Battery cell tab folding defect detection method and system and storage medium

Abstract

The invention provides a battery cell tab turnover defect detection method, which comprises the following steps: acquiring a first image set and a second image set of a tab lamination of a battery cell, and fusing images in the first image set and images in the second image set respectively to obtain a first target image and a second target image; identifying a tab folding defect in the first target image and a folding defect in the second target image based on a trained neural network model; dividing the first target image and the second target image into respective corresponding partitions based on a preset region division rule; and obtaining a tab folding detection result according to the identified tab folding defect in the first target image and the identified tab folding defect in the second target image in combination with a preset partition caliper rule. The technical problem that in the prior art, the detection rate of tab folding defects is low is solved.

Description

Technical Field

[0001] This application relates to the field of battery testing technology, specifically to a method, system, and storage medium for detecting cell tab folding defects. Background Technology

[0002] Lithium-ion batteries are the core power unit for electric vehicles, energy storage systems, and consumer electronics products. Their manufacturing quality directly affects the performance, safety, and lifespan of the final product. The battery cell is the smallest unit that makes up a lithium-ion battery, and the tabs, as the key metal conductive electrodes that electrically connect the internal electrode materials of the cell to the external circuitry, are prone to defects such as folding and wrinkling during cell production due to their thinness and low strength. Therefore, it is necessary to detect tab folding defects to determine whether the battery cell is qualified.

[0003] For example, the invention patent application number 202410030247.1, entitled "Method and Device for Detecting Electrode Defects," discloses a method for detecting electrode defects by acquiring two side images of the electrode stack to identify folding defects. However, the inventors found that using only side images to identify folding defects can easily lead to missed detection results. Summary of the Invention

[0004] In view of this, this application provides a method, system and storage medium for detecting cell tab flipping defects, in order to solve the technical problem of low detection rate of tab flipping defects in the prior art.

[0005] In a first aspect, this application provides a method for detecting cell tab flipping defects, including: A first image set and a second image set of the electrode stack of the battery cell are obtained, and the images in the first image set and the images in the second image set are fused together to obtain a first target image and a second target image; The trained neural network model is used to identify the tab folding defect in the first target image and the folding defect in the second target image. Based on preset region division rules, the first target image and the second target image are divided into their respective corresponding partitions; Based on the identified tab folding defects in the first target image and the second target image, and combined with the preset partitioning rules, the tab folding detection results are obtained.

[0006] Secondly, this application provides a battery cell tab flipping defect detection system, comprising: The acquisition module is used to acquire a first image set and a second image set of the electrode stack of the battery cell, and fuse the images in the first image set and the images in the second image set to obtain a first target image and a second target image; The recognition module is used to identify the tab folding defect in the first target image and the folding defect in the second target image based on the trained neural network model. The partitioning module is used to divide the first target image and the second target image into corresponding partitions based on preset region division rules. The determination module is used to obtain the electrode folding detection result based on the electrode folding defects in the first target image and the second target image, combined with the preset partitioning rules.

[0007] Thirdly, this application provides a computer device, including: a memory and a processor, which are interconnected and communicate with each other. The memory stores computer instructions, and the processor executes the computer instructions to realize the method for detecting cell tab flipping defects of the first aspect.

[0008] Fourthly, this application provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the cell tab flipping defect detection method of the first aspect.

[0009] The battery cell tab folding defect detection method, system, and storage medium provided in this application have at least the following beneficial effects: By simultaneously acquiring and fusing images of the outer corners of the tab stack into a high-quality target image, the side and front surfaces of the tab stack are fully covered, significantly improving the completeness and accuracy of defect identification. This solves the problem of missed defects when relying solely on side images for detection, greatly reducing the false negative rate. The integration of a neural network model and zoning rules balances intelligent defect identification with standardized judgment, improving the accuracy and consistency of detection results. Combined with the zoning rule mechanism, defects of different degrees and locations are treated differently, avoiding false negatives or misjudgments, and enhancing the adaptability and reliability of the detection. Attached Figure Description

