An automatic identification method for component microcrack defects

Abstract

The application relates to the technical field of image processing, in particular to a kind of automatic identification method of component hidden crack defect, method includes: obtaining the gradient value and gradient direction of each pixel point in battery piece EL image, the average gradient value of each column of pixel points is calculated to obtain average curve, the column where the peak point of average curve is located is taken as grid line column;According to the gradient direction of each pixel point on the grid line column, the abnormal grid line point is positioned, the gradient value of abnormal grid line point is corrected, the gradient value sequence of grid line column is obtained, each gradient value in gradient value sequence is subtracted gradient mean value, and the local gradient distribution of grid line column is obtained;According to the local gradient distribution of each grid line column, global gradient distribution is constructed;The difference graph of EL image real-time gradient distribution and global gradient distribution is threshold segmented, and hidden crack defect is obtained.Through the technical scheme of the application, the hidden crack defect on photovoltaic module can be accurately identified.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an automatic identification method for hidden crack defects in components. Background Technology

[0002] Photovoltaic modules, also known as photovoltaic cells, are important equipment for photovoltaic power generation. Microcracks in photovoltaic modules can cause partial damage to the cells or loss of current, and may even lead to cell open circuits and affect the power output of the module. Therefore, the detection of microcracks in photovoltaic modules has always been a difficult problem in the field of photovoltaic power generation.

[0003] In existing technologies, EL (Electro-luminescent) imaging is often used to detect defects in solar panels. EL imaging utilizes the electroluminescence principle of crystalline silicon, combined with a high-resolution infrared camera to capture near-infrared images of crystalline silicon. Software then analyzes the acquired images to detect whether there are hidden cracks or defects in the solar panel.

[0004] Currently, patent application CN117576093A discloses a method for detecting microcracks in solar cells based on EL images, including: acquiring an original EL image set; stacking the original EL image set at different resolution scales to generate a multi-scale image pyramid; enhancing the image resolution of the multi-scale image pyramid to generate a super-resolution corrected image; confirming the connected regions of the super-resolution corrected image to generate image connected region data; erasing the main grid lines of the image connected region data to obtain a standard image of the EL region; and using a pixel anomaly score evaluation formula to analyze the microcrack region of the standard image of the EL region to generate a microcrack region projection map.

[0005] The above method obtains a standard image of the EL region by erasing the main grid lines of the image connected area data. However, it ignores the gradient changes caused by illumination interference during the acquisition of the EL image, which means that there is also noise in the standard image of the EL region, resulting in inaccurate identification of hidden crack defects. Summary of the Invention

[0006] To address the technical problem of inaccurate identification of microcrack defects, this application provides an automatic identification method for microcrack defects in photovoltaic modules, which can accurately identify microcrack defects on photovoltaic modules.

[0007] In a first aspect, this application provides an automatic identification method for microcrack defects in a component. The identification method includes: acquiring the gradient value and gradient direction of each pixel in an EL image of a battery cell; calculating the average gradient value of each column of pixels to obtain an average curve; and taking the column containing the peak point of the average curve as a grid line column; performing anomaly detection on the gradient direction of each pixel in any grid line column to obtain abnormal grid line points; correcting the gradient value of the abnormal grid line points to obtain a gradient value sequence of the grid line column; subtracting the gradient mean from each gradient value in the gradient value sequence to obtain the local gradient distribution of the grid line column; constructing a global gradient distribution based on the local gradient distribution of each grid line column; acquiring the real-time gradient distribution of the EL image; and performing threshold segmentation on the difference map between the real-time gradient distribution and the global gradient distribution to obtain the microcrack defect.

[0008] Preferably, obtaining abnormal gate line points includes: calculating the absolute value of the difference in gradient direction between each pixel point on any gate line column and the previous adjacent pixel point; and marking the pixel point as an abnormal gate line point in response to the absolute value of the difference between any pixel point being greater than the average value of the absolute values ​​of the differences between all pixel points.

[0009] When a hidden crack defect exists in any pixel on the grid line column, the gradient direction and gradient value generated by the hidden crack defect will cause a large change in the gradient direction of the pixel. Therefore, abnormal grid line points can be accurately located based on the gradient direction of each pixel on the grid line column.

