaptive threshold is: ; Where T is the adaptive threshold, k is the coefficient, and MAD is the median absolute deviation. The median is the grayscale value in the residual image; the absolute deviation of the median is the median of the absolute values ​​of the differences between the grayscale value of a pixel and the median in the residual image.

[0009] The above scheme uses an adaptive threshold for image binarization, which can more accurately segment the image.

[0010] Optionally, the trend significance is the p-trend value of each gray value sequence obtained by using the Mann-Kendall trend analysis test.

[0011] The above scheme can determine the trend of grayscale value sequence by detecting the significance of the trend.

[0012] Optionally, the plurality of incident angles include 30°, 45°, and 60°.

[0013] By setting multiple oblique incidence angles, the sensitivity of thickness difference to image grayscale response can be enhanced, making minute coating unevenness appear as more obvious grayscale changes in the image, thereby significantly improving detection contrast and accuracy.

[0014] Optionally, the plurality of orientations include two directions, left and right, along the length of the filter, and two directions, front and back, perpendicular to the length of the filter.

[0015] By setting different orientations, the uniformity of the filter coating can be analyzed more accurately and comprehensively, avoiding the errors of measurement under a single orientation.

[0016] Optionally, it also includes: if the proportion of abnormal points with abnormal coating uniformity is greater than a set proportion, then the filter is determined to have abnormal coating uniformity and the filter is unqualified.

[0017] Optionally, after acquiring the transmission image, image denoising and image enhancement processing are also performed on the transmission image.

[0018] Optionally, the light source is determined based on the wavelength of the light transmitted through the filter.

[0019] In the second aspect, a visual inspection system for coating quality of spin-coated filters includes: processor; The memory stores computer instructions for visual inspection of the coating quality of spin-coated filters. When the computer instructions are executed by the processor, the system performs the aforementioned visual inspection method for the coating quality of spin-coated filters.

[0020] The beneficial effects of this invention are as follows: The present invention analyzes the grayscale images of the transmission images of filters at different incident angles under multiple orientations. By comprehensively analyzing the trend significance of the residual images between the reference image and the standard image at multiple incident angles under different orientations, it can determine whether the corresponding pixel is a real anomaly. This allows for accurate determination of the uniformity of the coating layer on the filter surface, improving detection sensitivity and reducing the problem of missed detections. Attached Figure Description

[0021] Figure 1 This schematically illustrates the steps of the visual inspection method for coating quality of spin-coated filters in this embodiment; Figure 2 This schematic diagram illustrates the optical path from the light source to the filter. Figure 3 This schematic diagram illustrates the structural block diagram of the coating quality visual inspection system for spin-coated filters in this embodiment; Reference numerals: 1. Coating layer; 2. Background plate. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] This invention addresses the issue of coating uniformity on spin-coated filters, and provides a method for detecting whether the coating layer is uniform.

[0024] Specifically, such as Figure 1 As shown, the visual inspection method for coating quality of spin-coated filters in this embodiment includes the following steps: Step S1: Obtain grayscale images of multiple transmitted images of the light source after transmission at at least three incident angles in multiple orientations.

[0025] In this embodiment, it is considered that uneven coating of the filter will lead to local thickness differences in the filter, specifically manifested as normal areas, excessively thin areas, and excessively thick areas. Since the coating layer is usually a weakly absorbing dielectric, when illuminated by a light source, its transmitted light intensity decreases exponentially with changes in thickness, which can be approximated by the Lambert-Beer law. The Lambert-Beer law is prior art and will not be elaborated here.

[0026] Different incident angles can lengthen the propagation path of the light source within the coating layer on the filter surface, such as... Figure 2 As shown, light rays are incident on point O in the coated area 1 of the filter along the incident angle θ1, forming a propagation path at point O and forming a transmission image on the background plate 2.

[0027] In this embodiment, by setting different incident angles, the optical path difference and transmission angle difference of different thickness areas are amplified, which significantly enhances the transmittance difference between thick and thin areas, so that the small thickness changes present a clearer grayscale contrast in the transmission image, and improves the detectability of coating unevenness and the stability of image response.

[0028] Therefore, in this embodiment, the spin-coated filter to be tested is placed on a sample stage whose rotation orientation and incident angle can be precisely controlled. Light emitted from the light source passes through the filter at a specific incident angle, and the transmitted image is captured by an image acquisition device (e.g., a high-resolution industrial CCD or CMOS camera). Each transmitted image is then converted to grayscale to obtain a grayscale image; where the grayscale value at each pixel represents the transmission intensity at that location.

[0029] It should be noted that under perpendicular incident light conditions, the propagation path of light in the coating layer is relatively short, and slight thickness differences often fail to produce obvious grayscale changes in the image, reducing detection sensitivity. Therefore, in this embodiment, an oblique incident illumination method is adopted, so that the light shines obliquely on the filter surface from one side in a non-perpendicular direction.

[0030] In this embodiment, the incident angle is preferably set within the range of 30° to 60°, balancing image enhancement and system feasibility. Specifically, the incident angle includes at least 30°, 45°, and 60°. By changing the incident angle, changes in optical path difference caused by uneven coating thickness or minor surface deformation can be more sensitively excited and observed, thereby enhancing the prominence of defect features.

