A method, device and storage medium for detecting a dot line defect on a display screen image

Abstract

The application discloses a display screen image point line defect detection method and device and a storage medium, and is used for improving display screen defect detection precision. A shooting image of a display screen to be detected is acquired; a positioning pixel point is randomly selected from the shooting image, and a positioning circular region is generated according to the positioning pixel point; a gray gradient vector of a sample pixel point in the positioning circular region to a corresponding positioning pixel point is calculated, and a mean value of the gray gradient vector is calculated; a Euclidean distance between the sample pixel point at the mean value of the gray gradient vector and each positioning pixel point is calculated to generate an abnormal contour region; a region elimination treatment is performed on the shooting image, a difference image is generated according to a gray mean value of a remaining region and the shooting image; a pre-segmentation extraction is performed on the difference image to generate a defect region to be detected; an array projection evaluation algorithm treatment is performed on the defect region to be detected to generate a suspected line defect region; and feature screening is performed on the suspected line defect region to generate a point defect and line defect set.

Description

Technical Field

[0001] This application relates to the field of display screen inspection, and more particularly to a method, apparatus, and storage medium for detecting dot and line defects on display screen images. Background Technology

[0002] With technological innovation, new display technologies are constantly being updated and iterated. As a core link in the display industry chain, display quality inspection technology continues to receive attention, especially the Mura defect detection of micron-level pixels on the surface of new displays (MicroLED, foldable screens, etc.), which has become a key factor determining the quality of display products. The optical Demura system uses a high-resolution industrial camera to capture images of the lit display and extract sub-pixel unit images from the captured images. It can detect Mura defects on the panel, perform high-precision color / brightness compensation on the display, thereby optimizing display quality and improving yield.

[0003] However, as the pixel integration of new displays increases and the display layer structure becomes more complex, existing display defects (especially point and line defects) are more susceptible to changes due to the increased pixel integration and layer structure. These defects fall under the category of new display defects, and conventional display defect detection methods are becoming less effective. To address these new defects, targeted defect detection is typically required. However, this targeted detection usually covers the entire display area, significantly increasing the computational load and making it more susceptible to interference, thus reducing accuracy. Therefore, it is often necessary to pre-detect defect contours in captured images of the display to construct a background image, thereby filtering out normal areas and retaining suspected defect areas. Targeted defect detection is then performed on these suspected defect areas to reduce computational load and interference. However, currently, there is no specific defect contour detection method for new displays, further reducing the accuracy of display defect detection. Summary of the Invention

[0004] This application discloses a method, apparatus, and storage medium for detecting dot and line defects on a display screen image, which can improve the accuracy of display screen defect detection.

[0005] In a first aspect, this application provides a method for detecting dot and line defects on a display screen image, comprising: acquiring an image of the illuminated display screen under test using an industrial camera to obtain a captured image of the display screen under test; Multiple positioning pixels are randomly selected from the captured image, and a positioning circular area is generated based on the multiple positioning pixels; Calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixel, and calculate the mean of the gray-level gradient vectors based on the calculated gray-level gradient vectors. Calculate the Euclidean distance between the sample pixel at the mean of the gray-level gradient vector and each localized pixel, so as to generate multiple abnormal contour regions based on multiple Euclidean distances; The captured image is processed by filtering out regions based on multiple abnormal contour regions, and the gray-scale mean of the remaining regions is calculated. A difference image is generated based on the gray-scale mean of the remaining regions and the captured image. Defect regions are pre-segmented and extracted from the difference image to generate multiple defect regions to be tested; Multiple defect regions to be tested are processed by an array projection estimation algorithm to generate suspected line defect regions; Feature filtering is performed on suspected line defect areas to generate sets of point defects and line defects.

[0006] Optionally, the step of processing multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions specifically includes: The two-dimensional image matrix information of multiple defect areas to be tested is compressed by row and column respectively to generate a set of one-dimensional grayscale statistical values. When the one-dimensional grayscale statistical value is greater than the preset threshold, the row and column corresponding to the one-dimensional grayscale statistical value are determined as suspected defect rows and columns; The suspected defect rows are integrated to generate suspected line defect areas.

[0007] Optionally, after processing multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions, and before performing feature filtering on the suspected line defect regions to generate a set of point defects and line defects, the detection method further includes: The suspected defect rows and columns that are segmented are first closed in the horizontal and vertical directions, and then convex hull is applied to them to connect the segmented suspected line defect regions into a whole.

[0008] Optionally, after the steps of feature filtering of suspected line defect areas and generating sets of point defects and line defects, the detection method further includes: Completely overlapping point and line defects are removed using grayscale detection. When the remaining point defect corresponds to an empty point defect area, the current point defect is determined to be empty, and the subsequent integration stage is skipped. When the remaining point defects correspond to point defect regions that are not empty regions, calculate the number of point defect regions and the number of line defect regions corresponding to line defects. Select the line defect area and the mean of the line defect area sequentially according to the index sequence; Calculate the shortest distance from all point defect regions to line defect regions, and generate a set of shortest distances for region contours. Record the point defect regions in the set of shortest distances to the region outline that are less than the preset distance value, and generate the first set of point defect regions. Calculate the difference between the mean value of the point defect area and the mean value of the line defect area for each point defect area, and generate a set of differences; Record point defect regions in the difference set that are less than the preset grayscale threshold for merging point lines, and generate a second set of point defect regions. A set of intersection point defects is generated based on the first set of defect regions and the second set of defect regions. The set of intersection point defects is then used to merge point and line defects to generate real point defects.

