A high-precision detection method for microscopic quality of optical films

Abstract

The present invention relates to the field of image processing technology, and in particular to a high-precision detection method for the microscopic quality of an optical film, comprising: obtaining a microscopic grayscale image of the optical film; analyzing the periodic characteristics of the grayscale value according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image, and obtaining the size of the sliding window; obtaining a structural correlation factor according to the grayscale difference between adjacent sliding windows of different sliding lengths in different directions of each sliding window in the microscopic grayscale image; obtaining a microscopic feature evaluation of each sliding window according to the distribution characteristics of the pixels belonging to the target in each adjacent sliding window corresponding to each sliding window and the corresponding sliding length, combined with the structural correlation factor; analyzing the defect feature degree of the optical film according to the microscopic feature evaluation of each sliding window in the microscopic grayscale image, and obtaining a quality evaluation result. The present invention can obtain more accurate microscopic quality detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a high-precision detection method for the microscopic quality of an optical film. Background Art

[0002] Optical thin film technology is one of the core technologies of modern optics, and plays an increasingly important role in cutting-edge scientific fields such as information, biology, and aerospace. Therefore, quality inspection of optical thin films is particularly important. The microstructure of optical thin films directly determines their optical properties, including transmittance, reflectivity, and dispersion characteristics. Defects in the film may lead to a decline in optical performance and affect the overall efficiency and accuracy of the equipment or system. Therefore, microscopic quality inspection of optical thin films is of great significance. Through precise inspection and analysis, the reliability and durability of optical thin films can be guaranteed.

[0003] At present, the microscopic quality of optical films is detected by machine vision methods. The microscopic quality problems of optical films are mainly manifested in partial defects in the microscopic structure of the film. Considering that the microstructure of general optical films has certain regularity, the local area with defects is screened by comparing the grayscale difference between the local area and the standard local area. However, in this method, the local area is set according to experience, and only the grayscale difference is compared and analyzed, which makes the defect analysis results less accurate, thereby resulting in lower accuracy of the microscopic quality detection results. Summary of the invention

[0004] In order to solve the technical problem that the defect analysis results of the existing methods are relatively inaccurate, which leads to low accuracy of the microscopic quality detection results, the purpose of the present invention is to provide a high-precision detection method for the microscopic quality of optical films. The technical scheme adopted is as follows:

[0005] Acquire a microscopic grayscale image of the optical film; analyze the periodic characteristics of the grayscale values ​​according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image, and obtain the size of the sliding window;

[0006] The microscopic grayscale image is processed by using the sliding window of the size, and the structural correlation factor between each sliding window and each adjacent sliding window is obtained according to the grayscale similarity distribution between adjacent sliding windows of different sliding lengths in different directions of each sliding window in the microscopic grayscale image and the difference distribution of local grayscale fluctuations;

[0007] According to the distribution characteristics of the pixel points belonging to the target in each adjacent sliding window corresponding to each sliding window and the corresponding sliding length, combined with the structural correlation factor, a microscopic feature evaluation of each sliding window is obtained;

[0008] The defect feature degree of the optical film is analyzed based on the microscopic feature evaluation of each sliding window in the microscopic grayscale image to obtain the quality evaluation result.

[0009] Preferably, the step of analyzing the periodic characteristics of the grayscale values ​​according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image to obtain the size of the sliding window specifically includes:

[0010] For any row or column of pixels, the period value of the row or column is obtained according to the grayscale change characteristics of all pixels in the row or column;

[0011] Based on the periodic values ​​of a preset number of rows in the microscopic grayscale image, the first side length of the sliding window is determined, and based on the periodic values ​​of a preset number of columns in the microscopic grayscale image, the second side length of the sliding window is determined, and the size of the sliding window is the product of the first side length and the second side length.

[0012] Preferably, obtaining the period value of the row or column according to the grayscale change characteristics of all pixels in the row or column specifically includes:

[0013] The grayscale values ​​of the pixels in any row or column form a pixel grayscale sequence, the autocorrelation function of the pixel grayscale sequence is calculated, and the period value of the row or column corresponding to the pixel grayscale sequence is determined based on the interval distance between the peaks of the autocorrelation function.

[0014] Preferably, the structural correlation factor between each sliding window and each adjacent sliding window is obtained according to the grayscale similarity distribution between adjacent sliding windows with different sliding lengths in different directions of each sliding window in the microscopic grayscale image and the difference distribution of local grayscale fluctuations, specifically including:

[0015] For any sliding window in the microscopic grayscale image, adjacent sliding windows with different sliding lengths in different directions of the sliding window are all recorded as reference windows of the sliding window, and the sliding length in the horizontal direction is not less than the first side length, and the sliding length in the vertical direction is not less than the second side length;

[0016] According to the difference characteristic distribution of the grayscale values ​​of the pixels at the corresponding positions of the sliding window and each reference window, the grayscale correlation feature between the sliding window and each reference window is obtained;

[0017] According to the difference between the grayscale distribution characteristics of each pixel point in the sliding window in the local range and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range, the aggregation fluctuation characteristics between the sliding window and each reference window are obtained;

[0018] The structural correlation factor between the sliding window and each reference window is obtained according to the grayscale correlation feature and the aggregation fluctuation feature, wherein the grayscale correlation feature is positively correlated with the structural correlation factor, and the aggregation fluctuation feature is negatively correlated with the structural correlation factor.