[0010] Figure 1 A schematic diagram of one image acquisition module setup is shown; Figure 2 A schematic diagram of a method for detecting cell tab flipping defects is shown; Figure 3 A schematic diagram showing a list of labels for a defect belonging to the folding type is provided. Figure 4 A schematic diagram of partitioning a target image is shown; Figure 5 A schematic diagram of a battery cell tab flipping defect detection system is shown. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] Example 1 See Figure 1 As shown, two adjustable-focus image acquisition modules 101 are positioned to be aimed at the two outer corners of the electrode stack 103 of the battery cell 102, respectively, so that the acquired image simultaneously includes both the side and front sides of the electrode stack. It should be noted that... Figure 1 Only one tab is shown, but the battery cell has two tabs, a cathode and an anode. The other tab, though not shown, should be understood in the same way.

[0013] See Figure 2 As shown, this application provides a method for detecting cell tab flipping defects, which includes the following steps.

[0014] S201. Obtain a first image set and a second image set of the electrode stack of the battery cell, and fuse the images in the first image set and the images in the second image set respectively to obtain a first target image and a second target image; S202. Identify the tab folding defect in the first target image and the folding defect in the second target image based on the trained neural network model; S203. Based on preset region division rules, divide the first target image and the second target image into their respective corresponding partitions; S204. Based on the identified tab folding defects in the first target image and the second target image, and combined with the preset partitioning rules, the tab folding detection result is obtained.

[0015] In this application, the first and second image sets are acquired using a camera with adjustable focal length, such as a liquid lens, aimed at the two outer corners of the tab stack. Since the lens is aimed at the outer corners of the tab stack, the depth of field from the outer corners of the tab stack to the lens, the depth of field from different positions on the sides of the tab stack to the lens, and the depth of field from different positions on the front side of the tab stack to the lens are different. If a fixed-focus camera is used, the acquired images will not clearly show the outer corners, inner corners, sides, and front side of the tab stack due to the different depths of field. Therefore, an adjustable-focus camera is used to acquire images at different focal lengths, and image fusion processing is performed to ensure that the fused image clearly shows the outer corners, inner corners, sides, and front side of the tab stack. A trained neural network model is used to analyze the first and second target images respectively, accurately identifying defects belonging to the tab folding type and extracting core information related to the defects. The labels corresponding to defects belonging to the folding type are as follows: Figure 3 As shown, according to the preset region division rules, dedicated partitions (such as strong detection area, standard control area, shielded area, etc.) are defined for the two target images, clarifying the detection priority and judgment criteria of different regions, laying the foundation for subsequent accurate judgment. Combining the defect information identified by the neural network with the preset partition standard control rules, a comprehensive judgment is made on whether there is a folding defect in the tab stack, and the final detection result is output.

[0016] In this application, by simultaneously acquiring and fusing images of the outer corners of the tab stack into a high-quality target image, the side and front surfaces of the tab stack are fully covered, significantly improving the completeness and accuracy of defect identification. This solves the problem of missed defects in traditional methods that rely solely on side images, greatly reducing the false negative rate. The integration of a neural network model and zoning rules balances intelligent defect identification with standardized judgment, improving the accuracy and consistency of detection results. Furthermore, the zoning rule mechanism differentiates defects of different degrees and locations, avoiding missed detections or misjudgments and enhancing the adaptability and reliability of the detection process.

[0017] Furthermore, if the first target image includes the left side and the front side of the tab stack, then the second target image includes the right side and the front side of the tab stack. If the first target image contains the right side and the front side of the tab stack, then the second target image contains the left side and the front side of the tab stack.

[0018] The viewpoint correspondence between the first and second target images is clearly defined, with two complementary scenarios: In the first scenario, the first target image covers the left and front sides of the tab stack, while the second target image covers the right and front sides; in the second scenario, the first target image covers the right and front sides, while the second target image covers the left and front sides. By fixing the viewpoint coverage of the two target images, it is ensured that there are no blind spots on the left, right, and front sides of the tab stack, and all critical areas where folding defects may occur are captured by the images.

[0019] Clearly define mandatory coverage requirements for image viewing angles to prevent defects from going undetected due to confusing or omitted perspectives, thereby further reducing the risk of missed detections. Standardize the viewing angle rules for image acquisition to facilitate subsequent image fusion, partitioning, and defect identification and judgment, improving detection stability and repeatability.

