A burr detection method for solid-state battery electrode sheet cutting edges

Abstract

The present application belongs to the technical field of battery pole piece burr identification, and particularly relates to a burr detection method for solid-state battery pole piece cutting edge. Through an effectiveness screening mechanism, the present application can accurately distinguish between effective image acquisition and ineffective image acquisition, trigger a compensation mechanism for ineffective images, and through interpolation compensation or secondary image acquisition strategies, ensure the integrity and accuracy of image data. In image splicing, a splicing method based on spatial position coding is adopted, combined with linear interpolation and weighted fusion processing, effectively eliminating the influence of splicing seams and overlapping areas. Through an edge feature extraction method, the cutting edge contour is outlined. Finally, based on the cutting edge contour, combined with local curvature change analysis and geometric feature extraction, the identification and quantitative analysis of the burr area are realized. Not only the distribution position of the burr is determined, but also the burr defect grade is evaluated through a weighted scoring mechanism, providing corresponding priority guidance for subsequent repair work.

Description

Technical Field

[0001] This invention belongs to the field of battery electrode burr identification technology, specifically relating to a burr detection method for the cutting edge of solid-state battery electrodes. Background Technology

[0002] With the widespread application of solid-state batteries in new energy vehicles, energy storage equipment and other fields, the production process has put forward higher requirements for the manufacturing precision and consistency of electrode sheets. During the cutting and forming process of solid-state battery electrode sheets, defects such as burrs, chipping or micro-cracks are easily generated at the edges of the electrode sheets due to reasons such as tool wear, fluctuation of cutting parameters, equipment vibration or release of internal stress of materials. These defects not only affect the geometric dimensional accuracy of the electrode sheets, but may also cause problems such as short circuits and poor interface contact during the stacking or cell assembly process, thereby reducing the reliability and service life of the battery.

[0003] Currently, the commonly used burr detection methods in the industry mainly rely on manual visual inspection or traditional image processing methods. Manual inspection suffers from low efficiency, strong subjectivity, and difficulty in adapting to the needs of high-speed production. Traditional image processing methods are prone to false detection or missed detection when the electrode surface is highly reflective, the edge contrast is low, and the background is complex. The detection accuracy and stability are difficult to meet the requirements of high-speed automated production lines. Based on this, this solution provides a burr detection method for the cutting edge of solid-state battery electrode to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a burr detection method for the cutting edge of solid-state battery electrode sheets, which can effectively overcome the shortcomings of low efficiency and strong subjectivity of manual inspection and the susceptibility of traditional image processing methods to environmental interference.

[0005] The specific technical solution adopted by this invention is as follows:

[0006] A method for detecting burrs on the cut edges of solid-state battery electrode sheets includes:

[0007] Multiple rows of staggered detection points are arranged along the width of the battery electrode, and image information of each detection point is collected along the cutting edge of the battery electrode to form a benchmark image dataset.

[0008] The detection point images in the benchmark image dataset are filtered for validity, and valid and invalid images are identified. The detection points corresponding to invalid images are marked and the image compensation mechanism is triggered.

[0009] The detection points corresponding to invalid acquired images are compensated according to the image compensation mechanism, and the compensated images are re-filtered for validity until all detection point images are valid.

[0010] The valid images are stitched together to form a complete continuous image of the electrode cutting edge, and edge features are extracted from the continuous image of the electrode cutting edge to output the cutting edge contour.

[0011] Burr region identification is performed based on the cut edge contour. The height, width and distribution density of the burrs are extracted to perform quantitative analysis of the burr region, and the distribution location and burr defect level of the burr region are determined.

[0012] In a preferred embodiment, the step of forming the reference image dataset includes:

[0013] Multiple rows of detection points are set along the width of the battery electrode, with the detection points in adjacent rows being staggered and the spacing between adjacent detection points in a single row being consistent.

[0014] Each row of detection points acquires images synchronously from a perspective perpendicular to the cropping edge, and the acquisition frequency of each detection point remains consistent.

[0015] The original images acquired at each detection point are preprocessed to generate standardized detection point images. The preprocessing steps include grayscale correction, noise filtering, and distortion correction.

[0016] The preprocessed detection point images are encoded and stored according to the spatial location of the detection points, and a benchmark image dataset containing timestamps, location coordinates and image data is constructed.

[0017] In a preferred embodiment, the step of filtering the detection point images in the benchmark image dataset to identify valid and invalid acquired images includes:

[0018] The edge gradient magnitude of the detection point image is calculated, and the average edge gradient magnitude of each region in the same detection point image is calculated to obtain the sharpness index of the detection point image.

[0019] The standard deviation of the gray value distribution in the detection point image is statistically analyzed and output as a uniformity index.

[0020] A threshold judgment method is used to judge the sharpness index and uniformity index of the detection point image respectively. The threshold includes a first screening threshold corresponding to the sharpness index and a second screening threshold corresponding to the uniformity index.

[0021] If the sharpness index of the detection point image is greater than the first screening threshold and the uniformity index is greater than the second screening threshold, the image is determined to be a valid acquisition image; otherwise, it is marked as an invalid acquisition image.

