Image-based gas diffusion layer sealant line integrity identification and judgment method

Abstract

The invention discloses an image-based gas diffusion layer sealant line integrity identification and judgment method, and relates to the field of image identification and fuel cell membrane electrode production and manufacturing. Through the image recognition and processing method, the glue line coating condition of the gas diffusion layer is recognized, the non-uniform glue line coating condition is recognized, an inspection means is provided for glue line coating inspection, and the problem that the glue line integrity inspection is low in automation degree is solved; by proposing the concept of the effective coating area of the glue line, a quantitative evaluation method for the coating effect of the glue line is provided, and the problem that manual inspection standards are not uniform is solved.

Description

Technical Field

[0001] The invention relates to the field of image recognition and fuel cell membrane electrode production and manufacturing, and in particular to an image-based gas diffusion layer sealant line integrity recognition and determination method. Background Art

[0002] As a potential small-scale hydrogen energy consumption and energy conversion device, fuel cells have shown great application potential in many fields such as automobiles, shipping, and fixed power generation equipment. The membrane electrode is the main energy conversion site of the fuel cell. It is a thin layer of porous structure formed by the close bonding of the proton exchange membrane, the catalyst layer, and the gas diffusion layer. The proton exchange membrane coated with the catalyst is fragile and easy to wrinkle, and requires external support to maintain the mechanical structure of the certificate. At the same time, membrane electrode leakage is an important evaluation indicator of the fuel cell stack, and the sealing of glue can effectively prevent leakage. The use of glue wire to reasonably seal the gas diffusion layer can effectively solve the above urgent needs.

[0003] The gas diffusion layer is an important component of the fuel cell, providing structural support and heat and mass transfer. Using a completely closed sealing glue line to achieve effective bonding between the gas diffusion layer and the proton exchange membrane can effectively fix the shape of the proton exchange membrane and improve the sealing of the fuel cell membrane electrode. In actual production, uneven glue line coating is one of the important factors affecting the degree of carbon paper bonding, and is also an important reason for the increase in the waste rate of the gas diffusion layer.

[0004] Image processing is a technology that has developed rapidly in recent years. Image signals are easy to obtain, and the processing difficulty is low and the processing speed is fast. It has been widely used in the industrial field. For example, the patent application with the publication number CN111600051A discloses a proton exchange membrane fuel cell fault diagnosis method based on image processing, including: obtaining one-dimensional flooding and dehydration fault voltage data from different PEMFC systems; converting the one-dimensional data into a two-dimensional image; extracting two image features of the image grayscale decrease rate and the pixel grayscale upward fluctuation ratio from the two-dimensional image; and using the K-means clustering algorithm to classify and diagnose the image features. After the voltage signal is converted into an image, the image pixel value range is large, the change is obvious, and it is easy to extract features for distinction, and the fault diagnosis has higher accuracy. For another example, Zhang Sixiang et al. proposed a method for detecting thermal battery assembly defects based on image analysis: by analyzing the standard single cell and defective single cell of the thermal battery, the internal structural characteristics of the thermal battery are self-compared to realize the overall inversion of the single cell in the thermal battery, the missing current collector of the single cell, and the wrong assembly order detection, and a non-destructive detection system is established. Template matching is used to segment the battery stack in the battery image taken by the X-ray machine, and vertical scanning is performed on different positions of the battery stack to generate grayscale and position waveforms. Defects are judged by self-comparison and multi-position structural feature parameters. The final accuracy of non-destructive battery detection is 94%, and the accuracy of defective battery detection is 96%. (Zhang Sixiang, Wang Zhe, Zhang Peng, et al. Non-destructive detection method of thermal battery defects based on image analysis [J]. Sensors and Microsystems, 2021, 40(6): 4.)