[0009] Optionally, after the steps of performing region filtering on the captured image based on multiple abnormal contour regions, calculating the gray-scale mean of the remaining regions, generating a difference image based on the gray-scale mean of the remaining regions and the captured image, and before the steps of pre-segmenting and extracting defect regions from the difference image to generate multiple defect regions to be tested, the detection method further includes: Multiple kernel factors are generated based on the type of defect detected in the display screen under test; Multiple kernel factors are used to process the difference image using a convolutional emphasis algorithm.

[0010] Optionally, after the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular area based on the multiple positioning pixels, the detection method further includes: Set up a darkroom acquisition environment, fix the display screen to be tested on the acquisition platform, and turn on the bar side light source; An industrial camera is used to capture images of the unlit display screen under test, thus obtaining images of the dust removal process. The captured images are processed to remove dust based on the dust removal process.

[0011] Optionally, after the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular area based on the multiple positioning pixels, the detection method further includes: The segmentation threshold is calculated based on the captured image to generate the maximum segmentation threshold. The captured image is subjected to threshold segmentation based on the maximum segmentation threshold to generate the minimum envelope of the segmented region. The corner coordinates of the minimum envelope are mapped to the reference coordinates to construct a system of matrix equations and generate a homography matrix. The captured image is subjected to matrix transformation using a homography matrix to correct the image.

[0012] Optionally, after the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular area based on the multiple positioning pixels, the detection method further includes: Calculate the spatial domain standard deviation and the range standard deviation of the captured images; Calculate the spatial domain weight and the value domain weight based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image; Construct nonlinear filter weights based on spatial domain weights and value domain weights; The captured image is subjected to nonlinear filtering based on the nonlinear filtering weights.

[0013] Secondly, this application provides a device for detecting dot and line defects on a display screen image, comprising: a first acquisition unit, used to acquire an image of a lit display screen under test by means of an industrial camera, so as to acquire a captured image of the display screen under test; The selection unit is used to randomly select multiple positioning pixels from the captured image and generate a positioning circular area based on the multiple positioning pixels; The first calculation unit is used to calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixels, and to calculate the mean of the gray-level gradient vectors based on the calculated multiple gray-level gradient vectors. The first generation unit is used to calculate the Euclidean distance between the sample pixel point located at the mean of the gray-level gradient vector and each localized pixel point, so as to generate multiple abnormal contour regions based on multiple Euclidean distances. The second calculation unit is used to perform region filtering on the captured image based on multiple abnormal contour regions, calculate the average gray value of the remaining regions, and generate a difference image based on the average gray value of the remaining regions and the captured image. The second generation unit is used to pre-segment and extract defect regions from the difference image to generate multiple defect regions to be tested. The third generation unit is used to process multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions. The fourth generation unit is used to perform feature filtering on suspected line defect areas and generate a set of point defects and line defects.

[0014] Optionally, the third generation unit includes: performing row compression and column compression on the two-dimensional image matrix information of multiple defect regions to be tested, respectively, to generate a set of one-dimensional grayscale statistical values; When the one-dimensional grayscale statistical value is greater than the preset threshold, the row and column corresponding to the one-dimensional grayscale statistical value are determined as suspected defect rows and columns; The suspected defect rows are integrated to generate suspected line defect areas.

[0015] Optionally, after the third generation unit and before the fourth generation unit, the detection device further includes a connection processing unit, which performs a closing operation on the suspected defect rows and columns with blocks in the horizontal and vertical directions and then performs a convex hull operation on them, thereby connecting the various segmented suspected line defect regions into a whole.

[0016] Optionally, after the fourth generation unit, the detection device further includes: a removal unit for removing completely overlapping point defects and line defects by grayscale detection; a determination unit for determining that the current point defect is empty and skipping the subsequent integration stage when the point defect region corresponding to the remaining point defect is an empty region; a third calculation unit for calculating the number of point defect regions and the number of line defect regions corresponding to the line defects when the point defect region corresponding to the remaining point defect is not an empty region; a selection unit for sequentially selecting line defect regions and the average value of line defect regions according to the index sequence; and a fifth generation unit for calculating the shortest distance of the region contour from all point defect regions to the line defect regions, generating... The system generates a set of shortest distances for the region contours; a sixth generation unit records point defect regions in the set of shortest distances for the region contours that are less than a preset distance value, generating a first set of point defect regions; a seventh generation unit calculates the difference between the average point defect region value and the average line defect region value for each point defect region, generating a set of differences; an eighth generation unit records point defect regions in the set of differences that are less than a preset grayscale threshold for merging point and line defects, generating a second set of point defect regions; and a ninth generation unit generates an intersection point defect set based on the first and second set of point defect regions, and uses the intersection point defect set to merge point and line defects to generate real point defects.

[0017] Optionally, after the second calculation unit and before the second generation unit, the detection device further includes: a tenth generation unit, used to generate multiple kernel factors according to the detection defect type of the display screen under test; and a convolution unit, used to perform convolutional emphasis algorithm processing on the difference image using the multiple kernel factors respectively.

[0018] Optionally, after the first acquisition unit and before the selection unit, the detection device further includes: an activation unit for setting the darkroom acquisition environment, fixing the display screen under test on the acquisition platform, and activating the strip side light source; a second acquisition unit for acquiring images of the unlit display screen under test using an industrial camera to obtain a dust removal image of the display screen under test; and a dust removal process for performing dust removal processing on the captured image based on the dust removal image.

[0019] Optionally, after the first acquisition unit and before the selection unit, the detection device further includes: a fourth calculation unit, used to calculate a segmentation threshold based on the captured image to generate a maximum segmentation threshold; a segmentation unit, used to perform threshold segmentation processing on the captured image based on the maximum segmentation threshold to generate the minimum envelope of the segmented region; a construction unit, used to map the corner coordinates of the minimum envelope to the reference coordinates to construct a system of matrix equations and generate a homography matrix; and a correction unit, used to perform matrix transformation on the captured image using the homography matrix to correct the captured image.

