An OLED dispensing coating specification detection method, storage medium and device

Abstract

The present application relates to OLED point glue detection technical field, especially in kind, a kind of OLED point glue coating specification detection method, storage medium, equipment, it includes the following steps: step S10, point cloud data acquisition;Step S20, based on the transmission belt point cloud elimination of straight-through filtering;Step S30, based on the product point cloud area segmentation of point cloud registration;Step S40, based on the point glue coating area segmentation of Euclidean clustering;Step S50, based on the point glue coating specification detection of point cloud minimum bounding box has been used to judge whether to pass detection.The present application designs relevant algorithm can realize efficient and accurate OLED point glue coating specification detection, improves production efficiency and quality control level.

Description

Technical Field

[0001] This invention relates to the field of OLED dispensing inspection technology, and in particular to a method, storage medium, and equipment for inspecting OLED dispensing coating specifications. Background Technology

[0002] Dispensing coating refers to the process of applying adhesive to the junction of OLED display panels and COF connectors to prevent foreign objects and impurities from entering the circuitry. This technology not only affects product performance and reliability but also directly relates to production efficiency and yield. Therefore, developing efficient and accurate inspection technologies has become one of the core requirements for improving product quality. Currently, dispensing inspection technology mainly relies on manual inspection, depending on human experience to judge dispensing quality. This method is not only inefficient but also prone to human error, failing to meet the high precision and efficiency requirements of modern production.

[0003] Existing defects in existing technologies (shortcomings and deficiencies of existing technologies or products): With the advancement of machine vision and intelligent technologies, inspection systems based on traditional 2D vision and image processing are gradually being applied to actual production. However, since most dispensing coating materials are transparent or high-transmittance adhesives, traditional 2D vision and image processing-based inspection methods can only detect obvious defects, resulting in a high rate of missed detection for defects such as size specifications and minor adhesive breaks. This indicates that current automatic inspection methods still have significant room for improvement in terms of efficiency and accuracy. Therefore, this invention aims to propose a 3D point cloud-based method for detecting defects in OLED dispensing coatings to address the problems existing in current methods. Summary of the Invention

[0004] This invention addresses the problems of existing technologies by providing a method, storage medium, and equipment for detecting OLED dispensing coating specifications. It also designs related algorithms to achieve efficient and accurate detection of OLED dispensing coating specifications, thereby improving production efficiency and quality control.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] This invention provides a method for detecting the specifications of OLED dispensing coating, characterized by comprising the following steps:

[0007] Step S10: Point cloud data acquisition;

[0008] Step S20: Conveyor belt point cloud removal based on pass-through filtering;

[0009] Step S30: Segmentation of product point cloud regions based on point cloud registration;

[0010] Step S40: Segmentation of dispensing application region based on Euclidean clustering;

[0011] Step S50: The dispensing and coating specification detection based on the minimum bounding box of the point cloud has been used to determine whether the detection has passed.

[0012] The specific method for product point cloud region segmentation based on point cloud registration in step S30 is as follows: Step S31, point cloud template creation; Step S32, coarse point cloud registration based on SAC-IA; Step S33, fine point cloud registration based on ICP; Step S34, obtaining a new point cloud; Based on the registration result, the boundary range is set using the point cloud template, and the point cloud contained outside the boundary is removed, and the resulting point cloud is the product point cloud region.

[0013] The specific method for segmenting the dispensing coating area based on Euclidean clustering in step S40 is as follows: Step S41, setting the search radius; Step S42, selecting the size range of the cluster; Step S43, selecting the search method, choosing KD-tree.

[0014] The point cloud data acquisition method in step S10 is as follows: point cloud data acquisition is to obtain the three-dimensional point cloud information of the product by a 3D linear scan camera, record millions of three-dimensional points in a short time, and form a point cloud dataset containing shape, contour and surface details.

[0015] The method for removing conveyor belt point clouds based on through filtering in step S20 includes the following steps:

[0016] Step S21: Set filtering parameters: First, define the spatial range of the pass-through filter;

[0017] Step S22: Apply the filter: Use the pass-through filtering algorithm to filter the point cloud data; remove all points that are not within the specified range and keep only the points that meet the conditions.