[0005] There are still few methods for fuel cell component detection based on image recognition, and there is a large room for application in this area. Summary of the invention

[0006] In order to enhance the reliability of the overall sealing of the fuel cell membrane electrode, to make up for the shortcomings of the prior art, and to accurately identify the gas diffusion layer with incomplete glue line coating, the present invention proposes a method for determining the integrity of the gas diffusion layer sealant glue line based on image recognition, which realizes the extraction of features such as the length and angle of the glue line in the gas diffusion layer image through an algorithm, and uses the extracted parameters to determine the distribution of the glue line. The method is simple to operate, the measured data is accurate and reliable, the calculation speed is fast, and it can effectively monitor the defects of the gas diffusion layer glue line.

[0007] The present invention can be implemented by the following technical solutions:

[0008] The present invention provides an image-based method for identifying and determining the integrity of a gas diffusion layer sealant line. First, a gas diffusion layer with a completed sealant line coating is placed on a fixed workstation, and a camera with calibrated parameters is used to take a sealant line picture. The preprocessed picture is used as the input of an image recognition model. The image in the ROI (Region Of Interest) is selected using the Otsu method to select a suitable grayscale threshold, and then the grayscale threshold method, Canny edge recognition, and probabilistic Hough transform method are used to identify the specific position of the sealant line. Finally, the percentage of the characteristic region area to the total area is calculated to determine whether the sealant line coating meets the requirements. The method specifically includes the following steps:

[0009] (1) Fix the relative position of the camera and the workstation to ensure that the camera can completely capture the gas diffusion layer on the workstation into the image, and perform orthogonal calibration on the camera;

[0010] (2) taking an image of the gas diffusion layer sprayed with the sealant line at the work station, and defining the target area belonging to the gas diffusion layer in the image;

[0011] (3) For pixels in the target area, the grayscale threshold is selected using the Otsu method, and the sealant line and the gas diffusion layer are divided according to the brightness;

[0012] (4) Using the probabilistic Hough transform method, the starting point and the end point of each sealant line are identified. The sealant line with the starting point and the end point at both ends is identified as one, and the effective packaging area of ​​the area enclosed by each sealant line is calculated;

[0013] (5) Taking the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant line as a reference, calculate the ratio of the effective packaging area in step (4) to the standard packaging area, and use this ratio as a criterion for judging the qualified rate of the sealant line coating.

[0014] Preferably, in step (1), the method used for orthodontic calibration of the camera is the Zhang Zhengyou calibration method.

[0015] Specifically, camera orthodontic calibration is achieved according to the following steps:

[0016] Step S11: Select a black and white rectangular chessboard as a calibration object, wherein the color blocks in the chessboard are uniform in size and the black and white blocks are distributed in an orderly manner;

[0017] Step S12: using the camera to take at least three photos of the calibration object at different spatial positions and different postures, where the calibration object is completely included in the photos;

[0018] Step S13: Identify the internal feature points of the chessboard and calculate the camera's internal and external parameter matrices;

[0019] Step S14: Use the intrinsic parameter matrix to achieve orthodontics of the camera photo, and use the extrinsic parameter matrix to determine the spatial position of the camera.

[0020] Furthermore, an external light source for supplementing light at the workstation is provided near the workstation;

[0021] In step (2), before taking the image, the position of the external light source is adjusted so that the brightness difference between the sealant line and the gas diffusion layer in the image taken by the camera is obvious.

[0022] Preferably, in step (2), the image is converted into a grayscale image before delineating the target area.

[0023] Convert the image into a grayscale image and use the following formula to calculate the grayscale value of the image:

[0024] For an 8-bit RGB image, there are three color channels: red, blue, and green. To convert an RGB color image into a grayscale image, it is necessary to average the brightness values ​​of the three RGB channels in the image. The brightness values ​​of the three RGB channels are recorded as Lr, Lb, and L respectively, so the grayscale value L gray The calculation method is as follows:

[0025]

[0026] Furthermore, in step (3), when the grayscale threshold is selected using the Otsu method, the specific steps are as follows:

[0027] Step S31: Calculate the grayscale histogram and inter-class variance:

[0028] Step S311: Count the grayscale histogram of pixel distribution in the target area:

[0029] For an 8-bit grayscale image, the grayscale value range is 0 to 255, and the total number of pixels in the image is recorded as N. Then the pixel frequency p(i) with grayscale value i is:

[0030]

[0031] Where n(i) is the number of pixels with gray value i;

[0032] Step S312: Calculate the total average gray value μ(g), the calculation formula is:

[0033]

[0034] Step S313: traverse possible grayscale thresholds t:

[0035] The grayscale values ​​in the image are divided into two categories: greater than t and less than or equal to t. The weight values ​​of the two categories are w 1 (t), w 2(t) and the average gray value μ1(t), μ2(t) are calculated as follows:

[0036]

[0037] w 2 (t) = 1-w 1 (t),

[0038]

[0039] Step S314: Calculate the inter-class variance

[0040]

[0041] Step S32: Select the gray value corresponding to the maximum inter-class variance and use it as the gray threshold Lt hresh , binarize the pixels in the image and calculate the pixel value Lpixel in the target area using the following method:

[0042]

[0043] More specifically, in step (4), when the probabilistic Hough transform method is used to identify the starting point and the end point of the sealant line on each side, the specific steps include:

[0044] Step S41: converting all pixels outside the target area into black pixels;

[0045] Step S42: Identify the sealant line in the image, obtain the coordinates of the starting point and the ending point of the sealant line, and distinguish the position of the identified sealant line.

[0046] Furthermore, when calculating the effective packaging area of ​​the area enclosed by each sealant line, first determine the positions of the four vertices of the outer quadrilateral corresponding to the identified sealant line; then calculate the center point of the outer quadrilateral, connect the center point with the starting point and end point of all identified sealant lines to form several triangles, and calculate the sum of the areas of all triangles, which is the effective packaging area.

[0047] Preferably, in step (5), if the ratio of the effective packaging area to the standard packaging area is greater than 98%, the sealant line coating is judged to be qualified.

[0048] The present invention also provides an image-based gas diffusion layer sealant line integrity recognition and determination device, comprising:

[0049] A photographing unit, used for photographing an image of a gas diffusion layer sprayed with a sealant line at a work station;

[0050] An image processing unit, used to use the image captured by the capturing unit as an input of an image recognition model, delineate a target area in the image belonging to the gas diffusion layer, and select a grayscale threshold for pixels in the target area using the Otsu method, delineate the sealant line and the gas diffusion layer according to different brightness, and identify the starting point and the ending point of the sealant line on each side;

[0051] A calculation unit is used to calculate the effective packaging area of ​​the area enclosed by each sealant line, identify the sealant line with a starting point and an end point at both ends as one; and calculate the ratio of the effective packaging area to the standard packaging area based on the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant lines;

[0052] The result output unit is used to input the calculation result in the calculation unit and make a judgment, and output the judgment result of whether the sealant line coating is qualified.

[0053] Beneficial effects of the present invention:

[0054] (1) Through image recognition and processing methods, the coating condition of the rubber line of the gas diffusion layer is identified, and the uneven coating of the rubber line is identified, which provides an inspection method for the coating inspection of the rubber line and overcomes the problem of low automation level of the rubber line integrity inspection;

[0055] (2) By proposing the concept of effective coating area of ​​glue line, a quantitative evaluation method for glue line coating effect is provided, overcoming the problem of inconsistent manual inspection standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 The present invention is a flow chart of the method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image.

[0057] Figure 2 Schematic diagram of the calibration board.

[0058] Figure 3 Schematic diagram of the shooting workbench; among them, ① is the lighting equipment, ② is the place where the gas diffusion layer is placed, and ③ is the camera equipment.

[0059] Figure 4 It is the ROI segmentation map.

[0060] Figure 5 This is the result of glue line identification.

[0061] Figure 6 Schematic diagram for calculating effective package area.

[0062] Figure 7 This is a schematic diagram of unqualified glue wire packaging. DETAILED DESCRIPTION

[0063] The process of the gas diffusion layer sealant line integrity recognition and determination method based on an image of the present invention is as follows: Figure 1 shown.