[0020] Optionally, after the first acquisition unit and before the selection unit, the detection device further includes: a fifth calculation unit for calculating the spatial domain standard deviation and the value domain standard deviation of the captured image; a sixth calculation unit for calculating the spatial domain weight and the value domain weight based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image; a construction unit for constructing nonlinear filtering weights based on the spatial domain weights and the value domain weights; and a filtering unit for performing nonlinear filtering on the captured image based on the nonlinear filtering weights.

[0021] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: In this application, an industrial camera is first used to capture images of the illuminated display screen under test to obtain images of the display screen. Multiple positioning pixels are randomly selected from the captured images, and a positioning circular region is generated based on these pixels. The gray-level gradient vectors from multiple sample pixels within the positioning circular region to their corresponding positioning pixels are calculated, and the mean of these gray-level gradient vectors is calculated. The Euclidean distance between the sample pixel located at the mean of the gray-level gradient vector and each positioning pixel is calculated to generate multiple abnormal contour regions based on these Euclidean distances. The captured image is then processed by region filtering based on these multiple abnormal contour regions, and the mean gray-level of the remaining regions is calculated. A difference image is generated based on the mean gray-level of the remaining regions and the captured image. The difference image is then pre-segmented and extracted to extract defect regions, generating multiple defect regions to be tested. An array projection estimation algorithm is applied to these multiple defect regions to generate suspected line defect regions. Feature filtering is then performed on the suspected line defect regions to generate sets of point defects and line defects.

[0022] From the captured image on the display screen, a positioning circular region is determined for several positioning pixels. A grayscale gradient vector is generated for each sample pixel within this region. This grayscale gradient vector is vector data generated based on the grayscale values ​​and positional orientation of the sample pixels and the positioning pixels. When a sample pixel has a defect, this grayscale gradient vector will fluctuate, thus affecting the grayscale gradient vector of the entire positioning circular region. By identifying the sample pixels with the average grayscale gradient vector of this positioning circular region, abnormal pixels are initially screened from these sample pixels, generating multiple abnormal contour regions. These abnormal contour regions serve as the initial defect contours. These abnormal contour regions are then removed, and the non-abnormal grayscale values ​​of the remaining regions are calculated. After difference processing with the original captured image, the contours of the defective regions are further highlighted. Next, the generated difference image is segmented and processed using an array projection estimation algorithm. This involves compressing pixel data in both the row and column directions, generating one-dimensional grayscale statistics for each row and column of pixels. These one-dimensional grayscale statistics are then analyzed to identify suspected line defect regions. Finally, targeted defect feature detection is performed on these suspected line defect regions. This method uses a deep clustering search algorithm to pre-detect the initial contours of defects in captured images, which can effectively construct the background image. This allows subsequent detection to focus only on areas suspected of line defects, greatly reducing the computational load and interference, and improving the detection accuracy of display defects. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the method for detecting dot and line defects on the display screen image of this application; Figure 2 A schematic diagram of the method for generating suspected line defect regions for this application; Figure 3 This is a schematic diagram of the method for handling suspected line defect areas in this application; Figure 4 This is a schematic diagram of the method for detecting defects in the novel display screen of this application; Figure 5 This is a schematic diagram of the method for enhancing defect features in this application; Figure 6 Another schematic diagram of a preprocessing method for images captured for the display screen of this application; Figure 7A schematic diagram of the correction method for images captured on the display screen of this application; Figure 8 A schematic diagram of a filtering method for images captured by the display screen of this application; Figure 9 This is a schematic diagram of the device for detecting dot and line defects on the display screen image of this application; Figure 10 This is a schematic diagram of the electronic device used in this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0027] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0029] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0031] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0032] The method described in this application can be applied to servers, devices, terminals, or other devices with logical processing capabilities; therefore, this application does not limit its application. For ease of description, the following description uses a terminal as the executing entity.

[0033] Please see Figure 1 This application provides an embodiment of a method for detecting dot and line defects on a display screen image, comprising: 101. Use an industrial camera to capture images of the lit display screen under test to obtain images of the display screen under test.

[0034] After the terminal has installed the acquisition platform in the intelligent manufacturing darkroom and adjusted the acquisition environment and other hardware of the industrial camera, the bar light source is turned off, and the display screen under test is driven to light up by the PG. Then, the display screen under test is photographed and the captured image is generated.

[0035] 102. Randomly select multiple positioning pixels from the captured image, and generate a positioning circle region based on the multiple positioning pixels.

[0036] 103. Calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixel, and calculate the mean of the gray-level gradient vectors based on the calculated gray-level gradient vectors.

[0037] In this embodiment, a deep clustering algorithm is used to pre-detect the defect contours of the initial image. The terminal first randomly selects a pixel in the grayscale sample matrix of the captured image and sets a radius d to form an initial circular range (positioning circle region). Next, the vectors from each sample point within the initial circular range (positioning circle region) to the center of the circle and the mean of these vectors are calculated respectively.

[0038] 104. Calculate the Euclidean distance between the sample pixel at the mean of the gray-level gradient vector and each localized pixel, so as to generate multiple abnormal contour regions based on multiple Euclidean distances.

[0039] Next, the terminal specifies a corresponding error T and compares the Euclidean distance between the pixel at the mean of the grayscale gradient vector and the located pixel x with the error T. Iteration stops when this error condition is met; that is, if the Euclidean distance is greater than the error T, the sample pixel is determined to be an edge point of the abnormal contour. Otherwise, steps 102 and 103 are repeated for the points obtained from the mean of the vector, thus integrating the edge points of the abnormal contours to generate multiple abnormal contour regions. The following is the mean of the grayscale gradient vector. The calculation formula is as follows:

[0040] Where h is the preset bandwidth parameter for capturing images, c is the preset normalization constant, and K is the kernel function. Let G be the probability density gradient estimate of the kernel function K, and G be the kernel function corresponding to the silhouette function of K. Let G be the probability density gradient estimate.