[0010] Preferably, correcting the gradient values ​​of abnormal gate line points includes: obtaining the left and right neighboring points of the abnormal gate line point; taking the point with the maximum gradient value among the left neighboring points as the left hidden crack point, taking the point with the maximum gradient value among the right neighboring points as the right hidden crack point, taking the average gradient direction of the left and right hidden crack points as the hidden crack direction of the abnormal gate line point, and taking the average gradient value of the left and right hidden crack points as the hidden crack gradient of the abnormal gate line point; the horizontal gradient value of the abnormal gate line point is the sum of the first horizontal gradient and the second horizontal gradient, the vertical gradient value of the abnormal gate line point is the sum of the first vertical gradient and the second vertical gradient, and the sum of the squares of the first horizontal gradient and the second vertical gradient is equal to the hidden crack gradient, the angle formed by the first horizontal gradient and the second vertical gradient is the hidden crack direction, and completing the construction of the equation system; solving the equation system to obtain the second horizontal gradient and the second vertical gradient, and taking the sum of the squares of the second horizontal gradient and the second vertical gradient as the correction result.

[0011] The gradient value and gradient direction of abnormal grid line points can be regarded as the superposition result of microcrack defect features and grid line features. The process of correcting the gradient value of abnormal grid line points is to decompose the gradient value of abnormal grid line points into microcrack defect feature gradient value and grid line feature gradient value, thereby eliminating the influence of microcrack defects on the gradient value sequence and enabling the gradient value sequence to accurately reflect the gradient value changes caused by uneven illumination.

[0012] Preferably, the left neighbor points are the preset number of pixels on the upper and lower sides of the row where the abnormal gate line point is located, in the column adjacent to the left of the abnormal gate line point.

[0013] Preferably, abnormal grid line points The corresponding system of equations is:

[0014] ;

[0015] in, and These are the first horizontal gradient and the first vertical gradient, respectively. and These are the second horizontal gradient and the second vertical gradient, respectively. and These are abnormal grid line points. The horizontal and vertical gradient values, and The gradient and direction of the microcrack are given.

[0016] Preferably, the global gradient distribution includes the local gradient distribution of each grid line column and the local gradient distribution of each pixel column between adjacent grid line columns. Methods for obtaining local gradient distribution include: obtaining each grid line column and pixel column. The horizontal distance, whereby the horizontal distance is between the grid line column and the pixel column. The absolute value of the difference in the number of columns between them; the difference between the proportion of 1 and abnormal grid points in any grid line column is used as the confidence level of each grid line column; the product of the confidence level and the Gaussian weight of the horizontal distance is used as the comprehensive weight, and the local gradient distribution of each grid line column is weighted and summed to obtain the pixel column. The local gradient distribution.

[0017] Preferably, pixel column Local gradient distribution for:

[0018] ; The number of grid lines, For grid lines Horizontal distance, For Gaussian kernel function, For grid lines Confidence level, This is the sum of the confidence levels for each grid line column. For grid lines The local gradient distribution.

[0019] The number of abnormal gate line points in the combined gate line column, and the number of gate line columns and pixel columns. The horizontal distance determines the comprehensive weight of each grid line column, accurately obtains the local gradient distribution of each pixel column, and thus obtains the accurate global gradient distribution.

[0020] Preferably, obtaining the real-time gradient distribution of the EL image includes: calculating the average gradient value of all pixels in each column of the EL image, and subtracting the average gradient value of the column from the gradient value of the pixel to obtain the real-time gradient distribution.

[0021] Preferably, the threshold segmentation includes: determining a segmentation threshold for the difference map using the maximum inter-class variance method, and performing threshold segmentation on the difference map using the segmentation threshold.

[0022] Preferably, pixels The gradient value is: Pixel The gradient direction is: ;in, For multiple scales, For scale Next pixel gradient value, For scale Next pixel The gradient direction.

[0023] Larger-scale Sobel operators can capture the features of subtle cracks around pixels. To capture the features of subtle cracks around each pixel, Sobel operators of multiple scales are used to calculate the gradient value and gradient direction of each pixel.