[0064] The determination method comprises the following steps: firstly, placing the gas diffusion layer with the glue line coated on a fixed workstation, taking a glue line picture with a camera calibrated with parameters, taking the pre-processed picture as the input of the image recognition model, selecting a suitable grayscale threshold for the image in the ROI (Region Of Interest, target region) using the Otsu method, then identifying the specific position of the glue line using the grayscale threshold method, Canny edge recognition, and probabilistic Hough transform method, and finally calculating the percentage of the characteristic region area to the total area to determine whether the glue line coating meets the requirements, specifically comprising the following steps:

[0065] (1) Fix the relative position of the camera and the workstation to ensure that the camera can completely capture the gas diffusion layer on the workstation into the image, and perform orthogonal calibration on the camera;

[0066] (2) taking an image of the gas diffusion layer sprayed with the sealant line at the work station, and defining the target area belonging to the gas diffusion layer in the image;

[0067] (3) For pixels in the target area, the grayscale threshold is selected using the Otsu method, and the sealant line and the gas diffusion layer are divided according to the brightness;

[0068] (4) Using the probabilistic Hough transform method, the starting point and the end point of each sealant line are identified. The sealant line with the starting point and the end point at both ends is identified as one, and the effective packaging area of ​​the area enclosed by each sealant line is calculated;

[0069] (5) Taking the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant line as a reference, calculate the ratio of the effective packaging area in step (4) to the standard packaging area, and use this ratio as a criterion for judging the qualified rate of the sealant line coating.

[0070] Specifically, in step (1), the method used for orthodontic calibration of the camera is the Zhang Zhengyou calibration method.

[0071] Specifically, camera orthodontic calibration is achieved according to the following steps:

[0072] Step S11: Select a black and white rectangular chessboard as a calibration object, wherein the color blocks in the chessboard are uniform in size and the black and white blocks are distributed in an orderly manner;

[0073] Step S12: using the camera to take at least three photos of the calibration object at different spatial positions and different postures, where the calibration object is completely included in the photos;

[0074] Step S13: Identify the internal feature points of the chessboard and calculate the camera's internal and external parameter matrices;

[0075] Step S14: Use the intrinsic parameter matrix to achieve orthodontics of the camera photo, and use the extrinsic parameter matrix to determine the spatial position of the camera.

[0076] Furthermore, an external light source for supplementing light at the workstation is provided near the workstation;

[0077] In step (2), before taking the image, the position of the external light source is adjusted so that the brightness difference between the sealant line and the gas diffusion layer in the image taken by the camera is obvious.

[0078] Preferably, in step (2), the image is converted into a grayscale image before delineating the target area.

[0079] Convert the image into a grayscale image and use the following formula to calculate the grayscale value of the image:

[0080] For an 8-bit RGB image, there are three color channels: red, blue, and green. To convert an RGB color image into a grayscale image, it is necessary to average the brightness values ​​of the three RGB channels in the image. The brightness values ​​of the three RGB channels are recorded as Lr, Lb, and Lg respectively, so the grayscale value L gray The calculation method is as follows:

[0081]

[0082] Furthermore, in step (3), when the grayscale threshold is selected using the Otsu method, the specific steps are as follows:

[0083] Step S31: Calculate the grayscale histogram and inter-class variance:

[0084] Step S311: Count the grayscale histogram of pixel distribution in the target area:

[0085] For an 8-bit grayscale image, the grayscale value range is 0 to 255, and the total number of pixels in the image is recorded as N. Then the pixel frequency p(i) with grayscale value i is:

[0086]

[0087] Where n(i) is the number of pixels with gray value i;

[0088] Step S312: Calculate the total average gray value μ(g), the calculation formula is:

[0089]

[0090] Step S313: traverse possible grayscale thresholds t:

[0091] The grayscale values ​​in the image are divided into two categories: greater than t and less than or equal to t. The weight values ​​of the two categories are w 1 (t), w 2 (t) and the average gray value μ1(t), μ2(t) are calculated as follows:

[0092]

[0093] w 2 (t) = 1-w 1 (t),

[0094]

[0095] Step S314: Calculate the inter-class variance

[0096]

[0097] Step S32: Select the gray value corresponding to the maximum inter-class variance and use it as the gray threshold Lthres h , binarize the pixels in the image and calculate the pixel value Lpixel in the target area using the following method:

[0098]

[0099] More specifically, in step (4), when the probabilistic Hough transform method is used to identify the starting point and the end point of the sealant line on each side, the specific steps include:

[0100] Step S41: converting all pixels outside the target area into black pixels;

[0101] Step S42: Identify the sealant line in the image, obtain the coordinates of the starting point and the ending point of the sealant line, and distinguish the position of the identified sealant line.

