An automatic plate selecting method and system based on surface defect identification of glass-magnesium plate

Abstract

The application provides an automatic plate selecting method and system based on surface defect identification of glass magnesium plates, and relates to the technical fields of image processing and intelligent manufacturing. The method comprises the following steps: obtaining an original surface image of a glass magnesium plate to be identified and an original surface image of a glass magnesium plate without surface defects; respectively pre-processing the original surface images to obtain a target surface image and a standard surface image; preliminarily identifying the surface of the glass magnesium plate to be identified through an image difference algorithm, and preliminarily judging whether the surface of the glass magnesium plate to be identified has defects; if yes, finely identifying the surface of the glass magnesium plate to be identified through a deep learning algorithm; otherwise, re-obtaining the original surface image; finely judging whether the surface of the glass magnesium plate to be identified has defects; if yes, calculating the surface defect score of the glass magnesium plate to be identified; otherwise, determining the glass magnesium plate to be identified as a perfect glass magnesium plate, and re-obtaining the original surface image; and automatically selecting the glass magnesium plate to be identified according to the surface defect score.

Description

Technical Field

[0001] This invention relates to the fields of image processing and intelligent manufacturing technology, and in particular to an automatic board selection method and system based on the identification of surface defects in magnesium oxide boards. Background Technology

[0002] With the continuous development of industrial automation and intelligent manufacturing technologies, the building materials industry is increasingly demanding higher standards for product quality inspection. Especially in the production of structural materials such as magnesium oxide boards, achieving rapid identification and grading of surface defects has become a crucial aspect of improving product yield, reducing labor costs, and promoting intelligent manufacturing.

[0003] In existing technologies, traditional methods for identifying surface defects in magnesium oxide boards mainly rely on manual visual inspection, which is easily affected by the operator's subjective judgment, resulting in low detection efficiency and large identification errors, making it difficult to meet the requirements of high-precision and high-efficiency quality control.

[0004] Furthermore, traditional automatic board selection methods lack a precise assessment mechanism for the impact of minor surface defects on magnesium oxide boards, misjudging a large number of boards with minor flaws but still usable as scrap, which not only wastes materials but also significantly reduces resource utilization. Summary of the Invention

[0005] To address the shortcomings of existing methods for identifying surface defects in magnesium oxide boards, which primarily rely on manual visual inspection and are susceptible to subjective judgment by operators, resulting in low detection efficiency, significant identification errors, and an inability to meet the demands for high-precision and high-efficiency quality control, and to prevent traditional automatic board selection methods from misjudging the impact of minor surface defects on magnesium oxide boards as scrap, leading to material waste and significantly reduced resource utilization, this invention provides an automatic board selection method and system based on surface defect identification of magnesium oxide boards.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First aspect:

[0008] An automatic board selection method based on surface defect identification of magnesium oxide board provided by this invention includes:

[0009] S1: Obtain the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects;

[0010] S2: Preprocess the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects to obtain the target surface image and the standard surface image respectively.

[0011] S3: Based on the target surface image and the standard surface image, use the image difference algorithm to perform preliminary identification of the surface of the magnesium oxide board to be identified, and preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified; if yes, proceed to S4; otherwise, return to S1.

[0012] S4: Based on the target surface image, use a deep learning algorithm to perform fine identification of the surface of the magnesium oxide board to be identified, and determine whether there are defects on the surface of the magnesium oxide board to be identified; if so, proceed to S5; otherwise, determine the magnesium oxide board to be identified as an intact magnesium oxide board and return to S1.

[0013] S5: Calculate the surface defect score of the magnesium oxide board to be identified;

[0014] S6: Automatically select the magnesium oxide board to be identified based on the surface defect score.

[0015] The second aspect:

[0016] An automatic board selection system based on surface defect identification of magnesium oxide board provided in this invention includes:

[0017] processor;

[0018] A memory storing computer-readable instructions, which, when executed by the processor, implement the automatic board selection method based on the identification of surface defects in magnesium oxide boards as described in the first aspect.

[0019] Third aspect:

[0020] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the automatic board selection method based on the identification of surface defects of magnesium oxide board as described in the first aspect.

[0021] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0022] In this embodiment of the invention, an image difference algorithm is used to perform preliminary identification of the surface of the magnesium oxide board to be identified, and to preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified. Then, a deep learning algorithm is used to perform fine identification of the surface of the magnesium oxide board to be identified, and to finely determine whether there are defects on the surface of the magnesium oxide board to be identified. This eliminates reliance on manual visual inspection, is less affected by the subjective judgment of the operator, has high detection efficiency, and small identification error, and can meet the requirements of high precision and high efficiency quality control. By calculating the surface defect score of the magnesium oxide board to be identified, the board to be identified is automatically selected based on the surface defect score. For magnesium oxide boards with minor surface defects, there is a fine evaluation mechanism for the degree of defect impact, which will not misjudge a large number of boards with minor flaws but still usable value as scrap, thereby reducing material waste and improving resource utilization efficiency. Attached Figure Description

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

[0024] Figure 1 A flowchart illustrating an automatic board selection method based on surface defect identification of magnesium oxide board provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of an automatic board selection system based on surface defect identification of magnesium oxide board, provided as an embodiment of the present invention. Detailed Implementation

[0026] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0027] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0028] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0029] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0030] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0031] Reference manual attached Figure 1 The diagram shows a flowchart of an automatic board selection method based on surface defect identification of magnesium oxide board provided by an embodiment of the present invention.