[0024] The technical solution of this application has the following beneficial technical effects:

[0025] EL images of solar cells are acquired. The grid line region in the solar cell exhibits significant gradient features, while the pixel value changes relatively smoothly. Therefore, the average gradient value of each column of pixels is calculated to obtain an average curve, and the peak point of the average curve is used to accurately locate the grid line column in the solar cell. Considering the significant gradient features of the grid line region, anomaly detection is performed on the gradient direction of each pixel in the grid line column to identify abnormal grid line points. The gradient value of the abnormal grid line points is considered as the superposition result of the grid line gradient features and the gradient features of hidden crack defects. The gradient values ​​of the abnormal grid line points are corrected to eliminate the influence of the gradient features of hidden crack defects, resulting in a gradient value sequence of the grid line column. This allows for accurate acquisition of the local gradient distribution of the grid line column. This local gradient distribution accurately reflects the gradient value changes caused by uneven illumination in the absence of hidden crack defects. A global gradient distribution is constructed based on the local gradient distribution of each grid line column, and the difference map between the real-time gradient distribution of the EL image and the global gradient distribution is thresholded to identify hidden crack defects. This eliminates the influence of noise information such as uneven illumination on the identification of hidden crack defects, achieving accurate identification of hidden crack defects. Attached Figure Description

[0026] Figure 1 This is a flowchart of an automatic identification method for component microcrack defects according to an embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] It should be understood that when the terms "first," "second," etc., are used in the claims, description, and drawings of this application, they are only used to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" used in the description and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0029] According to a first aspect of this application, this application provides an automatic identification method for microcrack defects in photovoltaic modules. Figure 1 This is a flowchart of an automatic identification method for microcrack defects in components according to an embodiment of this application. Figure 1 As shown, the automatic identification method for component microcrack defects includes steps S101 to S104, which are described in detail below.

[0030] S101, obtain the gradient value and gradient direction of each pixel in the EL image of the battery cell, calculate the average gradient value of each column of pixels to obtain the average curve, and take the column where the peak point of the average curve is located as the grid line column.

[0031] In one embodiment, acquiring an EL image of a photovoltaic cell can be accomplished by scanning or photographing the photovoltaic cell using an EL image detection device or an image acquisition device. The Sobel operator is then used to detect each pixel in the EL image, obtaining the gradient value and gradient direction of each pixel.

[0032] Specifically, the Sobel operator includes a horizontal Sobel operator and a vertical Sobel operator; the horizontal Sobel operator with a scale of 3×3 is... It can acquire 3×3 scale pixels. Gradient value in the horizontal direction The Sobel operator with a vertical scale of 3×3 is: It can acquire pixels on a 3×3 scale. gradient value in the vertical direction Then the number of pixels at a 3×3 scale gradient value Pixels at a 3×3 scale gradient direction .

[0033] The scale of the Sobel operator mentioned above is 3×3, for each pixel. The surrounding strong edges have a noticeable response. Since hidden cracks in EL images are often minute cracks, it is necessary to obtain pixel points. To analyze the characteristics of the surrounding fine cracks, this application uses 3×3 and 5×5 scale Sobel operators to calculate the gradient value and gradient direction of each pixel. Specifically, the 5×5 scale Sobel operator for the horizontal direction is... The Sobel operator with a vertical scale of 5×5 is: .

[0034] Specifically, pixels The gradient value is: Pixel The gradient direction is: ; For multiple scales, For scale Next pixel gradient value, For scale Next pixel The gradient direction. Here The value can be 2, including two scales: 3×3 and 5×5.

[0035] Thus, after obtaining the gradient value and gradient direction of each pixel, the average gradient value of each column of pixels is calculated to obtain the average curve. The horizontal axis of the average curve is each column in the EL image, and the vertical axis is the average gradient value of each column. Peak point detection is performed on the average curve, and the column where the peak point is located is taken as the grid line column.

[0036] Thus, since the cell area in a photovoltaic module is a region where pixel values ​​change gradually, while the grid line area is a region where pixel values ​​change drastically and has obvious gradient characteristics, the grid line can be accurately positioned by statistically analyzing the average gradient value of each column.

[0037] S102, perform anomaly detection on the gradient direction of each pixel on any grid line column to obtain abnormal grid line points, and correct the gradient values ​​of the abnormal grid line points to obtain the gradient value sequence of the grid line column. Subtract the gradient mean from each gradient value in the gradient value sequence to obtain the local gradient distribution of the grid line column.