[0019] Preferably, obtaining the grayscale correlation feature between the sliding window and each reference window according to the difference feature distribution of the grayscale values ​​of the pixels at the corresponding positions of the sliding window and each reference window specifically includes:

[0020] For any reference window, the grayscale correlation feature between the sliding window and the reference window is determined based on the negative correlation coefficient of the grayscale value difference between the sliding window and the reference window at each corresponding pixel at the same position.

[0021] Preferably, obtaining the aggregated fluctuation characteristics between the sliding window and each reference window according to the difference between the grayscale distribution characteristics of each pixel point in the sliding window in the local range and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range specifically includes:

[0022] For any reference window, a difference window is determined based on the grayscale value difference of the pixels at corresponding positions between the sliding window and the reference window;

[0023] Determine a neighborhood difference factor for each position based on the overall level of all grayscale value differences within a preset neighborhood of each position in the difference window;

[0024] The aggregation fluctuation characteristics between the sliding window corresponding to the difference window and the reference window are determined by combining the fluctuation degree of the neighborhood difference factor and the fluctuation degree of the gray value difference at each position in the difference window.

[0025] Preferably, obtaining the aggregated fluctuation characteristics between the sliding window and each reference window according to the difference between the grayscale distribution characteristics of each pixel point in the sliding window in the local range and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range specifically includes:

[0026] For any reference window, a difference window is determined based on the grayscale value difference of the pixels at corresponding positions between the sliding window and the reference window;

[0027] Determine a neighborhood difference factor for each position based on the overall level of all grayscale value differences within a preset neighborhood of each position in the difference window;

[0028] The aggregation fluctuation characteristics between the sliding window corresponding to the difference window and the reference window are determined by combining the fluctuation degree of the neighborhood difference factor and the fluctuation degree of the gray value difference at each position in the difference window.

[0029] Preferably, obtaining the reference weight of the reference window according to the feature proportion of the pixel points belonging to the target in the reference window and the sliding length of the reference window relative to the sliding window specifically includes:

[0030] The number of pixels in the reference window whose grayscale values ​​are greater than the grayscale mean of all pixels is obtained, and the reference weight of the reference window is determined based on the proportion of the number of pixels in the reference window and the sliding length of the reference window. The proportion of the number of pixels in the reference window is positively correlated with the reference weight, and the sliding length is negatively correlated with the reference weight.

[0031] Preferably, the step of screening each reference window of the sliding window by using the reference structure characteristic value to obtain an associated window of the sliding window specifically includes:

[0032] A set number of reference windows of the sliding window are selected in descending order of the reference structure characteristic values ​​as associated windows of the sliding window.

[0033] Preferably, the step of evaluating and analyzing the defect feature degree of the optical film according to the microscopic features of each sliding window in the microscopic grayscale image to obtain the quality evaluation result specifically includes:

[0034] Fusing the microscopic feature evaluations of all sliding windows in the microscopic grayscale image, determining a defect judgment factor of the microscopic grayscale image, wherein the defect judgment factor is a normalized value;

[0035] When the defect judgment factor is greater than or equal to the preset defect threshold, the quality evaluation result of the optical film is unqualified; when the defect judgment factor is less than the preset defect threshold, the quality evaluation result of the optical film is qualified.

[0036] The embodiments of the present invention have at least the following beneficial effects:

[0037] The present invention firstly obtains the size of the sliding window corresponding to the local area of ​​the periodic distribution characteristics of the pixel points in the rows and columns in the collected microscopic grayscale image, and determines the size of the window to ensure that the microstructure of the optical film must exist in the window, which is conducive to the subsequent detection of the abnormal area of ​​the optical film, and can approximately characterize the single periodic structural characteristics in the microstructure of the optical film. Then, by comparing the differences between the sliding windows, the difference characteristics of the microstructure are obtained, and the structural correlation factor is quantified. It is not limited to the complete microstructure, and the misjudgment caused by the normal deviation between the complete microstructure is avoided, and the detection accuracy is improved. At the same time, when performing feature comparisons between different windows, adjacent windows of different lengths and distances are considered, so that the feature correlation evaluation is more accurate. Further, the degree of defects that may exist in each window is quantified in combination with the feature distribution in the adjacent windows, and a microscopic feature evaluation is obtained, so that the defect feature degree finally determined adaptively can more accurately and effectively evaluate the microstructure quality results of the optical film. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a flow chart of the steps of a high-precision detection method for microscopic quality of an optical film provided by the present invention;

[0040] Figure 2 is a schematic diagram of a grayscale image of the surface of an optical film provided by the present invention;

[0041] Figure 3 is a schematic diagram of a microscopic grayscale image of an optical film provided by the present invention;

[0042] Figure 4 is a schematic diagram of a structural adhesion defect in the optical film provided by the present invention;

[0043] Figure 5 is a schematic diagram of a structure defect in the optical film provided by the present invention;

[0044] Figure 6 is a flowchart of the steps of the method for obtaining the structure association factor provided by the present invention;

[0045] Figure 7 It is a flow chart of the steps of the method for obtaining microscopic feature evaluation provided by the present invention. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the high-precision detection method for the microscopic quality of an optical film proposed by the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0047] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0048] The specific scheme of the high-precision detection method for microscopic quality of optical films provided by the present invention is described in detail below with reference to the accompanying drawings.

[0049] See also Figure 1 , which shows a flowchart of a method for high-precision detection of microscopic quality of an optical film provided by an embodiment of the present invention, the method comprising the following steps:

[0050] Step S100, obtaining a microscopic grayscale image of the optical film; analyzing the periodic characteristics of the grayscale values ​​according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image, and obtaining the size of the sliding window.