[0102] Furthermore, when calculating the effective packaging area of ​​the area enclosed by each sealant line, first determine the positions of the four vertices of the outer quadrilateral corresponding to the identified sealant line; then calculate the center point of the outer quadrilateral, connect the center point with the starting point and end point of all identified sealant lines to form several triangles, and calculate the sum of the areas of all triangles, which is the effective packaging area.

[0103] Preferably, in step (5), if the ratio of the effective packaging area to the standard packaging area is greater than 98%, the sealant line coating is judged to be qualified.

[0104] The present invention also provides an image-based gas diffusion layer sealant line integrity recognition and determination device, comprising:

[0105] A photographing unit, used for photographing an image of a gas diffusion layer sprayed with a sealant line at a work station;

[0106] An image processing unit, used to use the image captured by the capturing unit as an input of an image recognition model, delineate a target area in the image belonging to the gas diffusion layer, and select a grayscale threshold for pixels in the target area using the Otsu method, delineate the sealant line and the gas diffusion layer according to different brightness, and identify the starting point and the ending point of the sealant line on each side;

[0107] A calculation unit is used to calculate the effective packaging area of ​​the area enclosed by each sealant line, identify the sealant line with a starting point and an end point at both ends as one; and calculate the ratio of the effective packaging area to the standard packaging area based on the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant lines;

[0108] The result output unit is used to input the calculation result in the calculation unit and make a judgment, and output the judgment result of whether the sealant line coating is qualified.

[0109] The embodiments of the present invention are described in detail with reference to the following embodiments as examples.

[0110] Example 1

[0111] like Figure 2 As shown, the camera calibration is first performed. A checkerboard calibration plate is made. The calibration plate has a flat black and white checkerboard with a length and width of 30cm and 21cm respectively as a standard calibration prop. The size of a single grid is 3cm×3cm. For any grid, the grids in the four directions of up, down, left, and right have different colors from the grid. The colors of the grids in the entire chessboard are only black and white.

[0112] Use a camera to photograph the chessboard from enough angles. All black and white squares in the chessboard need to be completely captured in the acquired pictures. Calibrate the camera's intrinsic parameters. Use the camera calibration function in the opencv open source function package to calculate the camera's intrinsic parameter matrix. Specify a fixed position as the origin of the camera, and specify the directions of each coordinate axis in the three-dimensional Cartesian coordinate system. Calculate the spatial position of the camera's optical center in the Cartesian coordinate system. Calculate the Euler angle of the camera direction relative to the positive direction of the x-axis in the standard Cartesian coordinate system. The three Euler angles are denoted as θr o w、θpitc h , θyaw; retain the camera's internal and external parameter matrix. In the subsequent camera-based image parameter acquisition process, it is necessary to use the internal and external parameter matrix to calibrate the image first, and then process the image.

[0113] After calibration, fix the position of the gas diffusion layer:

[0114] (1) placing the gas diffusion layer coated with the glue line in a fixture to ensure that the fixture does not block the image of the gas diffusion layer in the image captured by the camera;

[0115] (2) Place the gas diffusion layer coated with the glue line, making sure that the side of the gas diffusion layer coated with the glue line faces the camera;

[0116] (3) Adjust the position of the lighting fixture to ensure that there is a clear brightness difference between the unsprayed glue line and the sprayed glue line in the gas diffusion layer in the field of view. In the grayscale image with 8-bit color depth, the brightness value difference should be greater than 30;

[0117] (4) The spatial position of the fixed lamp and camera relative to the gas diffusion layer fixture;

[0118] like Figure 3 As shown, the calibrated camera is installed in the device shown in the figure to shoot the gas diffusion layer sprayed with glue lines;

[0119] like Figure 4 As shown, in this embodiment, an image of the gas diffusion layer is taken, and its ROI is obtained as shown in the figure.