[0032] This invention provides an automatic board selection method based on the identification of surface defects in magnesium oxide (MgO) boards. This method can be implemented by an automatic board selection device based on the identification of surface defects in MgO boards, which can be a terminal or a server. The processing flow of the automatic board selection method based on the identification of surface defects in MgO boards may include the following steps:

[0033] S1: Obtain the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects.

[0034] Specifically, an industrial camera system is installed above the magnesium oxide board conveyor line. Using a high-resolution area scan camera or line scan camera, combined with a matching ring light source or strip light source, images of each board surface are scanned in a uniform and stable lighting environment. The images of the magnesium oxide boards to be identified are acquired in real-time during production, while images of high-quality boards without surface defects, selected through manual verification, are acquired using the same method. A standard image sample library is then established for subsequent image differential processing.

[0035] In this embodiment of the invention, by configuring an industrial camera system and a standard image sample library, not only is the accuracy, real-time performance and automation level of defect detection greatly improved, but it also provides strong data support for subsequent intelligent identification, model upgrades and defect tracing, which is an indispensable key link in realizing an intelligent manufacturing quality control system.

[0036] S2: Preprocess the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects to obtain the target surface image and the standard surface image.

[0037] In this embodiment of the invention, by preprocessing the original image, the structural alignment, grayscale consistency and edge clarity of the image are significantly improved, providing a stable, standardized and interference-resistant data input source for subsequent defect identification and image difference analysis, thereby effectively ensuring the recognition accuracy and robustness of the overall system under complex working conditions.

[0038] In one possible implementation, S2 specifically includes sub-steps S201 to S204:

[0039] S201: Spatially align the original surface image of the magnesium oxide board to be identified with the original surface image of the magnesium oxide board without surface defects using an elastic registration algorithm.

[0040] It should be noted that elastic registration is an image processing technique used to precisely align two images in space. It considers not only overall rigid transformations between images (such as translation, rotation, and scaling) but also allows for non-rigid elastic deformation in local regions, thus achieving more flexible and detailed registration results. This algorithm extracts key feature points (such as corners or edges) from the image and combines them with deformation models such as B-splines or thin plate splines (TPS) to perform local adaptive transformations on the target image to match the structure of the reference image to the greatest extent possible. It is commonly used in high-precision applications such as medical imaging, defect detection, and surface comparison.

[0041] Specifically, key feature points (such as edges, corners, or texture patterns) of the two images are first extracted. Then, based on global rigid registration, a multi-scale matching strategy is used to introduce a non-rigid deformation model (such as B-splines or thin plate splines) to adaptively transform the local area to correct minor geometric distortions caused by shooting angle, surface deformation, or lighting differences, thereby ensuring accurate alignment between the images in spatial structure.

[0042] S202: Perform grayscale processing on the original surface image of the spatially aligned magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects, respectively, to obtain a first grayscale image and a second grayscale image.

[0043] Optionally, the original surface image of the spatially aligned magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects are processed into grayscale according to the following formula to obtain a first grayscale image and a second grayscale image:

[0044]

[0045] Where S(x,y) represents the gray value of the first grayscale image at coordinates (x,y), and R S (x,y) represents the red channel value at coordinates (x,y) of the original surface image of the magnesium oxide board to be identified after spatial alignment. S (x,y) represents the green channel value at coordinates (x,y) of the original surface image of the magnesium oxide board to be identified after spatial alignment. S (x,y) represents the blue channel value at coordinates (x,y) of the original surface image of the magnesium oxide board to be identified after spatial alignment, T(x,y) represents the grayscale value at coordinates (x,y) of the second grayscale image, and R T (x,y) represents the red channel value at coordinates (x,y) of the original surface image of the spatially aligned, defect-free magnesium oxide board. T(x,y) represents the green channel value at coordinates (x,y) of the original surface image of the spatially aligned, defect-free magnesium oxide board. T (x,y) represents the blue channel value at coordinates (x,y) of the original surface image of the spatially aligned magnesium oxide board without surface defects.

[0046] S203: Perform grayscale normalization processing on the first grayscale image and the second grayscale image respectively.

[0047] Optionally, the first grayscale image and the second grayscale image are subjected to grayscale normalization processing according to the following formula:

[0048]

[0049]

[0050] in, μ represents the gray value of the first grayscale image after grayscale normalization at coordinates (x, y). S σ represents the average grayscale value in the first grayscale image. S This represents the standard deviation of gray levels in the first grayscale image. μ represents the gray value of the second grayscale image at coordinates (x, y) after grayscale normalization. T σ represents the average grayscale value in the second grayscale image. T This represents the standard deviation of gray levels in the second grayscale image.

[0051] S204: Using the LOG filtering algorithm, the first grayscale image and the second grayscale image after grayscale normalization are denoised to obtain the target surface image and the standard surface image.

[0052] It's important to note that the LOG filtering algorithm (Laplacian of Gaussian) is an image processing method that combines Gaussian smoothing and Laplacian edge enhancement, primarily used for edge detection and noise suppression in images. Its core idea is to first smooth the image using a Gaussian function to reduce high-frequency noise interference, and then perform second-order derivative operations on the image using the Laplacian operator to enhance edge regions with dramatic grayscale changes. This algorithm effectively highlights structural contours in images and is particularly suitable for edge extraction and detail enhancement against complex backgrounds. It is a key preprocessing step in many image recognition and differential detection algorithms.