[0051] First, obtain the optical film to be tested, and use an atomic force microscope to scan the microstructure of the optical film. Specifically, place the optical film to be tested in an atomic force microscope for scanning. The atomic force microscope probe moves on the surface of the optical film, and the surface topology information is obtained by sensing the force change on the surface. The atomic force microscope obtains the microstructure of the optical film according to the movement of the probe, and obtains the microscopic image of the optical film. The microscopic grayscale image of the optical film is obtained by preprocessing the microscopic image. The preprocessing operation may include graying, filtering, etc. The image preprocessing operation is a well-known technology and will not be introduced in detail here. Figure 2 As shown in FIG. 1 , it is a grayscale image of the surface of the optical film to be tested in this embodiment. Figure 3 As shown, it is a microscopic grayscale image of the optical film to be tested in this embodiment.

[0052] Depend on Figure 3It can be seen that the overall distribution of the microstructure of the optical film shows a certain regularity and periodicity, and then the pixel distribution characteristics of the local area in each independent period can be found to compare and analyze the possibility of defects in the local area. However, the size of the local area for feature analysis is particularly important. If the local area is too large, the defects in the local area may be less obvious, affecting the accuracy of defect recognition. If the local area is too small, the amount of grayscale information in the local area is small, which will also reduce the accuracy of defect recognition.

[0053] Based on this, this embodiment analyzes the periodic characteristics of gray values ​​through the gray value distribution of each row of pixels and the gray value distribution of each column of pixels to adaptively determine the local range size of the periodic distribution in the microstructure of the optical film.

[0054] Specifically, it is necessary to first analyze the periodic distribution of the grayscale values ​​of each row of pixels and the periodic distribution of the grayscale values ​​of each column of pixels in the microscopic grayscale image, that is, for any row or column of pixels, the periodic value of the row or column is obtained according to the grayscale change characteristics of all pixels in the row or column.

[0055] The grayscale values ​​of the pixels in any row or column constitute a pixel grayscale sequence, and then each row in the microscopic grayscale image corresponds to a pixel grayscale sequence, and each column also corresponds to a pixel grayscale sequence. The pixel points of the pixel grayscale sequences corresponding to all rows are arranged in the same manner, for example, the grayscale values ​​are obtained from left to right to construct the sequence, or the grayscale values ​​are obtained from right to left to construct the sequence. For the same reason, the pixel points of the pixel grayscale sequences corresponding to all columns are also arranged in the same manner, for example, the grayscale values ​​are obtained from top to bottom to construct the sequence.

[0056] Then, based on the periodicity analysis of each pixel grayscale sequence, the periodicity value of each pixel grayscale sequence can be obtained. In this embodiment, the periodicity analysis is performed using the autocorrelation function. Specifically, the autocorrelation function of the pixel grayscale sequence is calculated, and the periodicity value of the row or column corresponding to the pixel grayscale sequence is determined based on the interval distance between the peaks of the autocorrelation function. The average of the horizontal coordinate distances between each two adjacent peaks of the autocorrelation function is used as the periodicity value of the row or column corresponding to the pixel grayscale sequence.

[0057] In other embodiments, Fourier transform, periodogram method and other methods may be used to perform periodic analysis to obtain the periodic value of the grayscale sequence of pixels in each row and the periodic value of the grayscale sequence of pixels in each column in the microscopic grayscale image. The periodic value characterizes the periodic distribution characteristics of the grayscale values ​​of pixels in each row and the periodic distribution characteristics of the grayscale values ​​of pixels in each column.

[0058] Furthermore, the local range size for feature analysis can be adaptively determined based on the periodic distribution characteristics of each row and the periodic distribution characteristics of each column, that is, the first side length of the sliding window can be determined based on the periodic values ​​of a preset number of rows in the microscopic grayscale image, and the second side length of the sliding window can be determined based on the periodic values ​​of a preset number of columns in the microscopic grayscale image. The size of the sliding window is the product of the first side length and the second side length.

[0059] In this embodiment, the horizontal length of the local range is determined by comprehensively considering the distribution of periodic characteristics of a certain number of rows, and the vertical length of the local range is determined by comprehensively considering the distribution of periodic characteristics of a certain number of columns in the same way, that is, the value of the preset number is half of the side length of the microscopic grayscale image. Then, the mean of the periodic values ​​corresponding to all selected rows is used as the first side length, and the mean of the periodic values ​​corresponding to all selected columns is used as the second side length. Furthermore, when the calculated side length value is not an integer, it is necessary to perform a rounding operation downward.

[0060] It should be noted that if the microscopic grayscale image is not a square, half of the shortest side length in the microscopic grayscale image is obtained as the preset number. That is, this embodiment analyzes the grayscale periodicity characteristics of some rows and some columns to adaptively determine the size of the local range. In other embodiments, the implementer may also adaptively determine the size of the local range based on the grayscale periodicity characteristics of all rows and all columns.

[0061] So far, the first side length represents the length of the sliding window corresponding to the local range in the horizontal direction, and the second side length represents the length of the sliding window corresponding to the local range in the vertical direction.

[0062] Step S200, using the sliding window of the size to process the microscopic grayscale image, according to the grayscale similarity distribution between adjacent sliding windows of different sliding lengths in different directions of each sliding window in the microscopic grayscale image and the difference distribution of local grayscale fluctuations, obtain the structural correlation factor between each sliding window and each adjacent sliding window.