[0020] It should be noted that this application uses the first target image and the second target image as the basis for defect detection. Of course, the first and second target images can also be further fused and stitched to obtain a complete panoramic image including the two sides and the front side of the tab stack, which can then be used as the basis for defect detection and judgment. Subsequent partitioning can also be adapted accordingly. In short, the method of identifying folding defects using panoramic images can be derived from the basic deformation of identifying folding defects using two images, and it should also be included in the protection scope of this application, which will not be elaborated further. In this regard, clearly defining the mandatory coverage requirement of the image viewpoint ensures the integrity of information on the left and right sides and the front side, providing data support for the comprehensive judgment of cross-viewpoint defects.

[0021] Furthermore, based on preset region division rules, the first target image and the second target image are divided into their respective corresponding partitions, including: Identify the outer and inner angles of the tab stack in the image, and generate the outer and inner angle lines; The region where the inner corner line is close to the outer corner line is defined as the strong detection zone, based on a first preset threshold. The area near the outer corner of the strong inspection zone with a second preset threshold is defined as the first caliper zone. The area between the third and fourth preset thresholds in the direction away from the inner corner line is defined as the second caliper area; The area outside the mandatory inspection area, the first inspection area, and the second inspection area is designated as the shielded area.

[0022] In this application, the outer and inner angles of the electrode tabs in the image are first identified, and corresponding angle lines are generated. Then, based on these angle lines, a strong detection zone, a first calibration zone, a second calibration zone, and a shielding zone are defined. Specifically, the first preset threshold area near the outer angle line is the strong detection zone; the second preset threshold area near the outer angle line is the first calibration zone; the third to fourth preset threshold areas extending outward from the outer angle line are the second calibration zone; and the remaining areas are the shielding zone. See also... Figure 4 As shown, line 3 on the left is the outer corner line, line 1 on the right is the inner corner line, the area between line 1 and line 2 on the right is the strong detection area, the area between line 2 and line 3 on the right is the first calibrator area, the area between line 1 on the left and line 2 on the left is the second calibrator area, and the area between line 2 on the left and line 4 on the left is the shielded area. Of course, the area to the left of line 1 and the area to the right of line 1 are also shielded areas. It should be noted that the area to the left of line 1 is designated as the shielded area because a corresponding second calibrator area will be defined in another target image, and no duplicate detection is performed in this image. Furthermore, it should be noted that the first preset threshold is usually set to one-quarter of the length of the left and right sides, the third preset threshold is one-eighth of the length of the front side, and the fourth preset threshold is one-half of the length of the front side. For example, D(Right 2 line - Right 1 line) = D(Right 3 line - Right 2 line) = 1 / 2 D(Left 3 line - Right 3 line) = 1 / 2 D(Left 1 line - Left 3 line) = 1 / 4 D(Left 3 line - Right 1 line) = 2(Left 2 line - Left 3 line). In practical applications, the first, second, third, and fourth preset thresholds can be determined based on the actual electrode size and detection requirements.

[0023] Dividing the image into functionally defined regions facilitates differentiated detection strategies for different areas. The high-detection zone corresponds to the critical locations where defects are most likely to occur in the tab structure, improving the targeting and sensitivity of the detection. Simply put, folding curves near the root area can cause the tab to fold into the cell; folding defects detected in the high-detection zone cannot be missed. Defects in the shielding area, however, can be ignored because the shielding is further trimmed during subsequent battery manufacturing processes, so folding defects in the shielding area are tolerable. By setting up caliper zones and shielding zones, the false alarm rate can be reasonably controlled while ensuring detection accuracy, reducing invalid calculations, and balancing detection precision and speed.

[0024] Furthermore, the identified tab folding defects in the first target image indicate: defects belonging to the folding type in the first target image, as well as the confidence information, size data, and location information of the defects; The identified tab folding defect in the second target image indicates: the folding type defect in the second target image, along with the defect's confidence level, size data, and location information.