[0041] 105. Perform region filtering on the captured image based on multiple abnormal contour regions, calculate the average gray value of the remaining regions, and generate a difference image based on the average gray value of the remaining regions and the captured image.

[0042] After the terminal acquires multiple abnormal contour regions, it removes these regions from the original captured image, calculates the grayscale mean of the remaining image region, and generates a blank image with the same type, dimensions, and size as the original captured image. The grayscale mean is then filled into the blank image. Finally, the captured image and the filled blank image are compared to generate a difference image.

[0043] 106. Perform pre-segmentation and extraction of defect regions on the difference image to generate multiple defect regions to be tested.

[0044] After generating the difference image, defect enhancement processing can be performed on the difference image, which will be described in detail in subsequent embodiments. After the above steps are completed, the suspected defect regions in the difference image need to be segmented and extracted.

[0045] 107. Perform array projection estimation algorithm processing on multiple defect areas to be tested to generate suspected line defect areas.

[0046] Then, an array projection estimation algorithm is applied to each segmented suspected defect region to compress the two-dimensional image information into a one-dimensional projection vector. The one-dimensional projection vector is then analyzed to generate suspected line defect regions. The specific generation method will be described in detail in subsequent embodiments.

[0047] 108. Perform feature filtering on suspected line defect areas to generate a set of point defects and line defects.

[0048] After generating the suspected line defect region, the terminal can perform targeted defect detection processing on the suspected line defect region, generating a set of point defects and line defects. Specifically, in this embodiment, the terminal uses a feature quantity analysis method to filter defects (through area, slope, roundness, contrast, energy, etc.), and the obtained point defects are recorded as Dot_defect and line defects as Line_defect.

[0049] In this embodiment, an industrial camera is first used to capture images of the illuminated display screen under test to obtain images of the display screen. Multiple positioning pixels are randomly selected from the captured image, and a positioning circular region is generated based on these pixels. The gray-level gradient vectors from multiple sample pixels within the positioning circular region to their corresponding positioning pixels are calculated, and the mean of these gray-level gradient vectors is calculated. The Euclidean distance between the sample pixel located at the mean of the gray-level gradient vector and each positioning pixel is calculated to generate multiple abnormal contour regions based on these Euclidean distances. The captured image is then processed by region filtering based on these multiple abnormal contour regions, and the mean gray-level of the remaining regions is calculated. A difference image is generated based on the mean gray-level of the remaining regions and the captured image. The difference image is then pre-segmented and extracted to extract defect regions, generating multiple defect regions to be tested. An array projection estimation algorithm is applied to these multiple defect regions to generate suspected line defect regions. Feature filtering is then performed on the suspected line defect regions to generate sets of point defects and line defects.

[0050] From the captured image on the display screen, a positioning circular region is determined for several positioning pixels. A grayscale gradient vector is generated for each sample pixel within this region. This grayscale gradient vector is vector data generated based on the grayscale values ​​and positional orientation of the sample pixels and the positioning pixels. When a sample pixel has a defect, this grayscale gradient vector will fluctuate, thus affecting the grayscale gradient vector of the entire positioning circular region. By identifying the sample pixels with the average grayscale gradient vector of this positioning circular region, abnormal pixels are initially screened from these sample pixels, generating multiple abnormal contour regions. These abnormal contour regions serve as the initial defect contours. These abnormal contour regions are then removed, and the non-abnormal grayscale values ​​of the remaining regions are calculated. After difference processing with the original captured image, the contours of the defective regions are further highlighted. Next, the generated difference image is segmented and processed using an array projection estimation algorithm. This involves compressing pixel data in both the row and column directions, generating one-dimensional grayscale statistics for each row and column of pixels. These one-dimensional grayscale statistics are then analyzed to identify suspected line defect regions. Finally, targeted defect feature detection is performed on these suspected line defect regions. This method uses a deep clustering search algorithm to pre-detect the initial contours of defects in captured images, which can effectively construct the background image. This allows subsequent detection to focus only on areas suspected of line defects, greatly reducing the computational load and interference, and improving the detection accuracy of display defects.

[0051] Please see Figure 2 This application provides an embodiment of a method for generating suspected line defect regions, comprising: 201. Perform row compression and column compression on the two-dimensional image matrix information of multiple defect areas to be tested, respectively, to generate a set of one-dimensional grayscale statistical values.

[0052] 202. When the one-dimensional grayscale statistical value is greater than the preset threshold, the row and column corresponding to the one-dimensional grayscale statistical value are determined as suspected defect rows and columns.

[0053] 203. Integrate the suspected defect rows and columns to generate suspected line defect areas.

[0054] In this embodiment, the terminal first compresses the two-dimensional image (captured image) matrix information into a one-dimensional grayscale statistical value in a certain column or row, generating a set of one-dimensional grayscale statistical values. This embodiment mainly includes the grayscale mean, grayscale standard deviation, and covariance. These parameters need to be selected based on the display characteristics of the screen under test, focusing on features most affected by defects, rather than simply selecting a few specific image features for compression. After compression, the differences in defect features in each row and column can be more accurately divided, improving the accuracy of defect contour region localization. After generating the set of one-dimensional grayscale statistical values, the values ​​in the one-dimensional projection vector are compared with a set threshold to determine suspected defect rows and columns, and a suspected line defect region is finally generated based on these rows and columns. Grayscale statistical values ​​greater than the set threshold are retained. Here, the horizontal and vertical array projections are set to H_Projection(r) and V_Projection(c), with the following formula:

[0055] Where: I(r+r',c+c) is the image grayscale value, and n is the total number of pixels in the current row or column. Let r be the initial row coordinate of the current row, r be the offset coordinate of the current row, and c be the initial column coordinate of the current column. This is the offset coordinate of the current column.