[0063] In the microstructure of optical films, the defects that affect the evaluation results of microscopic quality are mainly manifested in structural missing, structural adhesion, etc. Figure 4 , which is a schematic diagram of the structural adhesion defects of optical films, such as Figure 5The figure shows a schematic diagram of a structural missing defect in an optical film. Since the size of the sliding window obtained in this embodiment approximately represents the local range corresponding to a single periodic structure, under normal circumstances, there are no defects in the local range corresponding to each single periodic structure, and the grayscale feature distribution representation is relatively similar. When defects appear in a single periodic structure, there are large difference features between adjacent single periodic structures. Therefore, the microscopic grayscale image is processed using a sliding window of a determined size. The grayscale features and local grayscale fluctuations between adjacent sliding windows can be compared and analyzed to characterize the degree of feature similarity between the two.

[0064] Based on this, the microscopic grayscale image is first processed using a sliding window. Specifically, in this embodiment, each time the sliding window slides, the length of the sliding window is equal to the first side in the horizontal direction and the second side in the vertical direction, that is, two adjacent sliding windows do not overlap. In particular, when a complete sliding window cannot be obtained for an edge portion, it can be overlapped with an already obtained sliding window. In other embodiments, processing can also be performed by padding with zeros or the like.

[0065] Then, the defect performance corresponding to each sliding window is analyzed respectively. The feature analysis of defect performance first requires comparing the degree of each feature performance between each sliding window and the adjacent sliding windows.

[0066] like Figure 6 As shown, in this embodiment, the method for obtaining the structure correlation factor can be implemented by steps S201 to S204.

[0067] Step S201, for any sliding window in the micro grayscale image, adjacent sliding windows with different sliding lengths in different directions are recorded as reference windows of the sliding window, and the sliding length in the horizontal direction is not less than the first side length, and the sliding length in the vertical direction is not less than the second side length.

[0068] In this embodiment, appropriate reference local ranges are selected according to different sliding lengths to perform difference comparison analysis on the single periodic structural distribution characteristics in each sliding window, which can fully consider the periodicity and regularity of the structural distribution characteristics of the optical film, making the result of the feature comparison more convincing and more accurate.

[0069] Specifically, taking any sliding window as an example, sliding windows with different sliding steps in the 0°, 90°, 180° and 270° directions of the sliding window are respectively obtained as reference windows of the any sliding window, and the reference window and the sliding window are in an adjacent position relationship on the microscopic grayscale image.

[0070] The specific method for obtaining different sliding step lengths is as follows: in the horizontal direction, starting from the first side length, the step length is increased by one in sequence to obtain 4 different sliding step lengths. For example, the 4 sliding step lengths in the 0° direction can be expressed as N, N+1, N+2 and N+3, where N represents the first side length. Based on the same method, the 4 sliding step lengths in the 180° direction can be obtained. In the vertical direction, starting from the second side length, the step length is increased by one in sequence to obtain 4 different sliding step lengths. For example, the 4 sliding step lengths in the 90° direction can be expressed as M, M+1, M+2 and M+3, where M represents the second side length. Based on the same method, the 4 sliding step lengths in the 270° direction can be obtained.

[0071] Based on the above steps, 4 sliding windows can be obtained in each direction, and then 16 reference windows adjacent to the sliding window in the center can be obtained. In other embodiments, the implementer can select the number of sliding windows in each direction according to the specific implementation scenario, but it is necessary to ensure that the number of sliding windows in all directions is equal.

[0072] It should be noted that, under normal circumstances, the features between adjacent sliding windows are relatively similar, so the degree of defect manifestation can generally be analyzed by comparing the feature differences between the reference window and the sliding window corresponding to the sliding step length of the first side length N and the sliding step length of the second side length M. In order to identify defects more accurately, considering that the adhesion defects in optical films are generally the edge parts of the single-period structure, the feature comparison results in more local ranges are analyzed through a certain sliding to obtain a more accurate degree of defect manifestation.

[0073] Step S202, obtaining the grayscale correlation feature between the sliding window and each reference window according to the difference feature distribution of the grayscale values ​​of the pixels at the corresponding positions of the sliding window and each reference window.

[0074] Each sliding window in the microscopic grayscale image can be approximately understood as a single-period structure in the microstructure of the optical film. Under normal circumstances, the characteristics of different single-period structures of the optical film are relatively similar. When defects exist, the characteristics of different single-period structures of the optical film are relatively different. Therefore, by comparing and analyzing the grayscale difference characteristics between each sliding window and each corresponding reference window, the similarity in grayscale can be quantified to obtain grayscale correlation characteristics.

[0075] Specifically, for any reference window, the grayscale correlation feature between the sliding window and the reference window is determined based on the negative correlation coefficient of the grayscale value difference of each pixel point at the same position of the sliding window and the reference window. The negative correlation coefficient can be represented by the form of a reciprocal, a negative exponential power, etc. As a specific example, the grayscale correlation feature between the i-th sliding window and the corresponding n-th reference window in the microscopic grayscale image is It can be expressed as:

[0076] ,in represents the gray value of the t-th pixel in the i-th sliding window, Represents the gray value of the t-th pixel in the n-th reference window corresponding to the ith sliding window, Indicates the total number of pixels contained in the sliding window, that is, the size of the sliding window, N represents the length of the first side, and M represents the length of the second side. It is a preset hyperparameter. In order to prevent the denominator from taking the value of 0, it is taken as 0.1 in this embodiment. The implementer can set it according to the specific implementation scenario.