[0018] In step S31, the same sensor is used to collect point clouds on the production line, so that the point cloud template and the registered point cloud have the same size and density. Then, point cloud processing software is used to cut the glue coating area of ​​the point cloud data to obtain the point cloud template.

[0019] The method for coarse registration of point clouds based on SAC-IA in step S32 is as follows:

[0020] Step S321, Solve for the rotation matrix:

[0021] The rotation matrix R can be obtained by selecting point pairs (p s ,p t ) Calculated; Assume p s and p t These are points in the source point cloud and the target point cloud, respectively. The rotation matrix can be calculated using the following steps:

[0022] Calculate the centroid:

[0023]

[0024] Calculate the covariance matrix H

[0025]

[0026] Calculate the rotation matrix R using Singular Value Decomposition (SVD)

[0027] [U,S,V T ] = SVD(H), R = VU T

[0028] Step S322, Solve for the translation matrix t:

[0029] The formula for calculating the translation matrix is ​​as follows:

[0030] t = c t -R·c s .

[0031] The point cloud fine registration method based on ICP in step S33 is as follows:

[0032] Step S331: Set the maximum number of iterations and the convergence threshold;

[0033] Step S332: Input the pose matrix obtained from coarse registration as the initial registration matrix;

[0034] Step S333: Nearest neighbor matching. Find the nearest point from the target point cloud and match the two.

[0035] Step S334: Calculate the covariance matrix H:

[0036]

[0037] Step S335, Calculate R * =VU T ,t * =u q -Ru q R = R * R,t=t+t * ;

[0038] Step S336: Perform a rigid transformation on the source point cloud until the maximum number of iterations or the convergence threshold set in the first step is met, and output the final rotation matrix and translation matrix.

[0039] Step S337: Obtain the new point cloud:

[0040] P s'=RP s +t;

[0041] Based on the registration results, the boundary range is set using the point cloud template, and the point cloud outside the boundary is removed. The resulting point cloud is the product point cloud region.

[0042] The method used in step S50 to determine whether the dispensing and coating specifications based on the minimum bounding box of the point cloud are passed is as follows:

[0043] Step S511: Principal component analysis based on PCA;

[0044] The basic idea of ​​PCA is to find the direction of maximum mutation, which can be achieved through the following steps:

[0045] Data centralization: The coordinates of data points are first averaged to centralize them. Specifically, this is done by subtracting the mean of all points in the point cloud. The specific principle and calculation method are as follows:

[0046] Suppose there are n points X = {x1, x2, ..., xn} in the point cloud. n If each point contains d dimensions, then the mean is:

[0047]

[0048] The centralized point cloud is:

[0049] X centered =X-μ

[0050] Step S512: Calculate the covariance matrix: Calculate the covariance matrix using the centered point cloud data. The covariance matrix reflects the relationships between various features of the data. The specific principle and calculation method are as follows:

[0051]

[0052] Step S513: Eigenvalue Decomposition: Perform eigenvalue decomposition on the covariance matrix. Eigenvalues ​​reflect the degree of variation in different directions, while eigenvectors represent the corresponding directions. The principle and formula for eigenvalue decomposition of the covariance matrix are as follows:

[0053] Cv i =λ i v i

[0054] Where, λ i It is an eigenvalue, v i It is the corresponding feature vector.

[0055] Step S514: Select principal components: Based on the magnitude of the eigenvalues, select the first few eigenvectors as principal components, and select the eigenvector corresponding to the largest eigenvalue.

[0056] S52: Minimum bounding box calculation;

[0057] Using the principal direction and centroid obtained from the above steps, the point cloud is transformed into coordinates based on the origin, and the coordinate axes are made to coincide with the principal direction. Then, the maximum and minimum values ​​of the three-axis coordinates of the transformed point cloud, Max(x,y,z) and Min(x,y,z), are calculated. Eight coordinate points are obtained by arranging the maximum and minimum values, and these eight coordinate points are the vertices of the minimum bounding box.