[0056] Please see Figure 3 This application provides an embodiment of a method for processing suspected line defect areas, including: 301. Perform closing operations on the suspected defect rows and columns that are segmented in the horizontal and vertical directions, and then perform convex hull operations on them to connect the segmented suspected line defect regions into a whole.

[0057] If the projection vector value is greater than the specified threshold, it indicates that the row or column is a suspected defect row or column, and the area is defined as a suspected line defect area. Then, the terminal performs a closing operation on the suspected defects in the block in the horizontal and vertical directions within the specified range, and then performs a convex hull operation on them, so as to connect the various segmented line defects into a whole, that is, to perform the initial integration of repeated defect areas and reduce the amount of product calculation for subsequent defect integration.

[0058] Please see Figure 4 This application provides an embodiment of a novel method for detecting defects in a display screen, comprising: 401. Remove completely overlapping point and line defects using grayscale detection.

[0059] The terminal merges and integrates point and line defects in the Dot_defect set and the Line_defect set. First, it enables grayscale judgment to remove completely overlapping point and line defects, that is, it directly subtracts the line defect area from the point defect area to obtain the remaining point defect area.

[0060] 402. When the area corresponding to the remaining point defect is an empty area, the current point defect is determined to be empty, and the subsequent integration stage is skipped.

[0061] Next, the terminal verifies whether the remaining point defect region is an empty region. If the point defect region corresponding to the remaining point defect is an empty region, the subsequent integration stage is skipped, indicating that the point defect set Dot_defect is empty. Otherwise, the algorithm continues to execute the judgment.

[0062] 403. When the remaining point defect regions are not empty, calculate the number of point defect regions and the number of line defect regions corresponding to line defects.

[0063] When the remaining point defect regions are not empty, calculate the number of remaining point defect regions and the number of line defect regions, and calculate the mean and variance of each region, denoted as Dot_Mean, Dot_Std, Line_Mean and Line_Std respectively.

[0064] 404. Select the line defect area and the mean of the line defect area in sequence according to the index sequence.

[0065] 405. Calculate the shortest distance from all point defect areas to line defect areas, and generate a set of shortest distances for the regional contours.

[0066] 406. Record the point defect regions in the shortest distance set of the region outline that are less than the preset distance value, and generate the first point defect region set.

[0067] The terminal selects the line defect region and the line defect mean in sequence according to the index sequence. Then, it first calculates the distance S from the set of all point defects to the line defect, as shown below:

[0068] Where S is the shortest distance between region contours (point defect region and line defect contour), and not the center distance between contour regions. This is a line defect area. Let J be the j-th defect region in a set of n point defect regions.

[0069] Next, the terminal record... The distance between points and lines is less than the preset threshold for merging. index subscript Generate the first set of defect regions.

[0070] 407. Calculate the difference between the mean value of the point defect area and the mean value of the line defect area for each point defect area, and generate a set of differences.

[0071] 408. Record the point defect regions in the difference set that are less than the preset grayscale threshold for merging point lines, and generate the second point defect region set.

[0072] The terminal calculates the difference between the average grayscale of each defect area at the point defect and the average grayscale of the line defect, and records the difference when it is less than the set grayscale threshold for point-line merging. index subscript This generates a set of second-point defect regions.

[0073] 409. Generate an intersection point defect set based on the first point defect region set and the second point defect region set, and use the intersection point defect set to merge point and line defects to generate real point defects.

[0074] Traverse and search the above index subscripts and The intersection index of the same value is found to be intersection_IND, and the defect corresponding to the intersection_IND+1 index portion is removed from the initial point defect region.

[0075] Next, select the next line defect area and the average line defect from the line defects. Repeat steps 405 to 409 until all line defects are selected. The algorithm is then merged. Finally, the remaining point defects are the real point defects, and the algorithm is merged.

[0076] Please see Figure 5 This application provides another embodiment of a method for enhancing defect features, comprising: 501. Generate multiple kernel factors based on the detected defect type of the display screen under test.

[0077] 502. Use multiple kernel factors to process the difference image using a convolutional emphasis algorithm.

[0078] In existing technologies, contrast enhancement algorithms can be used to enhance defects in captured images of displays. However, these algorithms are more susceptible to interference than other methods, often requiring enhancement for specific types of defects in new displays, and processing the entire display area, which significantly increases the computational load for defect detection. To reduce computational load and improve defect detection accuracy, this embodiment processes the difference image after calculation. First, a new adaptive contrast enhancement kernel is constructed in the temporal domain. This operator generates different kernel factors (including weight, distance, polarity, edge filling mode, etc. in this embodiment) based on the detected defect type. This enhancement kernel performs convolutional emphasis processing on the captured image, avoiding interference from irrelevant neighborhoods and focusing more on the dissimilarity of a given region. Even under uneven lighting conditions, the resulting convolutional enhancement exhibits good robustness. The following is a simplified representation of this convolutional kernel:

[0079] Here, N represents the enhancement contrast distance. By setting different enhancement distances, different kernel factors are formed. By performing convolution operations between this adaptive filtering kernel matrix and the image, suspected defects in the image can be significantly enhanced, while non-defective parts of the background can be weakened.

[0080] In this embodiment, the adaptive filter processes the image using a convolutional difference algorithm, which can automatically generate different filter kernels. Even when the external lighting conditions are uneven, the convolutional enhancement effect obtained after filtering will also show good robustness.