[0038] In one embodiment, compared to the cell region, the grid line region has obvious gradient characteristics. Environmental noise such as uneven illumination has little impact on the gradient characteristics of the grid line region, while microcracks have a greater impact on the gradient characteristics of the grid line region. Therefore, it is possible to accurately locate abnormal grid line points on the grid line column caused by microcracks. Thus, taking the grid line column as the research object, the local gradient distribution of the grid line column is obtained. This local gradient distribution can effectively eliminate the influence of microcracks and accurately reflect the gradient value changes on the grid line column caused by uneven illumination.

[0039] After obtaining the grid lines on the photovoltaic module, the gradient values ​​on each grid line are analyzed. Under the premise that there is no uneven illumination or microcrack defects, the gradient values ​​of each pixel on the grid line should be consistent. Therefore, for any grid line, after eliminating the influence of microcrack defects on the gradient values ​​on the grid line, the gradient value changes caused by uneven illumination on the grid line can be obtained.

[0040] Understandably, microcracks often manifest as minute cracks, and small gradient values ​​can lead to misidentification of microcracks. Therefore, it is necessary to obtain the gradient value changes caused by uneven illumination in the EL image. After eliminating these changes, accurate identification of microcracks can be achieved.

[0041] When any pixel on the grid line column has a hidden crack defect, the gradient direction and gradient value generated by the hidden crack defect will cause the gradient direction and gradient value of the pixel to change. Therefore, abnormal grid line points can be accurately located based on the gradient direction of each pixel on the grid line column. The abnormal grid line points are the pixels on the grid line column that have hidden crack defects.

[0042] Specifically, obtaining abnormal gate line points includes: calculating the absolute value of the difference in gradient direction between each pixel point on any gate line column and the previous adjacent pixel point; in response to the absolute value of the difference between any pixel point being greater than the average of the absolute values ​​of the differences between all pixel points, marking the pixel point as an abnormal gate line point.

[0043] The gradient value and gradient direction of abnormal grid line points can be regarded as the superposition result of microcrack defect features and grid line features. The process of correcting the gradient value of abnormal grid line points is the process of splitting the gradient value of abnormal grid line points into microcrack defect feature gradient value and grid line feature gradient value. The grid line feature gradient value is the correction result.

[0044] Specifically, correcting the gradient values ​​of abnormal gate line points includes: obtaining the left and right neighboring points of the abnormal gate line point; taking the point with the maximum gradient value among the left neighboring points as the left hidden crack point, taking the point with the maximum gradient value among the right neighboring points as the right hidden crack point, taking the average gradient direction of the left and right hidden crack points as the hidden crack direction of the abnormal gate line point, and taking the average gradient value of the left and right hidden crack points as the hidden crack gradient of the abnormal gate line point; the horizontal gradient value of the abnormal gate line point is the sum of the first and second horizontal gradients, the vertical gradient value of the abnormal gate line point is the sum of the first and second vertical gradients, and the sum of the squares of the first and second vertical gradients is equal to the hidden crack gradient, the angle formed by the first and second horizontal gradients is the hidden crack direction, and completing the construction of the equation system; solving the equation system to obtain the second horizontal gradient and the second vertical gradient, and taking the sum of the squares of the second horizontal gradient and the second vertical gradient as the correction result.

[0045] The left neighbor of the abnormal grid line point is a preset number of pixels above and below the row where the abnormal grid line point is located in the column to the left of the abnormal grid line point; similarly, the right neighbor of the abnormal grid line point is a preset number of pixels above and below the row where the abnormal grid line point is located in the column to the right of the abnormal grid line point; the preset number is 2.

[0046] Wherein, the first horizontal gradient and the first vertical gradient are the horizontal and vertical gradient values ​​of the microcrack defect feature, respectively; the second horizontal gradient and the second vertical gradient are the horizontal and vertical gradient values ​​of the grid line feature, respectively; and abnormal grid line points... The corresponding system of equations is:

[0047] ;

[0048] in, and These are the first horizontal gradient and the first vertical gradient, respectively. and These are the second horizontal gradient and the second vertical gradient, respectively. and These are abnormal grid line points. The horizontal and vertical gradient values, and Let represent the microcrack gradient and the microcrack direction. In the above system of equations, , , and Since all quantities are known, the second horizontal gradient can be solved. Second vertical gradient The value of is then used to obtain the correction result of the gradient value of the abnormal gate line point.