[0031] The aforementioned multiple orientations include two directions along the length of the filter (left and right) and two directions perpendicular to the length of the filter (front and back). This involves placing the light source at different orientations to illuminate the filter, thereby determining the consistency of pixels at the same location under each orientation and avoiding the problem of single-orientation errors.

[0032] Of course, as other implementation methods, the above-mentioned orientation can also be selected according to the actual situation.

[0033] The light source is determined based on the wavelength of the light transmitted through the filter.

[0034] For example, an ultraviolet filter transmits ultraviolet light while blocking infrared and visible light; therefore, the light source can be ultraviolet light, whose wavelength range can effectively penetrate the filter and interact with the coating layer to facilitate subsequent defect analysis.

[0035] After acquiring the original transmission images, preprocessing can be performed on each transmission image to improve the signal-to-noise ratio and enhance image details. Preprocessing steps can include image denoising and image enhancement. Specifically, algorithms such as Gaussian filtering or median filtering can be used to remove random noise from the image; subsequently, methods such as histogram equalization can be used to enhance the image and improve its overall contrast. After preprocessing, grayscale images are obtained from various positions and angles.

[0036] It should be noted that the thicker the coating layer on the filter surface or the larger the incident angle, the more significant the attenuation of transmitted light intensity (oblique incidence amplifies the effect of thickness difference in the optical path (increasing the path difference), making the response curve of the exponential term to the path steeper); conversely, the thinner the coating layer on the filter surface, the higher the gray level appears in the transmitted image on the white background. In particular, when using oblique incidence, the same thickness difference Δd will produce a larger gray level difference ΔI due to the amplified optical path difference ΔL. Therefore, analyzing the same pixel under different incident angles can enhance the distinguishability of abnormal areas in the image. However, due to the nonlinear characteristics of exponential decay, the increase in gray level difference ΔI is not linear or infinitely amplified; it has an upper limit and is controlled by other parameters (such as the coefficient in the exponential decay law and the incident light intensity).

[0037] Step S2: Perform preliminary identification of suspected defect areas in the grayscale image under each orientation.

[0038] The process for obtaining suspected defective areas is as follows: Step S21: The difference map between each grayscale image and the standard background grayscale image is used as a residual map; the residual map is binarized to obtain the corresponding binary map.

[0039] As a preferred embodiment, the standard background grayscale image can be obtained through one of the following two methods: First, a defect-free standard filter is used to acquire its transmission image under the exact same acquisition conditions as the filter under test (including orientation, incident angle, illumination, etc.), and this image is used as the standard background grayscale image.

[0040] Secondly, without placing any filters, or in a confirmed defect-free area of ​​a filter, multiple images are acquired, and then these images are averaged at the pixel level to eliminate random noise and generate a smooth background image.

[0041] The pixel values ​​in the residual plot above reflect the difference in optical transmission between the filter under test and the ideal reference. Potential defective areas usually appear as patches of abnormal brightness on the residual plot.

[0042] In this embodiment, the residual image is binarized to obtain a corresponding binary image. Specifically, the process is as follows: pixels with gray values ​​greater than the adaptive threshold in the residual image are set to 1, and pixels with gray values ​​less than or equal to the adaptive threshold are set to 0.

[0043] The adaptive threshold is: ; Where T is the adaptive threshold, k is the coefficient, and MAD is the median absolute deviation. The median is the grayscale value in the grayscale image; the median absolute deviation is the median of the absolute values ​​of the differences between the grayscale value of a pixel and the median in the grayscale image.

[0044] According to the above formula, when the coefficient... When the threshold increases, the adaptive threshold Correspondingly, increasing the coefficient reduces the detection sensitivity, suppressing more noise but potentially missing weak defects; conversely, decreasing the coefficient... When decreasing, adaptive threshold The sensitivity of detection is improved by reducing the risk of infection.

[0045] As a preferred option, the coefficient The range of values ​​can be In a preferred embodiment, k can be 3.

[0046] The binarization process described above is to achieve high robustness in detecting abnormal regions in the residual image, enabling accurate detection of abnormal pixels and providing more stable segmentation performance under actual working conditions.

[0047] Step S22: Perform connected component analysis on each binary graph to obtain the suspected defect regions of each binary graph.

[0048] Specifically, connectivity analysis is performed on each binary graph. This analysis allows spatially adjacent foreground points to be aggregated into independent regions, which are then identified as potential defect areas.

[0049] Step S3: Perform trend analysis on each suspected defect area to obtain the trend direction statistics for the corresponding orientation. If the trend direction statistics are consistent in all orientations, the coating layer in the suspected area has an abnormality in uniformity. Otherwise, the suspected defect area is affected by inconsistent tension.

[0050] The presence of non-uniformity in the coating layer is an inherent physical property of filters, and its grayscale variation trend in the image (e.g., a gradual transition from bright to dark) does not change with the viewing orientation. Conversely, physical deformations of the filter surface caused by inconsistent tension (such as micro-wrinkles or warping) and the resulting artifacts of brightness and darkness after interacting with incident light will show a significant change in the apparent grayscale variation trend with rotation of the viewing orientation.