[0081] Please see Figure 6 This application provides an embodiment of a preprocessing method for images captured by a display screen, comprising: 601. Set up a darkroom acquisition environment, fix the display screen to be tested on the acquisition platform, and turn on the strip side light source.

[0082] 602. Use an industrial camera to capture images of the unlit display screen under test to obtain dust removal images of the display screen under test.

[0083] 603. Perform dust removal processing on the captured images based on the dust removal images.

[0084] In this embodiment, after the terminal installs the acquisition platform in the intelligent manufacturing darkroom and adjusts the acquisition environment and other hardware of the industrial camera, it adjusts the strip side light source and takes pictures of the display screen under test through the industrial camera to obtain a dust removal image. Subsequently, the dust removal image is used to perform dust removal processing on all the captured images. It is necessary to detect dust defects in the dust removal image and then remove dust defects from all the captured images to reduce the impact of subsequent detection of display defects.

[0085] Please see Figure 7 This application provides an embodiment of a method for correcting images captured by a display screen, comprising: 701. Calculate the segmentation threshold based on the captured image to generate the maximum segmentation threshold.

[0086] In this embodiment, the Otsu threshold segmentation algorithm is used to effectively separate the target region and background of the captured image. First, the segmentation threshold is calculated based on the captured image to generate the maximum segmentation threshold, as shown in the following algorithm formula:

[0087] in: and These represent the proportions of the foreground and background in the image, respectively. and These are the average gray levels of the image foreground and background, respectively. Let the average gray level of the image be the value of the image. When the above formula... The segmentation threshold is obtained when the maximum value is reached.

[0088] 702. Perform threshold segmentation on the captured image based on the maximum segmentation threshold to generate the minimum envelope of the segmented region.

[0089] The terminal uses the aforementioned maximum segmentation threshold. The captured image is segmented and judged, specifically by calculating the minimum bounding box of the segmented region (the effective display area of ​​the display screen under test in the captured image) and determining the coordinates of each corner point of the minimum bounding box.

[0090] 703. Map the corner coordinates of the minimum envelope to the reference coordinates, construct a system of matrix equations, and generate the homography matrix.

[0091] The terminal maps the coordinates of the image edge corner points (the coordinates of each corner point of the minimum envelope) to the ideal coordinate point pairs one by one, constructs a system of matrix equations, and uses SVD decomposition to calculate the homography matrix:

[0092] in, That is, the homography matrix. These are the image transformation point pairs before and after correction.

[0093] 704. Use a homography matrix to perform matrix transformation on the captured image to correct the captured image.

[0094] After the terminal obtains the homography matrix, it stores the homography transformation matrix and inverse matrix in the dictionary dataset based on the above mapping point pairs. It then parses the dictionary dataset by referencing the illuminated images of each gray level to obtain the homography matrix, and finally performs matrix transformation on the captured image.

[0095] Please see Figure 8 This application provides an embodiment of a filtering method for images captured by a display screen, comprising: 801. Calculate the spatial domain standard deviation and the range standard deviation of the captured image.

[0096] To eliminate noise in captured images, the terminal needs to filter the images. Specifically, this involves algorithmic detection of the corrected images. First, nonlinear filtering is applied to the captured images. This algorithm effectively combines the principles of pixel spatial proximity and pixel value similarity, eliminating some noise while preserving edge details. The spatial domain standard deviation of the captured images is calculated first. Sum range standard deviation .

[0097] 802. Calculate the spatial domain weight and the value domain weight based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image.

[0098] Next, the terminal calculates the spatial domain weights based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image. Sum range weight .

[0099] The formula for calculating spatial domain weights is as follows:

[0100] The formula for calculating the range weight is as follows:

[0101] in, It is the center coordinate point of the template window for capturing images. These are the coordinates of other points within the template window where the image was captured. It is the standard deviation of the spatial domain. It is the standard deviation of the range. , It is the pixel value corresponding to the coordinates of the template window where the image was captured.

[0102] 803. Construct nonlinear filter weights based on spatial domain weights and value domain weights.

[0103] After the terminal obtains the spatial domain weights and the range weights, it then uses the spatial domain weights... Sum range weight The nonlinear filter weights are constructed using the following formula:

[0104] 804. Perform nonlinear filtering on the captured image based on the nonlinear filtering weights.

[0105] Once the nonlinear filter weights are calculated, they can be used. Nonlinear filtering is applied to the captured images.

[0106] Please see Figure 9 This application provides an embodiment of a device for detecting dot and line defects on a display screen image, comprising: The first acquisition unit 901 is used to acquire images of the lit display screen under test using an industrial camera to obtain captured images of the display screen under test. The selection unit 902 is used to randomly select multiple positioning pixels from the captured image and generate a positioning circle region based on the multiple positioning pixels; The first calculation unit 903 is used to calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixels, and to calculate the average gray-level gradient vector based on the calculated multiple gray-level gradient vectors. The first generation unit 904 is used to calculate the Euclidean distance between the sample pixel point located at the mean of the gray-level gradient vector and each localized pixel point, so as to generate multiple abnormal contour regions based on multiple Euclidean distances. The second calculation unit 905 is used to perform region filtering on the captured image based on multiple abnormal contour regions, calculate the average gray value of the remaining regions, and generate a difference image based on the average gray value of the remaining regions and the captured image. The second generation unit 906 is used to pre-segment and extract the defect region from the difference image to generate multiple defect regions to be tested. The third generation unit 907 is used to process multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions. The fourth generation unit 908 is used to perform feature screening on suspected line defect areas and generate a set of point defects and line defects.

[0107] Optionally, the third generation unit 907 includes: performing row compression and column compression on the two-dimensional image matrix information of multiple defect regions to be tested, respectively, to generate a set of one-dimensional grayscale statistical values; When the one-dimensional grayscale statistical value is greater than the preset threshold, the row and column corresponding to the one-dimensional grayscale statistical value are determined as suspected defect rows and columns; The suspected defect rows are integrated to generate suspected line defect areas.