[0049] After correcting the gradient values ​​of all abnormal grid points on any grid line column, the gradient value sequence of the grid line column is obtained. The gradient mean in the gradient value sequence is calculated, and the gradient mean is subtracted from each gradient value in the gradient value sequence to obtain the local gradient distribution of the grid line column. The local gradient distribution can reflect the gradient value changes on the grid line column caused by uneven illumination.

[0050] Thus, based on the grid lines with obvious gradient characteristics in photovoltaic modules as the research object, the gradient values ​​of abnormal grid line points are corrected to obtain a gradient value sequence. This gradient value sequence eliminates the influence of microcrack defects on the gradient values ​​and can accurately reflect the gradient value changes caused by uneven illumination, thereby obtaining the local gradient distribution of each grid line column.

[0051] S103, construct the global gradient distribution based on the local gradient distribution of each grid line column.

[0052] In one embodiment, after obtaining the local gradient distribution of each grid line column, the local gradient distribution of each pixel column within the cell region between adjacent grid line columns is obtained, thereby obtaining the global gradient distribution of the cell.

[0053] Specifically, the global gradient distribution includes the local gradient distribution of each grid line column, and the local gradient distribution of each pixel column between adjacent grid line columns. Methods for obtaining local gradient distribution include: obtaining each grid line column and pixel column. The horizontal distance, whereby the horizontal distance is between the grid line column and the pixel column. The absolute value of the difference in the number of columns between them; the difference between the proportion of 1 and abnormal grid points in any grid line column is used as the confidence level of each grid line column; the product of the confidence level and the Gaussian weight of the horizontal distance is used as the comprehensive weight, and the local gradient distribution of each grid line column is weighted and summed to obtain the pixel column. The local gradient distribution.

[0054] Among them, pixel column Local gradient distribution for:

[0055] ; The number of grid lines, For grid lines Horizontal distance, For Gaussian kernel function, For grid lines Confidence level, This is the sum of the confidence levels for each grid line column. For grid lines The local gradient distribution.

[0056] The Gaussian kernel function is a well-known technique in Gaussian filtering and will not be described in detail here. In this embodiment, the standard deviation of the Gaussian kernel is equal to the distance between adjacent grid lines.

[0057] Understandably, the more abnormal grid points a grid line contains, the more pixels in that grid line need gradient value correction. To ensure accurate acquisition of the local gradient distribution of each pixel column, a smaller weight is assigned to that grid line column, corresponding to... If a grid line column and a pixel column The larger the horizontal distance, the greater the distance between each grid line column and the pixel column. The smaller the influence of the local gradient distribution, the smaller the weight should be assigned to that grid line column, i.e., corresponding to... The number of abnormal gate line points in the combined gate line column, and the number of gate line columns and pixel columns. The horizontal distance determines the overall weight of each grid line column. This allows for the accurate acquisition of the local gradient distribution of each pixel column, thereby obtaining the accurate global gradient distribution.

[0058] S104: Obtain the real-time gradient distribution of the EL image, and perform threshold segmentation on the difference map between the real-time gradient distribution and the global gradient distribution to obtain the hidden crack defect.

[0059] In one embodiment, obtaining the real-time gradient distribution of the EL image includes: calculating the average gradient value of all pixels in each column of the EL image, and subtracting the average gradient value of the column from the gradient value of each pixel to obtain the real-time gradient distribution.

[0060] Understandably, the real-time gradient distribution is the result of the combined effects of uneven illumination and microcrack defects, while the global gradient distribution eliminates the influence of uneven illumination on the gradient value. Therefore, by obtaining the difference map between the real-time gradient distribution and the global gradient distribution, and using the Otsu's method to determine the segmentation threshold of the difference map, and then using this segmentation threshold to perform threshold segmentation on the difference map, microcrack defects can be obtained, thus realizing the automatic identification of microcrack defects in photovoltaic modules.

[0061] It should be noted that for other specifications of solar cells, there will also be grid line areas in the row direction, that is, there will be grid line rows in the solar cell EL image. In other embodiments, the average gradient value of each row of pixels can be calculated to obtain the average curve, and then the grid line rows in the solar cell EL image can be located. Based on the local gradient distribution of the grid line rows, a global gradient distribution can be constructed to achieve automatic identification of hidden crack defects.