[0051] Therefore, in this embodiment, a trend test is performed on the suspected defective areas. The specific process is as follows: For each suspected defect area detected in each direction, a trend test is performed to obtain a statistical quantity that quantifies the gray-scale change trend of the area, namely the trend direction statistical quantity.

[0052] In this embodiment, the trend direction statistic is quantified using Spearman's rank correlation coefficient.

[0053] Specifically, for all pixels within a suspected defect area, their coordinates can be extracted. and the corresponding grayscale value Calculate separately Coordinate sequence and gray value Spearman rank correlation coefficient of the sequence ,as well as Coordinate sequence and gray value Spearman rank correlation coefficient of the sequence .vector It can be used as a statistical measure of trend direction, with its direction roughly pointing in the direction of increasing gray value in the region, and its magnitude reflecting the significance of the trend.

[0054] After obtaining the trend direction statistics of suspected defect areas at the same location under all different orientations, a consistency judgment is performed. The judgment logic is as follows: If the trend direction statistics are consistent across all orientations, the suspected defect area is determined to have abnormal coating uniformity and unqualified coating quality. Conversely, if the trend direction statistics are inconsistent across different orientations, the suspected defect area is determined to be an artifact caused by inconsistent tension, rather than a real coating defect, and further verification is required.

[0055] The above-mentioned consistent judgment can be that the corresponding vectors have basically the same direction and the angle between them is less than a preset angle threshold.

[0056] This step effectively filters out misjudgments caused by installation or material stress, significantly improving the accuracy of coating quality inspection.

[0057] If all suspected defective areas are determined to be tension interference, the coating quality of the filter is initially deemed acceptable, but further verification is required.

[0058] The present invention employs a multi-directional, multi-angle image acquisition strategy, combined with adaptive image segmentation based on statistical characteristics and innovative trend analysis, to accurately identify coating uniformity defects in spin-coated filters and effectively eliminate interference from physical factors such as inconsistent tension, ultimately achieving a reliable assessment of filter quality.

[0059] This invention also provides a visual inspection system for the coating quality of spin-coated filters. For example... Figure 3 As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the above-described visual inspection method for coating quality of spin-coated filters according to the present invention.

[0060] 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 therefore will not be described in detail here.

[0061] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0062] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0063] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A visual inspection method for the coating quality of spin-coated filters, characterized in that, include: Acquire grayscale images of multiple transmitted images of the light source after transmission through the filter at at least three incident angles in multiple orientations; The difference between each grayscale image and the background grayscale image is used as a residual image; the residual image is binarized to obtain the corresponding binary image. Connectivity analysis was performed on each binary graph to identify potential defect regions. A trend test was performed on each suspected defect area to obtain the trend direction statistics for the corresponding location; If the trend statistics are consistent across all directions, then the suspected area has an abnormality in coating uniformity. Conversely, the suspected defective area is affected by inconsistent tension.

2. The visual inspection method for coating quality of spin-coated filters according to claim 1, characterized in that, The binarization process of the residual image includes: setting pixels with gray values ​​greater than an adaptive threshold to 1, and setting pixels with gray values ​​less than or equal to the adaptive threshold to 0; the adaptive threshold is: ； Where T is the adaptive threshold, k is the coefficient, and MAD is the median absolute deviation. The median is the grayscale value in the residual image; the absolute deviation of the median is the median of the absolute values ​​of the differences between the grayscale value of a pixel and the median in the residual image.

3. The visual inspection method for coating quality of spin-coated filters according to claim 2, characterized in that, The trend direction statistics are quantified using the Spearman rank correlation coefficient.

4. The visual inspection method for coating quality of spin-coated filters according to claim 1, characterized in that, The background grayscale image is an image acquired under the same conditions using a defect-free standard filter, or a background image generated by averaging multiple acquisitions.

5. The visual inspection method for coating quality of spin-coated filters according to claim 2, characterized in that, The multiple incident angles include 30°, 45°, and 60°.

6. The visual inspection method for coating quality of spin-coated filters according to claim 2, characterized in that, The multiple orientations include two directions along the length of the filter: left and right, and two directions perpendicular to the length of the filter: front and back.

7. The visual inspection method for coating quality of spin-coated filters according to claim 2, characterized in that, Also includes: If the ratio of the area of ​​the defect region corresponding to the uniformity abnormality to the area of ​​the grayscale image is greater than a set value, then the coating quality of the filter is unqualified.

8. The visual inspection method for coating quality of spin-coated filters according to claim 1, characterized in that, After acquiring the transmission image, image denoising and image enhancement processing are performed on the transmission image.

9. The visual inspection method for coating quality of spin-coated filters according to claim 1, characterized in that, The light source is determined based on the wavelength of the light transmitted through the filter.

10. A visual inspection system for the coating quality of spin-coated filters, characterized in that, include: processor; A memory storing computer instructions for visual inspection of the coating quality of a spin-coated filter, wherein when the computer instructions are executed by the processor, the system performs the visual inspection method for the coating quality of a spin-coated filter according to any one of claims 1-9.

Images