[0120] like Figure 5 As shown in the figure, the grayscale threshold function of the Otsu method and the probability Hough transform function in the opencv open source program are used to identify the rubber lines coated around the GDL (gas diffusion layer). A total of four target rubber lines are obtained. The coordinate information of the starting point and the end point is a two-dimensional Cartesian coordinate system. The positive direction of the x-axis is from left to right, the positive direction of the y-axis is from top to bottom, the image origin is the upper left corner of the image, the length unit is one pixel length, and the area unit is one pixel area. The rubber lines identified in this example are named upper rubber line, left rubber line, right rubber line, and lower rubber line. The positions of the starting points and end points of the four rubber lines in the image coordinate system are: upper rubber line: (48, 39), (1291, 28); left rubber line: (48, 39), (48, 651); lower rubber line: (50, 653), (1291, 636); right rubber line: (1300, 26), (1300, 621). Calculate the linear algebraic equations of the four glue lines in the coordinate system respectively, and obtain the coordinates of the intersection points between the glue lines. The coordinates of the intersection points of the upper left corner, upper right corner, lower left corner, and lower right corner in the figure are (48, 39), (1300, 27.92), (48, 653.03), and (1300, 635.88) respectively. Averaging the four corner points, we can get the coordinates of the center point (674, 338.96). Calculate the enclosed area of ​​the four corner points, and we can get the total package area in the ROI area to be 764965.74 unit pixels.

[0121] like Figure 6As shown, link the glue line endpoints and the ROI center of gravity, calculate the area of ​​all the linked geometric figures, and calculate the ratio of this area to the overall ROI area. In this example, the calculation result of the effective packaging area is: 757216.96 pixels, and the effective packaging ratio is: 98.99%. This ratio is greater than 98%, so the glue line coating is qualified.

[0122] Example 2

[0123] The following are examples of situations where the glue line is unqualified:

[0124] The steps are the same as those in Example 1, except that the position of the glue line is changed.

[0125] like Figure 7 As shown in the figure, the outer frame is the frame line of the area of ​​interest, and the four lines parallel to the outer frame are the schematic lines of the identified glue line position. The right half of the upper glue line is missing, and the effective pixel area of ​​the glue line packaging area is 2,515,506, of which the effective packaging area represented by the oblique line area is 314,438.25, and the effective packaging area accounts for 87.5% of the total packaging area, which does not meet the requirements.

Claims

1. A method for identifying and determining the integrity of a gas diffusion layer sealant line based on an image, characterized in that: The following steps are involved: (1) Fix the relative position of the camera and the workstation to ensure that the camera can completely capture the gas diffusion layer on the workstation into the image, and perform orthogonal calibration on the camera; (2) taking an image of the gas diffusion layer sprayed with the sealant line at the work station, and defining the target area belonging to the gas diffusion layer in the image; (3) For pixels in the target area, the grayscale threshold is selected using the Otsu method, and the sealant line and the gas diffusion layer are divided according to the brightness; (4) Using the probabilistic Hough transform method, the starting point and the end point of each sealant line are identified. The sealant line with the starting point and the end point at both ends is identified as one, and the effective packaging area of ​​the area enclosed by each sealant line is calculated; (5) Taking the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant line as a reference, calculate the ratio of the effective packaging area in step (4) to the standard packaging area, and use this ratio as a criterion for judging the qualified rate of the sealant line coating.

2. The method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image according to claim 1, characterized in that: In step (1), the method used for orthodontic calibration of the camera is Zhang Zhengyou calibration method.