[0022] In a preferred embodiment, the step of compensating the detection points corresponding to invalid acquired images according to the image compensation mechanism includes:

[0023] Spatial location of invalid acquisition areas in invalid acquisition images is determined, and spatial locations of adjacent upper and lower row detection points are matched.

[0024] If valid images exist in the corresponding regions of adjacent upper and lower rows of detection points, interpolation compensation is performed using the image data from the corresponding regions of the adjacent rows of detection points to generate a completed image, which is then incorporated into the baseline image dataset.

[0025] If there are invalid images in the corresponding areas of adjacent upper and lower rows of detection points, the invalid areas of the current detection point and the adjacent rows of detection points are jointly marked, and all invalid areas under the joint marking are simultaneously re-sampled.

[0026] In a preferred embodiment, the step of performing image stitching processing on the valid images to form a complete continuous image of the electrode cutting edge includes:

[0027] Based on the spatial location encoding of the detection points, the effectively acquired images are aligned according to their coordinate order;

[0028] The aligned and valid acquired images are stitched together for the first time to generate a preliminary stitched image. The stitching seams and overlapping areas in the preliminary stitched image are detected, and the stitching deviation area is output. The stitching deviation area includes the insufficient seam area and the excessive overlap area.

[0029] In areas with insufficient seams, missing pixels are supplemented by linear interpolation to form an interpolated complete area;

[0030] In areas of excessive overlap, the sharpness and uniformity indices of the two valid acquired images corresponding to the overlapping area are weighted and fused, and the result is a fusion score. The valid acquired image with the higher fusion score is retained, while the corresponding area of ​​the valid acquired image with the lower fusion score is cropped and removed.

[0031] Based on the selected valid acquired images and the interpolated complete area, the initial stitched images are stitched together again to generate a complete continuous image of the electrode cutting edge.

[0032] In a preferred embodiment, the step of detecting seams and overlapping areas in the preliminary stitched image and outputting the stitching deviation area includes:

[0033] The initial stitched image is subjected to grayscale normalization processing, and the grayscale gradient change amplitude of pixels within a set width on both sides of the stitching seam is extracted.

[0034] When the magnitude of the pixel gradient change exceeds the gradient threshold and the number of consecutive pixels reaches the preset length, it is determined that there is a seam in the preliminary stitched image, and the area corresponding to the seam is marked as a seam insufficiency area.

[0035] When the magnitude of pixel gradient change is lower than the gradient threshold, the similarity of the effective acquired images on both sides of the stitching area is calculated, and the area with similarity higher than the similarity threshold is determined as an over-overlapping area.

[0036] In a preferred embodiment, the step of extracting edge features from the continuous image of the electrode cutting edge and outputting the cutting edge contour includes:

[0037] A sliding window is set on the continuous image of the electrode cutting edge to extract the gradient change amplitude of the local area of ​​the image segment by segment, and the edge direction is determined according to the direction of the maximum gradient amplitude.

[0038] According to the edge direction, the edge points in the continuous image of the electrode cutting edge are marked, and the marked edge points are connected in sequence to form a preliminary edge outline.

[0039] The initial edge contour is smoothed to eliminate jagged and burr-like interference in the edge contour, resulting in a smooth edge contour.

[0040] The smoothed edge contour is closed. If there is an unclosed region, a neighborhood search is performed on the endpoints of the unclosed region, and the nearest other endpoint is selected to connect them to complete the contour closure and form a complete trimming edge contour.

[0041] In a preferred embodiment, the step of identifying burr areas based on the cut edge contour, and quantitatively analyzing the burr areas by extracting the height, width, and distribution density of the burrs to determine the distribution location and burr defect level, includes:

[0042] Along the normal direction of the cut edge contour, the neighborhood area is scanned point by point at a set step interval. The local curvature change of the contour line in the neighborhood area is calculated, and protruding areas that exceed the normal range are identified as burr candidate points.

[0043] Geometric features are extracted from burr candidate points, including the height difference, width span, and curvature extrema within the candidate point's neighborhood.

[0044] The extracted height difference, width span, and curvature extreme values ​​are compared with the preset burr feature thresholds to filter out candidate points that meet the burr features and mark them as burr regions.

[0045] The distribution density of burrs is determined by calculating and outputting the number of burrs on the cut edge per unit length;

[0046] The distribution density, height difference, and width span of the burr area are weighted and scored to obtain a comprehensive burr defect score. The comprehensive burr defect score is then matched with a preset level standard to output the burr defect level.

[0047] The present invention also provides a burr detection system for cutting edges of solid-state battery electrodes, using the above-described burr detection method for cutting edges of solid-state battery electrodes, comprising:

[0048] The initialization module is used to arrange multiple rows of staggered detection points along the width of the battery electrode, and to collect image information of each detection point along the cutting edge of the battery electrode to form a reference image dataset.

[0049] The image filtering module is used to filter the detection point images in the benchmark image dataset for validity, identify valid and invalid acquired images, and mark the detection points corresponding to invalid acquired images and trigger the image compensation mechanism.