[0058] S53: Geometric parameter analysis;

[0059] Step S531: Geometric parameter analysis: Determine whether the adhesive application meets the predetermined specifications by analyzing the dimensions (length, width, and height) of the bounding box; the dimension calculation method is as follows:

[0060]

[0061] Step S532: Comparison and Judgment: Compare the dimensions of the bounding box with the preset coating parameters (l h ,w h ,h h Compare and check if it is within the compliance range. If the following conditions are met simultaneously, it meets the specifications:

[0062]

[0063] The present invention also provides a computer storage medium storing computer instructions, which, when invoked, are used to execute the aforementioned OLED dispensing coating specification detection method.

[0064] The present invention also provides an electronic device comprising: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the OLED dispensing coating specification detection method. The beneficial effects of the present invention are as follows:

[0065] This invention integrates a 3D line scan camera into an OLED production line. Combined with automated production, it can acquire high-precision 3D point cloud data without requiring equipment modifications, laying the foundation for automated inspection. Using a method based on direct-pass filtering and Euclidean clustering, it removes point clouds from the conveyor belt and flexible circuit boards, eliminating redundant point clouds while retaining the product's point cloud, thus segmenting the product panel point cloud. Then, Euclidean clustering is used to segment the product's dispensing area. The minimum bounding box of the dispensing area is calculated, and the size of the calculated minimum bounding box is compared with a set threshold. Products that simultaneously meet the length, width, and height conditions meet the specifications. This invention designs related algorithms to achieve efficient and accurate OLED dispensing specification inspection, improving production efficiency and quality control. Attached Figure Description

[0066] Figure 1 This is a flowchart of an OLED dispensing coating specification testing method according to the present invention.

[0067] Figure 2 A layout diagram of the device for applying the present invention. Detailed Implementation

[0068] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention. The present invention will be described in detail below with reference to the accompanying drawings.

[0069] Example 1

[0070] Embodiment 1 of this application provides a method for detecting the specifications of OLED dispensing coating, such as... Figures 1-2 As shown, it includes the following steps:

[0071] Step S10, Point Cloud Data Acquisition; Point cloud data acquisition involves obtaining the product's 3D point cloud information using a 3D linear scan camera. This process is as follows: Figure 2 As shown, a conveyor belt and product moving at a constant speed are used to capture detailed features of the surface of the scanned area at high resolution. The machine is able to record millions of three-dimensional points in a short time, forming a point cloud dataset containing shape, contour and surface details.

[0072] Step S20: Conveyor belt point cloud removal based on pass-through filtering; Conveyor belt point cloud removal based on pass-through filtering is a commonly used point cloud processing technique. This method sets a spatial range so that only point cloud data that meets specific conditions can be retained, thereby achieving effective removal of redundant point clouds; This application uses this method to remove unnecessary conveyor belt background information from the collected point cloud data.

[0073] Step S30: Segmentation of product point cloud regions based on point cloud registration;

[0074] Step S40: Segmentation of dispensing application region based on Euclidean clustering;

[0075] Step S50: The dispensing and coating specification detection based on the minimum bounding box of the point cloud has been used to determine whether the detection has passed.

[0076] The specific method for product point cloud region segmentation based on point cloud registration in step S30 is as follows: Step S31, point cloud template creation; Step S32, coarse point cloud registration based on SAC-IA; Step S33, fine point cloud registration based on ICP; Step S34, obtaining a new point cloud; Based on the registration result, the boundary range is set using the point cloud template, and the point cloud contained outside the boundary is removed, and the resulting point cloud is the product point cloud region.

[0077] The specific method for segmenting the dispensing coating area based on Euclidean clustering in step S40 is as follows: Step S41, setting the search radius; Step S42, selecting the size range of the cluster; Step S43, selecting the search method, choosing KD-tree.

[0078] The point cloud data acquisition method in step S10 is as follows: point cloud data acquisition is to obtain the three-dimensional point cloud information of the product by a 3D linear scan camera, record millions of three-dimensional points in a short time, and form a point cloud dataset containing shape, contour and surface details.