[0062] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0063] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for automatic identification of assembly crack defects, characterized in that, The identification method comprises: Obtaining gradient values and gradient directions of each pixel point in the battery piece EL image, calculating average gradient values of each column of pixel points to obtain an average curve, and taking a column where a peak point of the average curve is located as a grid line column; Performing abnormality detection on gradient directions of each pixel point in any grid line column to obtain abnormal grid line points, and correcting gradient values of the abnormal grid line points to obtain a gradient value sequence of the grid line column, subtracting gradient mean values from each gradient value in the gradient value sequence to obtain a local gradient distribution of the grid line column; Constructing a global gradient distribution according to the local gradient distributions of each grid line column; Obtaining a real-time gradient distribution of the EL image, performing threshold segmentation on a difference value graph of the real-time gradient distribution and the global gradient distribution to obtain a hidden crack defect; The correction of the gradient value of the abnormal grid line point comprises: obtaining a left neighborhood point and a right neighborhood point of the abnormal grid line point; taking a point with the maximum gradient value in the left neighborhood point as a left hidden crack point, and taking a point with the maximum gradient value in the right neighborhood point as a right hidden crack point; taking an average gradient direction of the left hidden crack point and the right hidden crack point as a hidden crack direction of the abnormal grid line point, and taking an average gradient value of the left hidden crack point and the right hidden crack point as a hidden crack gradient of the abnormal grid line point; a horizontal gradient value of the abnormal grid line point is a sum of a first horizontal gradient and a second horizontal gradient, a vertical gradient value of the abnormal grid line point is a sum of a first vertical gradient and a second vertical gradient, and a square sum of the first horizontal gradient and the second vertical gradient is equal to the hidden crack gradient; an included angle formed by the first horizontal gradient and the second vertical gradient is the hidden crack direction, and a system of equations is constructed; the system of equations is solved to obtain the second horizontal gradient and the second vertical gradient, and a square sum of the second horizontal gradient and the second vertical gradient is taken as a correction result; wherein the left neighborhood point is a pixel point with a preset number on both sides of a row where the abnormal grid line point is located in a neighboring column on the left of the abnormal grid line point; the abnormal grid line point The corresponding system of equations is: ； in, and These are the first horizontal gradient and the first vertical gradient, respectively. and These are the second horizontal gradient and the second vertical gradient, respectively. and These are abnormal grid line points. The horizontal and vertical gradient values, and The gradient and direction of the microcrack are given.

2. The method according to claim 1, wherein The obtaining of the abnormal grid line points comprises: Calculating absolute values of differences in gradient directions between each pixel point and an adjacent pixel point in any grid line column, and marking the pixel point as an abnormal grid line point in response to the absolute value of the difference being greater than an average value of absolute values of differences of all pixel points.

3. The method of claim 1, wherein the method further comprises: The global gradient distribution includes local gradient distributions of each gate line column and local gradient distributions of each pixel column between adjacent gate line columns The method for obtaining the local gradient distribution includes: Obtain each grid line column and pixel column The horizontal distance, whereby the horizontal distance is between the grid line column and the pixel column. The absolute value of the difference between the column numbers; Taking a difference between 1 and a proportion of the abnormal grid line points in any grid line column as a confidence degree of each grid line column; The product of the confidence score and the Gaussian weight of the horizontal distance is used as the comprehensive weight, and the local gradient distribution of each grid line column is summed using weights to obtain the pixel column. The local gradient distribution.

4. The method according to claim 3, wherein Pixel column The local gradient distribution Is: ; is the number of gate line columns, is the horizontal distance of a gate line column , is the Gaussian kernel function, is the confidence of a gate line column , is the sum of the confidences of the gate line columns, is the local gradient distribution of a gate line column .

5. The method of claim 1, wherein the method further comprises: The obtaining of the real-time gradient distribution of the EL image comprises: Calculating a gradient value mean of all pixel points in each column of the EL image, and subtracting the gradient value mean of the column from the gradient value of the pixel point to obtain the real-time gradient distribution.

6. The method of claim 1, wherein the method further comprises: The threshold segmentation comprises: determining a segmentation threshold of the difference value graph by using the maximum inter-class variance method, and performing threshold segmentation on the difference value graph by using the segmentation threshold.

7. The method of claim 1, wherein the method further comprises: pixel point a gradient value of a pixel point ; a gradient direction of a pixel point ; wherein is a number of scales, is a scale a gradient value of a pixel point is a scale a gradient direction of a pixel point ​​​

Images