[0050] The image compensation module is used to compensate the detection points corresponding to invalid acquired images according to the image compensation mechanism, and to re-filter the compensated images until all detection point images are valid.

[0051] The contour extraction module is used to perform image stitching on the valid images to form a complete continuous image of the pole piece cropping edge, and to extract edge features from the continuous image of the pole piece cropping edge to output the cropping edge contour.

[0052] The burr recognition module is used to identify burr areas based on the cut edge contour. It performs quantitative analysis of the burr area by extracting the height, width and distribution density of the burrs, and determines the distribution location of the burr area and the burr defect level.

[0053] And, an electronic device, the electronic device comprising:

[0054] At least one processor;

[0055] and a memory communicatively connected to the at least one processor;

[0056] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the above-described burr detection method for cutting edges of solid-state battery electrodes.

[0057] The technical effects achieved by this invention are as follows:

[0058] This invention constructs a benchmark image dataset by arranging multiple rows of staggered detection points along the width of the battery electrode, ensuring comprehensive coverage of the electrode edge. Then, a validity screening mechanism accurately distinguishes between valid and invalid images, triggering a compensation mechanism for invalid images. Through interpolation compensation or secondary acquisition strategies, the integrity and accuracy of the image data are maximized. During image stitching, a spatial location coding-based stitching method is adopted, combined with linear interpolation and weighted fusion processing, effectively eliminating the influence of stitching seams and overlapping areas. This generates high-quality continuous images of the electrode cut edges. Edge feature extraction is used to outline the cut edge contour, laying the foundation for subsequent burr identification. Finally, based on the cut edge contour, local curvature change analysis and geometric feature extraction are used to identify and quantify the burr region. This not only determines the distribution location of burrs but also assesses the burr defect level through a weighted scoring mechanism, providing corresponding priority guidance for subsequent repair work. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0060] Figure 2 This is a schematic diagram of the system modules of the present invention;

[0061] Figure 3 This is a schematic diagram of the electronic device structure of the present invention. Detailed Implementation

[0062] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0063] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0064] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0065] Please see Figure 1 As shown, the present invention provides a method for burr detection on the cutting edge of a solid-state battery electrode, comprising:

[0066] S1. Arrange multiple rows of staggered detection points along the width direction of the battery electrode, and collect image information of each detection point along the cutting edge of the battery electrode to form a reference image dataset.

[0067] In step S1, when inspecting the burrs on the cut edges of the battery electrode sheets, the distribution parameters of multiple rows of inspection points need to be set according to the actual specifications and inspection accuracy requirements of the battery electrode sheets. These distribution parameters include the number of inspection points per row, the staggered distance between adjacent rows, and the spacing between adjacent inspection points within a single row. This needs to be determined through experimental calibration based on actual inspection requirements to ensure coverage of the entire cut edge area without any blind spots in image acquisition. After the inspection points are arranged, a multi-row camera array is deployed at the corresponding inspection points for synchronous image acquisition. For example, the cameras are numbered 1-19 and installed in three staggered rows. The first row is numbered 2, 5, 8, 11, 14, 17; the second row is numbered 1, 4, 7, 10, 13, 16, 19; and the third row is numbered 3, 6, 9, 12, 15, 18. The second row of cameras is installed parallel to the incoming electrode sheets, and the first row of cameras... The first row of cameras is mounted at a +45° angle to the incoming electrode sheet, while the second row is mounted at a -45° angle. Two sets of optical modules (light source module 1 and light source module 2) are also configured to work with the cameras. These optical modules can be telecentric lenses, etc., with a diameter of 34mm, allowing for mechanical mounting while meeting design field of view and single-pixel accuracy requirements. Both light source modules 1 and 2 are high-brightness coaxial light sources. Light source module 1 is mounted between the first and second row image acquisition units at a +23° angle to the incoming electrode sheet film surface, while light source module 2 is mounted between the second and third row image acquisition units at a -23° angle to the incoming electrode sheet film surface. This achieves coordinated multi-angle illumination and imaging. After image acquisition, a baseline image dataset is compiled to provide a data foundation for subsequent feature extraction and burr recognition. The steps for forming the baseline image dataset include:

[0068] Multiple rows of detection points are set along the width of the battery electrode, with the detection points in adjacent rows being staggered and the spacing between adjacent detection points in a single row being consistent.

[0069] Each row of detection points acquires images synchronously from a perspective perpendicular to the cropping edge, and the acquisition frequency of each detection point remains consistent.

[0070] The original images acquired at each detection point are preprocessed to generate standardized detection point images. The preprocessing steps include grayscale correction, noise filtering, and distortion correction.

[0071] The preprocessed detection point images are stored according to the spatial location encoding of the detection points, and a benchmark image dataset containing timestamps, location coordinates and image data is constructed.