3. The method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image according to claim 2, characterized in that: Follow the steps below to perform camera orthodontic calibration: Step S11: Select a black and white rectangular chessboard as a calibration object, wherein the color blocks in the chessboard are uniform in size and the black and white blocks are distributed in an orderly manner; Step S12: using the camera to take at least three photos of the calibration object at different spatial positions and different postures, where the calibration object is completely included in the photos; Step S13: Identify the internal feature points of the chessboard and calculate the camera's internal and external parameter matrices; Step S14: Use the intrinsic parameter matrix to achieve orthodontics of the camera photo, and use the extrinsic parameter matrix to determine the spatial position of the camera.

4. The method for identifying and determining the integrity of a gas diffusion layer sealant line based on an image according to claim 1, characterized in that: An external light source for supplementing light to the workstation is also provided near the workstation; In step (2), before taking the image, the position of the external light source is adjusted so that the brightness difference between the sealant line and the gas diffusion layer in the image taken by the camera is obvious.

5. The method for identifying and determining the integrity of a gas diffusion layer sealant line based on an image according to claim 1, characterized in that: In step (2), the image is converted into a grayscale image before delineating the target area.

6. The method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image according to claim 5, characterized in that: In step (3), when the grayscale threshold is selected using the Otsu method, the specific steps are as follows: Step S31: Calculate the grayscale histogram and inter-class variance: Step S311: Count the grayscale histogram of pixel distribution in the target area: For an 8-bit grayscale image, the grayscale value range is 0 to 255, and the total number of pixels in the image is recorded as N. Then the pixel frequency p(i) with grayscale value i is: Where n(i) is the number of pixels with gray value i; Step S312: Calculate the total average gray value μ(g), the calculation formula is: Step S313: traverse possible grayscale thresholds t: The grayscale values ​​in the image are divided into two categories: greater than t and less than or equal to t. The weight values ​​w1(t), w2(t) of the two categories and the average grayscale values ​​μ1(t), μ2(t) are calculated as follows: w2(t)=1-w1(t), Step S314: Calculate the inter-class variance Step S32: Select the gray value corresponding to the maximum inter-class variance and use it as the gray threshold L thresh , binarize the pixels in the image and calculate the pixel value L in the target area pixel , calculated using the following method:

7. The method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image according to claim 6, characterized in that: In step (4), when the probabilistic Hough transform method is used to identify the starting point and the ending point of the sealant line on each side, the specific steps include: Step S41: converting all pixels outside the target area into black pixels; Step S42: Identify the sealant line in the image, obtain the coordinates of the starting point and the ending point of the sealant line, and distinguish the position of the identified sealant line.

8. The method for identifying and determining the integrity of the gas diffusion layer sealant line based on an image according to claim 7, characterized in that: When calculating the effective packaging area of ​​the area enclosed by each sealant line, first determine the positions of the four vertices of the outer quadrilateral corresponding to the identified sealant line; then calculate the center point of the outer quadrilateral, connect the center point with the starting point and end point of all identified sealant lines to form several triangles, and calculate the sum of the areas of all triangles, which is the effective packaging area.

9. The method for identifying and determining the integrity of a gas diffusion layer sealant line based on an image according to claim 1, characterized in that: In step (5), if the ratio of the effective packaging area to the standard packaging area is greater than 98%, the sealant line coating is determined to be qualified.

10. An image-based gas diffusion layer sealant line integrity recognition and determination device, characterized in that: include: A photographing unit, used for photographing an image of a gas diffusion layer sprayed with a sealant line at a work station; An image processing unit, used to use the image captured by the capturing unit as an input of an image recognition model, delineate a target area in the image belonging to the gas diffusion layer, and select a grayscale threshold for pixels in the target area using the Otsu method, delineate the sealant line and the gas diffusion layer according to different brightness, and identify the starting point and the ending point of the sealant line on each side; A calculation unit, used for calculating the effective packaging area of ​​the area enclosed by each sealant line, and identifying the sealant line with a starting point and an end point at both ends as one; Taking the effective packaging area of ​​the sealant line in the gas diffusion layer with complete sealant line as the benchmark, calculate the ratio of the effective packaging area to the standard packaging area; The result output unit is used to input the calculation result in the calculation unit and make a judgment, and output the judgment result of whether the sealant line coating is qualified.

Images