[0079] The method for removing conveyor belt point clouds based on through filtering in step S20 includes the following steps:

[0080] Step S21: Set Filtering Parameters: First, define the spatial range of the pass-through filter; this typically includes setting the minimum and maximum values ​​of the X, Y, and Z axes to limit the point cloud region to be retained. The practical application scenario of this invention is a product assembly line. The point cloud data outside the product mainly includes the point cloud of the conveyor belt and the point cloud of the region on both sides of the X-axis outside the conveyor belt. Therefore, the filtering parameters only need to be set for the X and Z axes. Removing the conveyor belt point cloud involves collecting the minimum Z-axis value zmin of the conveyor belt during stable operation; point clouds with z values ​​greater than zmin are removed. Removing the point cloud of the region on both sides of the X-axis outside the conveyor belt only requires collecting the X-axis interval of the conveyor belt point cloud during stable operation; points outside this interval are removed. Step S22: Apply the Filter: Use the pass-through filter algorithm to filter the point cloud data; remove all points outside the specified range, retaining only those that meet the conditions. Through the above steps, point cloud removal based on pass-through filtering not only improves the efficiency of data processing but also enhances the accuracy and reliability of subsequent analysis.

[0081] Specifically, the main principle of this application embodiment is that, given the transparent nature of OLED dispensing defects, traditional 2D visual images are insufficient to identify and measure the specifications of the dispensing coating. Therefore, 3D point cloud information is used instead of traditional 2D visual images as the input data for detection. The OLED panel before the dispensing coating process is flat and smooth, while the OLED panel after dispensing coating has a certain amount of adhesive residue. Therefore, based on the difference before and after coating, corresponding equipment and methods can be designed to achieve automated detection of dispensing coating defects. The equipment layout of this application embodiment is as follows: Figure 2 As shown, after completing the dispensing and UV curing processes, the OLED panel flows into the detection area of ​​this invention via a conveyor belt, where product information is collected by a 3D line scan camera.

[0082] This invention integrates a 3D line scan camera into an OLED production line. Combined with automated production lines, it can collect high-precision 3D point cloud data without modifying equipment, laying the foundation for automated inspection. Using a method based on direct-pass filtering and Euclidean clustering, it removes point clouds from the conveyor belt and flexible circuit boards, eliminating redundant point clouds while retaining the product point cloud, thus segmenting the product panel point cloud. Then, Euclidean clustering is used to segment the product's adhesive coating area. The minimum bounding box of the adhesive coating area is calculated, and the size of the calculated minimum bounding box is compared with a set threshold. Products that simultaneously meet the length, width, and height conditions meet the specifications.

[0083] This application embodiment has made relevant designs in terms of hardware device layout design and hardware selection. It uses a 3D line scan camera to collect 3D point cloud data of OLED products and designs relevant algorithms to achieve efficient and accurate OLED dispensing and coating specification detection, thereby improving production efficiency and quality control level.

[0084] In this embodiment of the application, the point cloud actually collected by the same sensor on the production line is used in step S31, so that the point cloud template and the registered point cloud have the same size and the same point cloud density; then, point cloud processing software is used to cut the dispensing area of ​​the point cloud data to obtain the point cloud template.

[0085] In this embodiment, SAC-IA (Sample Consensus Initial Alignment) based coarse point cloud registration is an effective point cloud alignment method, aiming to solve the registration problem between 3D point clouds acquired from different viewpoints or at different times. It can effectively handle noise and mismatch problems, thereby obtaining a relatively accurate initial transformation matrix. However, due to the limited number of sample point pairs, the initial transformation matrix may have certain errors. The SAC-IA based coarse point cloud registration method in step S32 is as follows:

[0086] Step S321, Solve for the rotation matrix:

[0087] The rotation matrix R can be obtained by selecting point pairs (p s ,p t )Calculated; Assume p s and p t These are points in the source point cloud and the target point cloud, respectively. The rotation matrix can be calculated using the following steps:

[0088] Calculate the centroid:

[0089]

[0090] Calculate the covariance matrix H:

[0091]

[0092] Calculate the rotation matrix R using Singular Value Decomposition (SVD):

[0093] [U,S,V T ] = SVD(H), R = VU T

[0094] Step S322, Solve for the translation matrix t:

[0095] The formula for calculating the translation matrix is ​​as follows:

[0096] t = c t -R·c s .