[0025] In this application, regardless of whether it is the first target image or the second target image, the recognition result contains four types of key information: Defect type: Determine whether it is a folding defect, exclude other irrelevant defects, and determine it based on the label of the detected defect; Confidence information: reflects the reliability of defect identification results and provides a basis for classification judgment; Size data: Quantifies the size of defects (folding range, extent, etc.) to support quantitative judgment; Location information: The specific coordinates of the defect in the target image, used to match the corresponding partition.

[0026] Outputting multi-dimensional defect information facilitates subsequent logical judgment and result analysis, providing complete data support for multi-dimensional judgments based on confidence level, region, and size, and avoiding misjudgments due to missing information. Clearly defined defect type screening criteria focus on folding defects, reducing interference from non-target defects on detection results. Quantitative output of confidence level, size, and location information enables multi-level judgments, improving the accuracy and adaptability of detection.

[0027] Furthermore, based on the preset partition card rules, the electrode folding detection results are obtained, including: Determine whether the confidence level of the identified defect is greater than or equal to the first preset confidence threshold; if so, determine it as an electrode tab fold defect. Otherwise, determine whether the confidence level of the identified defect is greater than or equal to the second preset confidence threshold; if not, filter out the identified defect. Otherwise, the target area is determined based on the location information of the identified defect, the corresponding card specification rules are determined, and a judgment result is generated by combining the size data of the identified defect.

[0028] A confidence-based hierarchical judgment logic has been established, which is a key process for defect screening and accurate judgment: Level 1 Judgment: If the defect confidence level is ≥ the first preset confidence threshold (high confidence level), it is directly judged as a tab fold defect; Second-level judgment: If the defect confidence level is less than the first preset confidence threshold, further determine whether it is greater than or equal to the second preset confidence threshold (medium confidence); if it is less than the second threshold (low confidence), filter the defect directly; Level 3 Judgment: If the defect confidence level is between the first and second thresholds (medium confidence level), the target area to which it belongs is determined based on the defect location information, the corresponding card rule of the partition is matched, and the final judgment result is generated by combining the defect size data.

[0029] Preliminary defect screening is achieved through confidence level grading, filtering out invalid information with low confidence to reduce false positives, while retaining defects with medium confidence for precise review, thus balancing detection sensitivity and accuracy. A multi-dimensional judgment logic based on confidence level, region, and size is established to avoid the limitations of single-dimensional judgment and improve the scientific rigor of detection results. The grading process is simple and clear, adapting to rapid detection needs, ensuring accuracy without sacrificing detection efficiency.

[0030] Furthermore, based on the location information of the identified defects, the target area is determined, the corresponding caliper rules are determined, and a judgment result is generated by combining the size data of the identified defects, including: If the target area is a high-inspection area, it is determined to be a tab folding defect; If the target area is the first gauge area, the size data of the identified defect will be compared with the preset first gauge threshold to generate a judgment result; If the target area is the second gauge area, the size data of the identified defect will be compared with the preset second gauge threshold to generate a judgment result; If the target area is a shielded area, the identified defects will be filtered out.

[0031] Specifically, the determination of the strong inspection area: if the defect is located in the strong inspection area, regardless of its size, it is directly determined to be a tab folding defect; First gauge zone judgment: If the defect is located in the first gauge zone, compare the defect size data with the preset first gauge threshold. If it exceeds the threshold, it is judged as a defect; otherwise, it is judged as qualified. Second gauge zone judgment: If the defect is located in the second gauge zone, the defect size data is compared with the preset second gauge threshold, and a qualified / unqualified defect judgment result is generated according to the comparison result; Shielding zone determination: If the defect is located in the shielding zone, the defect is directly filtered out and not included in the final determination result.

[0032] This system achieves differentiated quantitative judgment by region, prioritizing high-risk defects in the high-inspection zone, using threshold quantification in the standard-gauge zone for precise judgment, and eliminating irrelevant interference in the shielded zone, further improving testing accuracy. The presettable thresholds of the standard-gauge zones adapt to the testing of battery cell tabs of different specifications and quality requirements, enhancing versatility and flexibility. Clearly defined judgment criteria for each region avoid subjective human judgment, ensuring consistency and impartiality of test results.