[0053] Optionally, according to the following formula, the first grayscale image and the second grayscale image after grayscale normalization are denoised using the LOG filtering algorithm to obtain the target surface image and the standard surface image:

[0054]

[0055]

[0056] in, LOG(x,y) represents the gray value of the target surface image at coordinates (x,y), and LOG(x,y) represents the Laplacian-Gaussian convolution kernel at coordinates (x,y). This represents the convolution operation. This represents the grayscale value of the standard surface image at coordinates (x, y).

[0057] Optionally, the calculation process of the Laplace-Gaussian convolution kernel specifically includes:

[0058] Constructing the Gaussian kernel function:

[0059]

[0060] Among them, g σ (x,y) represents the Gaussian kernel function at coordinates (x,y), σ represents the standard deviation of the Gaussian kernel function, and exp represents the exponential function.

[0061] Calculate the first derivative with respect to x using the Gaussian kernel function:

[0062]

[0063] in, This represents partial derivative operations.

[0064] Calculate the second derivative with respect to x based on the first derivative of x:

[0065] .

[0066] Similarly, based on the Gaussian kernel function, calculate the first derivative with respect to y:

[0067] .

[0068] Calculate the second derivative with respect to y based on the first derivative:

[0069] .

[0070] Calculate the Laplace-Gaussian convolution kernel based on the second derivatives with respect to x and y: .

[0071] In this embodiment of the invention, by preprocessing the original surface image of the magnesium oxide board to be identified and the original image of the standard defect-free magnesium oxide board, and sequentially performing elastic registration, grayscale conversion, grayscale normalization and LOG filtering operations, the spatial structure alignment, grayscale feature standardization and edge detail enhancement of the images can be effectively achieved. This eliminates the interference caused by differences in shooting angle, surface deformation and illumination, significantly improves the feasibility of image comparison and the saliency of defect features, and provides a unified, standardized and interference-resistant input data foundation for subsequent image differential recognition and deep learning fine detection, thereby improving the recognition accuracy and robustness of the overall system.

[0072] S3: Based on the target surface image and the standard surface image, use an image difference algorithm to perform a preliminary identification of the surface of the magnesium oxide board to be identified, and preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified. If yes, proceed to S4. Otherwise, return to S1.

[0073] It's important to note that image difference algorithms are image processing methods that detect changing regions in an image by comparing the differences in pixel-level grayscale values ​​between two images. It typically subtracts the current image from a reference image pixel by pixel to obtain a difference map reflecting the differing regions, thereby identifying whether there are changes or defects in the target object. This algorithm is simple and efficient, and is widely used in motion detection, target tracking, and surface defect recognition, especially suitable for image comparison tasks with stable backgrounds and prominent changes. In defect detection, it can be used to quickly locate regions that are inconsistent with standard samples, achieving initial screening.

[0074] In this embodiment of the invention, by employing an image difference algorithm to perform pixel-level grayscale comparison between the target image and the standard image, local abnormal areas on the surface of the magnesium oxide board can be quickly and intuitively identified, achieving efficient preliminary judgment of surface defects. This method is simple in structure, responds rapidly, and is highly adaptable, making it particularly suitable for real-time screening of high-volume, high-rate boards in actual industrial production lines. It is a key component of the entire defect identification system that balances speed and accuracy.

[0075] In one possible implementation, step S3, which involves preliminary identification of the surface of the magnesium oxide board to be identified based on the target surface image and the standard surface image using an image difference algorithm, specifically includes sub-steps S301 and S302:

[0076] S301: Calculate the grayscale difference between the target surface image and the standard surface image using an image difference algorithm.

[0077] Optionally, the grayscale difference between the target surface image and the standard surface image can be calculated using the following formula:

[0078] Where D(x,y) represents the gray-level difference at coordinate (x,y), and ω(x,y) represents the weighting coefficient of the gray-level value at coordinate (x,y).

[0079] S302: Determine if the grayscale difference is greater than the grayscale threshold. If yes, preliminarily identify that there are defects on the surface of the magnesium oxide board to be identified. Otherwise, preliminarily identify that there are no defects on the surface of the magnesium oxide board to be identified.

[0080] It should be noted that those skilled in the art can set the grayscale threshold according to actual needs, and this invention does not limit it.

[0081] In this embodiment of the invention, by combining image difference calculation with pixel-level weight adjustment and setting a grayscale threshold for rapid judgment, not only is efficient preliminary identification of potential defects on the surface of magnesium oxide boards achieved, but also accurate, stable, and region-sensitive input is provided for subsequent fine recognition modules. This method is simple in structure, computationally lightweight, and has flexible parameters, greatly improving the practicality and industrial adaptability of the overall recognition system.

[0082] S4: Based on the target surface image, a deep learning algorithm is used to perform a detailed identification of the surface of the magnesium oxide board to be identified, precisely determining whether there are defects on the surface. If so, proceed to S5. Otherwise, the magnesium oxide board to be identified is determined to be an intact magnesium oxide board, and the process returns to S1.