[0077] It reflects the grayscale difference between the i-th sliding window and the n-th reference window at the same corresponding position. The smaller the grayscale difference, the more similar the grayscale of the sliding window and the reference window at the corresponding position, and the larger the value of the corresponding grayscale correlation feature. The grayscale correlation feature between the sliding window and the reference window characterizes the grayscale similarity of the single-period structure of the two through the grayscale difference of the corresponding position. The larger the value of the grayscale correlation feature, the more similar the grayscale distribution of the single-period structure of the two is, and the smaller the possibility of defects in the corresponding sliding window.

[0078] Step S203, obtaining the aggregation fluctuation characteristics between the sliding window and each reference window according to the difference between the grayscale distribution characteristics of each pixel point in the sliding window in the local range and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range.

[0079] Secondly, defect analysis is only performed on the grayscale difference distribution at the same position. Considering that each single-period structure of the optical film may have certain differences, the grayscale difference distribution at the same position may also be large, affecting the similarity and difference analysis results. Therefore, further grayscale difference analysis is performed on the fluctuation degree of grayscale difference at all corresponding positions in the sliding window and the reference window.

[0080] First, the grayscale difference between the sliding window and the reference window at the same corresponding position is obtained, that is, any reference window is still used as an example for explanation, and the difference window is determined based on the grayscale value difference of the pixel points at the corresponding position between the sliding window and the reference window. In this embodiment, the absolute value of the grayscale value difference of the pixel points at each corresponding position between the i-th sliding window and the corresponding n-th reference window is used as each element of the difference window, that is, the grayscale value difference at the t-th position in the difference window can be expressed as It can be understood that the obtained difference window is equal to the size of the sliding window and the reference window. In this embodiment, the elements in the difference window are called gray value differences, that is, the absolute value of the difference between the gray values ​​of the corresponding positions of the sliding window and the reference window.

[0081] Then, based on the overall level of all grayscale value differences in the preset neighborhood of each position in the difference window, the neighborhood difference factor of each position is determined. In this embodiment, for the difference window between the i-th sliding window and the n-th reference window, the mean of all grayscale value differences in the preset neighborhood of the t-th position is used as the neighborhood difference factor of the t-th position, and the mean of all grayscale value differences is used to represent the overall level of all grayscale value differences in the preset neighborhood. In other embodiments, the median, mode, etc. can also be used to represent the overall level.

[0082] The preset neighborhood includes 5 data values, that is, the 5 grayscale value differences with the closest spatial distance to each position in the difference window. Specifically, in a difference window, the spatial Euclidean distance between other positions and the t-th position is calculated respectively, and the grayscale value differences of the 5 positions with the smallest Euclidean distance constitute the preset neighborhood of the t-th position. The average of the grayscale value differences of these 5 positions is the neighborhood difference factor of the t-th position.

[0083] Furthermore, the aggregated fluctuation characteristics between the sliding window and the reference window corresponding to the difference window are determined by combining the fluctuation degree of the neighborhood difference factor and the fluctuation degree of the gray value difference at each position in the difference window. In this embodiment, the variance is used to reflect the fluctuation degree of a set of data. In other embodiments, the standard deviation, difference, etc. can also be used to reflect the fluctuation degree of a set of data.

[0084] As a specific example, the aggregate fluctuation feature between the i-th sliding window and the corresponding n-th reference window is It can be expressed as: ,in, represents the variance of the grayscale value differences at all positions in the difference window between the i-th sliding window and the corresponding n-th reference window, Represents the variance of the neighborhood difference factor for all positions within the difference window between the i-th sliding window and the corresponding n-th reference window.

[0085] The larger the value of is, the greater the volatility and discreteness of the grayscale difference between the corresponding positions in the sliding window and the reference window. The larger the value of , the greater the volatility and discreteness of the local neighborhood difference distribution of the corresponding positions in the two windows, and the larger the value of the corresponding aggregation fluctuation feature. The aggregation fluctuation feature characterizes the degree of fluctuation characteristics of the grayscale difference aggregation between the sliding window and the reference window.

[0086] At this point, a clustered fluctuation feature can be calculated between a sliding window and each corresponding reference window. On the one hand, it reflects the fluctuation degree of the grayscale difference at the corresponding position of the sliding window and the reference window. On the other hand, the grayscale difference fluctuation also reflects the possibility of defect anomalies in the sliding window.

[0087] Step S204, obtaining a structural correlation factor between the sliding window and each reference window according to the grayscale correlation feature and the aggregation fluctuation feature, wherein the grayscale correlation feature is positively correlated with the structural correlation factor, and the aggregation fluctuation feature is negatively correlated with the structural correlation factor.

[0088] The grayscale correlation feature reflects the feature similarity between the two windows by analyzing the grayscale differences at the corresponding positions of different windows. The aggregated fluctuation feature reflects the feature difference between the two windows by analyzing the grayscale fluctuations at the corresponding positions of different windows and the local grayscale difference fluctuations. Combining the feature distribution of different windows in these two aspects, the correlation of the structural distribution between the two windows can be quantified.