[0033] Furthermore, it also includes: If the identified defect covers the strong detection area and other areas outside the strong detection area, it is judged as a tab folding defect; If the identified defect does not cover the mandatory inspection area, but covers the shielded area and the first or second gauge area, then only the dimensional data of the identified defect within the first or second gauge area is retained, and a judgment result is generated by combining the first or second gauge threshold.

[0034] Special judgment rules for defects distributed across regions have been added, improving the overall judgment logic: Judgment across mandatory inspection areas: If a defect covers both the mandatory inspection area and other areas (first gauge area, second gauge area, shielding area), it is directly judged as a tab folding defect; Judgment across shielded and gauge zones: If a defect does not cover the mandatory inspection zone, but covers both the shielded zone and the first / second gauge zone, only the dimensional data of the defect within the first / second gauge zone is retained, invalid data within the shielded zone is removed, and then the judgment result is generated by combining the corresponding gauge threshold.

[0035] This solution addresses the challenge of identifying cross-regional defects, preventing missed or misjudged defects due to regional divisions and improving the completeness of the judgment logic. Data screening is performed on defects crossing shielded and gauge zones to remove invalid interference data, ensuring judgments are based on valid dimensional information and improving accuracy. The high-priority judgment principle for high-risk areas is further strengthened to ensure no cross-regional defects related to high-risk areas are overlooked, safeguarding cell quality and safety. Only defects within the valid area are evaluated to prevent overall misjudgment due to interference from non-critical areas.

[0036] Furthermore, it should be noted that the positional height of defects in the two target images can be compared to determine whether corresponding defects belong to the same electrode tab. For defects belonging to the same electrode tab, the reported points can be merged to avoid excessive reporting, which would affect the actual detection accuracy. Of course, confirming whether related folding defects belong to the same electrode tab based on a complete panoramic image obtained by stitching together the first and second target images would be more efficient. Although not specifically described in this application, the point merging scheme derived from the scheme of this application should also be included within the scope of protection of this application.

[0037] Example 2 This embodiment provides a battery cell tab folding defect detection system, which is applied to the battery cell tab folding defect detection method provided in Embodiment 1 above for illustration. See also Figure 5 As shown, the battery cell tab flipping defect detection system provided in this application may include the following multiple modules.

[0038] The acquisition module is used to acquire a first image set and a second image set of the electrode stack of the battery cell, and fuse the images in the first image set and the images in the second image set to obtain a first target image and a second target image; The recognition module is used to identify the tab folding defect in the first target image and the folding defect in the second target image based on the trained neural network model. The partitioning module is used to divide the first target image and the second target image into corresponding partitions based on preset region division rules. The determination module is used to obtain the electrode folding detection result based on the electrode folding defects in the first target image and the second target image, combined with the preset partitioning rules.

[0039] The battery cell tab flipping defect detection system provided in this application embodiment can be applied to the battery cell tab flipping defect detection method provided in Embodiment 1 above. For relevant details, please refer to the above method embodiment. Its implementation principle and technical effect are similar, and will not be repeated here.

[0040] It should be noted that the battery tab folding defect detection system provided in this embodiment is only illustrated by the above-mentioned division of functional modules / units when performing battery tab folding defect detection. In practical applications, the above functions can be assigned to different functional modules / units as needed, that is, the internal structure of the battery tab folding defect detection system can be divided into different functional modules / units to complete all or part of the functions described above. In addition, the implementation method of the battery tab folding defect detection method provided in the above-mentioned method embodiment 1 and the implementation method of the battery tab folding defect detection system provided in this embodiment 2 belong to the same concept. The specific implementation process of the battery tab folding defect detection system provided in this embodiment 2 is detailed in the above-mentioned method embodiment 1, and will not be repeated here.

[0041] Example 3 One embodiment of this application also provides a computer device, including but not limited to a processor and a memory, wherein the memory stores computer instructions, and the processor executes the computer instructions to implement the method described above.

[0042] One embodiment of this application also provides a computer-readable storage medium for storing a computer program that, when executed by a processor, implements the method described in the above-described method embodiments.