[0083] In this embodiment of the invention, a deep learning algorithm is used to perform fine recognition of the target image of the magnesium oxide board, effectively improving the system's ability to identify complex, subtle, and unstructured defects. This method has advantages such as high accuracy, strong adaptability, and scalability, and is a key step in achieving intelligent, highly reliable defect detection and quality control.

[0084] Optionally, the deep learning algorithm is specifically a deformable convolutional neural network.

[0085] It's important to note that Deformable Convolutional Neural Networks (DCNs) are improved models that introduce spatial adaptation mechanisms into traditional convolutional neural networks, aiming to enhance the network's ability to model geometric deformations and complex structures. By introducing learnable offsets into standard convolution operations, each convolutional kernel can dynamically adjust its sampling position based on the image content, thus more flexibly capturing information such as non-rigid deformations, local variations, and edge distortions of the target. DCNs are particularly suitable for tasks requiring strong spatial awareness, such as object detection and semantic segmentation, and have significant advantages in handling irregular shape features such as defects on the surface of magnesium oxide boards.

[0086] A deformable convolutional neural network consists of: a backbone layer, multiple pooling layers, multiple fully connected layers, deformable convolutional layers, and an output layer.

[0087] Optionally, the backbone network layer is either a ResNet network layer or a MobileNet network layer.

[0088] It's important to note that ResNet (Residual Network) is a deep convolutional neural network architecture that mitigates the vanishing gradient and degradation problems during deep neural network training by introducing shortcut connections. Its core idea is to allow the network to learn the residual between the input and output, rather than directly learning the output itself, thus enabling efficient information transfer within the network. Due to its high stability and superior performance, ResNet is widely used in tasks such as image recognition and classification.

[0089] It's worth noting that the MobileNet network layer is a lightweight convolutional neural network architecture designed for mobile and edge computing devices. By introducing depthwise separable convolution, it splits traditional convolution into "spatial convolution" and "channel convolution," significantly reducing computational cost and model parameters. MobileNet maintains high accuracy while greatly improving operational efficiency, making it suitable for image recognition and object detection tasks in resource-constrained scenarios.

[0090] The pooling layer includes: a first pooling layer and a second pooling layer.

[0091] It should be noted that the first pooling layer is the max pooling layer, and the second pooling layer is the average pooling layer.

[0092] The fully connected layer includes: a first fully connected layer and a second fully connected layer.

[0093] In this embodiment of the invention, a deformable convolutional neural network structure is employed, combined with an optional ResNet or MobileNet backbone network, pooling layers, fully connected layers, and a channel weighting mechanism, significantly improving the ability to finely model and spatially perceive complex and irregular defects on the surface of magnesium oxide boards. This structure possesses significant advantages such as high recognition accuracy, strong deployability, and wide adaptability, and is a key technological foundation for realizing the core recognition module of the intelligent board selection system.

[0094] In one possible implementation, step S4, which involves performing detailed identification of the surface of the magnesium oxide board to be identified based on the target surface image using a deep learning algorithm, specifically includes sub-steps S401 to S408:

[0095] S401: Input the target surface image into the backbone network layer to obtain the feature map.

[0096] S402: Input the feature map into the first pooling layer to obtain the first pooling feature map.

[0097] S403: Input the feature map into the second pooling layer to obtain the second pooling feature map.

[0098] S404: Concatenate the first pooling feature map and the second pooling feature map to obtain a concatenated pooling feature map.

[0099] S405: Input the spliced ​​pooled feature map into the first fully connected layer and the second fully connected layer in sequence to obtain the importance score of each channel in the feature map.

[0100] Optionally, the importance score for each channel in the feature map is as follows:

[0101]

[0102] Among them, S c denoted by , where c represents the importance score of the c-th channel in the feature map, Sigmoid() represents the Sigmoid activation function, W2 represents the weight matrix of the second fully connected layer, ReLU() represents the ReLU activation function, W1 represents the weight matrix of the first fully connected layer, and z represents the concatenated pooled feature map.

[0103] S406: Input the importance score into the deformable convolutional layer to obtain the feature value of each pixel position in the feature map.

[0104] Optionally, the feature values ​​at each pixel location in the feature map are specifically as follows:

[0105]

[0106] Among them, y p Let represent the feature value at the p-th pixel position in the feature map, C represent the total number of channels in the feature map, K represent the total number of grid points in the convolution kernel, and w k Let x represent the convolution kernel weights at the k-th grid point, x represent the feature map, and p represent the current pixel position. k Δp represents the pixel position at the k-th grid point. k Δm represents the offset at the k-th grid point. k Let λ represent the modulation parameter at the k-th grid point. k This represents the Mask weight value at the k-th grid point.

[0107] It should be noted that those skilled in the art can choose the method for calculating the Mask weight value according to actual needs, and this invention does not limit it.

[0108] Optionally, λ k There are two calculation methods:

[0109] The first method is based on the calculation of a polynomial mask (PM). This method constructs a spatial weight mask based on the relative distance between a pixel and the center point of the feature map, following the principle of "high weight in the central region and low weight in the edge region". The calculation formula is as follows:

[0110]

[0111] in, The x-axis coordinate of a pixel in the feature map is represented by W, the width of the feature map is represented by H, and the height of the feature map is represented by H. The vertical axis coordinate of a pixel in the feature map is represented by , and n represents the order of the polynomial controlling the decay rate.