[0072] Specifically, when compiling the detection point images, the images acquired by each detection point are first aligned by timestamp to ensure spatial consistency of multi-view images at the same time. Then, the acquired raw images are preprocessed, including grayscale correction, noise filtering, and distortion correction, to eliminate the effects of uneven illumination, sensor noise, and lens distortion on image quality. The preprocessed detection point images are then stored in an orderly manner according to the spatial location encoding of the detection points, so that each detection point image is associated with a unique location coordinate and timestamp, thereby forming a structured reference image dataset for subsequent use.

[0073] S2. Perform validity screening on the detection point images in the benchmark image dataset, identify valid and invalid acquired images, and mark the detection points corresponding to invalid acquired images and trigger the image compensation mechanism.

[0074] In step S2, after the baseline image dataset is determined, the detection point images it covers can be subjected to corresponding validity checks to ensure that the detection point images used for spur recognition have sufficient imaging quality. Specifically, the detection point images are classified into valid and invalid acquisition images. To ensure the integrity of the overall image, the detection points corresponding to invalid acquisition images are marked, and an image compensation mechanism is triggered to supplement missing or low-quality images. The step of validally filtering the detection point images in the baseline image dataset and identifying valid and invalid acquisition images includes:

[0075] The edge gradient magnitude of the detection point image is calculated, and the average edge gradient magnitude of each region in the same detection point image is calculated to obtain the sharpness index of the detection point image.

[0076] The standard deviation of the gray value distribution in the detection point image is statistically analyzed and output as a uniformity index.

[0077] A threshold judgment method is used to judge the sharpness index and uniformity index of the detection point image respectively. The threshold includes a first screening threshold corresponding to the sharpness index and a second screening threshold corresponding to the uniformity index.

[0078] If the sharpness index of the detection point image is greater than the first screening threshold and the uniformity index is greater than the second screening threshold, the image is determined to be a valid acquisition image; otherwise, it is marked as an invalid acquisition image.

[0079] Specifically, when identifying valid and invalid images, the edge gradient magnitude of each detection point image is first calculated. The edge gradient magnitude reflects the sharpness of the image edges. By averaging the edge gradient magnitudes of various regions in the same detection point image, the sharpness index of that detection point image can be obtained. The higher the sharpness index, the sharper the image edges and the better the imaging quality. Simultaneously, the standard deviation of the grayscale distribution in the detection point image is calculated and output as a uniformity index. The uniformity index reflects the uniformity of the image grayscale distribution. The higher the uniformity index, the more uniform the grayscale distribution of the corresponding detection point image, and the better its imaging quality. The more stable the quality, the better. Then, a threshold judgment method is used to judge the sharpness index and uniformity index of the detection point image. The thresholds include a first screening threshold corresponding to the sharpness index and a second screening threshold corresponding to the uniformity index. The first screening threshold and the second screening threshold can be set according to actual detection needs and experience, and are not limited here. If the sharpness index of the detection point image is greater than the first screening threshold and the uniformity index is greater than the second screening threshold, the detection point image will be determined as a valid acquisition image and can be used for subsequent operations such as spur identification. Otherwise, the corresponding detection point image will be marked as an invalid acquisition image and corresponding compensation processing is required.

[0080] S3. Based on the image compensation mechanism, the detection points corresponding to invalid acquired images are compensated, and the compensated images are re-filtered for validity until all detection point images are valid.

[0081] In step S3, after the invalid acquired images are output, an image compensation mechanism is used to compensate for them. To avoid omissions in subsequent glitch detection, all invalid acquired images are compensated until all detection point images meet the requirements for clarity and uniformity. The step of compensating the detection points corresponding to the invalid acquired images according to the image compensation mechanism includes:

[0082] Spatial location of invalid acquisition areas in invalid acquisition images is determined, and spatial locations of adjacent upper and lower row detection points are matched.

[0083] If valid images exist in the corresponding regions of adjacent upper and lower rows of detection points, interpolation compensation is performed using the image data from the corresponding regions of the adjacent rows of detection points to generate a completed image, which is then incorporated into the baseline image dataset.

[0084] If there are invalid images in the corresponding areas of adjacent upper and lower rows of detection points, the invalid areas of the current detection point and the adjacent rows of detection points are jointly marked, and all invalid areas under the joint marking are simultaneously re-sampled.

[0085] Specifically, during the image compensation mechanism, the spatial location of the invalid acquisition area is first calibrated to determine its specific coordinates within the entire detection point layout. Simultaneously, the spatial locations of the adjacent upper and lower rows of detection points are matched to define the reference area for image compensation. If valid acquisition images exist in the corresponding areas of adjacent upper and lower rows of detection points, interpolation compensation is performed on the invalid acquisition image based on the adjacent valid image data. Methods include bilinear interpolation or cubic spline interpolation. A completed image is generated based on the pixel information of the adjacent valid images, ensuring consistency with the surrounding valid images in terms of color, texture, and edges. The generated completed image is then incorporated into the baseline image dataset for subsequent use. If invalid acquisition images exist in the corresponding areas of adjacent upper and lower rows of detection points... The invalid regions of the current detection point and adjacent rows of detection points are jointly marked. The joint marking can more comprehensively reflect the distribution of invalid regions, which is convenient for subsequent unified processing. Then, based on the specific location and range of all invalid regions under the joint marking, a second sampling is performed. During the second sampling process, the image acquisition quality can be improved by adjusting the camera's exposure time, gain, etc., to ensure that the second-acquired image meets the requirements of sharpness and uniformity. After the compensation processing is completed, the compensated image will be effectively filtered again. The filtering method is the same as described above and will not be repeated here. It should be noted that if the completed image output by the second time using the difference compensation method still does not meet the requirements of sharpness and uniformity, the second sampling process will be directly triggered, skipping the interpolation attempt again to avoid invalid loops.