[0097] The point cloud fine registration method based on ICP in step S33 is as follows:

[0098] Step S331: Set the maximum number of iterations and the convergence threshold;

[0099] Step S332: Input the pose matrix obtained from coarse registration as the initial registration matrix;

[0100] Step S333: Nearest neighbor matching. Find the nearest point from the target point cloud and match the two.

[0101] Step S334: Calculate the covariance matrix H:

[0102]

[0103] Step S335, Calculate R * =VU T ,t * =u q -Ru q R = R * R,t=t+t * ;

[0104] Step S336: Perform a rigid transformation on the source point cloud until the maximum number of iterations or the convergence threshold set in the first step is met, and output the final rotation matrix and translation matrix.

[0105] Step S337: Obtain the new point cloud:

[0106] P s '=RP s +t;

[0107] Based on the registration results, the boundary range is set using the point cloud template, and the point cloud outside the boundary is removed. The resulting point cloud is the product point cloud region.

[0108] S40: Euclidean clustering-based dispensing region segmentation:

[0109] Analysis of the point clouds of the product panel and the coating point cloud reveals a significant difference in their Z-axis values. To effectively segment the dispensing and coating area, Euclidean clustering can be used to segment the coating point cloud. This method measures the Euclidean distance between points in the point cloud, grouping points with similar Z-values ​​together to accurately extract the point cloud data of the coating area. The specific process is as follows:

[0110] S41: Search radius setting:

[0111] In Euclidean clustering-based dispensing and coating region segmentation, setting the search radius is a crucial step, directly impacting the clustering effect and segmentation accuracy. Due to the high precision of linear 3D cameras and the large point cloud density they acquire, and the significant differences in the Z-axis values ​​between the product panel's point cloud and the coating point cloud, this invention employs cross-validation. Multiple clustering operations are performed using different radius values ​​to evaluate the stability and accuracy of the clustering results. Specifically, the radius setting principle utilizes the theoretical point cloud distance *m* and the theoretical dispensing height *n* of the camera as reference standards. Multiple clustering operations are performed within the range of *m±n*, and the value with the best performance is selected as the search radius.

[0112] S42: Cluster size range selection:

[0113] Select products that meet the dispensing coating specifications, and count the number of points in the point cloud coating area within the interval Q. Select 10% of the extreme value of Q as the redundancy for the cluster size interval. That is, the minimum number of points in the cluster is Qmin-10%Qmin, and the maximum number of points in the cluster is Qmax+10%Qmax.

[0114] S43: Search method selection: KD-tree.

[0115] In this embodiment, the dispensing coating specification detection based on the minimum bounding box of point cloud is a method for detecting dispensing coating using point cloud data. By analyzing the dimensions (length, width, and height) of the bounding box and comparing them with set values, it is determined whether the dispensing coating meets the predetermined specification standards. This method can effectively analyze the specification parameters of the dispensing coating; the dispensing coating specification detection based on the minimum bounding box of point cloud in step S50 is used to determine whether the detection passes.

[0116] Step S511: Principal component analysis based on PCA;

[0117] The basic idea of ​​PCA is to find the direction of maximum mutation, which can be achieved through the following steps:

[0118] Data centralization: The coordinates of data points are first averaged to centralize them. Specifically, this is done by subtracting the mean of all points in the point cloud. The specific principle and calculation method are as follows:

[0119] Suppose there are n points X = {x1, x2, ..., xn} in the point cloud. n If each point contains d dimensions, then the mean is:

[0120]

[0121] The centralized point cloud is:

[0122] X centered =X-μ

[0123] Step S512: Calculate the covariance matrix: Calculate the covariance matrix using the centered point cloud data. The covariance matrix reflects the relationships between various features of the data. The specific principle and calculation method are as follows:

[0124]

[0125] Step S513: Eigenvalue Decomposition: Perform eigenvalue decomposition on the covariance matrix. Eigenvalues ​​reflect the degree of variation in different directions, while eigenvectors represent the corresponding directions. The principle and formula for eigenvalue decomposition of the covariance matrix are as follows:

[0126] Cv i =λ i v i

[0127] Where, λ i It is an eigenvalue, v i It is the corresponding feature vector.