[0112] It's important to note that a polynomial attenuation mask (PM) is a spatial attention mechanism that weights pixels based on their relative distance to the image center. The weights decrease polynomially as the distance increases. The core idea is to assign higher response weights to the central image region and lower weights to edge regions, thereby emphasizing potentially more critical features in the central area. This mask is simple, computationally inexpensive, and suitable for image processing tasks with significant location sensitivity, such as prioritizing anomalies in the central region during defect detection.

[0113] The second method is based on the Gaussian Mask (GM) calculation method. This method simulates the decay characteristics of a two-dimensional Gaussian distribution, with the weight peak at the center of the feature map and decaying exponentially outwards. This better reflects the actual characteristic of the effective receptive field in CNNs, which exhibits a Gaussian-like distribution. The calculation formula is as follows:

[0114]

[0115] Where G represents the magnitude coefficient of the Gaussian function, exp represents the exponential function, σ1 represents the standard deviation of the Gaussian distribution along the horizontal axis, and σ2 represents the standard deviation of the Gaussian distribution along the vertical axis.

[0116] It's important to note that a Gaussian mask (GM) is a spatial weighting model based on a two-dimensional Gaussian distribution. It peaks at the image center and decays exponentially outwards. By simulating the actual response of the receptive field in a convolutional neural network, it assigns higher weights to the central region and lower weights to the edge regions, thereby enhancing the ability to perceive features in key areas. Gaussian masks exhibit smooth decay and continuous response, making them suitable for tasks involving fine feature extraction and localization in complex images.

[0117] S407: Input the feature value into the output layer to obtain the surface defect probability value of the magnesium oxide board to be identified.

[0118] Optionally, the surface defect probability value of the magnesium oxide board to be identified is specifically as follows:

[0119]

[0120] Where P represents the probability value of surface defects in the magnesium oxide board to be identified, and Y... b Indicates the type of the b-th defect, x p This represents the position of the p-th pixel in the feature map, where e is a constant. This represents the raw score of the p-th pixel position in the feature map for the type of the b-th defect, where B represents the total number of defect types.

[0121] S408: Determine whether the surface defect probability value is greater than the surface defect probability threshold. If yes, accurately identify that there are defects on the surface of the magnesium oxide board to be identified. Otherwise, accurately identify that there are no defects on the surface of the magnesium oxide board to be identified.

[0122] It should be noted that those skilled in the art can set the surface defect probability threshold according to actual needs, and this invention does not limit it.

[0123] In this embodiment of the invention, a deep learning recognition process consisting of a backbone network layer, a dual-channel pooling layer, a fully connected layer, a deformable convolutional layer, and an output layer is sequentially input into the target surface image. Combined with a channel importance scoring mechanism and various mask weight values, this method achieves precise modeling and high-precision judgment of surface defects in magnesium oxide boards. This method not only flexibly adapts to various irregular defect morphologies such as peeling, bulging, delamination, and cracks, but also accurately delineates defect areas through pixel-level defect probability maps. Combined with set thresholds, it enables automatic decision-making, greatly improving the system's ability to identify complex defects, its robustness, and its adaptability to engineering deployment, ensuring intelligent, precise, and highly reliable selection of magnesium oxide boards.

[0124] S5: Calculate the surface defect score of the magnesium oxide board to be identified.

[0125] In this embodiment of the invention, by calculating the surface defect score of the magnesium oxide board to be identified, an objective and quantitative impact assessment of the defects on the surface of the magnesium oxide board can be performed, providing an accurate and controllable grading basis for subsequent automatic board selection. This method enhances the practicality and intelligence level of the identification system and is a key component in realizing automated quality management.

[0126] In one possible implementation, S5 specifically includes sub-steps S501 to S504:

[0127] S501: Calculate the area percentage of defects in the magnesium oxide board to be identified.

[0128] Optionally, the surface defect score of the magnesium oxide board to be identified can be calculated according to the following formula:

[0129]

[0130] Among them, O 1,z A represents the area percentage of the z-th defect in the magnesium oxide board to be identified. defect,z A represents the pixel area of ​​the z-th defect in the magnesium oxide board to be identified. total This represents the total pixel area of ​​the target surface image.

[0131] S502: Set the severity factor of the defects in the magnesium oxide board to be identified based on the type of defects in the board.

[0132] It should be noted that the severity factor is a parameter used to measure the degree of impact of defects on the functionality and appearance quality of the magnesium oxide board. Since different types of defects have significantly different destructive effects on the board's performance—for example, structural defects such as cracks and voids usually pose a greater threat to the board's strength and durability, and therefore should be assigned higher factor values—and appearance defects such as surface stains and minor scratches have a smaller impact on structural performance, and therefore should be assigned lower factor values. The higher the factor value, the greater the potential harm of that type of defect.

[0133] Furthermore, those skilled in the art can set the severity factor according to actual needs, and this invention does not limit this.

[0134] S503: Set the position influence factor of the defect in the magnesium oxide board to be identified based on the center distance between the defect in the magnesium oxide board to be identified and the target surface image.

[0135] It should be noted that the location impact factor is a parameter used to measure the sensitivity of the overall quality of a defect's location on the board surface. In practical applications of magnesium oxide boards, the central area or connecting edges are typically critical functional areas. Defects in these areas are more likely to cause installation obstacles or structural safety hazards, and therefore should be assigned a higher factor value. In contrast, defects located at the edges have a relatively smaller impact on the overall quality and should therefore be assigned a lower factor value. The higher the factor value, the more sensitive the defect's location, and the more significant its impact on product quality.