[0089] Specifically, in this embodiment, the ratio of the grayscale correlation feature to the aggregate fluctuation feature between the i-th sliding window and the corresponding n-th reference window is used as the structural correlation factor between the i-th sliding window and the corresponding n-th reference window. , which can be expressed as: , represents the grayscale correlation feature between the i-th sliding window and the corresponding n-th reference window, Represents the grayscale correlation feature aggregation fluctuation feature between the i-th sliding window and the corresponding n-th reference window. The larger the value of the grayscale correlation feature, the more similar the grayscale feature distribution in the two windows is. The smaller the value of the aggregation fluctuation feature, the smaller the grayscale difference fluctuation in the corresponding position in the two windows is. The more similar the feature distribution in the two windows is, the larger the value of the corresponding structural correlation factor is. The structural correlation factor characterizes the feature similarity between the single-period structure corresponding to the sliding window and the reference window. The larger its value is, the greater the similarity between the single-period structures corresponding to the two windows is, and the smaller the possibility of defects in the local area corresponding to the sliding window is.

[0090] Step S300, obtaining a microscopic feature evaluation of each sliding window according to the distribution characteristics of the pixels belonging to the target in each adjacent sliding window corresponding to each sliding window and the corresponding sliding length, combined with the structural correlation factor.

[0091] The structural correlation factor reflects the structural similarity between the sliding window and the adjacent local area by comparing and analyzing the grayscale difference distribution and discrete distribution of the sliding window and the reference window in two aspects. Considering that the distance distribution between different reference windows and the sliding window is different, the distance between the sliding window and different reference windows is different relative to the period length. When the distance between two different single-period structures is closer, the different single-period structures are more consistent with the grayscale change period of the pixel points, and the corresponding local range is more referenceable when performing defect feature comparison analysis. At the same time, the higher the similarity feature, the more the proportion of structural pixels in the reference window is, the stronger the corresponding reference is. Based on this feature, different reference weights can be adaptively assigned to each reference window for each target proportion and sliding step size of the reference window for defect feature comparison analysis. The reference weight is used to perform weighted fusion of the structural correlation factor of each reference window corresponding to the sliding window, quantify the degree of defect anomaly that may exist in each sliding window, and obtain microscopic feature evaluation.

[0092] In this embodiment, if Figure 7 As shown, the method for obtaining the microscopic feature evaluation of the sliding window can be implemented by steps S301 to S304.

[0093] Step S301, for any reference window of the sliding window, a reference weight of the reference window is obtained according to the feature proportion of the pixel points belonging to the target in the reference window and the sliding length of the reference window relative to the sliding window.

[0094] The more pixels belonging to the effective structure in the local area corresponding to the reference window, the greater the effectiveness of the reference window, and the greater the reference window can be used as a reference for defect analysis and evaluation of the sliding window, and the greater the weight can be given to the reference window. Among them, the effective structure can be understood as the distribution of pixels that can reflect the local microstructure of the optical film, which is composed of Figure 3 It can be seen that the edge part in the microstructure often appears as a highlighted part, so the weight of the reference window can be assigned by analyzing the proportion of highlighted pixels in the reference window.

[0095] Specifically, it is first necessary to screen the highlighted pixels, that is, to obtain the number of pixels in the reference window whose grayscale values ​​are greater than the grayscale mean of all pixels. In the tth reference window, the grayscale mean of all pixels in the window is calculated, and the number of pixels in the tth reference window whose grayscale values ​​are greater than the grayscale mean is obtained. This number can characterize the distribution of the number of brighter pixels in the reference window, and then the ratio of the number of pixels in the tth reference window whose grayscale values ​​are greater than the grayscale mean to the number of all pixels in the reference window is recorded as the feature ratio of the tth reference window.

[0096] Furthermore, considering that the closer the distance to the sliding window is, the greater the similarity with the reference of the sliding window is, the greater the reference value of the sliding window for defect feature evaluation is, each reference window corresponding to the sliding window is obtained by translation with different sliding steps, and then the sliding step can characterize the spatial distance distribution between the reference window and the sliding window, that is, the reference size corresponding to each reference window can be quantified in combination with the feature proportion and the sliding step.

[0097] Based on the proportion of the number of pixels in the reference window and the sliding length of the reference window, the reference weight of the reference window is determined, the proportion of the number of pixels in the reference window is positively correlated with the reference weight, and the sliding length is negatively correlated with the reference weight. The positive correlation can be an additive relationship, a multiplicative relationship, etc., and the negative correlation can be characterized by a reciprocal form, a negative exponential power form, etc.

[0098] In this embodiment, any reference window corresponding to any sliding window is still used as an example for description. The reference weight of the nth reference window of the i-th sliding window is It can be expressed as ;in, is the number of valid structural pixels of the nth reference window of the i-th sliding window, that is, the number of pixels greater than the mean grayscale value. is the sliding step size corresponding to the nth reference window of the ith sliding window, is the sum of the valid structural pixels of the nth reference window, that is, the total number of all pixels in the nth reference window. is the feature ratio of the nth reference window.

[0099] The smaller the sliding step value, the closer the distance between the two windows is. The larger the feature proportion value, the larger the proportion of effective structures in the corresponding reference window. The larger the corresponding reference weight value, the greater the reference window is.

[0100] Step S302: combining the reference weight of the reference window and the corresponding structure association factor to obtain a reference structure feature value of the reference window.

[0101] Specifically, the reference structure eigenvalue of the nth reference window of the i-th sliding window is It can be expressed as ,in represents the reference weight of the nth reference window of the ith sliding window, It represents the structural correlation factor of the nth reference window of the ith sliding window, that is, the feature similarity and feature correlation degree of the reference window are weighted by the referenceability corresponding to the reference window. The larger the reference weight is, the larger the value of the structural correlation factor is, and the larger the value of the reference structure feature value of the corresponding reference window is, which means that the reference value of the reference window is greater, the similarity between the corresponding sliding window and the reference window is greater, and the possibility of defect anomaly in the sliding window is smaller.