[0108] Optionally, after the third generation unit 907 and before the fourth generation unit 908, the detection device further includes: a connection processing unit, used to perform a closing operation on the suspected defect rows and columns with blocks in the horizontal and vertical directions and then perform a convex hull operation on them, so as to connect the various segmented suspected line defect areas into a whole.

[0109] Optionally, after the fourth generation unit 908, the detection device further includes: a removal unit for removing completely overlapping point defects and line defects by grayscale detection; a determination unit for determining that the current point defect is empty and skipping the subsequent integration stage when the point defect region corresponding to the remaining point defect is an empty region; a third calculation unit for calculating the number of point defect regions and the number of line defect regions corresponding to the line defects when the point defect region corresponding to the remaining point defect is not an empty region; a selection unit for sequentially selecting line defect regions and the average of line defect regions according to the index sequence; and a fifth generation unit for calculating the shortest distance of the regional contours from all point defect regions to the line defect regions. The system generates a set of shortest distances for the region contours; a sixth generation unit records point defect regions in the set of shortest distances for the region contours that are less than a preset distance value, generating a first set of point defect regions; a seventh generation unit calculates the difference between the average point defect region value and the average line defect region value for each point defect region, generating a set of differences; an eighth generation unit records point defect regions in the set of differences that are less than a preset grayscale threshold for merging point and line defects, generating a second set of point defect regions; and a ninth generation unit generates a set of intersection point defects based on the first and second set of point defect regions, and uses the set of intersection point defects to merge point and line defects to generate realistic point defects.

[0110] Optionally, after the second calculation unit 905 and before the second generation unit 906, the detection device further includes: a tenth generation unit, used to generate multiple kernel factors according to the detection defect type of the display screen under test; and a convolution unit, used to perform convolution emphasis algorithm processing on the difference image using the multiple kernel factors respectively.

[0111] Optionally, after the first acquisition unit 901 and before the selection unit 902, the detection device further includes: an activation unit for setting the darkroom acquisition environment, fixing the display screen under test on the acquisition platform, and activating the strip side light source; a second acquisition unit for acquiring images of the unlit display screen under test using an industrial camera to obtain a dust removal image of the display screen under test; and a dust removal process for performing dust removal processing on the captured image based on the dust removal image.

[0112] Optionally, after the first acquisition unit 901 and before the selection unit 902, the detection device further includes: a fourth calculation unit, used to calculate a segmentation threshold based on the captured image to generate a maximum segmentation threshold; a segmentation unit, used to perform threshold segmentation processing on the captured image based on the maximum segmentation threshold to generate the minimum envelope of the segmented region; a construction unit, used to map the corner coordinates of the minimum envelope to the reference coordinates to construct a system of matrix equations and generate a homography matrix; and a correction unit, used to perform matrix transformation on the captured image using the homography matrix to correct the captured image.

[0113] Optionally, after the first acquisition unit 901 and before the selection unit 902, the detection device further includes: a fifth calculation unit for calculating the spatial domain standard deviation and the value domain standard deviation of the captured image; a sixth calculation unit for calculating the spatial domain weight and the value domain weight based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image; a construction unit for constructing nonlinear filtering weights based on the spatial domain weights and the value domain weights; and a filtering unit for performing nonlinear filtering on the captured image based on the nonlinear filtering weights.

[0114] Please see Figure 10 This application provides an electronic device, including: Processor 1001, memory 1002, input / output unit 1003 and bus 1004.

[0115] The processor 1001 is connected to the memory 1002, the input / output unit 1003, and the bus 1004.

[0116] The memory 1002 stores a program, and the processor 1001 calls the program to execute it, such as... Figure 1 , Figure 2 and Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 and Figure 8 The detection method in [the text].

[0117] This application provides a computer-readable storage medium on which a program is stored, and when the program is executed on a computer, it performs the following... Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 and Figure 8 The detection method in [the text].

[0118] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0122] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for detecting dot and line defects on a display screen image, characterized in that, include: An industrial camera is used to capture images of the illuminated display screen under test to obtain images of the display screen under test. Multiple positioning pixels are randomly selected from the captured image, and a positioning circular region is generated based on the multiple positioning pixels; Calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixel, and calculate the mean of the gray-level gradient vectors based on the calculated multiple gray-level gradient vectors; The Euclidean distance between the sample pixel at the mean of the gray-level gradient vector and each of the localized pixels is calculated respectively, so as to generate multiple abnormal contour regions based on the multiple Euclidean distances; The captured image is subjected to region filtering based on multiple abnormal contour regions, and the average gray value of the remaining regions is calculated. A difference image is generated based on the average gray value of the remaining regions and the captured image. The difference image is pre-segmented and extracted to generate multiple defect regions to be tested; An array projection estimation algorithm is applied to multiple defect regions to be tested to generate suspected line defect regions; Feature filtering is performed on the suspected line defect areas to generate a set of point defects and line defects; Completely overlapping point and line defects are removed using grayscale detection. When the remaining point defect corresponds to an empty point defect area, the current point defect is determined to be empty, and the subsequent integration stage is skipped. When the remaining point defect regions are not empty, calculate the number of point defect regions and the number of line defect regions corresponding to line defects. Select the line defect areas sequentially according to the index sequence; Calculate the shortest distance from all point defect regions to line defect regions, and generate a set of shortest distances for region contours. Record the point defect regions in the set of shortest distances to the region contour that are less than a preset distance value, and generate the first set of point defect regions. Calculate the difference between the mean value of the point defect area and the mean value of the line defect area for each point defect area, and generate a set of differences. Record the point defect regions in the difference set that are less than the preset grayscale threshold for merging point lines, and generate a second set of point defect regions. An intersection point defect set is generated based on the first point defect region set and the second point defect region set, and the intersection point defect set is used to merge point and line defects to generate real point defects.