[0136] Furthermore, those skilled in the art can set the magnitude of the location influence factor according to actual needs, and this invention does not limit this.

[0137] S504: Calculate the surface defect score of the magnesium oxide board to be identified based on the area ratio, severity factor, and location influence factor.

[0138] It should be noted that the surface defect score is a quantitative indicator that comprehensively reflects the degree of surface defects in magnesium oxide boards, used to assess the impact of defects on the overall quality of the board. This score is calculated by weighting the defect area ratio, the severity factor corresponding to the defect type, and the location factor of the defect on the board surface, resulting in a comprehensive score for grading. A higher score indicates a greater impact from the defects, and a lower quality grade for the magnesium oxide board.

[0139] Optionally, the surface defect score of the magnesium oxide board to be identified can be calculated according to the following formula:

[0140]

[0141] Wherein, Score represents the surface defect score of the magnesium oxide board to be identified, Z represents the total number of defects in the magnesium oxide board to be identified, w1 represents the weighting coefficient of the area percentage, w2 represents the weighting coefficient of the severity factor, and O 2,z w3 represents the severity factor of the z-th defect in the magnesium oxide board to be identified, w3 represents the weighting coefficient of the location influence factor, and O 3,z This represents the positional influence factor of the z-th defect in the magnesium oxide board to be identified.

[0142] In this embodiment of the invention, a multi-factor scoring mechanism based on area ratio, defect type severity, and defect location influence can be established to quantitatively assess surface defects in magnesium oxide boards. This scoring system has a clear structure, adjustable factors, and flexible application. It can serve as a direct basis for subsequent board selection and also support multi-level sorting strategies, defect tracing, and quality statistical analysis. It is a core module for achieving intelligent quality assessment and refined product grading control.

[0143] Furthermore, the types of defects to be identified in the magnesium oxide board include: structural defects, appearance defects, and functional defects.

[0144] It should be noted that structural defects refer to defects that compromise the overall structural integrity of the magnesium oxide board, affecting its load-bearing capacity and durability. Appearance defects refer to surface anomalies that primarily affect the visual appeal or surface flatness of the magnesium oxide board, but have a relatively minor impact on its mechanical properties. Functional defects refer to inherent or surface processing defects that may limit or cause the magnesium oxide board to malfunction, and are usually closely related to the quality of processes such as gluing and interlayer bonding.

[0145] Structural defects include cracks, voids, and broken edges or missing corners.

[0146] It should be noted that cracks refer to linear fracture lines appearing on the surface of the magnesium oxide board, usually caused by stress concentration or curing shrinkage, affecting the overall strength. Voids refer to areas where material has fallen off or is not filled on or near the surface, forming noticeable depressions. Broken edges and corners manifest as damage, chipping, or missing corners at the edges of the board, often caused by handling or cutting defects.

[0147] Appearance defects include stains, scratches, and peeling.

[0148] It should be noted that stains typically appear as adhesive residue, oil stains, or other unusual discoloration on the board surface, affecting the decorative effect. Scratches are shallow grooves or linear marks formed on the surface of the magnesium oxide board after mechanical contact or friction. Peeling refers to the localized lifting and peeling of the coating or covering layer on the board surface, commonly caused by improper maintenance or abnormal mixing ratios.

[0149] Functional defects include: insufficient adhesive, delamination, and bulging.

[0150] It should be noted that insufficient adhesive refers to uneven application of adhesive or a thin adhesive layer in a certain area of ​​the board, resulting in insufficient bonding strength. Delamination manifests as localized separation between board layers, leading to surface bulging or deformation. Bubbling is caused by uneven adhesive layer or material expansion, resulting in localized bulges on the board surface, which can easily affect the accuracy of installation or splicing.

[0151] In this embodiment of the invention, by classifying the surface defects of magnesium oxide boards into structural defects, appearance defects, and functional defects, and defining and categorizing each type of defect, a complete defect knowledge system is constructed. This provides a clear semantic foundation and processing basis for subsequent image recognition, scoring, board selection decisions, and quality traceability. This classification method possesses advantages such as strong engineering applicability, high trainability of recognition, and clear judgment strategies, making it a crucial component supporting the core capabilities of intelligent defect detection systems.

[0152] S6: Automatically select the magnesium oxide board to be identified based on the surface defect score.

[0153] In one possible implementation, S6 specifically refers to:

[0154] Determine if the surface defect score of the magnesium oxide board to be identified exceeds the surface defect score threshold. If so, the magnesium oxide board to be identified is classified as a scrapped board and discarded. Otherwise, the magnesium oxide board to be identified is classified as a defective board and retained for automatic selection.

[0155] It should be noted that those skilled in the art can set the surface defect scoring threshold according to actual needs, and this invention does not limit it.

[0156] In this embodiment of the invention, by constructing an automatic board selection mechanism based on surface defect scoring, the system can quickly and accurately classify and control the execution of actions on magnesium oxide boards according to set thresholds. This mechanism not only realizes the automated application of test results and reduces human intervention, but also supports flexible configuration, traceability, and intelligent processing. It is a key link in the intelligent board selection system to complete the closed loop from identification to execution, significantly improving board selection efficiency, material utilization, and quality control capabilities.