[0102] Step S303: Filter each reference window of the sliding window by using the reference structure characteristic value to obtain an associated window of the sliding window.

[0103] In this embodiment, a set number of reference windows of the sliding window are selected as associated windows of the sliding window in the order of the reference structure characteristic values ​​from large to small. In this embodiment, 5 reference windows of the sliding window are selected as associated windows for further feature analysis in the order of the reference structure characteristic values ​​from large to small, that is, the set number is 5. In other embodiments, the implementer can set it according to the specific implementation scenario. The larger the value of the reference structure characteristic value, the greater the reference value of the corresponding reference window, and the more similar the characteristic distribution between the sliding window is, the better the effect of defect feature evaluation is.

[0104] Step S304, determining the microscopic feature evaluation of the sliding window based on the negative correlation coefficient of the reference structure feature value of each associated window of the sliding window.

[0105] The correlation window represents the local area that is similar to the sliding window characteristics and has a higher reference value. The more similar the feature distribution between the sliding window and all other similar local areas is, the more it means that the feature distribution between different local areas approximately shows the regular distribution of the optical film, that is, the possibility of defects in the local area corresponding to the sliding window is lower.

[0106] In this embodiment, the microscopic feature evaluation of the i-th sliding window The calculation formula can be expressed as: ,in, represents the normalization function, represents an exponential function with the natural constant e as the base, represents the reference structure eigenvalue of the sth associated window of the ith sliding window, represents the number of associated windows of the i-th sliding window, and in this embodiment, the value is 5. is the microscopic feature evaluation of the i-th sliding window, which represents the possibility of defects in the i-th sliding window. The larger the value of is, the greater the feature similarity between the sliding window and the associated window is, and the lower the possibility of defect anomaly in the corresponding sliding window is.

[0107] Step S400, evaluating and analyzing the defect feature degree of the optical film according to the microscopic features of each sliding window in the microscopic grayscale image to obtain a quality evaluation result.

[0108] Considering that the more sliding windows with abnormal performance in the optical film microstructure, the more significant the abnormality of the sliding windows, the worse the quality of the corresponding optical film. Therefore, the quality of the optical film microstructure can be evaluated by analyzing the possibility of defect characteristics in the local range of each sliding window.

[0109] Specifically, the microscopic feature evaluations of all abnormal sliding windows in the microscopic grayscale image are integrated to determine the defect judgment factor of the microscopic grayscale image. The defect judgment factor is a normalized value. When the microscopic feature evaluation value corresponding to the sliding window is larger, the degree of the defective microscopic features exhibited in the sliding window is greater, and thus the possibility of defects in the local area corresponding to the sliding window is greater, and thus the quality of the corresponding optical film is worse.

[0110] Based on this, the negative correlation coefficient of the microscopic feature evaluation of all sliding windows in the microscopic grayscale image is determined as the defect feature degree. Specifically, the reciprocal of the cumulative sum of the microscopic feature evaluation of all sliding windows in the microscopic grayscale image is used as the defect feature degree, that is, the defect feature degree characterizes the possibility of defects in the microstructure of the optical film corresponding to the microscopic grayscale image and the degree of defect characterization. The larger the value, the greater the degree of defects in the microstructure of the optical film, the more serious the quality problem of the optical film, and the worse the quality of the optical film.

[0111] Further, when the defect judgment factor is greater than or equal to the preset defect threshold, it means that the microscopic structure of the optical film corresponding to the microscopic characteristic image has a large degree of defect, and the quality evaluation result of the optical film is unqualified; when the defect judgment factor is less than the preset defect threshold, it means that the microscopic structure of the optical film corresponding to the microscopic characteristic image has a small degree of defect, and the quality evaluation result of the optical film is qualified. In this embodiment, the defect threshold is 0.5.

[0112] The embodiment of the present application also provides a computer program product. When the computer program product is run on a computer device, the computer device can execute any one of the above-mentioned high-precision detection methods for microscopic quality of optical films.

[0113] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program code is stored. When the computer program code is executed on a computer device, the computer device can execute any one of the above-mentioned high-precision detection methods for microscopic quality of optical films.

[0114] In the embodiments provided in the present application, it should be understood that the provided computer device, computer program product and computer-readable storage medium are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the methods provided above and will not be repeated here.