[0128] Step S514: Select principal components: Based on the magnitude of the eigenvalues, select the first few eigenvectors as principal components, and select the eigenvector corresponding to the largest eigenvalue.

[0129] S52: Minimum bounding box calculation;

[0130] Using the principal direction and centroid obtained from the above steps, the point cloud is transformed into coordinates based on the origin, and the coordinate axes are made to coincide with the principal direction. Then, the maximum and minimum values ​​of the three-axis coordinates of the transformed point cloud, Max(x,y,z) and Min(x,y,z), are calculated. Eight coordinate points are obtained by arranging the maximum and minimum values, and these eight coordinate points are the vertices of the minimum bounding box.

[0131] S53: Geometric parameter analysis;

[0132] Step S531: Geometric parameter analysis: Determine whether the adhesive application meets the predetermined specifications by analyzing the dimensions (length, width, and height) of the bounding box; the dimension calculation method is as follows:

[0133]

[0134] Step S532: Comparison and Judgment: Compare the dimensions of the bounding box with the preset coating parameters (l h ,w h ,h h Compare and check if it is within the compliance range. If the following conditions are met simultaneously, it meets the specifications:

[0135]

[0136] The key points of the embodiments of this application are as follows:

[0137] Key Point 1: By deploying 3D line scan cameras in OLED production lines and combining them with automated production lines, high-precision 3D point cloud data can be collected without modifying equipment, laying the foundation for automated inspection.

[0138] Key point 2: Using a method based on direct filtering and Euclidean clustering, the point clouds of the conveyor belt and the flexible circuit board are removed. While retaining the product point cloud, redundant point clouds are removed, and the product panel point cloud is segmented. Then, Euclidean clustering is used to segment the product's adhesive coating area.

[0139] Key point 3: Solve for the minimum bounding box of the dispensing area. Compare the size of the minimum bounding box with the set threshold. Products that meet the length, width and height conditions simultaneously meet the specifications.

[0140] Example 2

[0141] Embodiment 2 of this application provides a computer storage medium storing computer instructions. When the computer instructions are invoked, they are used to execute the aforementioned OLED dispensing coating specification detection method.

[0142] Example 3

[0143] Embodiment 3 of this application provides an electronic device, which includes: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the OLED dispensing coating specification detection method.

[0144] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and modifications made to the above embodiments based on the present invention without departing from the scope of the present invention are within the scope of the present invention.

Claims

1. A method for detecting the specifications of OLED dispensing coating, characterized in that, Includes the following steps: Step S10: Point cloud data acquisition; Step S20: Conveyor belt point cloud removal based on pass-through filtering; Step S30: Segmentation of product point cloud regions based on point cloud registration; Step S40: Segmentation of dispensing application region based on Euclidean clustering; Step S50: The dispensing and coating specification detection based on the minimum bounding box of the point cloud has been used to determine whether the detection has passed. The specific method for product point cloud region segmentation based on point cloud registration in step S30 is as follows: Step S31, point cloud template creation; Step S32, coarse point cloud registration based on SAC-IA; Step S33, fine point cloud registration based on ICP; Step S34, obtaining a new point cloud; Based on the registration result, the boundary range is set using the point cloud template, and the point cloud contained outside the boundary is removed, and the resulting point cloud is the product point cloud region. The specific method for segmenting the dispensing coating area based on Euclidean clustering in step S40 is as follows: Step S41, setting the search radius; Step S42, selecting the size range of the cluster; Step S43, selecting the search method, choosing KD-tree. The method used in step S50 to determine whether the dispensing and coating specifications based on the minimum bounding box of the point cloud are passed is as follows: Step S511: Principal component analysis based on PCA; The basic idea of ​​PCA is to find the direction of maximum mutation, which can be achieved through the following steps: Data centralization: The coordinates of data points are first averaged to centralize them. Specifically, this is done by subtracting the mean of all points in the point cloud. The specific principle and calculation method are as follows: Assume there are n points in the point cloud. If each point contains d dimensions, then the mean is: The centralized point cloud is: Step S512: Calculate the covariance matrix: Calculate the covariance matrix using the centered point cloud data. The covariance matrix reflects the relationships between various features of the data. The specific principle and calculation method are as follows: Step S513: Eigenvalue Decomposition: Perform eigenvalue decomposition on the covariance matrix. Eigenvalues ​​reflect the degree of variation in different directions, while eigenvectors represent the corresponding directions. The principle and formula for eigenvalue decomposition of the covariance matrix are as follows: in, It is an eigenvalue. It is the corresponding feature vector. Step S514: Select principal components: Based on the magnitude of the eigenvalues, select the first few eigenvectors as principal components, and select the eigenvector corresponding to the largest eigenvalue. S52: Minimum bounding box calculation; Using the principal direction and centroid obtained through the above steps, the point cloud is transformed into coordinates based on the origin, with the coordinate axes coinciding with the principal direction; then, the maximum and minimum values ​​of the three-axis coordinates of the transformed point cloud are calculated (Max( x , y , z ) and Min( x , y , z ); Arrange the maximum and minimum values ​​into 8 coordinate points, and these 8 coordinate points are the vertices of the minimum bounding box; S53: Geometric parameter analysis; Step S531: Geometric parameter analysis: Determine whether the adhesive application meets the predetermined specifications by analyzing the length, width, and height of the enclosure box; the dimension calculation method is as follows: Step S532: Comparison and Judgment: Compare the dimensions of the bounding box with the preset coating parameters. Compare and check if it is within the compliance range. If the following conditions are met simultaneously, it complies with the specifications: 。 2. The OLED dispensing coating specification testing method according to claim 1, characterized in that, The point cloud data acquisition method in step S10 is as follows: point cloud data acquisition is to obtain the three-dimensional point cloud information of the product by using a 3D linear scan camera, and record millions of three-dimensional points in a short time to form a point cloud dataset containing shape, contour and surface details.