[0157] Reference manual attached Figure 2 The diagram shows a schematic of an automatic board selection system based on surface defect identification of magnesium oxide board provided by the present invention.

[0158] The present invention also provides an automatic board selection system 20 based on the identification of surface defects in magnesium oxide boards, applied to the above-mentioned automatic board selection method based on the identification of surface defects in magnesium oxide boards, comprising:

[0159] Processor 201;

[0160] The memory 202 stores computer-readable instructions, which, when executed by the processor 201, implement the automatic board selection method based on the identification of surface defects of magnesium oxide board as described in the method embodiment.

[0161] The automatic board selection system 20 based on the identification of surface defects of magnesium oxide board provided by the present invention can perform the above-mentioned automatic board selection method based on the identification of surface defects of magnesium oxide board and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0162] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0163] In this embodiment of the invention, an image difference algorithm is used to perform preliminary identification of the surface of the magnesium oxide board to be identified, and to preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified. Then, a deep learning algorithm is used to perform fine identification of the surface of the magnesium oxide board to be identified, and to finely determine whether there are defects on the surface of the magnesium oxide board to be identified. This eliminates reliance on manual visual inspection, is less affected by the subjective judgment of the operator, has high detection efficiency, and small identification error, and can meet the requirements of high precision and high efficiency quality control. By calculating the surface defect score of the magnesium oxide board to be identified, the board to be identified is automatically selected based on the surface defect score. For magnesium oxide boards with minor surface defects, there is a fine evaluation mechanism for the degree of defect impact, which will not misjudge a large number of boards with minor flaws but still usable value as scrap, thereby reducing material waste and improving resource utilization efficiency.

[0164] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0165] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0166] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0167] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0168] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0169] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0170] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0171] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0172] In the several embodiments provided by this invention, it should be understood that the disclosed devices, 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 device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0173] 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.

[0174] In addition, the functional units in the various embodiments of the present invention 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.

[0175] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a 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 invention. 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.

[0176] This invention provides a computer-readable storage medium storing a computer program thereon, characterized in that, when executed by a processor, the program implements the automatic board selection method based on the identification of surface defects in a magnesium oxide board as described in the method embodiment.

[0177] The present invention provides a computer-readable storage medium that can realize the steps and effects of the automatic board selection method based on the identification of surface defects of magnesium oxide board in the above-described method embodiments. To avoid repetition, the present invention will not repeat them.

[0178] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0179] In this embodiment of the invention, an image difference algorithm is used to perform preliminary identification of the surface of the magnesium oxide board to be identified, and to preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified. Then, a deep learning algorithm is used to perform fine identification of the surface of the magnesium oxide board to be identified, and to finely determine whether there are defects on the surface of the magnesium oxide board to be identified. This eliminates reliance on manual visual inspection, is less affected by the subjective judgment of the operator, has high detection efficiency, and small identification error, and can meet the requirements of high precision and high efficiency quality control. By calculating the surface defect score of the magnesium oxide board to be identified, the board to be identified is automatically selected based on the surface defect score. For magnesium oxide boards with minor surface defects, there is a fine evaluation mechanism for the degree of defect impact, which will not misjudge a large number of boards with minor flaws but still usable value as scrap, thereby reducing material waste and improving resource utilization efficiency.

[0180] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0181] The following points need to be explained:

[0182] (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0183] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the invention, i.e., these drawings are not drawn to scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element or there may be intermediate elements.