[0115] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A high-precision detection method for the microscopic quality of an optical film, characterized in that: The method comprises the following steps: Acquire a microscopic grayscale image of the optical film; analyze the periodic characteristics of the grayscale values ​​according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image, and obtain the size of the sliding window; The microscopic grayscale image is processed by using the sliding window of the size, and the structural correlation factor between each sliding window and each adjacent sliding window is obtained according to the grayscale similarity distribution between adjacent sliding windows of different sliding lengths in different directions of each sliding window in the microscopic grayscale image and the difference distribution of local grayscale fluctuations; According to the distribution characteristics of the pixel points belonging to the target in each adjacent sliding window corresponding to each sliding window and the corresponding sliding length, combined with the structural correlation factor, a microscopic feature evaluation of each sliding window is obtained; The defect characteristic degree of the optical film is analyzed according to the microscopic characteristic evaluation of each sliding window in the microscopic grayscale image to obtain the quality evaluation result; The method of analyzing the periodic characteristics of the grayscale values ​​according to the grayscale value distribution of each row of pixels and the grayscale value distribution of each column of pixels in the microscopic grayscale image to obtain the size of the sliding window specifically includes: For any row or column of pixels, the period value of the row or column is obtained according to the grayscale change characteristics of all pixels in the row or column; Determine a first side length of the sliding window based on period values ​​of a preset number of rows in the microscopic grayscale image, and determine a second side length of the sliding window based on period values ​​of a preset number of columns in the microscopic grayscale image, wherein the size of the sliding window is the product of the first side length and the second side length; The structural correlation factor between each sliding window and each adjacent sliding window is obtained according to the grayscale similarity distribution between adjacent sliding windows with different sliding lengths in different directions of each sliding window in the microscopic grayscale image and the difference distribution of local grayscale fluctuations, specifically including: For any sliding window in the microscopic grayscale image, adjacent sliding windows with different sliding lengths in different directions of the sliding window are all recorded as reference windows of the sliding window, and the sliding length in the horizontal direction is not less than the first side length, and the sliding length in the vertical direction is not less than the second side length; According to the difference characteristic distribution of the grayscale values ​​of the pixels at the corresponding positions of the sliding window and each reference window, the grayscale correlation feature between the sliding window and each reference window is obtained; According to the difference between the grayscale distribution characteristics of each pixel point in the sliding window in the local range and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range, the aggregation fluctuation characteristics between the sliding window and each reference window are obtained; According to the grayscale correlation feature and the aggregation fluctuation feature, a structural correlation factor between the sliding window and each reference window is obtained, wherein the grayscale correlation feature is positively correlated with the structural correlation factor, and the aggregation fluctuation feature is negatively correlated with the structural correlation factor; The grayscale correlation feature between the sliding window and each reference window is obtained according to the difference feature distribution of the grayscale values ​​of the pixels at the corresponding positions of the sliding window and each reference window, specifically including: For any reference window, based on the negative correlation coefficient of the grayscale value difference between the sliding window and the reference window at each corresponding pixel at the same position, determine the grayscale correlation feature between the sliding window and the reference window; The method of obtaining the aggregate fluctuation characteristics between the sliding window and each reference window according to the difference between the grayscale distribution characteristics of each pixel point in the local range in the sliding window and the grayscale distribution characteristics of the pixel point at the corresponding position in each reference window in the local range specifically includes: For any reference window, a difference window is determined based on the grayscale value difference of the pixels at corresponding positions between the sliding window and the reference window; Determine a neighborhood difference factor for each position based on the overall level of all grayscale value differences within a preset neighborhood of each position in the difference window; The aggregation fluctuation characteristics between the sliding window corresponding to the difference window and the reference window are determined by combining the fluctuation degree of the neighborhood difference factor and the fluctuation degree of the gray value difference at each position in the difference window.

2. The high-precision detection method for microscopic quality of an optical film according to claim 1, characterized in that: The step of obtaining the period value of the row or column according to the grayscale change characteristics of all pixels in the row or column specifically includes: The grayscale values ​​of the pixels in any row or column form a pixel grayscale sequence, the autocorrelation function of the pixel grayscale sequence is calculated, and the period value of the row or column corresponding to the pixel grayscale sequence is determined based on the interval distance between the peaks of the autocorrelation function.

3. The high-precision detection method for microscopic quality of an optical film according to claim 1, characterized in that: The step of obtaining the microscopic feature evaluation of each sliding window based on the pixel distribution characteristics and the corresponding sliding length of each adjacent sliding window corresponding to each sliding window and the structural correlation factor specifically includes: For any reference window of the sliding window, a reference weight of the reference window is obtained according to the feature proportion of the pixel points belonging to the target in the reference window and the sliding length of the reference window relative to the sliding window; Combining the reference weight of the reference window with the corresponding structural correlation factor, a reference structural characteristic value of the reference window is obtained; Each reference window of the sliding window is screened by using the reference structure characteristic value to obtain an associated window of the sliding window; and a microscopic characteristic evaluation of the sliding window is determined based on a negative correlation coefficient of the reference structure characteristic value of each associated window of the sliding window.

4. The high-precision detection method for microscopic quality of an optical film according to claim 3, characterized in that: The step of obtaining a reference weight of the reference window according to the feature proportion of the pixel points belonging to the target in the reference window and the sliding length of the reference window relative to the sliding window specifically includes: The number of pixels in the reference window whose grayscale values ​​are greater than the grayscale mean of all pixels in the reference window is obtained, and the reference weight of the reference window is determined based on the proportion of the number of pixels in the reference window and the sliding length of the reference window. The proportion of the number of pixels in the reference window is positively correlated with the reference weight, and the sliding length is negatively correlated with the reference weight.

5. The high-precision detection method for microscopic quality of an optical film according to claim 3, characterized in that: The using the reference structure characteristic value to screen each reference window of the sliding window to obtain the associated window of the sliding window specifically includes: A set number of reference windows of the sliding window are selected in descending order of the reference structure characteristic values ​​as associated windows of the sliding window.

6. The high-precision detection method for microscopic quality of an optical film according to claim 1, characterized in that: The step of evaluating and analyzing the defect feature degree of the optical film according to the microscopic feature of each sliding window in the microscopic grayscale image to obtain the quality evaluation result specifically includes: Fusing the microscopic feature evaluations of all sliding windows in the microscopic grayscale image, determining a defect judgment factor of the microscopic grayscale image, wherein the defect judgment factor is a normalized value; When the defect judgment factor is greater than or equal to the preset defect threshold, the quality evaluation result of the optical film is unqualified; when the defect judgment factor is less than the preset defect threshold, the quality evaluation result of the optical film is qualified.

Images