3. The OLED dispensing coating specification testing method according to claim 1, characterized in that, The conveyor belt point cloud removal method based on pass-through filtering in step S20 includes the following steps: Step S21: Set filtering parameters: First, define the spatial range of the pass-through filter; Step S22: Apply the filter: Use the pass-through filtering algorithm to filter the point cloud data; remove all points that are not within the specified range and keep only the points that meet the conditions.

4. The OLED dispensing coating specification testing method according to claim 1, characterized in that: In step S31, the same sensor is used to collect point clouds on the production line, so that the point cloud template and the registered point cloud have the same size and the same point cloud density. Next, point cloud processing software is used to cut the glue coating area of ​​the point cloud data to obtain a point cloud template.

5. The OLED dispensing coating specification testing method according to claim 1, characterized in that, The point cloud coarse registration method based on SAC-IA in step S32 is as follows: Step S321, Solve for the rotation matrix: Rotation matrix R By selecting point pairs Calculated; Assumed and These are points in the source point cloud and the target point cloud, respectively. The rotation matrix can be calculated using the following steps: Calculate the centroid: Calculate the covariance matrix H Calculate the rotation matrix using Singular Value Decomposition (SVD) R Step S322: Solve for the translation matrix t： The formula for calculating the translation matrix is ​​as follows: 。 6. The OLED dispensing coating specification testing method according to claim 1, characterized in that, The point cloud fine registration method based on ICP in step S33 is as follows: Step S331: Set the maximum number of iterations and the convergence threshold; Step S332: Input the pose matrix obtained from coarse registration as the initial registration matrix; Step S333: Nearest neighbor matching. Find the nearest point from the target point cloud and match the two. Step S334: Calculate the covariance matrix H： ； Step S335, Calculation ; Step S336: Perform a rigid transformation on the source point cloud until the maximum number of iterations or the convergence threshold set in the first step is met, and output the final rotation matrix and translation matrix. Step S337: Obtain the new point cloud: ； Based on the registration results, the boundary range is set using the point cloud template, and the point cloud outside the boundary is removed. The resulting point cloud is the product point cloud region.

7. A computer storage medium storing computer instructions, wherein the computer storage medium, when invoked, is used to execute an OLED dispensing coating specification detection method as described in any one of claims 1-6.

8. An electronic device, wherein, The electronic device includes: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform an OLED dispensing coating specification detection method as described in any one of claims 1-6.

Images