[0086] S4. Perform image stitching on the valid images to form a complete continuous image of the electrode cutting edge, and extract edge features from the continuous image of the electrode cutting edge to output the cutting edge contour.

[0087] In step S4, after the validity screening of the benchmark image dataset is completed, the valid acquired images can be stitched together to construct a complete continuous image of the electrode cutting edge. Then, by extracting edge features from the continuous image of the electrode cutting edge, the cutting edge contour can be output, providing corresponding geometric information support for subsequent burr recognition. The step of stitching the valid images to form a complete continuous image of the electrode cutting edge includes:

[0088] Based on the spatial location encoding of the detection points, the effectively acquired images are aligned according to their coordinate order;

[0089] The aligned and valid acquired images are stitched together for the first time to generate a preliminary stitched image. The stitching seams and overlapping areas in the preliminary stitched image are detected, and the stitching deviation area is output. The stitching deviation area includes the insufficient seam area and the excessive overlap area.

[0090] In areas with insufficient seams, missing pixels are supplemented by linear interpolation to form an interpolated complete area;

[0091] In areas of excessive overlap, the sharpness and uniformity indices of the two valid acquired images corresponding to the overlapping area are weighted and fused, and the result is a fusion score. The valid acquired image with the higher fusion score is retained, while the corresponding area of ​​the valid acquired image with the lower fusion score is cropped and removed.

[0092] Based on the selected valid acquired image and the interpolated complete area, the initial stitched image is stitched a second time to generate a complete continuous image of the electrode cutting edge;

[0093] Specifically, when outputting continuous images of the electrode cutting edges, the process first aligns all valid acquired images according to coordinate order based on the spatial position encoding of the detection points, ensuring the accuracy of the images in spatial position. Then, the aligned valid acquired images undergo an initial stitching operation to generate a preliminary stitched image. During this process, stitching seams and overlapping areas may appear in the preliminary stitched image, therefore, these need to be detected and the stitching deviation areas output. Stitching deviation areas are mainly divided into two types: insufficient seam areas and excessive overlap areas. For insufficient seam areas, a linear interpolation method is used to supplement missing pixels. Specifically, the missing pixels can be calculated based on the known pixel grayscale values ​​through a linear relationship. The grayscale values ​​of pixels are used to form interpolation and completion areas, making the seams of the effectively acquired images smoother and more natural. For areas with excessive overlap, the sharpness and uniformity of the two effectively acquired images corresponding to the overlapping area need to be weighted and fused to comprehensively consider the quality information of the two images and output a fusion score. Based on the fusion score, the effectively acquired image with the higher score is selected and retained, while the corresponding area of ​​the effectively acquired image with the lower fusion score is cropped and removed to ensure the quality of the stitched image. Finally, the initial stitched image is stitched a second time based on the selected effectively acquired image and the interpolation and completion areas to generate a complete continuous image of the electrode cropping edge.

[0094] Secondly, the steps of detecting seams and overlapping areas in the preliminary stitched image and outputting the stitching deviation area include:

[0095] The initial stitched image is subjected to grayscale normalization processing, and the grayscale gradient change amplitude of pixels within a set width on both sides of the stitching seam is extracted.

[0096] When the magnitude of the pixel gradient change exceeds the gradient threshold and the number of consecutive pixels reaches the preset length, it is determined that there is a seam in the preliminary stitched image, and the area corresponding to the seam is marked as a seam insufficiency area.

[0097] When the magnitude of the pixel gradient change is lower than the gradient threshold, the similarity of the effective acquired images on both sides of the stitching area is calculated, and the area with a similarity higher than the similarity threshold is determined to be an over-overlapping area.

[0098] In the above process, when determining the splicing deviation area, the initial spliced ​​image is first subjected to grayscale normalization to eliminate grayscale differences between different images caused by factors such as lighting. Then, the grayscale gradient change amplitude of pixels within a set width on both sides of the splicing seam is extracted. The position of the splicing seam is located by analyzing the gradient change of pixel grayscale near the splicing seam. That is, when the pixel gradient change amplitude exceeds the gradient threshold and the number of consecutive pixels reaches the preset length, it indicates that there is an edge change, which means that there is a seam in the initial spliced ​​image. At the same time, the area corresponding to this seam is marked as a seam deficiency area for subsequent targeted interpolation and completion processing. When the pixel gradient change amplitude is lower than the gradient threshold, it indicates that there is no obvious edge change near the splicing seam. At this time, the similarity of the effective acquired images on both sides of the splicing area is calculated. By comparing the similarity of the two images in terms of pixel distribution, texture features, etc. in the splicing area, the area with a similarity higher than the similarity threshold is determined as an over-overlapping area, so as to remove redundant images.