Claims

1. A method for detecting cell tab flipping defects, characterized in that, include: A first image set and a second image set of the electrode stack of the battery cell are obtained, and the images in the first image set and the images in the second image set are fused together to obtain a first target image and a second target image; The trained neural network model is used to identify the tab folding defect in the first target image and the folding defect in the second target image. Based on preset region division rules, the first target image and the second target image are divided into their respective corresponding partitions; Based on the identified tab folding defects in the first target image and the second target image, and combined with the preset partitioning rules, the tab folding detection result is obtained.

2. The method for detecting cell tab flipping defects according to claim 1, characterized in that: If the first target image includes the left side and the front side of the tab stack, then the second target image includes the right side and the front side of the tab stack. If the first target image includes the right side and front side of the tab stack, then the second target image includes the left side and front side of the tab stack.

3. The method for detecting cell tab folding defects according to claim 2, characterized in that, The method of dividing the first target image and the second target image into corresponding partitions based on preset region division rules includes: Identify the outer and inner angles of the electrode stack in the image, and generate the outer and inner angle lines; The region where the inner corner line is close to the outer corner line is defined as a strong detection zone, based on a first preset threshold. The region of the strong detection zone that is close to the second preset threshold in the direction of the outer corner line is defined as the first caliper zone; The area between a third preset threshold and a fourth preset threshold in the direction in which the outer corner line is away from the inner corner line is defined as the second gauge zone; The area outside the mandatory inspection area, the first gauge area, and the second gauge area is designated as the shielding area.

4. The method for detecting cell tab flipping defects according to claim 3, characterized in that: The identified tab folding defect in the first target image indicates: the folding type defect in the first target image, as well as the confidence information, size data and location information of the defect; The identified tab folding defect in the second target image indicates: the folding type defect in the second target image, along with the defect's confidence level, size data, and location information.

5. The method for detecting cell tab folding defects according to claim 4, characterized in that, The process of combining preset partitioning rules to obtain electrode folding detection results includes: Determine whether the confidence level of the identified defect is greater than or equal to the first preset confidence threshold; if so, determine it as an electrode tab fold defect. Otherwise, determine whether the confidence level of the identified defect is greater than or equal to the second preset confidence threshold; if not, filter the identified defect. Otherwise, the target area is determined based on the location information of the identified defect, the corresponding card size rule is determined, and a judgment result is generated by combining the size data of the identified defect.

6. The method for detecting cell tab folding defects according to claim 5, characterized in that, The step of determining the target area based on the location information of the identified defect, determining the corresponding caliper rule, and generating a judgment result by combining the size data of the identified defect includes: If the target area is the strong detection area, it is determined to be an electrode tab folding defect; If the target area is the first gauge area, the size data of the identified defect is compared with the preset first gauge threshold to generate a judgment result; If the target area is the second gauge area, the size data of the identified defect is compared with the preset second gauge threshold to generate a judgment result; If the target area is the shielded area, then the identified defects are filtered out.

7. The method for detecting cell tab folding defects according to claim 6, characterized in that, Also includes: If the identified defect covers the strong detection area and other areas outside the strong detection area, it is determined to be a tab folding defect; If the identified defect does not cover the mandatory inspection area, but covers the shielded area and the first gauge area or the second gauge area, then only the size data of the identified defect within the first gauge area or the second gauge area is retained, and a judgment result is generated by combining the first gauge threshold or the second gauge threshold.

8. A battery cell tab flipping defect detection system, characterized in that, include: The acquisition module is used to acquire a first image set and a second image set of the electrode stack of the battery cell, and fuse the images in the first image set and the images in the second image set to obtain a first target image and a second target image; The identification module is used to identify the tab folding defect in the first target image and the folding defect in the second target image based on the trained neural network model; The partitioning module is used to divide the first target image and the second target image into corresponding partitions based on preset region division rules. The determination module is used to obtain the electrode folding detection result based on the electrode folding defects in the first target image and the second target image, combined with the preset partitioning rules.

9. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to implement the cell tab flipping defect detection method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when executed by a processor, implement the cell tab flipping defect detection method as described in any one of claims 1-7.

Images