[0184] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0185] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An automatic board selection method based on surface defect identification of magnesium oxide board, characterized in that, include: S1: Obtain the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects; S2: Preprocess the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects to obtain the target surface image and the standard surface image respectively. S3: Based on the target surface image and the standard surface image, the surface of the magnesium oxide board to be identified is initially identified using an image difference algorithm to preliminarily determine whether there are defects on the surface of the magnesium oxide board to be identified; if so, proceed to S4; Otherwise, return to S1; S4: Based on the target surface image, a deep learning algorithm is used to perform fine identification of the surface of the magnesium oxide board to be identified, and to determine whether there are defects on the surface of the magnesium oxide board to be identified; if so, proceed to S5; Otherwise, the magnesium oxide board to be identified is determined to be an intact magnesium oxide board, and the process returns to S1; Specifically, the deep learning algorithm is a deformable convolutional neural network. The deformable convolutional neural network includes: a backbone network layer, multiple pooling layers, multiple fully connected layers, deformable convolutional layers, and an output layer; The pooling layer includes: a first pooling layer and a second pooling layer; The fully connected layer includes: a first fully connected layer and a second fully connected layer; Specifically, step S4 involves using a deep learning algorithm to perform detailed identification of the surface of the magnesium oxide board to be identified based on the target surface image. S401: Input the target surface image into the backbone network layer to obtain a feature map; S402: Input the feature map into the first pooling layer to obtain the first pooling feature map; S403: Input the feature map into the second pooling layer to obtain the second pooling feature map; S404: The first pooling feature map and the second pooling feature map are concatenated to obtain a concatenated pooling feature map; S405: The spliced ​​pooled feature map is sequentially input into the first fully connected layer and the second fully connected layer to obtain the importance score of each channel in the feature map: in, S c Represents the first in the feature graph c Importance score of each channel Sigmoid ( ) represents the Sigmoid activation function. W 2 represents the weight matrix of the second fully connected layer. ReLU ( ) represents the ReLU activation function. W 1 represents the weight matrix of the first fully connected layer. z This represents a spliced ​​pooled feature map; S406: Input the importance score into the deformable convolutional layer to obtain the feature value of each pixel position in the feature map: in, y p Represents the first in the feature graph p Feature values ​​at each pixel location C This represents the total number of channels in the feature map. K This represents the total number of grid points in the convolution kernel. w k Indicates the first k The convolution kernel weights at each grid point x Representing feature maps, p Indicates the current pixel position. p k Indicates the first k The pixel position at each grid point, Δ p k Indicates the first k The offset at each grid point, Δ m k Indicates the first k Modulation parameters at each grid point λ k Indicates the first k Mask weight values ​​at each grid point; Among them, the λ k The calculation methods include: a calculation method based on polynomial attenuation masks and a calculation method based on Gaussian attenuation masks; The calculation formula for the polynomial attenuation mask calculation method is as follows: in, The x-axis coordinates of pixels in the feature map are represented. W Indicates the width of the feature map. H Indicates the height of the feature map. This represents the y-axis coordinate of a pixel in the feature map. n The order of the polynomial controlling the decay rate; The calculation formula for the Gaussian attenuation mask-based calculation method is as follows: in, G The exp represents the magnitude coefficient of the Gaussian function, and exp represents the exponential function. σ 1 represents the standard deviation of the Gaussian distribution along the horizontal axis. σ 2 represents the standard deviation of the Gaussian distribution along the vertical axis; S407: Input the feature value into the output layer to obtain the surface defect probability value of the magnesium oxide board to be identified: in, P This represents the probability value of surface defects in the magnesium oxide board to be identified. Y b Indicates the first b The type of defect, x p Represents the first in the feature graph p The position of each pixel e Represents a constant. Represents the first in the feature graph p The position of the nth pixel is related to the... b The original score for the type of defect, B The total number of defect types; S408: Determine whether the surface defect probability value is greater than the surface defect probability threshold; if so, accurately identify that there is a defect on the surface of the magnesium oxide board to be identified; otherwise, accurately identify that there is no defect on the surface of the magnesium oxide board to be identified. S5: Calculate the surface defect score of the magnesium oxide board to be identified; S6: Based on the surface defect score, automatically select the magnesium oxide board to be identified.

2. The automatic board selection method based on surface defect identification of magnesium oxide board according to claim 1, characterized in that, S2 specifically includes: S201: Spatially align the original surface image of the magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects using an elastic registration algorithm; S202: Perform grayscale processing on the original surface image of the spatially aligned magnesium oxide board to be identified and the original surface image of the magnesium oxide board without surface defects, respectively, to obtain the first grayscale image and the second grayscale image. S203: Perform grayscale normalization processing on the first grayscale image and the second grayscale image respectively; S204: Using the LOG filtering algorithm, the first grayscale image and the second grayscale image after grayscale normalization are denoised to obtain the target surface image and the standard surface image.

3. The automatic board selection method based on surface defect identification of magnesium oxide board according to claim 1, characterized in that, The preliminary identification of the surface of the magnesium oxide board to be identified in step S3, based on the target surface image and the standard surface image, using an image difference algorithm, specifically includes: S301: Calculate the grayscale difference between the target surface image and the standard surface image using the image difference algorithm; S302: Determine whether the grayscale difference is greater than the grayscale threshold; if so, preliminarily identify that there is a defect on the surface of the magnesium oxide board to be identified; otherwise, preliminarily identify that there is no defect on the surface of the magnesium oxide board to be identified.

4. The automatic board selection method based on surface defect identification of magnesium oxide board according to claim 1, characterized in that, S5 specifically includes: S501: Calculate the area ratio of defects in the magnesium oxide board to be identified; S502: Based on the type of defect in the magnesium oxide board to be identified, set the severity factor of the defect in the magnesium oxide board to be identified; S503: Set the positional influence factor of the defect in the magnesium oxide board to be identified based on the center distance between the defect in the magnesium oxide board to be identified and the target surface image; S504: Calculate the surface defect score of the magnesium oxide board to be identified based on the area ratio, the severity factor, and the location influence factor.

5. The automatic board selection method based on surface defect identification of magnesium oxide board according to claim 4, characterized in that, The types of defects in the magnesium oxide board to be identified include: structural defects, appearance defects, and functional defects; The structural defects include: cracks, voids, and broken edges or missing corners; The appearance defects include: stains, scratches, and peeling; The functional defects include: insufficient adhesive, delamination, and bulging.

6. The automatic board selection method based on surface defect identification of magnesium oxide board according to claim 1, characterized in that, Specifically, S6 is: Determine whether the surface defect score of the magnesium oxide board to be identified is greater than the surface defect score threshold; if so, identify the magnesium oxide board to be identified as a scrap magnesium oxide board and remove it; otherwise, identify the magnesium oxide board to be identified as a defective magnesium oxide board and retain it for automatic board selection.

7. An automatic board selection system based on surface defect identification of magnesium oxide board, characterized in that, include: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the automatic board selection method based on surface defect identification of magnesium oxide board as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the automatic board selection method based on the identification of surface defects of magnesium oxide board as described in any one of claims 1 to 6.

Images