[0099] Secondly, the steps of extracting edge features from the continuous image of the electrode cropping edge and outputting the cropping edge contour include:

[0100] A sliding window is set on the continuous image of the electrode cutting edge to extract the gradient change amplitude of the local area of ​​the image segment by segment, and the edge direction is determined according to the direction of the maximum gradient amplitude.

[0101] According to the edge direction, the edge points in the continuous image of the electrode cutting edge are marked, and the marked edge points are connected in sequence to form a preliminary edge outline.

[0102] The initial edge contour is smoothed to eliminate jagged and burr-like interference in the edge contour, resulting in a smooth edge contour.

[0103] The smoothed edge contour is closed. If there is an unclosed region, a neighborhood search is performed on the endpoints of the unclosed region. The nearest other endpoint is selected and connected to complete the contour closure and form a complete clipping edge contour.

[0104] In the above process, by setting a sliding window on the continuous image of the pole piece cropping edge, the gradient change amplitude of the local area of ​​the image can be extracted segment by segment. The sliding window moves on the continuous image of the pole piece cropping edge according to a predetermined step size. At each window position, the gradient change amplitude of the image within the corresponding window is calculated. The gradient amplitude reflects the drastic change of pixel grayscale in the local area of ​​the image and can reflect edge information. The direction of the maximum gradient change amplitude can determine the edge direction, so as to guide the extraction direction of subsequent edge points. Then, according to the edge direction, the edge points in the continuous image of the pole piece cropping edge are marked, and the marked edge points are arranged according to their spatial order in the image. The edges are connected sequentially to form a preliminary edge contour. To make the contour smoother, the preliminary edge contour is smoothed using smoothing algorithms such as Gaussian filtering to eliminate jagged and burr-like interference. This results in a smooth edge contour. Then, a closure detection is performed on the smoothed edge contour to check for any unclosed regions. If an unclosed region exists, a neighborhood search is performed on the endpoints of the unclosed region. Within the neighborhood, the nearest other endpoint is selected and connected to complete the contour closure, ultimately forming a complete trimmed edge contour, providing the corresponding geometric information for subsequent burr detection.

[0105] S5. Based on the cut edge contour, identify the burr area, and perform quantitative analysis of the burr area by extracting the height, width and distribution density of the burr to determine the distribution location of the burr area and the burr defect level.

[0106] In step S5, after the trimmed edge contour is determined, the burr area can be identified. By extracting parameters such as the height, width, and distribution density of the burrs, a quantitative analysis of the burr area is performed to determine its specific location. Simultaneously, the burr defect level is assessed according to preset standards, providing a basis for subsequent burr repair. The steps of identifying the burr area based on the trimmed edge contour, extracting the height, width, and distribution density of the burrs, performing quantitative analysis to determine the distribution location and defect level of the burr area include:

[0107] Along the normal direction of the cut edge contour, the neighborhood area is scanned point by point at a set step interval. The local curvature change of the contour line in the neighborhood area is calculated, and protruding areas that exceed the normal range are identified as burr candidate points.

[0108] Geometric features are extracted from burr candidate points, including the height difference, width span, and curvature extrema within the candidate point's neighborhood.

[0109] The extracted height difference, width span, and curvature extreme values ​​are compared with the preset burr feature thresholds to filter out candidate points that meet the burr features and mark them as burr regions.

[0110] The distribution density of burrs is determined by calculating and outputting the number of burrs on the cut edge per unit length;

[0111] The distribution density, height difference, and width span of the burr area are weighted and scored to obtain a comprehensive burr defect score. The comprehensive burr defect score is then matched with a preset level standard to output the burr defect level.

[0112] Specifically, when identifying burr regions, the process first involves scanning the neighborhood point-by-point along the normal direction of the cut edge contour at pre-set step intervals. During this scanning, the local curvature change of the contour line within the neighborhood is calculated. This can be achieved by calculating the rate of change of the tangent angle between adjacent points or by using differential geometry to solve for the curvature value. Regions with abrupt curvature changes exceeding a set threshold are identified as burr candidate points. Then, geometric features are extracted from these candidate points, primarily focusing on the height difference within the candidate point's neighborhood (the vertical height difference between the candidate point and its surrounding area), the width span (the horizontal extension range of the burr), and the curvature extremum reflecting the degree of curvature in the burr shape. Finally, the extracted height difference, width span, and curvature extremum are compared with pre-set burr feature thresholds. If the geometric feature parameters of the above burr candidate points all meet the preset burr feature threshold range (the burr feature threshold range is determined by combining historical defect sample statistics and experimental verification to ensure coverage of the geometric characteristics of typical burrs), then the continuous area where the corresponding burr candidate points are located will be marked as a burr area. Then, by calculating and outputting the number of burrs on the unit length of the cut edge, the burr distribution density can be determined to reflect the density of burrs on the cut edge. Finally, the distribution density, height difference, and width span of the burr area are weighted and scored to obtain and output the comprehensive burr defect score. Finally, the comprehensive burr defect score is matched with the preset level standard (the level standard is set according to the product process requirements) to output the burr defect level, providing a corresponding judgment basis for subsequent burr repair work.