2. The detection method according to claim 1, characterized in that, The step of processing multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions specifically includes: The two-dimensional image matrix information of the plurality of defect regions to be tested is compressed by row and column respectively to generate a set of one-dimensional grayscale statistical values. When the one-dimensional grayscale statistical value is greater than the preset threshold, the row and column corresponding to the one-dimensional grayscale statistical value are determined as suspected defect rows and columns; The suspected defect rows and columns are integrated to generate suspected line defect regions.

3. The detection method according to claim 2, characterized in that, After the step of processing multiple defect regions to be tested using an array projection estimation algorithm to generate suspected line defect regions, and before the step of performing feature filtering on the suspected line defect regions to generate a set of point defects and line defects, the detection method further includes: The suspected defect rows and columns that are segmented are first closed in the horizontal and vertical directions, and then convex hull is applied to them to connect the segmented suspected line defect regions into a whole.

4. The detection method according to claim 1, characterized in that, After the steps of performing region filtering on the captured image based on multiple abnormal contour regions, calculating the average gray value of the remaining regions, generating a difference image based on the average gray value of the remaining regions and the captured image, and before the steps of pre-segmenting and extracting defect regions from the difference image to generate multiple defect regions to be tested, the detection method further includes: Multiple kernel factors are generated based on the detected defect type of the display screen under test; The difference image is processed by a convolutional emphasis algorithm using the multiple kernel factors.

5. The detection method according to any one of claims 1 to 4, characterized in that, After the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular region based on the multiple positioning pixels, the detection method further includes: Set up a darkroom acquisition environment, fix the display screen to be tested on the acquisition platform, and turn on the bar side light source; An industrial camera is used to capture images of the unlit display screen under test, thereby obtaining dust removal images of the display screen under test. The captured image is processed to remove dust based on the dust removal image.

6. The detection method according to any one of claims 1 to 4, characterized in that, After the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular region based on the multiple positioning pixels, the detection method further includes: The segmentation threshold is calculated based on the captured image to generate the maximum segmentation threshold. The captured image is subjected to threshold segmentation based on the maximum segmentation threshold to generate the minimum envelope of the segmented region. The corner coordinates of the minimum envelope are mapped to the reference coordinates to construct a system of matrix equations and generate a homography matrix. The homography matrix is ​​used to perform matrix transformation on the captured image to correct the captured image.

7. The detection method according to any one of claims 1 to 4, characterized in that, After the step of acquiring an image of the illuminated display screen using an industrial camera, and before the step of randomly selecting multiple positioning pixels from the captured image and generating a positioning circular region based on the multiple positioning pixels, the detection method further includes: Calculate the spatial domain standard deviation and the range standard deviation of the captured image; The spatial domain weight and the value domain weight are calculated based on the pixel coordinates, pixel grayscale, spatial domain standard deviation, and value domain standard deviation of the captured image. Construct nonlinear filter weights based on the spatial domain weights and value domain weights; The captured image is subjected to nonlinear filtering based on the nonlinear filtering weights.

8. A device for detecting dot and line defects on a display screen image, characterized in that, include: The first acquisition unit is used to acquire images of the illuminated display screen under test using an industrial camera, so as to obtain the captured images of the display screen under test. The selection unit is used to randomly select multiple positioning pixels from the captured image and generate a positioning circular region based on the multiple positioning pixels; The first calculation unit is used to calculate the gray-level gradient vector from multiple sample pixels within the positioning circle to the corresponding positioning pixels, and to calculate the mean value of the gray-level gradient vector based on the calculated multiple gray-level gradient vectors. The first generation unit is used to calculate the Euclidean distance between the sample pixel at the mean of the gray-level gradient vector and each of the localized pixels, so as to generate multiple abnormal contour regions based on the multiple Euclidean distances. The second calculation unit is used to perform region filtering processing on the captured image based on multiple abnormal contour regions, calculate the average gray value of the remaining regions, and generate a difference image based on the average gray value of the remaining regions and the captured image. The second generation unit is used to pre-segment and extract the defect region of the difference image to generate multiple defect regions to be tested. The third generation unit is used to process the array projection estimation algorithm on the multiple defect regions to be tested in order to generate suspected line defect regions. The fourth generation unit is used to perform feature filtering on the suspected line defect area and generate a set of point defects and line defects. The removal unit is used to remove completely overlapping point and line defects through grayscale detection; The determination unit is used to determine that the current point defect is empty and skip the subsequent integration stage when the point defect area corresponding to the remaining point defect is an empty area. The third calculation unit is used to calculate the number of point defect regions and the number of line defect regions corresponding to line defects when the remaining point defect regions are not empty regions. The selection unit is used to sequentially select line defect areas according to the index sequence. The fifth generation unit is used to calculate the shortest distance of the regional contour from all point defect regions to line defect regions and generate a set of shortest distances of regional contours. The sixth generation unit is used to record point defect regions in the set of shortest distances to the region contour that are less than a preset distance, and to generate the first set of point defect regions. The seventh generation unit is used to calculate the difference between the mean value of the point defect area and the mean value of the line defect area for each point defect area, and generate a set of differences. The eighth generation unit is used to record point defect regions in the difference set that are less than the preset grayscale threshold for merging point lines, and to generate a second set of point defect regions. The ninth generation unit is used to generate an intersection point defect set based on the first point defect region set and the second point defect region set, and to use the intersection point defect set to merge point and line defects to generate real point defects.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a program that, when executed on a computer, performs the detection method as described in any one of claims 1 to 7.

Images