[0113] Please see Figure 2 A burr detection system for cutting edges of solid-state battery electrodes, using the aforementioned burr detection method for cutting edges of solid-state battery electrodes, includes:

[0114] The initialization module is used to arrange multiple rows of staggered detection points along the width of the battery electrode, and to collect image information of each detection point along the cutting edge of the battery electrode to form a reference image dataset.

[0115] The image filtering module is used to filter the detection point images in the benchmark image dataset for validity, identify valid and invalid acquired images, and mark the detection points corresponding to invalid acquired images and trigger the image compensation mechanism.

[0116] The image compensation module is used to compensate the detection points corresponding to invalid acquired images according to the image compensation mechanism, and to re-filter the compensated images until all detection point images are valid.

[0117] The contour extraction module is used to perform image stitching on the valid images to form a complete continuous image of the pole piece cropping edge, and to extract edge features from the continuous image of the pole piece cropping edge to output the cropping edge contour.

[0118] The burr recognition module is used to identify burr areas based on the cut edge contour. It performs quantitative analysis of the burr area by extracting the height, width and distribution density of the burrs, and determines the distribution location of the burr area and the burr defect level.

[0119] The execution process of the above detection system corresponds exactly to the process of the aforementioned method, so it will not be repeated here.

[0120] Please see Figure 3 An electronic device, comprising:

[0121] At least one processor;

[0122] and memory that is communicatively connected to at least one processor;

[0123] The memory stores a computer program that can be executed by at least one processor, which enables the at least one processor to perform the above-described method for detecting burrs on the cutting edges of solid-state battery electrode sheets.

[0124] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for detecting burrs on the cutting edges of solid-state battery electrode sheets, characterized in that: include: Multiple rows of staggered detection points are arranged along the width of the battery electrode, and image information of each detection point is collected along the cutting edge of the battery electrode to form a benchmark image dataset. The detection point images in the benchmark image dataset are filtered for validity, and valid and invalid images are identified. The detection points corresponding to invalid images are marked and the image compensation mechanism is triggered. The detection points corresponding to invalid acquired images are compensated according to the image compensation mechanism, and the compensated images are re-filtered for validity until all detection point images are valid. The valid images are stitched together to form a complete continuous image of the electrode cutting edge, and edge features are extracted from the continuous image of the electrode cutting edge to output the cutting edge contour. Burr region identification is performed based on the cut edge contour. The height, width and distribution density of the burrs are extracted to perform quantitative analysis of the burr region, and the distribution location and burr defect level of the burr region are determined. The step of compensating the detection points corresponding to invalid acquired images according to the image compensation mechanism includes: Spatial location of invalid acquisition areas in invalid acquisition images is determined, and spatial locations of adjacent upper and lower row detection points are matched. If valid images exist in the corresponding regions of adjacent upper and lower rows of detection points, interpolation compensation is performed using the image data from the corresponding regions of the adjacent rows of detection points to generate a completed image, which is then incorporated into the baseline image dataset. If there are invalid images in the corresponding areas of adjacent upper and lower rows of detection points, the invalid areas of the current detection point and the adjacent rows of detection points are jointly marked, and all invalid areas under the joint marking are simultaneously re-sampled.

2. The method for burr detection on the cutting edge of a solid-state battery electrode as described in claim 1, characterized in that: The step of forming the benchmark image dataset includes: Multiple rows of detection points are set along the width of the battery electrode, with the detection points in adjacent rows being staggered and the spacing between adjacent detection points in a single row being consistent. Each row of detection points acquires images synchronously from a perspective perpendicular to the cropping edge, and the acquisition frequency of each detection point remains consistent. The original images acquired at each detection point are preprocessed to generate standardized detection point images. The preprocessing steps include grayscale correction, noise filtering, and distortion correction. The preprocessed detection point images are encoded and stored according to the spatial location of the detection points, and a benchmark image dataset containing timestamps, location coordinates and image data is constructed.

3. The method for burr detection on the cutting edge of a solid-state battery electrode sheet according to claim 1, characterized in that: The step of filtering the detection point images in the benchmark image dataset to identify valid and invalid acquired images includes: The edge gradient magnitude of the detection point image is calculated, and the average edge gradient magnitude of each region in the same detection point image is calculated to obtain the sharpness index of the detection point image. The standard deviation of the gray value distribution in the detection point image is statistically analyzed and output as a uniformity index. A threshold judgment method is used to judge the sharpness index and uniformity index of the detection point image respectively. The threshold includes a first screening threshold corresponding to the sharpness index and a second screening threshold corresponding to the uniformity index. If the sharpness index of the detection point image is greater than the first screening threshold and the uniformity index is greater than the second screening threshold, the image is determined to be a valid acquisition image; otherwise, it is marked as an invalid acquisition image.

4. The method for burr detection on the cutting edge of a solid-state battery electrode sheet according to claim 1, characterized in that: The step of performing image stitching processing on the valid images to form a complete continuous image of the electrode cutting edge includes: Based on the spatial location encoding of the detection points, the effectively acquired images are aligned according to their coordinate order; The aligned and effectively acquired images are stitched together for the first time to generate a preliminary stitched image. The stitching seams and overlapping areas in the preliminary stitched image are detected, and the stitching deviation area is output. The stitching deviation area includes the insufficient seam area and the excessive overlap area. In areas with insufficient seams, missing pixels are supplemented by linear interpolation to form an interpolated complete area; In areas of excessive overlap, the sharpness and uniformity of the two valid acquired images corresponding to the overlapping area are weighted and fused, and the result is a fusion score. The valid acquired image with the higher fusion score is retained, while the corresponding area of ​​the valid acquired image with the lower fusion score is cropped and removed. Based on the selected valid acquired images and the interpolated complete area, the initial stitched images are stitched together again to generate a complete continuous image of the electrode cutting edge.

5. A method for detecting burrs on the cutting edge of a solid-state battery electrode as described in claim 4, characterized in that: The step of detecting seams and overlapping areas in the preliminary stitched image and outputting the stitching deviation area includes: The initial stitched image is subjected to grayscale normalization processing, and the grayscale gradient change amplitude of pixels within a set width on both sides of the stitching seam is extracted. When the magnitude of the pixel gradient change exceeds the gradient threshold and the number of consecutive pixels reaches the preset length, it is determined that there is a seam in the initial stitched image, and the area corresponding to the seam is marked as a seam insufficiency area. When the magnitude of pixel gradient change is lower than the gradient threshold, the similarity of the effective acquired images on both sides of the stitching area is calculated, and the area with similarity higher than the similarity threshold is determined as an over-overlapping area.

6. The method for burr detection on the cutting edge of a solid-state battery electrode according to claim 1, characterized in that: The step of extracting edge features from the continuous image of the electrode cutting edge and outputting the cutting edge contour includes: A sliding window is set on the continuous image of the electrode cutting edge to extract the gradient change amplitude of the local area of ​​the image segment by segment, and the edge direction is determined according to the direction of the maximum gradient amplitude. According to the edge direction, the edge points in the continuous image of the electrode cutting edge are marked, and the marked edge points are connected in sequence to form a preliminary edge outline. The initial edge contour is smoothed to eliminate jagged and burr-like interference in the edge contour, resulting in a smooth edge contour. The smoothed edge contour is closed. If there is an unclosed region, a neighborhood search is performed on the endpoints of the unclosed region. The nearest other endpoint is selected and connected to complete the contour closure and form a complete trimming edge contour.

7. The method for detecting burrs on the cutting edge of a solid-state battery electrode as described in claim 1, characterized in that: The step of identifying burr areas based on the cut edge contour, and quantitatively analyzing the burr areas by extracting the height, width, and distribution density of the burrs to determine the distribution location and burr defect level, includes: Along the normal direction of the cut edge contour, the neighborhood area is scanned point by point at a set step interval. The local curvature change of the contour line in the neighborhood area is calculated, and protruding areas that exceed the normal range are identified as burr candidate points. Geometric features are extracted from burr candidate points, including the height difference, width span, and curvature extrema within the candidate point's neighborhood. The extracted height difference, width span, and curvature extreme values ​​are compared with the preset burr feature thresholds to filter out candidate points that meet the burr features and mark them as burr regions. The distribution density of burrs is determined by calculating and outputting the number of burrs on the cut edge per unit length; The distribution density, height difference, and width span of the burr area are weighted and scored to obtain a comprehensive burr defect score. The comprehensive burr defect score is then matched with a preset level standard to output the burr defect level.

8. A burr detection system for cutting edges of solid-state battery electrode sheets, characterized in that: The burr detection method for cutting edges of solid-state battery electrode sheets according to any one of claims 1 to 7 includes: The initialization module is used to arrange multiple rows of staggered detection points along the width of the battery electrode, and to collect image information of each detection point along the cutting edge of the battery electrode to form a reference image dataset. The image filtering module is used to filter the detection point images in the benchmark image dataset for validity, identify valid and invalid acquired images, and mark the detection points corresponding to invalid acquired images and trigger the image compensation mechanism. The image compensation module is used to compensate the detection points corresponding to invalid acquired images according to the image compensation mechanism, and to re-filter the compensated images until all detection point images are valid. The contour extraction module is used to perform image stitching on the valid images to form a complete continuous image of the pole piece cropping edge, and to extract edge features from the continuous image of the pole piece cropping edge to output the cropping edge contour. The burr recognition module is used to identify burr areas based on the cut edge contour. It performs quantitative analysis of the burr area by extracting the height, width and distribution density of the burrs, and determines the distribution location of the burr area and the burr defect level.

9. An electronic device, characterized in that: The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform the burr detection method for cutting edges of solid-state battery electrode sheets as described in any one of claims 1 to 7.

Images